Direct Current Transmission

Other Books by the Author
DIRECT CURRENT
TRANSMISSION
POWER SYSTEM STABILITY
,Volume i Elements of Stability Calculations, 1948, John Wiley & Sons,
Volume
Inc.
Volume
Power Circuit Breakers and Protective Relays, 1950, John

Wiley & Sons, Inc.
Volume
1956, John Wiley & Sons, Inc.,

(republished by Dover Publications, Inc., 1967)
ii
ni
EDWARD WILSON KIMBARK, Sc.
Synchronous Machines,
fI:
Bonneville Power AdministrationFellow I.E.E.E.

Portland, Oregon
ELECTRICAL TRANSMISSION OF POWER AND SIGNALS, 1949,

John Wiley & Sons, Inc. (Also published for the Asiatic market by
Toppan Co., Ltd., Tokyo, Japan, 1964)
f.
WILEY-INTERSCIENCE
a Division of
John.Wiley &
Inc.
New York London SydneySons,
Toronto
PREFACE
s
'
'
Copyright
1971 by John Wiley & Sons, Inc.
All rights reserved. Published simultaneously

in Canada.
j
s
Reproduction or translation of any part of this work beyond that

permitted by Sections 107 or 108 of the 1976 United States Copy
right Act without the permission of the copyright owner is unlaw
ful. Requests for permission or further information should be
addressed to the Permissions Department, John Wiley & Sons, Inc.
Library of Congress Catalog Card Number: 72-142717

ISBN 0-471-47580-7
1
Printed in the United States of America
10 9 8 7 6 5
The most exciting new technical development in electric power systems in

the last two decades is direct-current transmission. From 1950 to 1970, eight
direct-current links have gone into commercial operation in various parts of
the world. From the first of these links to the last, the voltage has increased
from 100 to 800 kV; the rated power, from 20 to 1440 MW; and the distance
from 96 to 1370 km (60 to 850 miles). Several other dc links are under construction or proposed.
Preceding and accompanying this rapid growth of direct current transmis
sion were developments in high-voltage, high-power valves, in control and
protective systems, in dc cables, and in insulation for overhead dc lines.
Industrial, governmental, and academic laboratories were involved in this
development. Dc transmission became a favored subject for research by
graduate students of electrical engineering.
The circumstances leading to the adoption of direct-current transmission
are diverse: long water crossings requiring submarine cables, frequency
changing, asynchronous operation of systems having the same nominal
frequency, large hydroelectric resources remote from load centers, long interregional ties, and transmission through congested metropolitan areas.
The rapid growth of dc transmission, combined with the diversity of reasons
for its use, assures for it a brilliant future and also points to the need for a new
and better book on the subject.
The art of dc transmission in the past two decades has been based on the
use of improved mercury arc valves. Consequently, this book, in endeavoring
to describe the present state of the art, is necessarily based largely on the
technology employing such valves. There are indications that mercury arc
valves have reached almost the peak of their development. At least, solidstate controllable valves (thyristors), though not yet used in any major dc
transmission project, are appearing as formidable contenders for future pro
jects. Fortunately, the technology of employing thyristors for dc transmission
differs more in dimensions than in principles from that of employing mercury
arc valves. Hence, it is not primarily the transition to thyristors that will in
:-1
:n
i
VI
PREFACE
time render this book obsolete, but rather the continuing rapid development
of all phases of the art.
My interest in direct current transmission was awakened in 1962 when the
Bonneville Power Administration (BPA) asked me to teach two courses in
this subject for their engineers. The lack of an adequate text book for those
courses determined me to write one. Since 1962, Ihave taught three more
courses on dc transmission, and have been involved in studies of various
aspects of this subject. Iam indebted to the BPA for the opportunity to teach
these courses, to work on problems in the field, and to confer with my col
leagues, as well as for access to the BPA's excellent library services. However,
Iwish to make clear that this book is not an official publication of the BPA
nor one sponsored by it. It has been a spare-time project. Ialone am respon
sible for its contents, including any errors which may inadvertently appear in
it.
The large amount of essential information now available on direct-current
transmission and the time required to organize it led to the decision to divide
the work into two volumes of which this is the first. The proposed contents of
the second volume are indicated on page xi.
Units of physical quantities used herein are those of the International System
(SI) recommended by the I.E.E.E. and I.E.C.
Iam indebted to various engineers at the BPA and elsewhere for supplying
information, especially to Dr. John J. Vithayathil for many enlightening
technical discussions. Iam indebted to my wife, Ruth Merrick Kimbark, for
typewriting much of the manuscript and pertinent correspondence and for her
valued advice and encouragement.
CONTENTS
1. GENERAL ASPECTS OF DC TRANSMISSION AND COMPARISON

OF IT WITH AC TRANSMISSION
1
1-1
1-2
1-3
1-4
1-5
1-6
Historical Sketch
Constitution of EHV AC and DC Links
Kinds of DC Links
HV DC Projects from 1954 to 1970
Limitations and Advantages of AC and DC Transmission
Summary of Advantages and Disadvantages of HV DC
Transmission
Principal Applications of DC Transmission
Economic Factors
The Future of DC Transmission
1-7
1-8
1-9
Bibliography
1
9
11
12
19
32
32
33
35
36 .
Edward Wilson Kimbark

Portland, Oregon
March, 1971
2. CONVERTER CIRCUITS
49
Valve Characteristic
Properties of Converter Circuits
Assumptions
Single-Phase Converters
Three-Phase Converters
Pulse Number
2-7 Additional Six-Pulse Converter Circuits
2-8 Choice of Best Circuit for HV DC Converters
2-9 Twelve-Pulse Cascade of Two Bridges
Problems
Bibliography
2-1
2-2
2-3
2-4
2-5
2-6
vii
,
/
49
50
51
51
56
61
62
65
67
68
70
Viii
CONTENTS
3. ANALYSIS OF THE BRIDGE CONVERTER
Analysis with Grid Control but no Overlap

Analysis with Grid Control and with Overlap less than 60
3-3 Analysis with Overlap Greater than 60
3-4 Complete Characteristics of Rectifier
3-5 Inversion
3-6 Series and Parallel Arrangements of Valves, Anodes, or
Bridges
3-7 Multibridge Converters
Problems
Bibliography
3-1
3-2
4. CONVERTER CHARTS
4-1
4-2
4-3
Chart 1with Rectangular Co-ordinates of Direct Current and

Voltage
Chart 2 with Rectangular Co-ordinates of Active and Re
active Power
Relations between the Two Charts
Problems
Bibliography
5. CONTROL
5-1
5-2
5-3
5-4
5-5
5-6
5-7
5-8
Grid Control
Basic Means of Control
Power Reversal
Limitations of Manual Control
Constant Current versus Constant Voltage
Desired Features of Control
Actual Control Characteristics
Constant-Minimum-Ignition-Angle Control
Constant-Current Control
Constant-Extinction-Angle Control
Stability of Control
Tap-Changer Control
Power Control and Current limits
Frequency Control
5-9
5-10
5-11
5-12
5-13
5-14
5-15 Multiterminal Lines
CONTENTS
71
73
80
92
103
105
112
115
123
126
129
129
138
146
146
147
148
148
152
153
154
156
157
158
5-16 Measuring Devices
187
192
194
Problems
Bibliography
198
6. MISOPERATION OF CONVERTERS
Malfunctions of Mercury-Arc Valves

Bypass Valves
Arcback
Short Circuit on a Rectifier
Commutation Failure
Arcthrough
6-7 Misfire
6-8 Quenching
6-9 Generalization of Inverter Faults and Certain Rectifier Faults
,
6-10 Consequential Faults in Rectifier
Problems
Bibliography
6-1
6-2
6-3
6-4
6-5
6-6
7. PROTECTION
7-1 General
7-2 DC Reactors
7-3 Voltage Oscillations and Valve Dampers
7-4 Current Oscillations and Anode Dampers
7-5 DC Line Oscillations and Line Dampers
7-6 Clearing Line Faults and Reenergizing the Line
7-7 Circuit Breakers
7-8 Overvoltage Protection
Problems
Bibliography
IX
198
199
206
220
222
227
228
229
230
231
233
234
235
235
235
247
260
270
272
280
282
291
292
164
165
167
174
179
180
182
183
8. HARMONICS AND FILTERS
8-1
8-2
8-3
8-4
8-5
Summary
Characteristic Harmonics
Uncharacteristic Harmonics
Troubles Caused by Harmonics
Definitions of Wave Distortion or Ripple
295
295
296
318
323
325
CONTENTS
8-6 Means of Reducing Harmonics

8-7 Telephone Interference
8-8 Harmonic Filters
Problems
Bibliography
332
333
343
386
386
TENTATIVE CONTENTS OF VOLUME II

9. GROUND RETURN
391
Advantages and Problems

The Current Field in the Earth Near an Electrode
The Current Field between the Electrodes
The Natural Current Field in the Earth
Compatability with Other Services
Design of Electrodes General
Design of Land Electrodes
Design of Sea and Shore Electrodes
Problems
Bibliography
391
393
417
419
423
443
445
465
476
478
APPENDICES
484
9-1
9-2
9-3
9-4
9-5
9-6
9-7
9-8
10. OVERHEAD LINES

i
11. DC CABLES
12. FORCED COMMUTATION
13. OPERATION OF A DC LINK AS PART OF AN AC SYSTEM
14. HIGH-POWER VALVES

15. CONVERTER TRANSFORMERS AND REACTORS
16. RADIO INTERFERENCE
A. Effective Value of Alternating Current of a Six-pulse Converter 484

B. Fundamental Current, Power, and Reactive Power of a Six-pulse
Converter
490
C. Inclusion of Direct Voltage Drops Due to Resistance and Arcs in
Converter Equations
494
INDEX
17. ASYNCHRONOUS TIES

18. MODELS AND SIMULATION
496
i.'
xi
ABBREVIATIONS
A
ac
A.C.S.R.
A.E.G.
A.G.
Ah
Amer. Power
Conf. Proc.
ASEA
Assn.
AWG
B.E. & A.I.R.A.
B.I.C.C.
B.I.L.
BPA
B.T.S.
Bull.
C
CAB
cal/g
CAT
C.C.
C.C.I.F.
ampere
alternating-current
aluminum cable, steel reinforced
Allgemeine Elektricitatsgesellschaft
Aktiengesellschaft
ampere-hour
American Power Conference Proceedings (Illinois
Institute of Technology, Chicago)
Allmanna Svenska Elektriska Aktiebolaget,
Sweden
Association
American Wire Gage
British Electrical and Allied Industries Re
search Association (later known as E.R.A.)
British Insulated Callenders Cables
basic insulation level
Bonneville Power Administration (U.S. Dep't.
of the Interior, Portland, Oregon)
Bell Telephone System
Bulletin
coulomb, Celsius (temperature scale, formerly
Centigrade)
consequential arcback
calorie per gram
consequential arcthrough
constant current
Comite Consultatif International T616phonique

(International Consultative Committee on Tele
phony).
Xlll
XIV
ABBREVIATIONS
ABBREVIATIONS
C.CXT.
C.C.I.T.T.
C.E.A.
C.E.G.B.
C.G.E.
C.I.G.R.E.
cm
Conf.
const.
cos
cosh
cot
coth
CP
Comite Consultatif International T61egraphique

(International Consultative Committee on Tele
graphy)
Comite Consultatif International T61ephonique
et T61egraphique (International Consultative
Committee on Telephony and Telegraphy),
Geneva, Switzerland
constant extinction angle
Central Electricity Generating Board, Great
Britain
Compagnie Generale d'Electricite, France
Conference Internationale des Grands Reseaux
Electriques a Haute Tension (International
Conference on Large High-Voltage Electrical
Systems), Paris
centimetre
Conference
constant
cosine
hyperbolic cosine
cotangent
hyperbolic cotangent
Elek. Stantsii
U.S.S.R.
electromotive force
Engg.
Engineering
Eq.
Eqs.
equation
equations
Electrical Research Association, Great Britain
E.R.A.
ETZ
E.u.M.
exp
F
Fe
Fe+ +
Fe(OH)2
ft
Gen. Elec. Rev.
GW
H
H
H+
H2
h
HV
csc
cosecant
dB
dc
deg.
Disc.
decibel
direct-current
degree (of angle)
discussion
free electron
Edison Electric Institute, New York
extra high voltage
Electrical
electrical degree
Electrical Engineering, formerly published by the
A.I.E.E.
Electrical World
Hz
ibid.
e"
ehv
Elec.
elec. deg.
Elec. Eng.
Elec. World
Elektricheskie Stantsii (Electric Powerplants),
EMF
Conference Paper (A.I.E.E. or I.E.E.E.)
E.E.I.
XV
Id.
I.E.C.
I.E.E.
I.E.E.E.
J
Jour.
K
kA
kg
Elektrotechnische Zeitschrift.
Elektrotechnik und Maschinenbau (Vienna)
exponential function
farad
iron atom
ferric ion
ferric hydroxide
feet
General Electric Review
gigawatt
henry
hydrogen atom
univalent positive hydrogen ion

hydrogen molecule
hour
high-voltage
hertz
Latin for "in the same place"
Island
International Electrotechnical Commission
Institution of Electrical Engineers (London)
Institute of Electrical and Electronic Engineers
(New York, U.S.A.), founded in January, 1964,
by merger of the A.I.E.E. and the I.R.E.
joule
Journal
Kelvin (temperature scale)
kiloampere
kilogram
1
Xvi
ABBREVIATIONS
kHz
km
kV
kVA
kvar
kW
k$
lb
lbf/in2
LC
In
log
mA
MCM
mH
MHD
MHz
mi
mm
MMF
ms
mV
MVA
ABBREVIATIONS
kilohertz
kilometre
kilovolt
rad/s
Ref.
pounds force per square inch

inductance-capacitance
natural logarithm
common logarithm
Rev.
Review
RLC
rms
resistance-inductance-capacitance
SCR
milliampere
thousands of circular- mils
millihenry
sec
S.E.V.
magnetohydrodynamic(s)
megahertz
mile
millimetre
magnetomotive force
S.I.L.
root-mean-square
silicon controlled rectifier

second (time), secant
Schweizerischer Elektrotechnischer Verein, also
known as Association Suisse des Electriciens
(Zurich)
surge-impedance loading
sin
sinh
SW
hyperbolic sine
tan
tangent
T.H.F.F.
TIF
Trans.
Trans, and Dist.
U.S.
U.S.A.
U.S.S.R.
V
telephone harmonic form factor

telephone influence factor
vs
versus
number
northwest
New Zealand
negative hydroxyl ion
W
w.r.t.
yd
with respect to
yr
year
Ontario
pP
microfarad
millisecond
millivolt
megavolt-ampere
MW
N
nF
nanofarad
newton
Nauchno-Izsledovatel'skii
Institut Postoyannovo
Toka, Izvestiya (Proceedings of the Direct

Current Research Institute), Leningrad.
NW
peak inverse voltage

Proceedings
publication
polyvinyl chloride
radian
radian per second
reference
rad
thousands of dollars
pound
megavar
N.Z.
OH"
Ont.
P.I.V.
Publ.
PVC
kilovar
kilowatt
megawatt
No.
Power Apparatus and Systems
Proc.
kilovolt-ampere
Mvar
N.I.I.P.T.
PA. &S.
XVII
sine
Southwest
Transactions
Transmission and Distribution
United States
United States of America

Union of Soviet Socialist Republics
volt
watt
yard
xviii
flS
pY
a
Q -m
ABBREVIATIONS
microsecond
microvolt
ohm
ohm-metre (unit of resistivity)
DIRECT CURRENT TRANSMISSION
General Aspects of DC Transmission

and Comparison of it with
AC Transmission
1-1 HISTORICAL SKETCH817181
Early Discoveries and Applications
Both electrical science and the practical applications of electricity began

with direct current. Alternating current came later.
The basic discoveries of Galvani, Volta, Oersted, Ohm, and Ampere per
tained to direct current. The first widespread practical application was dc
telegraphy powered by electrochemical batteries and using ground-return
circuits.
Electric lighting and power also began with direct current powered by
dynamos. First came carbon arc iamps operated in series at constant current
and fed from series-wound generators. Later came carbon-filament incan
descent lamps operated in parallel at constant voltage and supplied with
current from shunt-wound generators.
The first electric central station in the world, on Pearl Street, in New York,
was built by Thomas A. Edison and began operation in 1882. It supplied
direct current at 110 V through underground tubular mains to an area
roughly 1 mi(1.6 km) in radius. It hadEdison bipolar dc generators driven by
steam engines. Within a few years similar stations were in operation in the
central districts of most large cities throughout the world.
In view of the initial supremacy of direct current it is interesting to see why
it was almost completely superseded by alternating current and why direct
current is again being used for some high-voltage transmission lines.
* Superior numerals and, in some chapters, including this one, superior letters alone or
followed by numerals refer to items or. to groups of items in the bibliography at the end
of the chapter.
1-1 HISTORICAL SKETCH
GENERAL ASPECTS OF DC TRANSMISSION
Later Ascendancy of Alternating Current
The advent of the transformer, polyphase circuits, and the induction motor
in the 1880s and 1890s led to ac electric power systems.
The transformer, simple, rugged, and efficient, made possible the use of
different voltage levels for generation, transmission, distribution, and use.
Inparticular, it made long-distance, high-voltage power transmission possible.
The exploitation of water power, usually available at sites distant from
major load centers, gave impetus to such transmission.
The induction motor, especially the polyphase induction motor, is also
simple, rugged, and cheap and serves the majority of industrial and residential purposes. The commutators of dc motors and generators, in addition
to requiring maintenance, impose limitations on the voltage, speed, and size
of these machines. The voltage per bar of the commutator should not exceed
about 22 V lest excessive sparking occur. Thus a high voltage per commutator
requires many bars, resulting in a large diameter. A large diameter requires
a low speed in order that the commutator and windings may withstand the
centrifugal force. And a low-speed machine is heavier and more expensive
than a high-speed machine of equal rating. The advent of steam turbines,
which are best at high speed, gave a great advantage to ac generators.
When ac systems first appeared, there were heated arguments between the
proponents of dc and ac systems. Advocates of dc branded ac dangerous
because of the high voltages used. As a result of their advantages, however,
ac electric power systems became almost universal. Power was generated,
transmitted, distributed, and used as alternating current. If direct current was
needed for some particular purpose, such as adjustable-speed motor drives
or electrolytic processes, alternating current was converted to direct current
locally by motor-generator sets or synchronous converters or, later, by
mercury-arc rectifiers.
The last vestiges of dc distribution were the low-voltage networks in the .
centers of large cities and electric traction (streetcar, trolley bus, rapid transit,
interurban and suburban railways, and some long tunnels or mountainous
sections of main-line railways). Finally, however, low-voltage ac networks
replaced low-voltage dc networks, diesel locomotives replaced steam loco
motives and many electric locomotives, and gasoline or diesel buses replaced
most of the streetcars and interurban lines. Some dc rapid-transit systems still
remain. The victory of alternating current over direct current, however, was
almost complete.
Status of DC Transmission During the Ascendancy of AC Transmission
Despite the general acceptance of ac transmission, some engineers never

forgot the obvious advantages of dc transmission (discussed in Section 1-4).
f
j
j.
"
f
,
They proposed, however, not to replace ac but to supplement it with dc.

Specifically, they would superpose a dc transmission link on an ac system or
interconnect two ac systems by a dc transmission tie line. Generation, use,
and even most transmission and distribution, would remain by ac.
Such a dc transmission scheme requires that ac be converted to dc at the
sending end of the dc link and that dc be converted to ac at the receiving end.
The feasibility and advantageousness of the scheme depended on the develop
ment of suitable converters for the required high voltage and power. The
development of suitable converters is considered shortly. First, however, let
us turn aside to describe the Thury dc system.
The Thury System318
f
|
A system of hv dc transmission designed by a French engineer, Ren6

Thury, came into use at a time when ac systems were in their infancy, and it
persisted well into the era of ac predominance. This system is interesting both
as an engineering achievement and because of certain similarities to modern
hv dc systems. At the sending end of the transmission line a number of
series-wound dc generators, driven by prime movers, were connected in
series to generate the required high voltage, and at the receiving end, a com
parable number of series-wound dc motors, connected in series, drove lowvoltage dc or ac generators. The system operated at constant current. The
voltage of each machine in the hv series circuit was regulated by shifting the
brushes.
Since the series circuit was normally grounded at only one point, many of
the machine windings had a high potential with respect to ground. It was not
feasible to provide insulation between windings and frame for such voltages;
instead, the frames were insulated from ground by setting them in a floor of
asphalt over asphalt concrete, and were insulated from the driving or driven
machines by insulated couplings.
j
Switching and instrumentation were very simple. Each machine was pro
vided with a short-circuiting switch. A machine was taken out of service by
reducing its terminal voltage to zero and then short-circuiting it. It was
brought into service by the reverse of this procedure. An ammeter and a volt
meter were the only instruments required.
From 1880 to 1911 at least 19 Thury systems were installed in Europe,
principally for the use of water power. The most important of these was that
from Moutiers, .in the French Alps, to Lyons,33 installed in 1906 with a
route length of 112 mi (180 km) of which 2.8 mi (4.5 km) were in under
ground cable, the remainder being open-wire line. Initially, its rated power of
4.3 MW was transmitted at 57.6 kV, 75 A. This line was built as a reinforce
ment of an existing ac system and was integrated with it. The Moutiers plant
1-1
had four water turbines, each driving four generators of 3.6 kV each. At Lyons
the greater part of the power received by hv direct current was converted to
alternating current and the remainder to 600 V dc for the street railway. The
over-all efficiency was 70.5%, which was considered satisfactory for a hydro
electric system.
In 1911 a second hydroelectric plant at La Bridoire, situated about halfway
along the line and rated at 6 MW, was added (in series). The line current was
then doubled (to 150 A). In 1912 a third hydro plant, located at Bozel, 7 mi
(11 km) beyond Moutiers, and rated at 9 MW, was added, raising the total
generating capability on the line to 19.3 MW. The maximum circuit voltage
became 125 kV and the route length 140 mi (225 km). Operation of the line
continued until 1937, when it was dismantled. Thury himself died in 1938.
The Thury system performed reliably in spite of the large number of com
mutators in series. The limitations of dc machines, already mentioned, how
ever, made it unsuitable to the larger amounts of power that had come to be
required. Further development of hv dc transmission required better con
verters than motor-generator sets.
Development of a Practical Converter
A converter is basically nothing more than an assemblage of controlled

switches. The commutator of a dc motor, generator, or synchronous con
verter is such a device. The vibrating reed is an even simpler switching device,
used for stepping up direct voltage from a storage battery to a value suitable
to plate supply in automobile radios or as a chopper in stabilized dc ampli
fiers. In these two applications the input and output are dc, with ac in the
intermediate circuit, which is just the opposite of dc transmission inter
connecting two ac systems.
Two of the more serious attempts to develop a switching converter suitable
to hv dc transmission are the transverter and the Marx atmospheric-arc
converter.
The transverter, patented in 1920 by two British enineers, W. E. Highfield

and J. E. Calverley, consisted essentially of polyphase transformers commutated by synchronously rotating brush gear. It performed the three basic
operations of voltage transformation, phase multiplication, and commuta
tion and could be used either as a rectifier or as an inverter. Since the com
mutators were stationary and only the brush gear rotated, the problem of
centrifugal force was mitigated. Several experimental transverters were built,
the largest of which was rated at 2 MW, 20 A, 100 kV on the dc side, but none
has been used commercially.
The atm&spheric arc converter, devised by E. Marx of Braunschweig in
1932, is a switching device in which an arc between two like water-cooled
HISTORICAL SKETCH
main electrodes is ignited by a high-frequency spark getween auxiliary

electrodes in the path of the main arc and is extinguished after a current zero
by a blast of air or mixed gases that continually plays on the arc path. At one
time such converters could handle more power (40 MW) than any other
converter then available. The life of the electrodes, however, was short, and
the voltage drop across the arc was high (500 V). The loss in the arc, together
with the power required for ignition, air blast, and cooling, amounted to
2.5 to 3% of the transmitted power at each terminal. This is considerably
greater than the corresponding loss (about 0.3%) in mercury-arc converters.
Valves
The synchronously controlled switches heretofore described for use as

converters can conduct in either direction, andthe actual direction of current
depends on the controlled instants of closing and opening and on the
in the circuit. Generally, unidirectional conduction is desired. Devices

having inherent unidirectional conduction are herein called valves. Among
such devices are vacuum and vapor or gas-filled tubes having thermionic
cathodes, the mercury-vapor tube with mercury-pool cathode, and various
solid-state devices. In their simplest form, as diodes, they can be used as
rectifiers but not as inverters.
The mercury-arc rectifier with pool cathode is the most suitable for
handling large currents. It was invented by Peter Cooper Hewitt about 1903
and initially was made with a glass envelope. The steel tank appeared about
1908 to 1910.
In order for a valve to be used as an inverter, it must have a control
electrode that can prevent the valve from conducting, although the anode is
positive with respect to the cathode. Such an electrode (the grid) was added
to the vacuum tube (Fleming valve) by de Forest in 1906. The grid was added
to the mercury-vapor valve in 1928. It was then applied to the hot-cathode
glass-envelope valve, and the resulting triode was called the thyratron. Later
some thyratrons were made with teel jackets. Control grids were added to
mercury-arc valves with pool cathodes about 1930. Although the grid in the
vacuum tube can start, stop, and modulate the current through the tube, the
grid of a mercury-arc valve can only prevent conduction from starting. After
it has started the control grid can neither stop the current nor control its
magnitude. Conduction does not cease until the anode becomes negative ,
with respect to the cathode.
Actually the first control element used in a mercury-arc valve with pool
cathode was not a grid but an igniter, introduced in 1923. The resulting valve
is called an ignitron. The igniter is a rod the end of which dips into the mercury
pool. When current from an auxiliary source is sent through the igniter, an
arc is started. The igniter, like the grid, cannot stop conduction.
emfs
1-1
Present-day mercury-arc valves for high-voltage transmission, known as

excitrons, have, in addition to the anode and the mercury-pool cathode, an
ignition electrode for starting the arc, one or more excitation electrodes for
maintaining the arc, and a control grid that prevents the arc from reaching
the anode until it is desired that the valve begin to conduct. There are also
several grading electrodes placed between the control grid and the anode for
obtaining a more uniform potential gradient than would otherwise exist. The
grading electrodes are kept at the desired potentials by connecting them to
taps on an external resistance-capacitance potential divider the ends of which
are connected to the anode and control grid. This system of grading electrodes, invented by U. Lamm in 1939, has considerably increased the peak
inverse voltage that the valves can withstand.
Valves for hv dc transmission are invariably of single-phase construction,
in contrast to the polyphase valves with mercury-pool cathode formerly used
extensively in low-voltage rectifiers for industrial and railway application.
The development of valves for hv dc transmission has been carried out
since World War II principally by engineers in the U.S.S.R. and by the
Swedish firm of Allmanna Svenska Elektriska Aktiebolaget (ASEA), with
which Lamm is connected. A noteworthy feature of ASEA valves is the use of
several, usually four, anodes in multiple on single-phase valves. The current
ratings are 200 to 300 A per anode. Russian engineers have concentrated on
single-anode valves, which so far appear to have been less successful than the
ASEA valves.
About 1960, control electrodes were added to silicon diodes, giving siliconcontrolled rectifiers (SCRs), also called thyristors. At present these are not
capable of handling the highest voltages and powers required for hv dc
transmission. Their ratings have increased, however, with surprising rapidity,
and it seems certain that such valves will soon replace mercury-arc valves in
hv dc use.
Experimental DC Transmission Projects and First Commercial Lines
The initiative in exploring the use of mercury-arc valves for dc transmission

was taken by the General Electric Company. After two smaller experi
ments81,2 they proceeded in December 1936 to use direct current on a
17-mi (27-km) line between the Mechanicville hydroelectric plant of the
New York Power & Light Corporation and the General Electric factory in
Schenectady.815 The line carried 5.25 MW at 30 kV, 175 A. The converter at
each end of the line had 12 hot-cathode glass-envelope thyratrons in 6 series
pairs. The ac input at Mechanicville was at a frequency of 40 Hz, and the
output at Schenectady was at 60 Hz. Thus was demonstrated a feature of dc
transmission that has been important in several subsequent installations:
frequency conversion.
LC bridge circuit called the monocyclic square. Constant-current operation

was chosen because the hot-cathode tubes then used could not withstand the
high short-circuit currents expected to occur on a constant-voltage system.
After the more rugged steel-envelope mercury-poolignitron became available,
however, the line was converted in 1940 to constant-voltage operation. The
circuitry then used was basically the same as that of modern dc transmission
systems, fault currents being limited by control of valve ignition. The operation
of the line was discontinued in 1945 in the belief that nothing more would be
learned by continuing it. Perhaps an additional belief was that there was no
future in dc transmission.
Meanwhile, two 25/60-Hz frequency changers using controlled mercuryarc valves were installed in steel mills in the United States in 1943. The larger
of these, rated at 20 MW, was installed at the Edgar Thompson plant of the
Carnegie-Illinois Steel Company near Pittsburgh. The United States was
inactive in the field of dc transmission, however, for nearly 20 years.
A demonstration of dc transmission using grid-controlled steel-tank
mercury-arc conversion was given at Zurich, Switzerland, in 1939, at the
Fifth Swiss National Exhibition.84,5 Power of 0.5 MW at 50 kV, 10 A, was
sent 19 mi (30 km) from Wettingen power plant near Baden to Zurich over
a circuit of one conductor, partly overhead and partly in underground cable,
with earth return. In 1946, Brown-Boveri discontinued their work on hv dc
__
The line initially operated at constant current, the conversions from con
stant alternating voltage to constant current and vice versa being made by an
HISTORICAL SKETCH
transmission.
Two hv dc experiments were conducted in Germany during World War II
instance of the German Secretariat for Aviation.86,19 A 400-kV
three-phase line from the Alps to the Ruhr had already been planned, but the
at the
Secretariat intervened in favor of a hv dc cable line, which, it felt, would be

less vulnerable to air-raid damage. The Siemens-Schuckertwerke A.G. began
experiments in preparation for such a line. They transmitted 4 MW at
110 kV a distance of 3 mi (5 km) over an existing line from a station in the
Charlottenburg district of Berlin to one in the Moabit district.86 A second,
larger experiment was to be the transmission of 60 MW by means of a 70-mi
(110-km) 400-kV dc cable from the Elbe (near Dessau) to Marienfelde (near
Berlin).86,9 This experiment was to be conducted jointly by Siemens and the
A.E.G. The fortunes of war prevented completion of the project, and in 1945
such plant and pertinent documents as survived were taken to the U.S.S.R. as
reparations.
In Sweden, where the principal new hydroelectric sites are in the north and
the principal loads are in the south, hv transmission is required ; and, because
of the development of valves by the Swedish firm of ASEA, interest was
aroused in the possibility of a hv dc transmission system as an alternative to
ac. An experimental transmission between Mellerud and Trollhattan (36 mi)

began operation in 1944. It aided further development of valves by permitting
them to be tested under service conditions. The Swedish State Power Board
decided to use alternating -current for the north-to-south transmission
already mentioned. The results of the Mellerud-Trollhattan transmission,
however, encouraged the Board to proceed with hv dc transmission by sub
marine cable from the Swedish mainland to the island of Gotland, 96 km
(60 mi) offshore;0 This system, built by ASEA, began service in 1954 and
may be considered the first commercial hv dc transmission system. The line
transmits 20 MW at 100 kV through a single-conductor cable, with return
path through the sea and earth. Each converter has two valve groups rated
50 kV, 200 A, 10 MW, the groups being in series on the dc side. Each valve
has two anodes working in parallel.
Building the dc link was judged more economical than constructing
additional thermal power plants on the island. The distance is far too great
for ac cable transmission.
Power flow is normally from the mainland to Gotland but is sometimes in
the opposite direction. Much of the time when power is delivered to Gotland,
there are no generators in operation there, the only synchronous machine
being a condenser. Power is adjusted automatically to maintain rated
frequency (50 Hz) in Gotland.

The link is still in operation (1970) and has a good performance record.
One of the mercury-arc valves was replaced by an air-cooled thyristor assem
bly, which also has performed well. Plans have been announced for doubling
the voltage and power on the existing cable by the addition of a new thyristor
valve group to each terminal, thereby doubling the voltage.
In the U.S.S.R., where even greater distances than in Sweden separate the
potential hydroelectric sites from the principal industrial load areas, the use
of hv dc transmission was considered necessary,8 and an extensive program
of research and development was undertaken, begun as a part of the 5-yr plan
of industrial development for 1946 to 1950.
An experimental line between Moscow and Kashira (112 km or 60 mi,
30 MW, + 100 kV) began operation in December i95o.B10'11,13'16
It was basically an underground cable line, but at times sections of overhead
line were put into the circuit. Both bipolar metallic operation and monopolar,
ground-return operation were tried. Practical ground electrodes were
developed, and various kinds of valves and converter control were tested.
A Direct Current Institute312,14 was established in Leningrad, which since
1957 has published approximately one volume per year of articles on its
researches/3
A full-scale 474-km (294-mi) overhead line between a hydroelectric plant

at Volgograd, formerly called Stalingrad, and the Donets Basin was energized
1-2
CONSTITUTION OF EHV AC AND DC LINKS
at reduced voltage and power in 1962 and, beginning in 1965, was operated
at its full rating of +400 kV, 900 A, 720 MW.1 Other dc lines of lengths of
2000 to 2500 km and voltage of +750 kV are planned/6
1-2 CONSTITUTION OF EHV AC AND DC LINKS
ehv transmission links, superposed on a lower-voltage ac network, or inter
connecting two such networks, or connecting distant generating plants to an
ac network, are compared as to their principal components and the arrange
ments thereof, according to whether the line operates on ac or dc. The phrase
"transmission link" denotes the transmission line proper together with its
terminal and auxiliary equipment.
Figure 1a shows a single-circuit three-phase ac line. In general, such a line
in the categories already mentioned, one which might be competitive with
a dc link, requires transformers at both ends step-up transformers at the
sending end and step-down transformers at the receiving end although in
some cases they can be omitted at one or both ends. If the transformers are
operated as an integral part of the link, only low-voltage circuit breakers are
required.
Ac
system
system
Ac
system
Ac
system
nverter
Dc line
Ac
system
Ac
system
Ac
system
Ac
system
O-J
-o--c
Fig. 1. Constitution of ac and dc ehv links shown by single-line diagrams.
10
1-3
Most long overhead ac lines require series compensation of part of the

inductive reactance. In the figure, one bank of series capacitors for this pur
pose is shown at the middle of the line.
Three-phase lines cannot be operated, except for a very short time (less
than 1 sec) with one or two conductors open, because such operation causes
unbalanced voltages in the ac system and interference in parallel telephone
lines. Therefore three-pole switching is always used to clear permanent
faults, although such a fault may involve only one conductor. This being so,
two parallel three-phase circuits are required for reliable transmission (see
Fig. 16). Long two-circuit ac links are usually sectionalized by means of
intermediate switching stations for several reasons. Among these are (a)
limiting the decrease in stability power limit attributable to switching out one
circuit to clear a fault or for line maintenance, (b) limiting the overvoltage
when a line is energized from one end, (c) providing a place for the connection
of grounding transformers to limit the overvoltages of the unfaulted phases
with respect to ground when one phase is faulted to ground, and (d) for con
nection of intermediate loads or generation. Intermediate generation raises
the stability limit of the link. On many long ehv lines, shunt reactors are
required for limiting the voltage, especially at light loads, but they may be
required even at full load. These reactors are usually placed at intermediate
switching stations and are so indicated in Figure 16.
A representative single-circuit dc link is shown in Figure lc. The line itself
usually has two conductors, although some lines have only one, the return
path being in the earth or seawater or both. At both ends of the lines are
converters, the components of which are transformers and groups of mercuryarc valves. The converter at the sending end is called a rectifier, and that at
the receiving end an inverter. Either converter, however, can function as
rectifier or inverter, permitting power to be transmitted in either direction.
The ac line, of course, also has this reversibility.
Circuit breakers are installed only on the ac sides of the converters. These
breakers are not used for clearing faults on the dc line or most misoperations
of the valves, for these faults can be cleared more rapidly by grid control of
the valves. The breakers are required, however, for clearing faults in the
transformers or for taking the whole dc link out of service.
Harmonic filters and shunt capacitors for supplying reactive power to the
converters are connected to the ac sides of the converters. Large inductances
called dc smoothing reactors are connected in series with each pole of the dc
line.
Some writers claim that a two-conductor dc line provides the same re
liability as a two-circuit three-phase line having six line conductors, for either
conductor of the dc line can be used with ground return continuously or for
limited periods, say, a few days per year.
KINDS OF DC LINKS
11
If higher reliability is required of a dc line than that provided by two con

ductors, three or four conductors may be provided. One pole of a fourconductor line is shown in Figure Id, with two converters per terminal. The
bus-tie switches 1 are normally open. If a permanent fault occurred on the
lower conductor, the converters connected to it would be controlled so as to
bring the voltage and current on it to zero. Then switches 3 would be opened,
isolating the faulted line. Next the converter voltages would be raised to
equality with those of the respective adjacent converters, after which switches
1 would be closed. The capability of all converters would then be usable, and
the power normally carried by two conductors would then be carried by one.
The line loss would be four times its normal value, somewhat diminishing the
delivered power. The whole switching operation would take about 0.3 sec,
a time as short as that required for rapid reclosure on an ac line. Each pole
would be switched independently of the other.
Comparison of the ac and dc links shows that (a) the dc line proper is
simpler, having one or two conductors instead of three, but that (b), on the
other hand, the terminal equipment is more complex, having the groups of
valves and some auxiliary equipment that the ac line does not need.
1-3 KINDS OF DC LINKS
Direct-current links are classified as shown in Figure 2. )

The monopolar link has one conductor, usually of negative polarity, and
ground or sea return.
The bipolar link has two conductors one positive, the other negative.
Each ternfig||||has two converters of equal rated voltages in series on the dc
side. Thioints (junctions between converters) are grounded at one or
both ends, fillllh neutrals are grounded, the two poles can operate inde
pendently. Normally they operate at equal current ; then there is no ground
current. In the event of a fault on one conductor, the other conductor with
ground return can carry up to half of the rated load.
The rated voltage of a bipolar link is expressed as 100 kV, for example,
pronounced plus and minus 100 kV.
The homopolar link has two or more conductors all having the same polar
ity, usually negative, and always operates with ground return. In the event of a
fault on one conductor, the entire converter is available for connection to the
remaining conductor or conductors, which, having some overload capability,
can carry more than half of the rated power, and perhaps the whole rated
power, at the expense of increased line loss. In a bipolar scheme reconnection
of the whole converter to one pole of the line is more complicated and is
usually not feasible because of graded insulation. In this respect a homopolar.
line is preferable to a bipolar line in cases where continual ground current is
1
12

Inverter
Rectifier
CO
"o
o
o. 3
I
03
a
O
D.
O cO
(U
VI
no
no
o
H
I
.S
8
u
o
V
if
On
no
o
o
On
If
u-
On
fS
w
rf
Tf
On
r-
cd
/N
ID
NO
ry,
if
-cO
cO
oo
ry
If
CllO
ID
r11
Tf
ry
if
r-
<D
NO
00
w
<o
stage
00
<o P
'C
SP -o
[S
NO
*0-
CO
CD
cy
CD
ID
r-
CD
CN
ry
'-3
<s>
CO
<o
'irst
t-
CO
"o
a O fl
a 2
<3 j- ,<u
<a
.g
a y
c"N
!s
c3
-o a
X)
<0
<y
h
ID
ID
D
ID
On
00
CD ID
?y
ID 00
CD
60
c
a
e
O
I
os
=
(c)
it-
cS
On
S -S
kj
r'0*
S~
OO
DN
v
IT)
If
ID
00
OO
On
fS
ID
<N
r-
Tj"
ID
CD
<y
CD
ID
OO
<B.
cy

O
N
no
60
CN
(N
ro -0"
Tf
St
3
tH
no
D
D
On
00
II
>
N in
>"
Fig. 2. Kinds of dc links.
not deemed objectionable (see Chapter 9). An additional minor advantage is

the lower power loss due to corona. Negative polarity is preferred on overhead lines because of its smaller radio interference.
U
Q
>
Cascaded Groups
In each of these kinds of links there are usually several converters connected
in parallel on the ac side but in series on the dc side for obtaining the desired
OA >
CO :0
53 O
level of direct voltage from pole to ground. Each such converter consists of
a transformer bank and a group of valves.
1-4 HV DC PROJECTS FROM 1954 TO 1970
"O
The successful operation of the Gotland link awakened interest in dc

transmission in other countries. A list of the dc links in operation or under
construction in 1970 is given in Table 1. These links are situated in nine
a
a
CO
CO
*o
CO
*a
cO
bb
O
60

II
CO
i3
&
a
o
cO
a
*3
VI
>
p
o
>
eI
c3
Ph
c?
a a
o
CO
*3
CO
lej
14
different countries. Brief comments are made on these links, all of which
except Volgograd-Donbass were based wholly or mainly on ASEA
techniques.
Volgograd-Donbass Line1
When built, this was the longest dc line. It usually carries power from a
hydroelectric power plant on the Volga River at Volgograd to an industrial
and mining district in the Donets Basin. For such an operation generators
15
in the hydro plant may be disconnected from the ac bus and connected only
in the opposite direction
the inverter valve groups are connected to the ac bus.
It seems that the link did not offer any advantage in cost compared with an
ac link, but it was built to gain experience in dc transmission for longer higherpower lines that will be built in the future.
Each terminal has eight valve groups (four per pole), using single-anode
valves of Russian design, with two valves in series in each arm.
The year 1965 was called "the dc year" by the editor of Direct Current. Not
only was the Volgograd-Donbass link brought up to its designed voltage and
power, but also two additional dc transmission schemes (New Zealand and
Konti-Skan) and a frequency changer at Sakuma, Japan, went into operation.
A third transmission sdggme (Sardinia) was expected to go into operation, but
it was delayed untiHMlli&owinp year.
to a valve group of the rectifier. For power flow
English Channel Crossing11
The next link to go into service after Gotland was an interconnection

between the ac systems of England (Central Electricity Generating Board)
and France (Electricite de France) through two single-conductor submarine
cables. The distance (42 mi or 64 km) is shorter than that of the Gotland
scheme, but the rated power (160 MW) is eight times as great. Each valve
has four anodes, and each of two bridges (one per pole) is rated at 800 A,
100 kV, 80 MW. Like Gotland, the Channel Crossing scheme involves cross
ing water; but, unlike Gotland, it does not use the sea as a return conductor.
Because of concern with the effect of the direct current on ships'- compasses in
a channel having much shipping, two cables were laid close together, one
operating at + 100 kV with respect to ground and the other at 100 kV. The
midpoint (neutral) of the converters is grounded at one terminal only, so that
ground current cannot flow except briefly in the event of a cable fault.
This link interconnects two large ac systems but has a small power rating
compared with the capacity of either system. An ac link of this kind would
have been feasible except that it would be difficult to control. The British
power system has no automatic load-frequency control. Installation of such
a control for the sake of the interconnection would have been very expensive.
The dc link is an asynchronous tie between two systems of the same nominal
frequency (50 Hz). Its power flow is readily controlled to a set value.
The purpose of the interconnection is to take advantage of time-zone and
generation diversity. The direction of power flow varies. The French system
has a considerable amount of hydroelectric generation ; the British system has
practically none. In seasons in which the supply of water to the hydro plants
is ample, power can be exported to Great Britain. When water is scarce,
power can be imported from there.
The Channel link was plagued by troubles in its first few years of operation.
One of the transformers in the French terminal failed. The submarine cables
were broken several times by trawlers, and they could not be repaired soon
because of bad weather and rough seas. Since then the link has operated with
very little trouble.
1-4 HV DC projects from 1954 to 1970
New Zealand LinkJ

To meet the growing demand for power on the North Island, either addi
tional steam-electric power plants would have to be built there, or hydro
electric power plants would have to be built on the South Island, from which
the power would be transmitted electrically to the North Island. Submarine
cables 24 mi (39 km) long would be required across Cook Strait, which
separates the two islands. The hydroelectric alternative was more economical
and it was chosen. Direct-current transmission was selected as being more
feasible than ac for this long water crossing. Three dc cables are used (one
for each pole and a spare), but 11 ac cables would have been required (for
three three-phase circuits and two spares), which would have occupied a wide
belt of sea bed. The decision was made even before the English Channel
scheme was in operation.
The transmission system includes, in addition to the submarine cables,
335 mi (535 km) of overhead bipolar transmission line on the South Island
and 25 mi (40 km) on the North Island. It extends from Benmore power
plant on the South Island to Haywards Substation on the North Island, near
the city of Wellington. The power rating of 600 MW is considerable com
pared with the aggregate generation on either island then (1400 MW on the
North Island and less on the South Island) and slightly exceeds the rating of
the Benmore plant. The cost of the dc-transmission scheme was about twothirds of that of the ac-transmission scheme that was considered as an
alternative.
Ground return is used in emergencies when one pole of a converter or the
transmission circuit is faulted.
16
1-4
The valves, manufactured by ASEA, are rated at 1.2 kA, 125 kY, and have
four anodes.
Konti-Skan LinkK
This is an interconnection between Sweden and Denmark and thus,

through previously existing ac connections, between Germany and the rest of
Western Europe and the Scandinavian countries. It crosses the Kattegat by
way of the island of Laeso and has two cable sections and overhead sections
on the island and at each end.
The dc scheme was compared with an ac scheme having a shorter cable.
The cost of the two schemes was approximately equal, but the dc scheme
presented two advantages over the ac:
1. The dc line provides an asynchronous tie. The stability limit of the ac

scheme was estimated as 350 MW; the ultimate power capability of the dc
link was 500 MW. The need for expensive load-frequency regulation is
avoided.
2. The dc scheme can be built in two stages, and thus almost half of the
investment can be postponed. The first stage operates monopolarly with one
submarine cable and sea return at a power capability of 250 MW. In the
second stage the line will be a bipolar, metallic circuit for 500 MW, with
sea return used only in emergencies.
Four-anode, 1.1-kA, 125-kV valves are used.
Sakuma Frequency Changer

This station was put into operation in 1965, interconnecting the 50- and
60-Hz systems of Japan. It can transmit 300 MW in either direction. There
is no dc transmission line, the dc circuits being confined to the station. With
minor exceptions, the equipment and circuits are like thotee of a transmission
scheme. The valves are similar to those of the New Zealand and Konti-Skan
links.
Sardinian SchemeL
In order to use large deposits of low-grade coal on the Italian island of
Sardinia, a thermal power plant was built there, and a dc link was built con
necting it, by way of the French island of Corsica, to the Italian mainland
near La Spezia. This link consists mainly of submarine cable, with some
overhead line on Corsica and at the ends. A peculiarity of this scheme is that
the line has two conductors of the same polarity, with sea return. The polarity
is negative when power is transmitted from Sardinia to the mainland, which
HV DC PROJECTS FROM
1954 TO 1970
17
is the usual direction, although the opposite direction holds when the Sardin
ian plant is shut down. Power flow is regulated so as to keep constant fre
quency on the Sardinian ac system. The valves are similar to those of several
other schemes, are rated at 1.0 kA, 100 kV, and have four anodes.
Vancouver Island Scheme1*
This provides a dc connection between the mainland of the Canadian

province of British Columbia at Arnott, south of the mouth of the Fraser
River, and Vancouver Island. It is being built in stages of 78 MW each, with
an expected final power of 312 MW. It crosses the Strait of Georgia by
submarine cable anpring Island by overhead line. The four-anode
valves are rated at
kV.
This is the first schlBgP which a dc link operates in parallel with an
lWO
ac link.
Pacific Northwest-Pacific Southwest IntertiesN
The purpose of this scheme is to take advantage of seasonal diversity in

load and generation between the northwest area, comprising the states of
Washington and Oregon, and the southwest area, comprising southern
California and Arizona. The entire scheme includes two 500-kV ac circuits
with a total length of 905 mi (1450 km) from the Columbia River to the
vicinity of Los Angeles and two 400-kV bipolar dc circuits. The first dc
circuit is from Celilo substation near The Dalles, Oregon, to Sylmar sub
station, near Los Angeles. The second dc circuit is planned to be built from
Celilo to Mead substation near Hoover Dam at Boulder City, Nevada. The
power ratings of the ac lines are 1000 MW each and those of the dc lines
1440 MW each. A third dc line, the so-called "dc cross tie," from Sylmar to
Mead, about 270 mi (430 km), has been discussed, but there is no definite plan
for building it.
Each of the two dc lines exceeds any previous dc line in length and in
power rating, although the rated voltage is equal to that of the VolgogradDonbass line. The valve ratings are also greater, being 1.8 kA, 133 kV,
240 MW per group, with six anodes per valve.
The dc lines operate in parallel with a 60-Hz ac system. Because of the
great length of the ac lines, the stability of the ac system poses a considerable
problem, and it was necessary to use a high degree (average 65%) of series
compensation. A permanent bipolar fault on a fully-loaded dc line is one of
the severest disturbances that the ac system must withstand, although the
occurrence of such a fault is believed to be very infrequent.
18
Kingsnorth0
The Central Electricity Generating Board of Great Britain is interested in
the use of dc links for reinforcing an ac system in areas of high load density
without increasing the interrupting duty of ac circuit breakers. A trial installa
tion of this kind is the transmission of power by underground dc cable from
the Kingsnorth thermal power plant, situated on the south shore of the
Thames River estuary, to two substations in London. This is a bipolar scheme
having three cables: one for each pole and a neutral cable. Each pole goes to
a different substation, with the result that, although the whole scheme has
three terminals, each pole has only two terminals. The Beddington substation
is 37 mi (59 km) from Kingsnorth, and the Willesden substation is 14 mi
(23 km) beyond Beddington. Whenever the loads of the two substations are
unequal, there will be neutral current. This current is not allowed to flow in
the ground for fear of damage by electrolytic corrosion to some of the
many buried metallic structures found in a metropolitan area.
The rating of this scheme is 266 kV, 1.2 kA, 640 MW. There are four
groups of valves at Kingsnorth and two groups at each substation, each
group being rated at 133 kV, 1.2 kA, 160 MW.
1-5
LIMITATIONS AND ADVANTAGES OF AC AND DC TRANSMISSION
19
other. The distinctive feature of the Eel River station is that it is the first large
converter station designed to use thyristor valves initially and exclusively.
The rating of the station is 320 MW, 80 kV dc, 230 kV ac.
1-5 LIMITATIONS AND
ADggpGES OF AC AND DC TRANSMISSION
Noting the universal use of alternating current for electric power trans
mission, as well as for generation, distribution, and use, one naturally asks
what limitations ac transmission has that have led to the use of dc transmis
sion in some projects.
The limitations may beither technical something cannot be done or
economic it can be done more cheaply some other way. In most practical
cases the technical limitations are not reached, and economic limitations
dictate the final choice of design.
We are interested in limitations on the amount of transmitted power and
on the distance over which it can be transmitted. More exactly, we are inter
ested in the cheapest method by which a certain amount of power at a certain
load factor can be transmitted reliably over a certain distance. The power
depends on the current, voltage, power factor, and number of conductors.
Nelson River, Manitoba, Schemep

Current Limit
The Nelson River has a potential hydroelectric power development of

about 6500 MW, including some diversion of water from other streams. It has
been decided to develop this power and to transmit it to Winnipeg by direct
current. Bipolar +450-kV overhead dc lines were judged more economical
than 500-kV ac lines. Ultimately there will be several such bipolar circuits as
the development proceeds by stages. With two such lines, the transmission
capacity will be 3240 MW.
In response to the invitation for bids on terminal equipment for the first
stage, three proposals were received for thyristor converters and two
for mercury-arc-valve converters. The proposal for mercury-arc equipment
by the English Electric Company was accepted. Each valve group will
operate at 1.8 kA, 150 kV, 270 MW. There will be three groups in series per
The temperature of a conductor must be limited in order to avoid damage

to the conductor itself (permanently increased sag) or, in case of a cable, to
the insulation in contact with it. Hence the current in the conductor must be
limited in accordance with its duration and the ambient temperature. The
limiting current is seldom reached on long overhead ac lines because of other
limitations' being reached first, but on cables the current limit due to heating
isggmnortant. as shown later.
mBmle ac resistance of a conductor is somewhat higher than its dc resistance
because of skin effect, but the difference is not important in nonmagnetic
conductors of the usual diameters at the usual power frequencies.
pole.
Voltage Limits
Eel River (New Brunswick)
The normal working voltage and the overvoltages caused by switching

surges and lightning must be limited to values that will not cause puncture or
flashover of the insulation. On ehv overhead lines, switching surges, rather
than lightning, have become the more serious transient overvoltages, and on
ac lines attempts are made to limit them to peak values of two or three times
normal crest voltage. Switching surges on dc lines are lower than this, say,
This station provides an asynchronous tie between the 60-Hz ac systems of

Hydro Quebec and of New Brunswick. As at Sakuma, the dc circuits are
confined to the station. In contrast to Sakuma, the nominal frequencies of
the two ac systems are equal, although one can drift with relation to the
20
1-5
1.7 times normal voltage. On overhead lines, the maximum working voltage
or the minimum conductor size is limited also by loss and radio interference
due to corona. In current ac practice, radio interference during foul weather
(rain, snow, or fog) is usually the limiting factor. Here dc lines have a distinct
advantage in that radio interference is slightly decreased by foul weather,
while interference due to ac lines is greatly increased by foul weather. In
cables, where the limiting factor is usually the normal working voltage, the
insulation will withstand a direct voltage higher than the crest of alternating
voltage, which is already 1.4 times the rms value of the alternating voltage.
voltage (kV)
surge impedance loading (MW)
1.05
On long ehv ac overhead lines and on much shorter ac cables, the produc
tion and consumption of reactive power by the line itself constitutes a serious
problem. On a line having series inductance L and shunt capacitance C per
unit of length and operating voltage V and ciirrent I, the line produces
reactive power
Ql = coLI2
(2)
and consumes reactive power
230
345
500 .
43
130
300
830
700
.
1600
\. No
load
2. Natural load Pw
1.00
(1)
132
oljp line carrying its natural load, the magnitude of voltage is the same
everywhere, as shown in curve 2 in Figure 3, and the reactive power is zero
Reactive Power and Voltage Regulation
Qc = coCV2
21
->L.eavv load (15
0.95
20
Distance from sending end (elec deg)
10
30
Fig. 3. Voltage profiles of one-twelfth-wavelength Low-loss line with equal terminal

voltages E. Length at 60 Hz is 258 mi (416 km).
everywhere (curve 2 in Figure 4).

+0.5
per unit of length. The reactive power produced by the line equals that con
sumed by it, with no net production or consumption, if
Q
Pn
coCV2 = (oLI2
2. Natural load Pn
hence if
HW'2-
(3)
In this case the load impedance has the value Zs , known as the surge im
pedance of the line. The surge impedance of an overhead line with single
conductors is about 400 Q, and with bundle conductors, about 300 Q; that of
cables is only 15 to 25 Q.
The power carried by the line so loaded is
V2
P = VI = -
P+jQ
-0.5
10
20
Distance from sending end (elec deg)

Fig. 4. Flow of reactive power Q on the line in Figure 3.
(4)
and is called the surge impedance loading (SIL) or natural load. It is inde
pendent of distance and depends mainly on the voltage. Typical values for
three-phase overhead lines are as follows :
Most lines cannot be operated always at their natural loads, for the loads
vary with time. The most economical load on an overhead line is usually
greater than the natural load. If the load is greater than the natural load, net
reactive power is consumed by the line and must be supplied from one or
both ends. If equal voltages are maintained at both ends of the line, equal
amounts of reactive power are supplied from both ends (curve 3 in Figure 4),
22
1-5
and the voltage elsewhere sags, being least at the center of the line (curve 3 in
Figure 3). If the load on the line is less than the natural load, net reactive
power is produced by the line and is delivered to one or both ends. With
equal voltages at both ends, equal amounts of reactive power are delivered to
both ends (curve 1 in Figure 4), and the voltage everywhere else is higher than
at the ends, being greatest at the middle (curve 1 in Figure 3). In all cases, the
flow of reactive power through series inductive reactance is "downhill," that
is, in the direction of decreasing voltage.
Thus, to maintain constant equal voltages at both ends, reactive power
must be absorbed at light load and supplied at heavy load. The reactive
power required for a given variation of load increases with distance (see
Figure 5, curves 3 and 4).
j
(
j
)
23
to supply or consume the reactive power required for maintaining approxi

mately constant voltages. On lines up to 250 mi long, reactive power is
ordinarily supplied at the terminals. In the past, synchronous condensers
were commonly used for this purpose. They can control the voltage by either
supplying or consuming reactive power, as required. Nowadays shunt static
capacitors and reactors are found to be more economical. They are switched in
blocks.
Figure 5 shows some other disadvantageous characteristics of long, un
compensated transmission lines up to one-half wavelength (180). One is their
power limit. For any given length Ithe maximum power that can be trans
mitted on such a line is shown by curve 1. It is
Pm=*PHsec0
(5)
9 = IsJtC
where
3
2. Voc/E
!
f'
V(T=7xT=T).
P= 1.5P,
1
/ 5:Vm/forP= 1.5P,
3- Qt/Pn for P= 1.5P
-2
0
30
60
90
120
Length of line (elec deg)
150
180
:
Fig. 5. Characteristics of lossless line with equal terminal voltages E (except curve 2)
versus length up to one-half wavelength. Curve 1. Maximum power/natural power, PmjP.
Curve 2. Voltage at open end/sending-end voltage. Curve 3. Reactive power from both
ends/natural power for P=1.5P. Curve 4. Reactive power from both ends/natural
power for no load (P = 0). Curve 5. Voltage at middle/terminal voltage for P = 1.5P.
If we stipulate, instead of constant voltages at both ends, fixed voltage at

the sending end and fixed power factor, say, unity, at the receiving end, the
receiving-end voltage varies with load. For a unity-power-factor load, the
voltage is high at light load and low at heavy load. The amount of variation
increases with the length of line (up to a quarter wavelength). On a long line
the variation of voltage becomes intolerably great, and it becomes necessary
(6)
is the electrical length of the line, Ibeing the actual length and X the wave
length. For a greater load than the natural load, there is a maximum distance;
for example, for P = 1.5Pn this distance corresponds to a line angle of
sec"1 (PJPn) = sec-1 1.5 = 41.8 and is (41.8/360)2, or 360 mi at 60 Hz. As
the length of line approaches this value, the reactive power that must be
supplied to the line increases rapidly, as shown by curve 3, and the voltage at
the middle of the line drops rapidly, as shown by curve 5.
Another limitation of long lines is the high voltage at an open end (the
Ferranti effect), shown in curve 2. This is important when a line is being put
into service by first connecting one end of it to the main ac system, for it is
not feasible to close both ends at exactly the same moment.
Long-distance ac power transmission is feasible only with the use of series
and shunt compensation, applied at intervals along the line, as illustrated in
Figure 6. Series compensation of degree s reduces the effective series in
ductance from L by sL to (1 s)L and thus decreases the electrical length
Eq. (6)from /?/ to
s and at the same time decreases the surge
impedanceEq. (3) and increases the natural load by the same factor. The
reactive power produced by shunt capacitance of the line at light load may
still be excessive, requiring shunt compensation of part h of it. The effective
shunt capacitance is then reduced from C to (1 - h)C,and the electrical length
is reduced by the additional factor %/l h or by the total factor
= pi
/?/Vl
The surge impedance is altered by the factor

essentially unchanged if h s.
V(1 -y)/(l h) and may be
1
24
1-5
Fig. 6. Series and shunt reactive compensation for 750-mi (1200 km) 500-kV 60-Hz
single-circuit line delivering 1000 MW and having two 1780-MCM A.C.S.R. conductors
per phase. Sending-end voltage, 525 kV /0; receiving-end voltage, 500 kV / 29.5. The
series capacitors have an aggregate loading of 1507 Mvar.; shunt reactors, 1275 Mvar. ;
reactive power entering line from sending end, 193 Mvar. ; from receiving end, 355 Mvar.
Degree of series compensation, 73%; of shunt compensation, 58%. (Ref. D4, Appendix B,
Figure 7.)
In Figure 6 and in similar schemes shown in the Ref. D4 for other voltages
and distances, the compensation was chosen to limit the angle between termi
nal voltages to 30 and to limit the voltages at the ends and at compensating
points to not more than 1.05 times nominal voltage, except that the voltage
at an open end is limited to not more than 1.1 times normal voltage.
On representative long overhead ehv compensated lines operating at full
load, the total reactive power furnished from both ends of the line and from
intermediate series capacitors plus the reactive power consumed by shunt
1
1
500 kV
1ooo
I
&
700 kV
6O-Hz a
o
.2 v
F
200
400
Y
600
2000 MW-
Dc
800
1000'
Length of line (mi)
Fig. 7. Reactive power requirements of long ehv overhead ac and dc lines at full load
as a function of the length of line. (That of ac lines is from data in Ref. D4.)
reactors varies almost linearly with distance, as shown in Figure 7. It is

approximately 4.4 Mvar/MW kmi 2.8 Mvar/MW Mm.
A dc line itself requires no reactive power, and the voltage drop on the line
itself is merely the resistive drop RI. The converters at both ends of the line,
25
however, draw reactive power from the ac systems. It varies with the trans
mitted power and is approximately half of the latter at each end. It is in
dependent of the length of line. Usually shunt capacitors or synchronous
condensers are installed for supplying this reactive power.
Both ac and dc lines have the disadvantage of requiring adjustable supplies
(or sinks) of reactive power. For distances of more than 400 km (250 mi),
however, Figure 7 shows that the dc line requires less than the ac line.
On submarine or underground cables, the situation is different from that on
overhead lines. Cables are always operated at a load much below the surge
impedance load in order to avoid overheating. Consequently the reactive
power produced by charging the shunt capacitance greatly exceeds that con
sumed by the series inductance.02 In a 50- or 60-Hz cable, 25 to 50 mi (40 to
80 km) long, the charging current alone equals the rated current, leaving no
margin for load current. Shunt compensation theoretically could correct this
situation.03 Shunt reactors, however, would be required at, perhaps, 10-mi
(16-km) intervals. Since it is difficult to lay and repair submarine cable to
which shunt reactors are connected, the practical length of ac submarine
cables is only about 20 mi (30 km). Dc cables have no such limitation.
Stability
By the stability of an ac system is meant its ability to operate with all
synchronous machines in synchronism. If a long ac line is loaded to a certain
value, known as its steady-state stability limit, the synchronous machines at
the sending end accelerate and go out of synchronism with those at the
receiving end. This condition is analogous to a slipping belt or clutch in a
mechanical transmission system. The slipping electrodynamic system not
only fails to transmit the power that it should but also gives rise to objection
able fluctuations in voltage.
Even if a line is operated below its steady-state limit, the machines at the
sending and receiving ends may lose synchronism after some large disturbance,
notably a short circuit, unless the line is operated below its transient stability
limit, which is always lower than the steady-state limit. Practically speaking,
the steady-state stability limit is the transient stability limit for very small
disturbances.
The problem of stability or synchronous operation constitutes the most
serious limitation of a long ac transmission system.
The power transmitted from one machine to the other in a two-machine
lossless ac power system is given by
pMsin5
(7)
26
1-5
where Ex and E2 are the internal voltages of the two machines, <5 is the phase
difference of these voltages," and X is the reactance of the architrave of the
equivalent n circuit of the system joining the internal points. Each machine
is represented by an internal voltage "behind" an internal reactance. The
reactance X is very nearly the sum of the inductive reactances inside the two
machines, of the transmission line, and of the step-up and step-down trans
formers. An actual power system involving a long interconnecting line with
many generating stations at each end of the line may be represented reason
ably well by a two-machine equivalent system in which all the machines at the
receiving end of the line are replaced by one equivalent machine and all those
at the sending end by another.
A graph of power P as a function of phase difference 5 between internal
voltages is a sine wave. Maximum power occurs at 3 = 90 and is
A,= exe2
X
(8)
Pm is the steady-state stability limit. It is approximately equal to the square of

the operating voltage divided by the series reactance. In a long-distance
transmission system most of the reactance is in the line itself, and a much
smaller part is in the two terminal systems, consisting of machines, trans
formers, and local lines. The inductive reactance of a single-circuit 60-Hz
overhead line with single conductors is about 0.8 Q/mi (0.5 Q/km); with
double conductors, about ias great. The reactance of the line itself is pro
portional to the length of the line, and thus the power per circuit of a given
voltage, as limited by steady-state, stability, is inversely proportional to the
length.
The transient stability limit is lower than the steady-state limit, and, as a
rough guide, we may take the former as half of the latter, corresponding to
a phase difference of sin-1 0.5 = 30 in the initial steady state. (This value
was assumed in Figures 6 and 7.)
In an uncompensated line operating at its natural load the phase of the
voltage varies directly with the distance, going through one cycle (360) per
wavelength. A 30 difference, then, corresponds to
wavelength. On a 60-Hz
line this is 3100/12 = 258 mi (416 km). On a 50-Hz line it is 310 mi (500 km).
A lighter load can be transmitted farther : a heavier load not so far.
The distance to which the natural load can be transmitted stably can be
extended considerably by placing synchronous condensers or, better yet,
synchronous generators at various intermediate points of the transmission
system. If both generators and loads are scattered along the transmission
system, this method of long-distance transmission is called transmission by
displacement. As shown in Figure 8, the over-all transmission can be regarded
"
500 MW
27
rT~
500 MW
500 MW
500 MW
500 MW
500 MW
500 MW
500 MW
500 MW
500 MW
1000 MW
wO
500 MW
1000 MW
Fig. 8. Long-distance transmission by displacement regarded (a) as several short trans

mission systems in tandem and (b) as a long-distance straightaway transmission system
supported by several intermediate generators each having its local load. Losses are
neglected.
either as several short transmission systems in tandem or as a long trans

mission line supported by several intermediate generating stations, each
having its own local load.
Perhaps the most economical method of increasing the distance of straight
away ac transmission is by use of series capacitors, whose reactance compen
sates a part of the series inductive reactance of the line itself. The maximum
part that can be compensated feasibly or economically has not yet been
determined. Probably it is about 75%. By use of this assumed maximum series
compensation, the distance for stable 60-Hz transmission of the natural load
of an overhead line could be extended to 258/(1 0.75) = 1030 mi (1660 km).
Such amounts of series compensation have not yet been used: 35 to 50% is
more usual. For straightaway transmission of 1000 mi (1600 km) dc trans
mission would prove more economical than ac.
Another method of making very long ac power lines operate stably has
been proposed : it is to make the line electrically somewhat longer than onehalf wavelength.01'5 ~7 It will then behave as if it were a half-wavelength
shorter than it is. If the actual distance is less than one-half wavelength, the
electrical length may be artificially increased in either of two ways: (a) by
adding lumped LC sections at the ends or (b) by connecting shunt capacitors
at intervals along the line.
A dc transmission link in itself has no stability problem. Two separate ac
systems interconnected only by a dc line do not operate in synchronism, even
if their nominal frequencies are equal, and they can operate at different
nominal frequencies, for example, one at 50 and the other at 60 Hz. Each of
the separate ac systems may have its own internal stability problems. The
sustained interruption of the power on the dc line constitutes a mild threat "to
28
stability equal to that caused by loss of a large load in the sending-end

system and to loss of a generator in the receiving system. Alternating-current
systems are designed so as to be stable under such mild shocks.
If the two ac systems are interconnected by one or more ac lines in addition
to a dc line of comparable rating, sudden and sustained interruption to the
power on the dc line may result in a loss of synchronism between the two ac
systems. Therefore parallel operation of one dc line and one or more ac lines
is inadvisable unless the ac lines are strong enough to withstand the loss of the
dc line.
If, however, there are two or more dc lines in parallel with one or more ac
lines, the dc lines can be so arranged that if one of them is lost, the other dc
line or lines assumes its load. In such a case, there is no great stability
problem.
In this regard each pole of a bipolar line may be considered a separate line.
Circuit Breakers
Alternating-current circuit breakers take advantage of the current zeros

that occur twice per cycle. They are designed to increase the breakdown
strength of the arc path between contacts so rapidly that the arc. does not
restrike. Direct-current circuit breakers do not have this natural advantage
and therefore have to force the current to zero. So far no successful dc
circuit breaker has been built for the high voltages and high currents used in
dc transmission.
In simple two-terminal dc transmission, such as all projects in operation to
date have been, the lack of dc circuit breakers has not been felt, because
faults on the dc line or in the converters are cleared by using the control
grids of the converter valves to block the direct current temporarily. Experi
ence with ac transmission, however, has shown that most lines that initially
operate radially later became incorporated into an ac network. The lack
of a dc circuit breaker is a handicap to the tapping or networking of dc lines.
Reasonable proposals have been made for the operation of a three- or fourterminal line in which a faulted line section can be switched out by running
the voltage of the whole system to zero, opening switches to isolate the
faulted section, and then raising the voltage back to normal. The time of the
whole sequence of events would be approximately equal' to that now re
quired for rapid reclosure of ac circuit breakers.
Nevertheless, the lack of hv dc circuit breakers must be regarded as a
present limitation of hv dc transmission. It is likely that such circuit breakers
will be developed.*
* Some development work is described in Section 7-7.
1-5
29
Short-Circuit Current
The interconnection of ac systems through an ac line raises the shortcircuit currents, sometimes to an extent that exceeds the interrupting capa
bility of existing circuit breakers and requires their replacement by more
capable breakers. The interconnection of ac systems by a dc link, however,
does not increase short-circuit currents of the ac systems nearly so much, for
the dc line contributes no current to an ac short circuit beyond its rated
current.
On the other hand, the proper operation of a dc line terminal requires that
the short-circuit power of the ac system at the point of installation be several
(now at least five) times the rated power of the dc line, and sometimes this
requirement dictates increase of the ac short-circuit power by the provision of
synchronous condensers or additional ac connections.
The current in a short circuit on the dc line, after a momentary transient
due to a discharge of the shunt capacitance of the line, is limited by automatic
grid control to twice rated current. Nor do faults on the dc line draw ex
cessive currents from the ac systems.
Power per Conductor and per Circuit
Let us assume that an ac line and a dc line using the same conductors and
insulators are built. How does the power per conductor compare on the two
lines?
Assume that in each case the current is limited by temperature rise. Then
the direct current equals the rms alternating current.
Assume also that the insulators withstand the same crest voltage to ground
in each case. Then the direct voltage is 2 times the rms alternating voltage.
The dc power per conductor is
-J
(9)
Pi=VJ*
and the ac power per conductor is
Pa =
VJa COS 0
(10)
where Id and Ia are the curents per conductor, Vd and Va the conductor-toground voltages, and cos 0 the power factor. The ratio is
pd
VdId
_ Vd h
pa
VaIa cos 0
Vg
Ia
1
cos 0
_ 72
cos 0
(11)
Taking cos0 = 0.945, pjpa = 1.5.

Now compare a three-phase, three-conductor ac line with a bipolar two>
30
1-5
conductor dc line. The power capabilities of the respective circuits are
The converters required at both ends of a dc transmission link have proved

to be reliable but expensive. They also constitute a bottle neck to the power
transmissible, for the valves have but little overload capability.

Other terminal equipment on either ac or dc lines may limit the voltage or
current, hence the power; for example, the voltage and continuous current
rating of circuit breakers and the seal-off voltage of lightning arresters.
and the ratio is

=
Pa
3 2
. 31
Terminal Equipment
Pd ~ 2Pd
pa = 3p
=
Pa 3
(12)
Both lines can carry the same power. The dc line, however, is simpler and
cheaper, having two conductors instead of three. Consequently an overhead
line requires only |
as many insulators, and the towers are simpler, cheaper,
and narrower. A narrower right of way could be used.
Both lines have the same power loss per conductor. The percentage loss
of the dc line is only two-thirds that of the ac liner If the basis of comparison
is equal percentage loss, the power of the three-phase ac line is decreased to
2/3 that of the two-conductor dc line.
If cables are used instead of overhead line, the permissible working stress
(voltage per unit thickness of insulation) is higher for direct current than for
alternating current, and, in addition, the power factor for direct current is
unity and, for alternating current, considerably lower than that assumed
above. Both changes further favor direct current over alternating current by
increasing the ratio of dc power to ac power per conductor. The resulting
ratio might be from 5 to 10.
Because the power limit of overhead ac lines is. often determined by factors
other than conductor heating, the ratio of dc power per conductor to ac
power per conductor may be as high as 4.
Harmonics
The converters used with a dc line produce harmonic voltages and currents
on both ac and dc sides.These harmonics, especially in the extensive ac net
works, may cause interference with audio-frequency telephone lines. Filters
are required on the ac side of each converter for diminishing the magnitude
of harmonics in the ac networks. These increase the cost of the converter
stations. Fortunately the capacitors used in the filters also supply part of the
reactive power required by the converters. The cost of the filters and of the
additional reactive power supply should be regarded as a part of the cost of
a dc line terminal.
Control of Tie-Line Power
The power flow on tie lines interconnecting different areas under different
ownerships must be controlled in conformity with contractual obligations.
In addition, the frequency of the whole system, or the frequencies of the
parts connected asynchronously, must be controlled.
The control system is a little simpler if the tie lines operate on dc than if on
ac, but the difference is not important. This subject is discussed in Volume 2.
Ground Return
A two-conductor bipolar dc line is more reliable than a three-conductor ac
line, because, in the event of a fault on one conductor, the other conductor
can continue to operate with ground return during the period required for
repairing the fault. The operation of an ac line with ground return is not
feasible on account of the high impedance of such a circuit and the telephone
interference caused by such operation. Further information on dc ground
return is given in Chapter 9.
A monopolar dc line with earth return is still simpler than a three-phase ac
line and is equally reliable. It is especially suitable to submarine cable. A line
cart be built in stages with monopolar operation initially, later changed to
bipolar operation with doubling of the power rating.
Generating Units
Some hydroelectric generating stations connected to a load center through
long ac lines have generators with abnormally low transient reactance or
abnormally high moment of inertia specified in order to raise the stability
limit. These features raise the cost of the generators and would not be re
quired if dc transmission were used, for there would be no stability problem
with direct current. In addition, if such a station were connected to an ac
system only through dc lines, the speed of the prime movers could be allowed
to vary with the load or the head of water, perhaps giving a cheaper or a more
efficient prime mover, and the nominal frequency of the generator, no longer
confined to 50 or 60 Hz, could be chosen for best economy. Perhaps also, in
32
1-8
1-6 SUMMARY OF ADVANTAGES AND DISADVANTAGES OF HV DC

TRANSMISSION
impractical.
5. And, of course, combinations of these factors occurring in the same
project.
Advantages
Six of the first seven commercial installations, beginning with Gotland,

involve submarine cables. All but the first two of these include great lengths
of overhead line in addition to cables.
In the English Channel crossing and in the Konti-Skan scheme asyn
chronous operation was preferred because of the simplcity and economy of
control. Some installations of converters similar to those used for hv dc
transmission have been installed for frequency conversion with no dc line.
In the United States and the U.S.S.R. the principal interest in hv dc
transmission is for long overhead lines.
In Britain there is much interest in dc transmission by underground cable
through metropolitan areas, especially London. The first such scheme is
Kingsnorth. It is likely that such applications will be considered in large
cities in the United States in the future.
Greater power per conductor.

Simpler line construction.
Ground return can be used.
Hence each conductor can be operated as an independent circuit.
No charging current.
No skin effect.
Cables can be worked at a higher voltage gradient.
Line power factor is always unity; line does not require reactive compensa
'
tion.
Disadvantages
Converters are expensive.

Converters require much reactive power.
Converters generate harmonics, requiring filters.
Converters have little overload capability.
Lack of hv dc circuit breakers hampers multiterminal or network operation.
1
1-7 PRINCIPAL APPLICATIONS OF DC TRANSMISSION
The foregoing discussion of the advantages and shortcomings of

transmission indicates the following applications:
1. For cables crossing bodies of water wider than 20 mi (32 km).
hv
dc
33
2. For interconnecting ac systems having different frequencies or where

asynchronous operation is desired.
3. For transmitting large amounts of power over long distances by over
head lines.
4. In congested urban areas or elsewhere where it is difficult to acquire
right of way for overhead lines and where the lengths involved make ac cables
such a station, less harmonic filtering would be required. (The Volgograd

hydroelectric plant has no filters.) Altogether, the generating plant could be
designed for best economy. To date, however, no such plant has been built.
Less corona loss and radio interference, especially in foul weather, for a
certain conductor diameter and rms voltage.
Synchronous operation is not required.
Hence distance is not limited by stability.
May interconnect ac systems of different frequencies.
Lcjw short-circuit current on dc line.
Does not contribute to short-circuit current of ac system.
Tie-line power is easily controlled.
ECONOMIC FACTORS
\
1
|
I
1-8 ECONOMIC FACTORS
The cost per unit length of a dc line is lower than that of an ac line of the
same power capability and comparable reliability, but the cost of the terminal
equipment of a dc line is much more than that of an ac line. If we plot the
cost of transmitting a certain amount of power by one method or the other
as a function of the distance over which it is transmitted, the resulting graph
is similar to Figure 9. The vertical intercept of each curve is the cost of the
terminal equipment alone. The slope of each curve is the cost per unit length
of the line and of that accessory equipment which varies with the length.
The curve for ac transmission intersects that for dc transmission at an abscissa
called the break-even distance. If the transmission distance is shorter than the
break-even distance, ac transmission is cheaper than dc; if longer, dc is
cheaper than ac.
Estimates of the break-even distance of overhead lines, published in the
technical literature, range from 500 km (310 mi) to 1500 km (930 mi). Such
great variation can be explained, at least in part, by a simple modification
34
1-9
Break-even
|
-distance, -|
500 mi
|
-iTerminal costs
Tt
i
200
iii
400
600
Distance (mi)
1
800
1000
Fig. 9. Comparative costs of ac and dc overhead lines versus distance.
of Figure 9, shown in Figure 10. Here the cost of each line is assumed to
5% for the ac line and 10% for the dc line. The
true cost of each is assumed to be within the crosshatched area. (Greater
variation is assumed for the cost of dc transmission than for that of ac
because there has been less experience with dc than with ac.) Itis now apparent
that even such small variations in estimated costs make the estimated break
even distance vary over a range of 2 or 3 to I.
For cables the break-even distance is, of course, much shorter than for
overhead lines, lying between 15 and 30 mi (24 and 48 km) for submarine
cables and, perhaps, twice as far for underground cables.
The ordinate in Figures 9 and 10 might be either capital cost or annual
vary over a certain range,
Range of break-even distance
400
600
Distance (mi)
Fig. 10. Effect of variation of costs on break-even distance.
1000
THE FUTURE OF DC TRANSMISSION
35
cost; it might be for a given amount of power or per megawatt. In any case
the curves would have the same form.
In view of the relative novelty of hv dc transmission, there is a prospect
for a greater decrease in the unit cost of dc line terminals with increasing
experience and volume of production than in the cost of ac equipment. The
result would be to decrease the break-even distance.
An economic comparison between ac and dc transmission made by an
international working party of C.I.G.R.E.07 and based on 1965 costs showed
average break-even distances of 1000 km (600 mi) for transmitting 1080 or
2160 MW on two overhead circuits and 77 km (48 mi) for transmitting 1080
MW on two shunt-compensated underground cable circuits. An assumed
future 20% reduction in dc terminal costs reduced the break-even distance to
830 km (515 mi) for the overhead lines and to 64 km (40 mi) for the under
ground cables.
In the great majority of dc transmission schemes already built, other
factors than the costs assumed in such comparisions play a significant role.
These other factors are long water crossings, frequency conversion, and the
advantage of asynchronous ties between large ac systems.
1-9 THE FUTURE OF DC TRANSMISSION6
The increasing size and load density of metropolitan areas create problems
of right of way for hv overhead lines. The increased public demand for the
better appearance of electric lines and for the preservation of the natural
environment is putting pressure on the electric power companies for placing
transmission and distribution lines underground, out of sight, even where the
load density is not high. Dc cables are cheaper and more compact than ac
cables for the same power and are not so limited in the feasible distance of
transmission.
As Greber discerningly points out,E3 the basic problem of ac transmission
is that of inductive and capacitive reactance; the basic problem of dc trans
mission is switching.
It is the series inductive reactance of long overhead ac lines that causes the
synchronous stability limit. It is the shunt capacitive reactance of long ac
cables that overloads them with charging current. On long overhead lines,
the presence of both kinds of reactance causes excessive variation of voltage
with load.
Series and shunt compensation of reactance are used on long ac lines, but
they add to the cost and complexity of such lines. Reactive compensation is
not required on a dc line itself, but only on the ac side of the converters. This
fact gives an advantage to long dc lines over long ac lines. If cheaper and
36
BIBLIOGRAPHY
simpler means of compensation were developed, however, the economic

balance between ac and dc transmission would be shifted in favor of ac.
The switching problem on dc lines lies not only in the need for dc circuit
breakers but also in the converters, which are essentially a group of syn
chronously controlled switches. If cheaper, simpler, and more reliable
switches (perhaps, solid-state devices) were developed, not only would dc
networks be feasible, but also the converters would be cheaper than they are
now and more free from misoperations, such as arcbacks. Improved switches
would make the control of the reactive power of converters possible, per
mitting it to flow in or out of the converter, or neither, as desired. In addition,
dc transformers would be possible. They could operate on either of two
principles. One kind would be analogous to the vibrator power supplies now
used with battery-operated radios, but, of course, at a much greater power
level and with the vibrator replaced by a new kind of switch. The other kind
would rapidly switch capacitors so as to be charged in parallel and discharged
in series for voltage step-up or, vice versa, for step-down. Thus the develop
ment of superior switches could give great impetus to dc transmission.
Other impending developments could alter the picture in favor of direct
current. The new methods of power generation thermoelectric, magnetohydrodynamic,5 and by fuel cellinherently generate direct current. There
is some possibility that direct conversion from nuclear energy to hv direct
current might be developed.1 Cryogenic superconducting cables might
transmit direct current long distances at low voltage and high current with no
voltage drop and no power loss except that required to remove from the
cable the heat that leaked into it from its surroundings.9* Superconducting
dc generators and motors are being developed.4
The future of dc transmission looks bright.
'
2.
A. General
1. Power Transmission by Direct Current, by Ya. M. Chervonenkis, Moscow, 1948,

translated from the Russian by N. Kaner and published for the National Science
Foundation, Washington, D.C., by the Israel Program for Scientific Translation,
Jerusalem, 1963. Elementary and out-of-date.
* Refrigerated cables, either superconducting or resistive, are being actively investigated

for ac underground transmission as well as for dc.E2>7'8, 10-16 Direct current, however has
an inherent advantage in such cables.
Direct Current, a magazine published by Direct Current, Ltd. (asubsidiaryof Garraway

Ltd.), London, from June 1952 to February 1967, quarterly, except from March 1961
to October 1963, when it was monthly. New series published by Pergamon, Oxford,
beginning in April 1969, with editorial office at Manchester University, Department
of Electrical Engineering and Electronics.
i
f
j
s
\
1
j
j
j
)
I
BIBLIOGRAPHY
37
j
|
I
!
|
;
;
3. Nauchno-IzsledovateVskii Institut Postoyannovo Toka, Izvestiya (Proceedings of the

Direct Current Research Institute), Leningrad. Vol. 1 is dated 1957. Approximately
one volume per year has been published since then. Contains articles on both ac and
dc transmission, in Russian. Cited hereinafter as N.l.I.P.T.
4. " D.C. Power Transmission," a series of six articles published in Elec.Jour. (London),
Vols. 163 and 164 (1959-1960).
Part I, "Historical Development," by E. Openshaw Taylor, pp. 1227-1231, Dec. 4,
1959.
Part II, "Basic Principles," by E. Openshaw Taylor, pp. 22-27, Jan. 1, 1960.
Part III, "Rectifiers and Inverters," by R. Feinberg, pp. 294-299, Jan. 29, 1960.
Part IV, "Transmission Circuits," by A. L. Williams, pp. 619-626, Mar. 4, 1960.
Part V, "Operation and Control," by Gunnar Engstrom, pp. 1048-1055, Apr. 8,
1960.
Part VI, "Planning and Economics," by J. L. Egginton, pp. 1271-1280, May 6, 1960.
5. High Voltage Direct Current Power Transmission, by Colin Adamson and N. G.
Hingorani, Garraway, London, 1960, xvi + 284 pp.
6. High Voltage Direct Current Convenors and Systems, edited by B. J. Cory, Macdonald, London, 1965, xiii + 269 pp.
7. Conference on High Voltage D.C. Transmission, held at Manchester, Sept. 19-23, 1966,
I.E.E. Conference Publication 22, London, 2 parts. Part 1, Contributions, 454 pp.
Part 2, Discussions, 143 pp.
B. History
1. "Constant-Current D.C. Transmission," by C. H. Willis, B. D. Bedford, and F. R,
Elder, Elec. Eng., Vol. 54, pp. 102-108, January 1935. Disc., pp. 327-329 (March),
447-449 (April), and 882-883 (August).
2. "Power Transmission by Direct Current: Apparatus Used in 3000-kw 15,000-volt,
200-amp Pump-back Test," by B. D. Bedford, F. R. Elder, and D. H. Willis, Gen.
Elec. Rev., Vol. 39, pp. 220-224, May 1936. Tests preceding the MechanicvilleSchenectady experimental dc transmission.
3. "D.C. Transmission in France," Elec. World, Vol. 106, No. 19, pp. 1341-1342,
May 9, 1936. On 275-mi, 125-kV, 20-MW line from Moutiers to Lyon.
4. "The First Power Transmission at 50 kV D.C. with Mutators" (in French), by
P. Egloff and J. J. Felix, Electricity, Vol. 23, No. 58-59, pp. 237-240, July-August
1939. Baden-Zurich transmission.
5. "The D.C. Power Transmission at the Swiss National Exhibition" (in German),
by E. Kern, Bull, de V Association Suisse des Electriciens, Vol. 30, No. 17, pp. 481-482,
Aug. 18, 1939.
6. "H.V.D.C. Transmission," by F. Busemann, Elec. Times, Vol. Ill,Nos. 2881,2883,
2885, pp. 3640, 98-101, 169-170, Jan. 9 and 23, Feb. 6, 1947. Experience in Germany
during World War II,including Charlottenburg-Moabit experimental 110-kV dc trans
mission and Elbe-Berlin project. English translation from German report.
>
j
38
7. "D.C. Power Transmission Developments by the Siemens-Schuckert Concern in

Germany," by F. Busemann, B.E. & A.I.R.A., Report Z/T67, Nov. 24, 1947, 20 pp.
8. Item A1 above.
9. " Origin of the 440 kV D.C. H.V. Transmission Line Elbe-Berlin " (in German), by
R. Troger, ETZ, Vol. 69, pp. 261-272, August 1948.
10. " Experience of High-Voltage Direct Current Transmission" (in Russian), by A. M.
Nekrasov and M. R. Sonin, Elek. Stantsii, Vol. 26, No. 7, pp. 26-32, July 1955.
Kashira-Moscow experimental cable transmission.

11. " H.V.D.C. Transmission System "(in Russian), by V. P. PimenovandM. R. Sonin,
Elektrichestvo, No. 7, pp. 93-99, 1955. Kashira-Moscow link.
12.. "Institute's Activities in the Field of High-Voltage Direct Current Transmission of
Energy" (in Russian), by V. P. Pimenov, N.I.I.P.T., Vol. 1, pp. 7-20, 1957. Work of
of the Institute of DC Transmission.
13. " Results of the Operation of the Experimental Industrial Direct Current Transmission
Line, Kashira-Moscow " (in Russian), by M. R. Sonin, N.I.I.P.T., Vol. 2, pp. 5-21,
1957. Report of operating problems, December 1950 to May 1956.
14. " The Work of the Direct Current Institute," by V. P. Pimenov, Direct Current,
Vol. 3, No. 6, pp. 185-191, September 1957. Translated from N.I.I.P.T., Vol. 1.
15. "D-C Transmission: An American Viewpoint," by G. D. Breuer, M. M. Morack,
L. W. Morton, and C. A. Woodrow, A.I.E.E. Trans., Vol. 78, Part 3A, pp. 504-512,
August 1959. Disc., pp. 512-515. Includes information on Mechanicville-Schenectady
link.
16. " Work Done in the Soviet Union on High-Voltage Long-Distance D-C Power Trans
mission," by A. M. Nekrasov and A. V. Posse, A.I.E.E. Trans., Vol. 78, Part 3A,
pp. 515-521, August 1959. Disc., pp. 521-522. Kashira-Moscow experimental trans
mission, development work and plans for the Stalingrad-Donbass line.
17. "D.C. Power Transmission," Part Iof item A4 above.
18. "The History of D.C. Transmission," in Direct Current:
Part I, Vol. 6, pp. 260-263, December 1961.
Part II, Vol. 7, pp. 60-63, March 1962.
Part III, Vol. 7, pp. 228-231 and 250, September 1962.
Part IV, Vol. 8, pp. 2-5 and 27, January 1963.
Part V, Vol. 8, pp. 88-93 and 115, April 1963.
Includes a bibliography of 160 entries.
19. "Development of High Voltage D.C. Transmission at Siemens Schuckertwerke
up to 1945 " (in German), by M. Bosch and O. Schiele, Siemens Zeitschrift, Vol. 40,
pp. 672-681, September 1966.
BIBLIOGRAPHY
'
39
3. Comparison of Direct and Alternating Current for High-Voltage Electric Power Trans
mission, Edison Electric Institute Publication No. 62-901, 1962.
4. "High-Voltage DC Transmission," Advisory Committee Report No. 20, published
on pp. 289-313 of NationalPower Survey, a report by the Federal Power Commission,
part II, "Advisory Reports," U.S. Government Printing Office, Washington,
October 1964.
5. "Cost of Electrical Energy Transmission by AC and DC Extra High Voltage,"
Advisory Committee Report No. 16, pp. 189-203, loc. cit.
6. "High Capacity D.C. Transmission in the U.S.S.R.," by A. Berkovski, N. Chouprakov, T. Izrailevich,A. Kolpakova, and S. Rokotyan, I.E.E. Conference Publication 22,
H.V.D.C. Transmission, Manchester, Sept. 19-23, 1966, Part 2, Paper No. 94,
pp. 126-129. Also in Direct Current , Vol. 11, pp. 145-149, November 1966.
7. "A Technical and Economic Comparison between A.C. and D.C. Transmission,"
by W. Casson, C.I.G.R.E., 1968, Report 42/43-01, 42 pp.
D. Special Problems of AC Transmission
j
1
i
I
j
j
|
\
1
I
I
j
1. "On Normal Working Conditions of Compensated Lines with Half-Wave Charac

teristics " (in Russian), by A. A. Wolf and O. V. Shcherbachev, Elektrichestvo, No. 1,
pp. 57-60, 1940.
2. " Charging Current Limitations in Operation of High-voltage Cable Lines," by C. S.
Schifreen and W. C. Marble, A.I.E.E. Trans., Vol. 75, Part 3, pp. 803-812, October
1956. Disc., pp. 813-817.
3. "Long Cable LinesAlternating Current with Reactor Compensation for Direct
Current," by J. J. Dougherty and C. S. Schifreen, A.I.E.E. Trans., Vol. 81, Part 3,
pp. 169-178, June 1962. Disc., pp. 179-182.
4. "EHV AC Transmission Line Compensation," Advisory Committee Report No. 14,
prepared by a subcommittee of the Transmission and Interconnection Special Technical Committee, July 1963, published on pp. 141-172 of National Power Survey,
a report by the Federal Power Commission, Part II, "Advisory Reports," U.S.
Government Printing Office, Washington, October 1964.
5. "Half-Wavelength Power Transmission Lines," by F. J. Hubert and M. R. Gent,
I.E.E.E. Spectrum, Vol. 2, pp. 87-92, January 1965. Also in I.E.E.E. Trans, on
P.A. & S., Vol. 84, pp. 965-974, October 1965.
6. "Analysis of Natural Half-Wave-Length Power Transmission Lines," by F. S.
Prabhakara, K. Parthasarathy, and H. N. Ramachandra Rao, I.E.E.E. Trans, on
P.A. & S., Vol. 88, pp. 1787-1794, December 1969. Disc., p. 1794.
7. "Performance of Tuned Half-Wave-Length Power Transmission Lines," by authors
of last item, ibid., pp. 1795-1800. Disc., pp. 1800-1802.
C. Comparison Between AC and DC Transmission
1. "Comparison of Transmission Costs for High-Voltage AC and DC Power Trans

mission in Japan," by Sadao Saeki, Appendix II to "Report on the Work of the
Study Committee No. 10: D.C. Transmission at E.H.V.," C.I.G.R.E., 1956, Paper 507.
2. "Comparison of Transmission Costs for High Voltage A.C. and D.C. Systems"
(presented in the name of Study Committee No. 10, D.C. Transmission Lines at
E.H.V.), by F. J. Lane, Bo G. Rathsman, U. Lamm, and K. S. Smedsfelt, C.I.G.R.E.,
1956, Paper No. 417, 25 pp.
)
|
E. Future Prospects for DC Transmission

1. "Foreword," by Max Hoyaux, Direct Current, Vol. 2, pp. 133-134, September 1955.
"An 'Atomic Battery': Direct Conversion from Atomic Radiation to Electrical
Engergy," ibid., pp. 135-137.
2. "Applications of Superconductivity to the Generation and Distribution of Electric
Power," by Richard McFee, Elec. Eng., Vol. 81, pp. 122-129, February 1962.
Superconductors, heat insulation, transformers, cables, generators and motors, fuses,

circuit breakers, rectifiers, and refrigeration.
40
3. "Future Developments in H.V.D.C.," letter to the editor of Direct Current from

Henry Greber, published in Vol. 9, inside front cover, August 1964.
4. "Superconducting DC Generators and Motors," by David L. Atherton, I.E.E.E.
Spectrum, Vol. 1, pp. 67-71, December 1964.
5. " Survey of MHD Research," introduced by M. W. Thring, Direct Current, Vol. 10,
No. 1, pp. 40-59, February 1965. Survey of research in Britain, Australia, United
States, France, Japan, Switzerland, Sweden, and Poland on magnetohydrodynamics.
6. "Future Possibilities of H.V.D.C.," by J. H. M. Sykes, Chapter 10 of Cory, Ref.
A6 above, 1965.
7. "Superconducting Power Transmission May Be a Reality Within Ten Years," by
D. Atherton, Elec. News and Engg. (Don Mills, Ontario), Vol. 74, No. 11, pp. 52-55,
November 1965.
8. "Prospect of Employing Conductors at Low Temperature in Power Cables and in
Power Transformers," by K. J. R. Wilkinson, I.E.E. Proc., Vol. 113, No. 9, pp. 1509
1521, September 1966. Disc., Vol. 114, No. 12, pp. 1892-1898, December 1967.
Estimate of power saved if the conductor in a 760-MVA, 275-kV ac cable were,
alternatively, niobium at 4K, aluminum at 20K, or beryllium at 77K. Summary of
this paper and of discussion of it in Elec. Times, Vol. 151, No. 5, pp. 168-170, Feb. 2,
1967, under the title, "Prospect for Low Temperature Transmission."
9. " Superconducting Lines for the Transmission of Large Amounts of Electrical Power
over Great Distances," by R. L. Garwin and J. Matisoo, I.E.E.E. Proc., Vol. 55,
No. 4, pp. 538-545, April 1967. Preliminary design of 1000-km, 100-GW, 200-kV,
500-kA, dc line with Nb3Sn conductors refrigerated to 4K.
10. "Superconducting Power Cables," by D. R. Edwards and R. J. Slaughter, Elec.
Times, Vol. 152, No. 5, pp. 166-169, Aug. 3, 1967. Includes summary of historical
development.
"Design
for a 750 MVA Superconducting Power Cable," by E. C. Rogers and D.R.
11.
Edwards, Elec. Rev., Vol. 181, No. 10, pp. 348-351, Sept. 8, 1967. Study made by
B.I.C.C. for C.E.G.B. on design of three-phase, 33-kV, 13-kA, superconducting
cable. Conductors of 0.0025-cm niobium foil on 0.25-cm aluminum tubing, vacuum
12.
13.
14.
15.
dielectric, and liquid helium coolant.

"Cryogenic Power Transmission," by S. H. Minnich and G. R. Fox, Cryogenics,
Vol. 9, No. 3, pp. 165-176, June 1969. Based on studies by General Electric Company
for Edison Electric Institute and Tennessee Valley Authority. Considers both resis
tive cryogenic cable with stranded aluminum conductors in liquid nitrogen or
hydrogen and superconducting cable with niobium-coated tubes in liquid helium,
especially for three-phase ac.
"Low Temperatures and Electric Power," by B. J. Maddock, W. T. Norris, D. A.
Swift, and M. T. Taylor, Cryogenics, Vol. 9, No. 4, pp. 291-297, August 1969. Report
on a conference organized by the I.E.E. and held in London on March 24-26, 1969.
It included papers on electric power systems, refrigeration, conductor materials, dielec
trics, generators and motors, transformers, cables, energy storage, and transportation.
"French Develop Modular HVDC Thyristor Valve," Elec. Rev,, Vol. 185, No. 22,
pp. 790-791, Nov. 28, 1969.News item on research atC.G.E.'s Marcoussis Laboratory,
which includes work on low-temperature cables.
"Economics of Underground Transmission with Cryogenic Cables," by Peter
Graneau, I.E.E.E. Trans, on P.A. & S., Vol. 89, pp. 1-7, January 1970. Disc., pp.
14-16. Three cryogenic ac transmission lines, cooled, respectively, by liquid nitrogen,
BIBLIOGRAPHY
41
hydrogen, and helium, are compared with one another and with conventional pipetype cable. The nitrogen-cooled cable is found to be the most economical.
16. "Economic Assessment of a Liquid-Nitrogen-Cooled Cable," by S. B. Afshartous,
Peter Graneau, and John Jeanmonod, I.E.E.E. Trans, on P.A. & S., Vol. 89, pp.
8-13, January 1970. Disc., pp. 14-16. Cable of tubular aluminum conductors cooled
internallyby liquid nitrogen and supported by dielectric spacers inhigh-voltage vacuum
insulation.
F. Bibliographies
1. "Direct Current Bibliography 1," Direct Current, Vol. 1, pp. 50-52, September
1952. Covers years 1943 to 1952. Vol. 1, p. 97, March 1953, covers rest of 1952.
An
Annotated Bibliography of High Voltage Direct Current Transmission, 1932-1962,
2.
compiled by Eric Bromberg, I.E.E.E. Paper CP 63-388, January 1963. Also in D-C
Transmission, Publication S-155, pp. 76-214, I.E.E.E., June 1963.
3. An Annotated Bibliography of High Voltage Direct Current Transmission, 1963-1965,
complied by Eric Bromberg, I.E.E.E. Publication 31 S 60, iii-f 113 pp., I.E.E.E.,
New York, 1967.
4. High Voltage Direct Current Transmission: An Annotated Bibliography, 1966-68, com
piled by Val S. Lava, published by the Library, Bonneville Power Administration,
Portland, Oregon, December 1968, ii 90 pp.
5
G. Gotland Link
|
|
|
;
1. "The High Voltage D.C. Power Transmission from the Swedish Mainland to the
Swedish Island of Gotland," by Ake Rusck, B. G. Rathsman, and U. Glimstedt,
C.I.G.R.E., Report 406, 1950.
"D.C.
Transmission from Swedish Mainland to Island of Gotland, by A. Rusck,
2.
B. G. Rathsman, and U. Glimstedt, Engineer, Vol. 190, No. 4931, pp. 92-93, July 28,
1950.
3. " High-Voltage D.C. Power Transmission Pioneer Project," by U. Lamm, ASEA
Journal, Vol. 23, No. 12, pp. 172-174, December 1950.
4. " Submarine Cable Project Will Operate at 100 kV D.C.," by B. G. Rathsman, Electric
Light and Power, Vol. 29, No. 8, pp. 108-109, August 1951.
5. "Gotland H.V.D.C. Link: Present Progress," by B. G. Rathsman and U. Lamm,
Direct Current, Vol. 1, pp. 2-6, June 1952.
6. "Gotland D.C. Link: Layout of Plant," by I. Liden, Sy Sviden and E. Uhlmann,
Direct Current, Vol. 2, pp. 2-7, June, and pp. 34-39, September 1954.
7. "The First High Voltage D.C. Transmission wigh Static Converters: SomeNoteson
the Development," by U. Lamm, ASEA Journal, Vol. 27, No. 10, pp. 139-140,
October 1954. "The Gotland D.C. Link: The Layout of the Plant," by I. Liden
and E. Uhlmann, pp. 141-154.
8. "The D.C. Transmission to Gotland: Initial Experience," by S. Ekefalk, ASEA
Journal, No. 10, pp. 123-126, 1956.
H. English Channel Link
1. "The Possibilities of a Cross-Channel Power Link between the British and French
42
2.
3.
4.
5.
6.
7.
8.
9.

Supply Systems," by D. P. Sayers, M. E. Laborde, and F. J. Lane, I.E.E. Proc.,
Vol. 101, Part 1, pp. 284-297, September 1954. Disc., pp. 297-308.
"English Channel: Channel Cable," by J. H. M. Sykes, Engineer (London), Vol. 202,
No. 5253, pp. 433-434, Sept. 28, 1956.
"The Design of the D.C. Connection across the English Channel,"- by I. Lid6n,
ASEA Journal,Vol. 36, No. 6, pp. 70-74, 1958.
"The High Voltage D.C. Transmission Scheme across the English Channel," by
I. Lid6n, ASEA Journal, Vol. 33, No. 7-8, pp. 124-126, 1960.
"The Cross-Channel Cable: A Preliminary Survey," by the editor, Direct Current,
Vol. 6, No. 4, pp. 97-109, July 1961.
"D.C. Channel Link: Lydd Operational," Elec. Times, Vol. 140, No. 23, pp. 845-848,
Dec. 7, 1961.
"Anglo-French Power Link," Engineer, Vol. 212, pp. 950-953, Dec. 8, 1961.
"Cross Channel Power Link," Elec. Rev. (London), Vol. 169, no. 23, pp. 907-912,
Dec. 8, 1961.
"Some Problems in Connection with the Commissioning of the Lydd Converter
Station," by L. Csuros and G. S. H. Jarrett, Direct Current, Vol. 7, No. 5, pp. 114
121, May 1962.
10. "Operational Performance oftheDirect Current Cross Channel Link," by J. Malaval,

J. Clade, L. Csuros, and G. S. H. Jarrett, C.I.G.R.E., 1964, Paper No. 417, 11 pp.
and folded chart.
11. " Some Design Aspects of the Cross-Channel Power Link," by L.A. Harris, Chapter 8
of Cory, Ref. A6, 1965.
12. "The Performance of the Lydd Convertor of the Cross Channel Connection," by
G. S. H. Jarrett and L. Csuros, I.E.E. Conference Publication 22, H.V.D.C. Trans
mission, held at Manchester on Sept. 19-23, 1966, Part 1, Paper No. 2, pp. 17-20
and folded sheet.

13. " Special Operational Tests onthe Cross Channel Connection," by J. Clad6, R. M. H.
Middleton, and E. Uhlmann, ibid., Paper No. 3, pp. 21-25. Transient conditions, such
as starting, blocking, emergency power reversal, and ac faults.
14. " Service Experience with the Anglo-French D.C. Cross Channel Cable," by P. Fourcade and C. C. Barnes, ibid., Paper No. 4, pp. 26-29.
I. Volgograd-Donbass Link
1. "D.C. Transmission from Stalingrad Hydro-electric Station to Donbass" (in Rus
sian), by V. P. Pimenov, A. V. Posse, A. M. Reider, S. S. Rokotian, and V. E. Turetskii,
Elektricheskie Stantsii, No. 11, 1956, pp. 12-18.
2. "Transmission of Direct Current at HighVoltage according to Present-Day Concepts
and the Prospects for Their Application in the U.S.S.R." (in Russian), by N. M.
Mel'gunov, N.I.I.P.T., Vol. 1, pp. 21-38, 1957.
3. " The Transmission System Stalingrad Hydro-electric Station,Donbass "(in Russian),
by E. S. Grois, M. L. Zelikin, V. E. Turetskii, andE. A. Man'kin, Elektrichestvo,Vol.
77, No. 9, pp. 1-10 September 1957.
4. "Design Features of Stalingrad-Donbass 800 kV D.C. Line," by F. I. Butaev, E. S.
Grois, E. K. Levitski, E. A. Man'kin, A. V. Posse, A. A. Sakovitch, and V. E.
Turetski, Direct Current, Vol. 4, No. 2, pp. 59-66, September 1958.
BIBLIOGRAPHY
43
5. " 800 kV D.C. Transmission System Stalingrad-Donbass," by E. S. Groiss, A. V.

Posse, and V. E. Touretski, C.I.G.R.E., 1960, Paper No. 414, 17 pp.
6. " 800 kv D.C. Transmission System Stalingrad-Donbass," by E. S. Grois, A. V. Posse,
and V. E. Turetskii, Engineer, Vol. 210, No. 5450, pp. 66-68, July 8, 1960. Based on
C.I.G.R.E. paper.
7. " Some Problems of the Operation of the D.C. Transmission Line, Stalingrad Hydro
electric Station to Donbass " (inRussian),by N.M. Mel'gunov and V. M.Kviatkovskii,
Elektrichestvo, No. 3, 1961, pp. 14-17.
8. "The Initial Operating Stage of the Volgograd-Donbass D.C. Transmission," by
E. S. Grois, N.I.I.P.T., No. 9, pp. 5-28, 1962.
9. "Initial Period of Operation of the D.C. Transmission Line between Volgograd and
Donbass," by N. Chuprakov, A. Milutin, A. Posse, and V. Shashmurin, I.E.E.
Conference Publication No. 22, H.V.D.C. Transmission, Manchester, Sept. 19-23,
1966, Part 2, Paper No. 93, pp. 120-125. Also in Direct Current, Vol. 11, No. 4, pp.
142-145 and 148, November 1966.
10. " Operation of the Control and Protection System of the Volgograd-Donbass Link,"
by K. Gusakovsky, A. Posse, and A. Reider, I.E.E. Conference Publication No. 22,
ibid., Part 2, Paper No. 95, pp. 130-135.
11. "Operating Experience of the Volgograd-Donbass D.C. Transmission Line and Its
Applications to Extra High Voltage D.C. High Capacity Transmission," by A. M.
Berkovski, F. I. Butaev, E. S. Grois, A. V. Posse, S. S. Rokotyan, and P. E. Saudler,
C.I.G.R.E., Report 43-07, 1968, 7 pp.
J. New Zealand Link
1. "Report on the Possibilities of Interconnecting the Islands of New Zealand," by

F. J. Lane, Direct Current, Vol. 5, pp. 12-24, June 1960.
2. " The H.V.D.C. Interconnectionbetween the Islands of New Zealand," Direct Current,
Vol. 7, pp. 32-38, February 1962.
The following series of papers (items 3 to 17) was published in New Zealand Engineering,
the journal of the N.Z. Institution of Engineers, Wellington, 1965 to 1966:
3. " Economic Aspects of the Inter-islandTransmission Scheme," editorial by E. B. M.,
Vol. 20, No. 6, p. 211, June 1965.
4. "A Significant Achievement," by P. W. Blakely,Vol.20, No. 7, pp. 255-256, July 1965.
5. " Main Generating and Electrical Equipment of Benmore Power Station," by H. C.
Hitchcock, Vol. 20, No. 1, pp. 3-13, January 1965.
6. "A Direct Current Transmission Line: The Design and Construction of the 600 MW,
500 kV d.c. Line Between Benmore and Haywards," by T. A. J. Dicicensj VoL-20,
No. 4, pp. 121-129, April 1965.
N
7. "The Benmore Land Electrode," by D. G. Dell, Vol. 20, No. 5, pp. 165-175, May
1965.
8. "The North Island Sea Electrode," by D. G. Deli, Vol. 20, No. 6, pp. 213-222, June
1965.
9. " General Layout of the Haywards Terminal," by R. J. Fyfe, Vol. 20. No. 8, pp. 303310, August 1965.
10. "Outdoor Direct Current Equipment at the Haywards Terminal," by J. Noble,
Vol. 20, No. 9, 345-354, September 1965.
44
BIBLIOGRAPHY
11. "Valves, Valve House, and Indoor Equipment at the Converter Stations," by M. A.
Louden, Vol. 20, No. 10, pp. 393-402, October 1965.
12. "Layout of the Direct Current Switchyards," by D. G. Young, Vol. 20, No. 11,
pp. 472-478, November 1965.
13. "Harmonic Phomenena," by G.H. Robinson,Vol.21, No. 1,pp. 16-29,January 1966.
14. " Synchronous Condenser Installation at Haywards Substation," by L. S. Y. Gock,
Vol. 21, No. 1, pp. 29-35, January 1966.
15. "Power Line Carrier Communications," by F. R. Swan, Vol. 21, No. 2, pp. 45-55,
February 1966.
16. "Commissioning and Early Operating Experience," by H. R. Gunn, Vol. 21, No. 3,
pp. 93-101, March 1966.
17. "The 250 kV d.c. Submarine Power-Cable Interconnection," by A. L. Williams,
E. L. Davey, and J. N. Gibson, Vol. 21, No. 4, pp. 145-160, April 1966. Also pub
lished in I.E.E. Proc., Vol. 113, No. 1, pp. 121-133, January 1966.
18. "The New Zealand 500 kV High-Voltage Direct-Current Project," by P. W. Blakely,
Amer. Power Conf. Proc., Vol. 28, pp. 850-859, April 1966.
The following papers are from I.E.E. Conference Publication 22, Conference on High
Voltage D.C. Transmission, Sept. 19-23, 1966, Manchester, Part 1:
19. " Commissioning, and Early Operating Experience with the New Zealand HVDC
Inter-Island Transmission Scheme," by H. R. Gunn, Paper No. 5, pp. 30-38.
20. "Benmore Power Station: Special Features for H.V.D.C. Transmission," by H. C.
Hitchcock, Paper No. 19, pp. 101-103.
21. "The Synchronous Condenser Installation at Haywards Sub-station for the BenmoreHaywards H.V.D.C. Transmission Scheme," by L. S. Y. Gock, Paper No. 52, pp.
265-267.
22. "Communications (Power Line Carrier Systems)," by F. R. Swan, Paper No. 63,
pp. 306-311.
23. "The Cook Strait 250-kV Cables," by E. L. Davey, Paper No. 64, pp. 312-314.
24. "A Direct Current Transmission Line: The Design and Construction of the 600 MW,
500 kV D.C. Line Between Benmore and Haywards," by T. A. J. Dickens, Paper
No. 72, pp. 343-346.
25. "Some Features of New Zealand's Inter-island H.V.D.C. Transmission," by R. J.
Fyfe, M. A. Louden, J. Noble, and D. G. Young, Paper No. 78, pp. 375-383.
26. "The Benmore Land Electrodes for the Benmore-Haywards H.V.D.C. Transmission
Scheme," by D. G. Dell, Paper No. 82, pp. 415-418.
27. " The NorthIsland Sea Electrode for the Benmore-Haywards H.V.D.C. Transmission
Scheme," by D. G. Dell, Paper No. 85, pp. 427-430.
28. "Experience with Harmonics New Zealand H.V.D.C. Transmission Scheme," by
G. H. Robinson, Paper No. 89, pp. 442-444.
29. "Operational Experience of the Benmore-Haywards HVDC Transmission Scheme,"
by M. T. O'Brien, C.I.G.R.E., report 14-03, 1970, 11 pp.
K. Konti-Skan Link
1. "The Conti-Skan HVDC Project," by G. von Geijer, S. Smedsfelt, and L. Ahlgren,
I.E.E.E. Conference Paper CP 63-1056, January 1963.
45
2. "The Konti-Skan Project," by G. von Geijer, Direct Current. Vol. 8, pp. 149-51,
June 1963.
3. "The Konti-Skan H.V.D.C. Project," by G. von Geijer, S. Smedsfelt, L. Ahlgren,
and E. Andersen, C.I.G.R.E., 1964, Paper No. 408, 45 pp.
4. Direct Current, Vol. 10, pp. 10-12, February 1965. Unsigned news article.
5. " Operational Performance and Service Experience with the Konti-Skan and Gotland
H.V.D.C. Projects," by S. Smedsfelt, L. Ahlgren, and V. Mets, I.E.E. Conference
Publication 22, H.V.D.C. Transmission, held at Manchester, Sept. 19-23, 1966,
Part 1, Paper No. 1, pp. 11-16.
L. Sardinia.Link
Unsigned news articles :
1. Direct Current, Vol. 10, No. 1, pp. 13-14, February 1965.

Elec. Engineer (Australia), Vol. 44, No. 6, p. 25, June 1967.
Elec. Times, Vol. 151, pp. 257-259, Feb. 16, 1967.
Elec. World, Vol. 167, p. 21, Mar. 27, 1967.
Trans, and Dist., Vol. 19, No. 4, p. 32, April 1967.
Elec. News and Engg. (Don Mills, Ont.), Vol. 76, No. 4, p. 28, April 1967.
7. "The Sardinian-Italian Mainland H.V.D.C. Interconnection," by M. Natale, F. J.
Lane, and T. E. Calverley, I.E.E. Conference Proc. 22, H.V.D.C. Transmission, held
at Manchester on Sept. 19-23, 1966, Part 1, Paper No. 7, pp. 42-49.
8. " Testing and Operating Experience on the Sardinia-Italian Mainland D.C. Link," by
V. Ciallella, P. Grattarola, A. Taschini, C. J. B. Martin, and D. B. Willis, C.I.G.R.E.,
2.
3.
4.
5.
6.
1968, Paper 43-09, 21 pp.
M. Vancouver Island Link (British Columbia)

Unsigned news articles:
1. Direct Current, Vol. 10, No. 1, p. 7, February 1965.
2. ASEA Journal, Vol. 38, No. 10-12, pp. 165-166, 1965.
3. Elec. News and Engg. (Don Mills, Ontario), Vol. 76, No. 6, pp. 56-57, June 1967.
4. Elec. World, Vol. 170, No. 4, p. 18, July 22, 1968.
5. The Engineer, Vol. 226, No. 5871, p. 161, Aug. 2, 1968.
6. "D.C. ' First' Provides Power for the Future," by P. J. Croft, British Columbia Hydro
and Power Authority, Progress, Summer 1967, pp. 8-10.
7. "Major Features of the Vancounver Island 260-kV HVDC Submarine Link," by
H. M. Ellis and W. Chin, Amer. Power Conf. Proc., Vol. 30, pp. 1017-1034 April
1968.
8. "Vancouver Island HVDC Transmission," by Gordon H. Dunn and Lars A. Bergstrom, ASEA Journal, Vol. 42, No. 2-3, pp. 29-31, 1969.
N. Pacific Northwest-Southwest Intertie

1. "Task Force Backs 375 kVD.C. for BPA-California Tie," Elec. World, Vol. 156,
No. 26, pp. 34-36, Dec. 25, 1961.
46
BIBLIOGRAPHY
2. "Layout Arrangements for EHV-D-C Transmission Terminals," by E. M. Hunter

and W. E. Matson, Amer. Power Conf. Proc., Vol. 28, pp. 860-867, April 1966.
Celilo terminal at The Dalles, Oregon.
3. "Design of the Celilo-Sylmar 800-kV DC Line (BPA Section)," by R. F. Stevens,
I.E.E. Conf. Publication 22, H.V.D.C. Transmission, Manchester, Sept. 19-23, 1966,
Part 1, Paper No. 74, pp. 354-358. Also I.E.E.E. Trans, on P.A. & S., Vol. 86, No. 7,
pp. 916-920, July, 1967. Disc., pp. 921-922.
4. "The Celilo (The Dalles) Convertor Station for the Pacific H.V.D.C. Intertie," by
G. D. Breuer, E. M. Hunter, P. G. Engstrom, and R. F. Stevens, ibid., Paper No. 80,
pp. 394-402.
5. "Technology and Economics of EHVD.C. with Application to PNW-PSW Intertie,"
by R. J. Mather and E. F. Weitzel, Bonneville Power Administration report, presented
at the symposium of the Assn. of Amer. Railroads, Denver, Nov. 15-16, 1966.
6. "The Pacific Northwest-Pacific Southwest Intertie," by E. V. Lindseth, Civil Engg.,
Vol. 36, No. 12, pp. 46-47, December 1966.
7. "Long Distance Transmission of H.V.D.C. in Western U.S.A." (in German), by
H. Dommel, S.E.V. Bull. (Switzerland), Vol. 58, pp. 60-68, Jan 21, 1967.
8. "Reclamation's Mead Substation," by N. B. Bennett, Jr., Power Engg., Vol. 72,
No. 1, pp. 32-35, January 1968. Southern terminal of second dc intertie near Hoover
Dam.
9. "Some Design Considerations of the Celilo-Sylmar 800 kV D-C Line (Los Angeles
Department of Water and Power Section)," by L. L. Burnside and W. M. Mahoney,
Western Water & Power Symposium Proc., Los Angeles, Apr. 8-9, 1968, pp. D25-33.
10. " 800 kV DC Transmission Lines of the Bonneville Power Administration," by S. A.
Annestrand, E. J. Harrington, M. N. Marjerrison, and R. F. Stevens, Western Water
& Power Symposium Proc., Los Angeles, Apr. 8-9, 1968, pp. D35-52.
11. "Final Design Criteria Established for E.H.V. D.C. Transmission Line and Terminal
Facilities," by J. L. Mulloy and Edward York, Jr., Amer. Power Conf. Proc., Vol. 30,
pp. 1035-1044, Chicago, April 1968.
O. Kingsnorth Link
Unsigned news articles:

1. Engineering, Vol. 201, p. 484, Mar. 11, 1966.
2. Elec. Rev., Vol. 178, No. 10, pp. 378-379, Mar. 11, 1966.
3. Elec. World, Vol. 165, No. 15, pp. 19-20, Apr. 11, 1966. ,
4. Direct Current, Vol. 11, p. 53, May 1966.
5. English Electric Jour., Vol. 21, No. 3, pp. 2-3, May-June 1966.
6. Elec. Times, Vol. 149, No. 10, pp. 361-362, Mar. 10, 1966.
7. Elec. Times, Vol. 150, p. 479, Sept. 29. 1966.
8. " Kingsnorth-London D.C. Transmission Interconnector," by W. Casson, J.E.E.
Conference Publication 22, High Voltage DC Transmission, Manchester, Sept- 19-23,
1966, Part 1, Paper No. 9, pp. 56-57.
9. "The Kingsnorth, Beddington, Willesden D.C. Link" by T. E. Calverley, F. H. Last,
A. Gavrilovic, and C. W. Mott, C.I.G.R.E., Report 43-04, 1968, 14 pp.
47
P. Nelson River, Manitoba

1. "The 450 kV Direct Current Transmission System for the Nelson River Project,"
by L.A. Bateman, L. S. Butler, and R. W. Haywood, C.I.G.R.E., Report 43-02, 1966,
9 pp.
2. "The Nelson River Transmission System," by E. M. Scott, Trans. Canadian Elec.
Assn., Vol. 6, Part 2, Paper No. 67-SP 131, Mar. 21, 1967, 16 pp.
3. "The Selection of 450-kV HVDC Transmission for the Nelson River," by E. M.
Scott, American Power Conference Proc., Vol. 29, pp. 966-977, Chicago, April 1967.
Abstracted in Elec. Rev., Vol. 180, pp. 711-712, May 12, 1967.
4. " Why 450-kV H.V.D.C. Was Selected for the NelsonRiver Transmission Medium,"
by E. M. Scott, Elec. News and Engg., Vol. 76, No. 6, pp. 50-55, June 1967.
5. "English Electric Wins Nelson River Contract," Elec. Rev. (London), Vol. 181,
No. 7, p. 229, Aug. 18, 1967.
6. "The 450 kV Direct Current Transmission System for the Nelson River Project,"
by L. A. Bateman, R. W. Haywood, and L. S. Butler, C.I.G.R.E., Report 43-02,
1968, 9 pp.
7. "Manitoba's Kettle Simmers Year 'Round," Engg. News Record, Vol. 181, No. 7,
pp. 34-36, 41, Aug. 15, 1967.
8. " Nelson River D.C. Transmission Project," by L.A. Bateman, R. W. Haywood, and
R. F. Brooks, I.E.E.E. Trans, on P.A. & S., Vol. 88, pp. 688-693, May 1969. Disc.,
pp. 693-694. Also in I.E.E.E. Publ. 68 C57-PWR, October 1968.
Q. Miscellaneous Projects
1. "The Exploitations of Yugoslav Water Resources and the Possibility of Using
H.V.D.C. Transmission," by H. von Bertele, Direct Current, Vol. 2, No. 5, pp. 107109, June 1955.
2. "The Present Status of High-Voltage D.C. Power Transmission in Japan," by Naohei
Yamada, Appendix Ito "Report on the Work of the Study Committee No. 10:
D.C. Transmission at E.H.V.," C.I.G.R.E., 1956, Paper No. 407, pp. 2-7.
3. "Electric Power Transmission by H.V.D.C. Submarine Cables across the Adriatic
Sea from Yugoslavia to Southern Italy," by M. Visentini, A. Asta, and F. Trimani,
C.I.G.R.E., Report 210, 1958, Vol. 2, 22 pp.
4. "The Introduction of H.V.D.C. Transmission into a Predominantly A.C. Network,"
by W. Casson, F. H. Last, and K. W. Huddart, Elec. Rev., Vol. 178, No. 8, pp. 290295, Feb. 25, 1966.
5. "The Economics of D.C. Transmission Applied to an Interconnected System," by
W. Casson, F. H. Last, and K. W. Huddart, I.E.E. Conf. Publication 22, H.V.D.C.
Transmission, Manchester, Sept. 19-23, 1966, Part 1, Paper No. 13, pp. 75-83. Rein
forcement of an ac system with dc links which do not increase required circuit-breaker
interrupting ratings.
6. "High Capacity D.C. Transmission in theU.S.S.R.," by A. Berkovsky, N. Chouprakov, T. Izrailevich, A. Kolpakova, and S. Rokotjan, ibid., Part 2, Paper No. 94, pp.
126-129.
7. "Introductory Lecture," by F. J. Lane, ibid., Part 2, pp. 7-25. Describes proposed
transmission of 4500 MW by 500-kV dc from Churchill Falls, Labrador, to Boston
and New York, U.S.A., by the Atlantic route, comprising 1575 miles of land line and
92 miles of submarine cable.
48
8. "Preliminary Studies of Power Transmission from the Churchill Falls Development,"

by G. W. Clayton and D. T. McGillis, Trans. Canadian Elec. Assn., Vol. 6, Part 2,
Paper No. 67-SP 132, March, 1967. Also in Amer. Power Conf. Proc., Vol. 29, pp.
954-965, Chicago, April 1967.
9. "Problems in Designing a Direct Voltage Power Transmission System," by S. P.
Jackson,7.,... Region 3 Conf. Record, 1967, pp. 309-314. 300-mi submarine cable
network.
10. " Underwater D.C. Line Proposed in Alaska," Elec. World, Vol. 169, No. 6, p. 20,
February 1968. Line from Snettisham powerplant to Juneau by submarine cable is
proposed.
11. " SeveralAspects of the System Studies of the DCAlternative for Power Transferfrom
Churchill Falls," by G. A. Baril, A. Lacoste, H. Persoz, J. D. Ainsworth, and J. P.
Bowles, C.I.G.R.E., 1968, Paper No. 43-06, 18 pp.
12. " High Voltage Direct Current Plans for an Integrated Power System in the U.S.A.,"
by D. B. Giesner, ' Direct Current (new series), Vol. 1, No. 3, pp. 121-123, February
1970. Based on Transmission Study 190, U.S. Department of the Interior, 1968.
Converter Circuits
It was indicated in Chapter 1 that the best kind of valve for use in the
converters for HV dc transmission is the mercury-arc valve with control grid.
In this chapter we examine various ways in which groups of such valves and
transformers can be connected to form a converter, and we determine which
of several possible circuits is best for hv dc transmission.
2-1 VALVE CHARACTERISTIC
The symbol for a controlled mercury-arc valve is shown in Figure la.

Symbols for uncontrolled and controlled valves of any type (mercury-arc or
solid-state) are shown in Figures lb and c, respectively.
Voltage Current -
(b)
Anodej Cathode
Current
Voltage
Anode
voltage
Forward
Envelope
Anode
Control grid
Cathode
Inverse
Wii
Z-l
Current
Forward
Control
(grid or gate)
Current
Anode "'Cathode
(c)
Fig. 1. (a) Symbol for mercury-arc
Inverse
(d)
valve with control grid; (b) symbol for any uncon

trolled valve; (c) symbol for any controlled valve; (d) idealized valve characteristic.
49
50
2-4
CONVERTER CIRCUITS
A valve normally conducts in only one direction, from anode to cathode;

and while it is conducting there is a small drop of voltage across it. Such
current and voltage are in the forward direction and are taken as positive.
The forward voltage drop in a mercury-arc valve is in the range of 20 to 50 V,
the higher values (40 to 50 V) pertaining to hv valves. Barring abnormal
operation (arcbacks), the valve can sustain a comparatively high voltage in
the negative or inverse direction without conducting, except for a negligible
leakage current (milliamperes).
The rated current for valves for hv dc may be hundreds or thousands of
amperes, and the rated peak inverse voltage, 50 to 150 kV. In comparison,
the inverse leakage current and the forward voltage drop are negligible.
Hence the idealized voltage-current characteristic of a diode, shown by the
solid heavy line in Figure Id, is adequate. It consists of two half-axes:
1. Positive (forward) current at zero voltage
2. Negative (inverse) voltage at zero current
In a valve having a control grid at a sufficiently negative voltage with

respect to the cathode, the current is prevented from starting, although the
anode may be positive. The valve may then operate on the branch shown as
a broken heavy line in Figure Id.
In this chapter, valves are assumed to have no control grids, from which it
follows that the converters operate only as rectifiers with no ignition delay.
This mode of operation affords a comparison of the various converter
circuits that is valid also for rectifier and inverter operation with grid
control.
It follows from the characteristics of uncontrolled valves that (1) if the

cathodes of several valves are connected together, the common potential of
these cathodes is equal to that of the most positive anode, and (2) if the
anodes of several valves are connected together, the common potential of
these anodes is equal to that of the most negative cathode.
2-2 PROPERTIES OF CONVERTER CIRCUITS
For each circuit considered, we find the wave forms of the voltages and
currents and their magnitudes in terms of the direct voltage Vd and direct
current Id . Fromthese data we find the required volt-ampere ratings of valves
and transformers in terms of the dc power Pd = VdId.
The volt-ampere rating of a valve is taken as the product of its average
current and its peak inverse voltage (PIV), and the rating of a transformer
winding is the product of its rms voltage and rms current.
SINGLE-PHASE CONVERTERS
51
2-3 ASSUMPTIONS
In addition to idealizing the valves, we idealize the ac source, the trans

former, and the dc sink (load), as follows:
The ac source has no impedance and delivers constant voltage of sinu
soidal wave form and constant frequency. If polyphase, it delivers balanced
voltages.
The transformers have no leakage impedance nor exciting admittance.
The dc load has infinite inductance, from which it follows that the direct
current is constant, that is, free from ripple. This assumption is justified by
the fact that hv dc converters have large dc smoothing reactors (about
1 H), and it is reasonably accurate for converters having six or more pulses
per cycle, as those used for hv dc transmission have, although it is a poor
assumption for single-phase converters. Although the current is assumed
free from ripple, the direct voltage on the valve side of the smoothing reactor
has ripple. The dc load is shown on our circuit diagrams as a reactor in
series with an emf of constant voltage, which is equal to the average value of
the ripply voltage on the valve side of the reactor.
2-4 SINGLE-PHASE CONVERTERS
Half-wave Rectifier
This is the simplest rectifier, having only one valve (Figure 2). The current
is inherently intermittent, and therefore the circuit cannot be analyzed in
i .
W
Ac
vi
Vl
d
=- Dc
'
vd
Fig. 2. Half-wave rectifier circuit.
accordance with the assumption of steady current, although all other converter
circuits can be and are so analyzed. Both direct current and direct voltage
pulsate at the same frequency as the alternating voltage. If a transformer is
placed between the ac source and the valve, as is usually necessary for obtain
ing the desired magnitude of direct voltage, the direct current passes through
the secondary winding of the transformer, and its mmf may be sufficient to
saturate the iron core and make the primary current excessive. The circuit is
useful only for very small amounts of power.
52
CONVERTER CIRCUITS
T: 1: 1
Full-wave Rectifier
This has two valves and one transformer with center-tapped secondary
winding (Figure 3). The wave forms are shown in Figure 4. In Figure 4a, the
line-to-neutral secondary voltages et and e2 , having a phase difference of
one-half period (180), are plotted. The anode voltages of valves 1 and 2 with
respect to neutral point N are equal to et and e2 , respectively. The common
cathode voltage of both valves, being equal to the higher of the anode
voltages, consists of the positive half waves of ex and e2 . This curve, redrawn
in Figure 1b, represents also the instantaneous direct voltage vd on the valve
side of the smoothing reactor. The average direct voltage Vd is also shown.'
The difference vd Vd, which appears across the reactor, is represented by
the vertical shading in Figure lb. Its average value is zero, corresponding to
equal positive and negative areas between the curve and the horizontal line.
When valve 2 is conducting, the full secondary voltage et e2 appears
across valve 1; when valve 1 is conducting, e2 e1 appears across valve 2.
Figure 4c shows the voltage vx across valve 1. The valve currents, which are
also the currents in the halves of the secondary winding, are shown in Figure
4d. They are rectangular pulses of height Id and length 180. The mmf of the
entire secondary winding is porportional to ix i2 and has an average value
of zero; in other words, there is no dc component of mmf, hence no tendency
to saturate the core. The primary mmf must oppose the secondary mmf
(Tip = ix - i2), so that the primary current ip has the form shown in
Figure 4e.
Now let us compute the numerical values of the various circuit quantities.
The filtered direct voltage Vd is the average value of vd, and the latter consists of the positive halves of sine waves having crest value Em and frequency
/ cojln. Let 9 = cot.
2f
Vd = 71
71/2 .
2E.
71/2
Em cos 9 dd = 71-| sin
2E
= 0.637F.
e =
TEm sin a> t
Fig. 3. Single-phase full-wave rectifier circuit.
(a)
(b)
>
Vd
(c)
(1)
(d)
Em = -Vd=\.51lVd
(2)
The peak-to-peak ripple is Em = 1.571Vd, and its frequency is 2f where /

is the frequency of the ac source.
The peak inverse valve voltage is 2Em = 3A42Vd.
The transformer voltages are sinusoidal by assumption. The voltage
across each half of the secondary winding has crest value Em and rms value
0.7()lEm = l.ll\Vd. The primary voltage has crest value TEm and rms value
0J01TEm = 1.111TVd , where T is the transformer turns ratio.
The crest value of current in each valve and in each half of the secondary
1P
(e)
UlT
IV?
->- bit
Fig. 4. Wave forms of the circuit of Figure 3 : (a)

transformer secondary voltages et
and e2; (b) unfiltered and filtered direct voltages vd and V; (c) voltage
across valve 1,
vii (d) valve currents or secondary currents h and i2; (e) primary current
i.
54
2-4
CONVERTER CIRCUITS
winding is 7d, the average value is I

J2, and the rms value is (/d)2/2 =
0.707/d. The primary current is +/d/r during one half of the cycle and
IJT during the other half; its rms value is 7d.
The aggregate volt-ampere rating of the valves is 2 x 3.142Fd x 0.500/d =
3.1427>d.
The volt-ampere rating of the whole secondary winding is 2 x l.lllFd x

0.707Id = 1.571Pd; that of the primary winding is l.lllXVd x IJT = l.lllPd.
The foregoing values are listed in Table 1 (page 66) for comparison
with other converter circuits.
Bridge Rectifier
If the two valves in Figure 3 were reversed, so that the anodes instead of
the cathodes were at a common potential, nothing would be changed except
the direction of the direct current and voltage. Note, further, that the
same transformer could feed two pairs of valves, one pair connected as
in Figure 3 and the other pair reversed, as just discussed. Each pair of valves
could feed a separate load with a common return to neutral point N; in other
words, the rectifier would now be feeding a three-wire dc wire system. If the
currents on the positive and negative sides were equal, the neutral conductor
and the center tap on the transformer secondary winding would no longer be
needed. The result of such omission is the single-phase bridge rectifier, shown
in Figure 5. It is also known as a two-way circuit.
SINGLE-PHASE CONVERTERS
55
The number of valves, the direct voltage Vd, and the dc power Pd of the
full-wave rectifier have been doubled, but the P.I.V. of each valve is un
changed. Two valves (1 and T) conduct in series for one-half cycle, then the
other two (2 and 2') in the next half cycle. The valve and load currents are
unchanged. The transformer primary current is doubled. The transformer
secondary currents now have the same wave form as the primary current,
and thus the required secondary volt-ampere rating is equal to that of the
primary, being less than doubled.
The wave forms are shown in Figure 6, and numerical values are given in
Table 1.
Although the bridge circuit may appear more complicated than the fullwave circuit because it has four valves instead of two, the secondary winding
is used more effectively, and the P.I.V. of each valve has been halved for a
given dc output voltage. The former feature is important in high-power
vs=vn/T
(a)
(b)
nwr-
(c)
Dc
(d)
Vd
Tip
h h
ls ~
(e)
Fig. 5. Single-phase bridge rectifier circuit.
.If
I1*
03t
Fig. 6. Wave forms of the circuit of Figure 5 : (a) primary and secondary voltages v
and vs\ (b) unfiltered and filtered direct voltages vd and VA\ (c) voltage across valve 1, vi;
(d) valve currents i2 and i2\ (e) primary and secondary currents ip and is.
56
2-5
CONVERTER CIRCUITS
converters, and the latter in high-voltage ones, where the valve P.I.V. may
be a limiting factor.
THREE-PHASE CONVERTERS
57
3
z*=i r cos 0 dd = -{3/sin. 6V/3
I = V3
- = 0.828
(3)
Em = 1.209Fd
(4)
Em(1 - cos 60) = y = 0.605 Vd
(5)
Em
a,a
Xj 0
whence
Jo
7E
The peak-to-peak ripple is
2-5 THREE-PHASE CONVERTERS
Large amounts of power are taken from a three-phase circuit in preference

to a single-phase circuit. Besides, in polyphase converters, the ripple in
direct voltage is smaller in magnitude and higher in frequency than in singlephase converters and therefore easier to filter. The same is true of harmonics
on the ac side, but this is not shown until Chapter 8.
with ripple frequency 3f
Three-phase One-way Rectifier
This is the simplest three-phase converter circuit (Figure 7). It is not

practical as shown, because the direct current in the secondary windings
Dc
Vd
(c)
-<
Fig. 7. Three-phase one-way rectifier circuit.
saturates the transformer cores. This may be avoided by replacing the Y con
nection by the zig-zag connection, in which the dc mmfs of the two secondary
windings on the same core cancel out. The circuit as shown, however, is useful
as a step in explaining several other connections, and therefore it is analyzed
as it stands.
The three secondary voltages ea , eb , ec form a balanced three-phase set, as
shown in Figure 8a, and the anode voltages with respect to neutral point N
are equal to the corresponding secondary voltages. The common cathode
voltage vd coincides with the upper envelope of this set of voltages, as shown
by the heavy line. The average direct voltage Vd is given by
(d)
UdlT
1
3 Ci)t
(e)
lpa
1IdIT
\
-t)f
Fig. 8. Wave forms in the circuit of Figure 7. (a) secondary

voltages, ea, eb, ec, unfiltered
direct voltage v and filtered direct voltage Vd; (6) voltage
across valve I; (c) valve
currents iu i2, i3, which are also transformer secondary currents; (d)
condensed representa
tion of valve currents ; (e) one primary winding current ipa.
58
2-5
CONVERTER CIRCUITS
The wave form of voltage across valve 1, shown in Figure 8a and b, consists
segment, while
of three segments, each lasting one-third cycle or 120. In one
other
two segments,
the
zero;
is
in
the valve is conducting, the valve voltage
to one of the
equal
voltage
inverse
has
an
while the valve is nonconducting, it
by
the height
shown
is
8a
it
Figure
In
line-to-line voltages, first eab, then eac.
axis. The
horizontal
a
from
redrawn
b
is
it
of the vertical shading; in Figure 8
P.I.V. is
(6)
y/3tn = %/3 x 1.209K, = 2.094K,
and rms value
The transformer secondary voltage has crest value E,
= 1.209 Vd x 0.707 = 0.855Vd.
Each valve conducts during the one-third cycle when the associated
rectangular
secondary voltage is the highest one. The wave of valve current is a
and its
is
value
IJ3,
average
Its
8c).
(Figure
120
length
pulse of height Id and
same
the
is
secondary
current
rms value is IJsj3 = 0.511ld. The transformer
second
the
from
differs
8)
(Figure
as the valve current. The primary current
average value is zero. Its
ary current in having no dc component; that is, its
rms value is
EJ\/2
_ 72/0.471/,
lTT2Y7JT'IY
3
rV V3/ 3\ 3/
3T
The aggregate valve rating is

3 x 2.094 Vd x
0.333Id = 2.094Pd
The volt-ampere ratings of the transformer bank are

3 x 0.855Kd x 0.577/d 1.48 IP,
Secondary;
Primary;
3 x 0.855TVd x
= L209Pd
(8)
(9)
(10)
Three-phase Two-way, or Three-phase Bridge Rectifier
the circuit in
This is known in Europe as the Graetz circuit (Figure 9), In
except
as
before
operates
Figure 7, if the three valves are reversed, the circuit
The
same
reversed.
are
that the directions of direct current and direct voltage
each,
valves
three
of
groups
transformer secondary windings may feed two
similarly except for
one group connected as in Figure 7, the other connected
load, the two loads
dc
separate
reversal of the valves. Each group may feed a
to the trans
connected
constituting a three-wire load with its neutral point
neutral con
the
currents,
equal
former neutral. Now, if the two loads have
point of the
neutral
the
Since
ductor carries no current and may be omitted.
THREE-PHASE CONVERTERS
59
*4
r~-1
'c
O.
*5
m.
Li2
Ffg. 9. Three-phase two-way, three-phase bridge, or Graetz

rectifier circuit.
transformer windings is no longer necessary, those windings can be connected

in A instead of in Y if desired. The Y connection is shown in Figure 9.
The relation between the three-phase single-way and two-way circuits
(Figures 7 and 9) is like that between the single-phase single-way and
two-way
circuits (Figures 3 and 5). Again, for a given alternating voltage, the direct
voltage and power are doubled, but the P.I.V. is not altered. In terms of
Vd,
it is halved and becomes 1.041Vd. This makes the circuit advantageous
if
high direct voltage and high power are required.
Other advantages appear in the transformer bank. There is no direct
current in the windings, and the rms current is less than twice that of
the
single-way connection, giving more efficient use of the windings.
The wave forms are shown in Figure 10. The transformer secondary line-toneutral voltages are shown in Figure 10a. These are also the voltages of the
anodes of the lower group of valves and of the cathodes of the upper group,
all with respect to neutral point N. The common cathode voltage of the
lower group of valves is the upper envelope of the transformer voltages, as it
was for the single-way circuit in Figure 7. The common anode voltage of
the
upper group of valves is the lower envelope. The difference in ordinates
between the upper and lower envelopes is the instantaneous direct voltage
vd on the valve side of the smoothing reactor. This is replotted in Figure
\0b as the envelope of the line-to-line voltages. The voltage across valve 1 is
also shown (Figure 10c).
It is immediately to be noted that the ripple of the direct voltage is
of
frequency 6/, twice that of the one-way connection, and the magnitude of
the
ripple is smaller. The reason is that the scallops of the lower envelope
are
shifted one-sixth cycle from those of the upper envelope.
60
2-6
CONVERTER CIRCUITS
eb
ec
eb
PULSE NUMBER
61
The peak-to-peak ripple is

(1- cos 30o)V3m = 0.134 x 1.047Fd = 0.140Fd
(a)
(13)
The rms line-to-neutral secondary voltage is
0.707m = 0.707
(b)
V3E;
(c)
VaE;
(d)
w t
.2
4
<
ia
(?)
~5l
cot
v3
= q.428 Vd.
The load current is always carried by two valves in series, one from the
upper half bridge and one from the lower. Each valve conducts for one-third
cycle, as in the one-way circuit. Commutation in one group, however, is stag
gered with respect to commutation inthe other group ;considering both groups,
commutation occurs every one-sixth cycle (60). In Figure 9, as well as in the
diagrams of other converter circuits, the valves are numbered in the order in
which they fire (begin to conduct). Commutation occurs from valve 1 to
valve 3, then from 2 to 4, from 3 to 5, from 4 to 6, from 5 to 1, and from
6 to 2. The current wave forms are shown in condensed fashion in Figure 10d.
The current in each phase of the Y-connected secondary windings is the
difference of the currents of two valves, the numbers of which differ by 3 ; for
example, ia = it iA (Figure lOe). Its rms value is Id\j2j3 = 0.816/d.
The aggregate valve rating is 6 x 1.047Fd x I
J3 = 2.094/V
The aggregate volt-ampere rating of the transformer secondary windings
is 3 x 0.428 Fd x 0.816/d= 1.047Pd. The rating of the primary windings is
the same as that of the secondaries.
If the secondary windings are connected in A instead of Y, the wave shape
times that of
of current is different (Figure 10/), and its rms value is 1/
current in the Y. The primary line currents have the same wave shape as
secondary Y currents if the transformer connection is YY or AA, and the
same wave shape as secondary A currents if the transformer connection is
\/3
(f)
Fig. 10. Wave forms of the circuit of Figure 9 : (a) secondary line-to-neutral voltages ea
eb, ec and, in heavy lines, unfiltered voltages of positive and negative dc poles with respect
to transformer neutral point; (b) secondary line-to-line voltages and, in heavy line,
unfiltered direct pole-to-pole voltage vd; (c) secondary line-to-line voltages and, in lower
heavy line, voltage t?i across valve 1; (d) condensed representation of valve currents;
(e) transformer secondary current ia=iiu', (/) primary alternating line current
iA = dc ~ h)lT.
The average direct voltage
Va
SEm
Vd is given by
n/6
COS
(11)
whence the P.I.V. is
SEm = -Vd=\MlVd
/
(12)
AY or YA.
2-6 PULSE NUMBER
The pulse number of a converter is the number of pulsations (cycles of

ripple) of the direct voltage per cycle of alternating voltage. The circuits so
far considered in this chapter have the following pulse numbers :
Circuit
Single-phase half-wave
Single-phase full-wave
Single-phase bridge
Three-phase one-way
Three-phase two-way or bridge
Pulse Number p
1
2
2
3
62
CONVERTER CIRCUITS
It is shown in Chapter 8, on harmonics, that the orders of harmonics in the

direct voltage are given by pq, where q is an integer, and that the orders of
harmonics in the alternating current are pq 1. The higher the pulse number,
therefore, the higher the lowest frequencies of harmonics. As we have seen in
the analyses of the foregoing circuits, the higher the pulse number, the lower
the peak-to-peak amplitude of the ripple. The higher the frequency and the
lower the amplitude of a harmonic, the easier to eliminate it substantially by
filtering. The higher the pulse number of a converter, the better from the
standpoint of harmonics. This advantage, however, may be offset by the
increased complexity of transformer connections.
,/
2-7 ADDITIONAL SIX-PULSE CONVERTER CIRCUITS
We have seen that the three-phase two-way circuit may be derived by

combining two one-way circuits in a particular way. There are at least two
other ways in which two three-phase one-way circuits can be combined to
form six-pulse circuits.

Cascade of Two Three-phase Rectifiers
This connection is shown in Figure 11.
As in the three-phase bridge circuit, the two three-phase groups of valves
are in series on the dc side, but they are not adjacent, and each group has a
common cathode connection instead of one group having a common cathode
connection and the other a common anode connection, as the bridge circuit
has (Figure 9). The transformer secondary windings are connected in double
Y with 180 phase difference between one Y and the other. This transformer
connection is more complicated than that used with the three-phase bridge
circuit, and the secondary windings must have a greater aggregate voltampere rating. This circuit has no advantage over the bridge unless it is
desired to use valves having one mercury-pool cathode in an envelope (steel
tank or glass tube) with three anodes.
Fig. 11,
Cascade of two three-phase one-way rectifiers,
Dc
Parallel Connections with Interphase Transformer

The two three-phase groups of valves are again fed by double-Y-connected
secondary windings, but the two groups are in parallel on the dc side instead
of in series (Figure 12). They cannot be paralleled directly, because the threepulse ripple of one group is staggered with respect to the ripple of the other
group. Instead, one dc pole of one group is connected directly to the like pole
of the other group and to one pole of the dc line. The opposite poles are
Fig. 12. Y-Y-interphase rectifier circuit;,

I.T. = interphase
transformer.
64
2-8
CONVERTER CIRCUITS
connected to opposite ends of an autotransformer whose center tap is con

nected to the other pole of the dc line. The autotransformer is called an inter
phase transformer. It is usually connected to the neutral points of the Y's and
to the negative pole of the dc line so as to permit the use of a single-cathode
six-anode valve. The instantaneous voltage of the center tap is, of course,
equal to the average of the instantaneous voltages of the two ends of the
winding. Consequently, the instantaneous direct line voltage is the average
of the voltages of the two three-phase groups and has a six-pulse ripple.
This connection has the same disadvantages as the preceding one and, in
addition, requires the interphase transformer and gives a lower direct voltage.
It is not suitable for hv dc work.
CHOICE OF BEST CIRCUIT
65
Cascade of Three Single-phase Full-wave Rectifiers

At first sight it appears that this circuit (Figure 14) might give a higher
direct voltage for a given P.I.V. on the valves than either the bridge circuit
or the cascade of two three-phase rectifiers, but it does not. Moreover,
the transformer utilization is poor.
Six-phase Diametrical Connection
This appears similar to the preceding connection. With the interphase

transformer omitted and the neutral points of the two Y's solidly connected,
the transformer secondary windings form a six-phase star connection
(Figure 13). This can be made up from one center-tapped winding per core
instead of two separate windings per core. Now, however, each valve con
ducts for one-sixth of a cycle instead of one-third cycle, as in the three
preceding connections. Consequently, the transformer utilization is poorer.
A valve having one pool cathode and six anodes in one envelope can be used.
l
Fig. 14. Cascade of three single-phase full-wave rectifiers.
2-8 CHOICE OF BEST CIRCUIT FOR HV DIRECT-CURRENT
CONVERTERS
Fig. 13. Six-phase diametrical rectifier circuit.
A comparison of the nine different converter circuits discussed above is

presented in Table 1. Circuits 4 to 8 are six-pulse circuits having six valves
each and a three-phase ac supply. Let us confine our attention to these,
because the advantages of three-phase ac circuits and a high pulse number for
high-power converters are apparent.
Study of Table 1and the circuit diagrams reveals that the best converter
circuit for hv dc transmission is the three-phase bridge. It has the following
advantages:
2-9
-ffi. "5
cn
cs
o
cj)
J. -a
cq
.
U o
IO
cs
O co
O CO
""i
o o
o o
N N,
o
o
VD
>3
u
a>
.is
"N
O W">
5-1
>
n?
t-
o tO vo
V> n
P-
v->
00
CB
00 CO
r- o

r- o
fN
N, N
n >-n
o
--
O co
o>
60
T3
-Tf CN
oo
00
5b
rf 5=
t~-
'
fck.
00 00
o o
o d
c-i
PS
r-
WO
Tf
o
En
ft, ft,
00
oo
<'
00
3;
ii
3"
O O
fsj
i
I
3
o o
OS
wo wo
wo wo
oo
Os
so o
u"i
Os
<N <N
Tf
23
CO
n, >n
o
00 CO
i>
00

<N
-N
*5
o o
o o
w->
w->
cs <s
o o
t
o o
CO
W"l
co
O co
r-i
w-l w-l
ri
r-
w->
ON 00
<N
2 8*
oo
g co
'
.-O
En
t>
O CO
O co
En
Cl
WN 00
pa
rs
J
3
OS CO
n "*5
o NO
oh
1. For a given direct voltage, the P.I.V. of the valves is only half that of
En
d ci
do
-'
Tj- OS
Os Q
r4
ft- ft,
<S
o o
o o
*-
J-
Cl
o
>
s
N N
r- o
o o
r- o
O
>n
En
0h
67
any of the other six-pulse circuits except 5 and 8, in which it is equal. Conse
quently for a given P.I.V., the direct voltage is twice that of some other
O"
&o
T3
cq
rm
ON
Ft
23
"n
o rvn
cq
En
i--
oo r-
00
g<
>
rl- Tf
(S
W1
$s
En
tJ-
En
-e-
CS ' i'
O fN]
N
En
a
Is
\o
3"
o
"7
"n >n
CN
TWELVE-PULSE CASCADE
circuits.
2. For a given power throughput, the volt-ampere rating of the transformer
secondary winding is less than that in any other circuit.
3. The volt-ampere rating of the transformer primary windings is equal to
or less than that of the other circuits.
4. The transformer connections are the simplest. Double or center-tapped
secondary windings are not required. This simplcity is important for obtaining
a sturdy and reliable design for operation at very high voltages having super
posed high-frequency transient voltages caused by commutation.
5. The secondary windings may be connected either in Y or in A. The
advantages of this will appear in Section 2-9.
6. The aggregate volt-ampere rating of the valves is lower than that of
circuit 8 and equal to that of the rest.
7. A very important property of the bridge circuit is that arcbacks* can be
suppressed by grid control and a bypass valve. Arcback is a random phenomenon that is unlikely to occur in more than one valve at the same time. Thus,
although grid control is ineffective in the valve with the arcback, it may
reasonably be expected that in all other valves of the same bridge such control
will be effective and may be used to prevent any of these valves from reigniting
after the current in them has once become zero. In the bridge connection
there are two valves in series across the dc line and two in series opposition
across each pair of ac terminals. If all valves but the maloperating one are
blocked by grid control, that one has no circuit through which current can be
furnished to it. This is in contrast to the six-phase star and Y-Y-interphase
connections, in which there is a path through the ac source, the maloperating
valve, and the dc load.
Because of these advantages, the bridge circuit is universally used for highpower hv ac-dc converters.
sr,
2-9 TWELVE-PULSE CASCADE OF TWO BRIDGES
<a
i>
ft ea
O
g>
+- S?
G
>*
x>
'5
S
!>
..
iS
E "2p
.o
O
- ggB
S"3 2 1/1
5>>H
-
ft
"3
W5
Jo
its
Ifil3
.2
J 73
*5 Q
>
>H
"
<a
<o
s 6e cS -ft
a i?
& S 2
>
53
*S 9
8
<HH
In this connection (Figure 15), the dc ports (pairs of terminals) of two

bridges are in series, doubling the direct voltages, and the ac ports are in
parallel, doubling the alternating current. If the transformer bank of one
bridge is connected YY and that of the other, YA, the pulsations of direct
* An arcback is a malfunctioning in which a valve conducts in the reverse direction. See

Chapter 6 for a discussion of its causes and suppression.
68
PROBLEMS
CONVERTER CIRCUITS
4. Analyze a half-wave rectifier circuit like that in Figure 2 except that it

has a resistance-inductance loadwith no counter-EMF. Findwave shapes
*and average and rms values of the current and the several voltages
as functions of coCR.
5. Find the magnitudes of the harmonics, in terms of Id, in the secondary
currents of a single-phase full-wave rectifier (Figures 3 and Ad).
6. Find the magnitudes of the harmonic currents, in terms of Id, in the
secondary current of a single-phase bridge rectifier (Figures 5 and 6e).
7. Calculate the aggregate volt-ampere rating of zig-zag-connected secon
dary windings if used instead of Y-connected windings in the threephase one-way circuit in Figure 7.
8. Analyze a three-phase one-way rectifier like that in Figure 7 except that
the transformer is connected A primary,zig-zag secondary. Find wave
shapes and crest, average, and rms values.
3 4> Ac
11T.
5J[
9V
7 y.
11::
3 y.
Dc
-=
21'.
12::
43:
i
Fig. 15. Twelve-pulse cascade of two three-phase bridges.
voltage in one bridge are staggered with respect to the other, giving a 12-pulse
converter. Under balanced conditions, the twelfth harmonic is the lowest one
in the direct voltage, and the eleventh and thirteenth are the lowest pair in the
alternating current.
The two bridges could be fed from one three-winding transformer bank
instead of from two two-winding banks.

For increasing the direct voltage further, more bridges can be added,
preferably in pairs, but the pulse number is seldom, if ever, made greater than
12, because the complexity of the transformer connections is believed to
offset the advantage of a higher pulse number.
secondary current of a three-phase one-way rectifier (Figures 7 and 8c).

10. Find the wave shape and the magnitudes of the harmonic currents, in
terms of Id, in the primary line current of a three-phase one-way
rectifier (Figures 7 and 8e).
secondary current of a three-phase two-way rectifier (Figures 9 and
lOe).
12. Calculate the rms value of the primary line current in Figure 9, in terms
13.
14.
PROBLEMS
1. Prove that the average voltage across the smoothing reactor is zero if
the reactor has no resistance.
2. Prove that the half-wave rectifier in Figure 2 can conduct only inter
mittently.
3. Analyze the half-wave rectifier circuit in Figure 2, finding the wave
shapes and the average and rms values of the current and the sev
eral voltages as functions of EJVd.
69
15.
16.
17.
18.
19.
of Id, from the wave shape in Figure 10/.

Find the magnitudes of harmonics in the primary line currents of a
three-phase two-way rectifier having a A-Y-connected transformer by
two methods: (1) analysis of the wave form of the primary line current
and (2) adding the harmonics of two phases of the A-connected primary
winding displaced from one another by one-sixth of the fundamental
period.
Analyze the cascade circuit in Figure 11.
Analyze the double-Y-interphase circuit in Figure 12.
Analyze the six-phase diametrical connection in Figure 13.
Analyze the cascade circuit in Figure 14.
Derive the wave shape of the voltage across the interphase transformer
in the double-Y connection. Which is the principal harmonic component
of this voltage ? Consider the frequency of the ac line to be the funda
mental.
Which of the six-pulse circuits considered in this chapter permit the use
of a common-envelope common-cathode six-anode valve ?
70
CONVERTER CIRCUITS
Draw the circuit diagram and the wave shapes, and find the peak,
rms, and average voltages and currents and the aggregate volt-amperes
of each set of transformer windings of converters described as follows:
20. Four-phase star, one-way, four valves.
21. Two-phase, center-tapped, one-way, with interphase transformer.
22. Four-phase, two-way, eight valves.
23. Cascade of two single-phase full-wave converters, fed from a two-phase
line.
BIBLIOGRAPHY
Hingorani, Chapter 2, " Type of Converter Circuits and Valve Connections."
2. Power Transmission by Direct Current, by Ya. M. Chervonenkis, 1948, translated from
Russian to English by the Israel Program for Scientific Translations, Jerusalem, 1963.
Chapter IV, "High-Voltage Rectifier Circuits," pp. 32-44.
3. "D.C. Power Transmission," Part II, "Basic Principles," by E. Openshaw Taylor,
Elec. Jour. (London), Vol. 164, pp. 25-26, Jan. 1, 1960.
4. Rectifier Circuits: Theory and Design, by Johannes Schaefer, Wiley, New York, 1965.
5. High Voltage Direct Current Convenors and Systems, edited by B. J. Cory, Macdonald,
London, 1965. Chapter 2, "H.V.D.C. Transmission Circuits," by B. J. Cory, pp.
14-21.
3
Analysis of the Bridge Converter
In Chapter 2 several converter circuits, including the three-phase bridge,

were analyzed under the assumptions of no grid control and no overlap;
and the superiority of the three-phase bridge circuit for use in hv dc power
transmission was shown. In the present chapter a better analysis of the bridge
circuit is made in which grid control and overlap are accounted for.
The assumption made in Chapter 2 regarding the ac source must be
modified thus:
1. Power source (or sink) consisting of balanced sinusoidal emfs of constant
voltage and frequency in series with equal lossless inductances.
The inductance, which really is chiefly the leakage inductance of the trans
formers, is the cause of the overlap. The other assumptions made in Chapter 2
are retained. They are :
2. Constant ripple-free direct current.
3. Valves which have no forward resistance and infinite inverse resistance.
4. Ignition of valves at equal intervals of one-sixth cycle (60).
Figure 1shows the bridge circuit with the notation adopted. The instantaneous
line-to-neutral emfs of the source are taken as
ea = Em cos (cat + 60)

eb = Em cos (cot - 60)
ec = Em cos (cot 180)
(la)
(16)
(lc)
corresponding to the horizontal projections of the rotating vectors shown in

Figure 2 and to the waves shown in Figure 3.
The corresponding line-to-line emfs are
= ea-ec =
3 Em cos (cot + 30)
ea =
Em cos (cot - 90)
eba = eb
= fZEmsm<ot
eCb = ec
eb
\/3 Em cos (cot + 150)

71
(2a)
(26)
(2c)
72
ANALYSIS OF THE BRIDGE CONVERTER
3-1
ANALYSIS WITH GRID CONTROL BUT NO OVERLAP
Ld=
-nsw
l>3
2.3
wb 5
Vb
Lc
Vd
Vd
Id
-V4
-120'
ve
120'
180'
240'
0)t
V2
Fig. 1. Bridge converterschematic circuit for analysis. The valves are numbered in their
firing order.
Fig. 3. Waves of instantaneous e'mfs of ac source: line to neutral, ea> eb, ec; line to line,
aci &bo>
Continuity of the currents is obtained in our solution of the circuits by

making the final values of currents for one form of the circuit determine the
corresponding initial values of currents for the next consecutive form of the
circuit.
Reference
3-1 ANALYSIS WITH GRID CONTROL BUT NO OVERLAP
Fig. 2. Source-EMF vectors at t = 0.
The circuit in Figure 1, viewed in the light of assumptions 1 and 2, contains

a three-phase voltage source and a dc source. In our analysis of this circuit,
the effects of the two sources are superposed. Superposition is valid in a
linear circuit. A circuit with valves is piece-wise linear. That is, with a certain
set of valves conducting, it is a linear circuit; with another set of valves
conducting, it is a different linear circuit. At the instant when any valve
begins or ceases conducting, the circuit changes. The solutions of two con
secutive circuits are pieced together at the instants of change so as to satisfy
conditions of continuity, which, in this case, are the continuity of the current
in the dc source, because of the smoothing reactor Ld, and of the currents in
the ac source, because of the inductance Lc in each phase.
The number of conducting valves may be two, three, or four according to

the overlap (see Figure 4). For each of these numbers, there are six sets of
conducting valves, each limited to consecutive valve numbers. Let us first
analyze those cases in which only two valves conduct simultaneously (Figure
4a). Each set consists of a valve of the upper row (Figure 1) and one of the
lower row. The transfer of current from one valve to another in the same row
is called commutation. If no more than two valves ever conduct, commutation
must occur instantly, that is, without overlap.' Such commutation is im
possible if the ac source has any inductance. The case, therefore, must be
regarded as a limiting one, which is approached if the product LcId becomes
very small.
Figure 5 shows the circuit in Figure 1, with nonconducting valves omitted,

when only valves 1 and 2 are conducting. The current source causes a steady
current Id in the series circuit n2N\pn, producing no voltage drops because
dldjdt = 0. The alternating-voltage source sets up open-circuit voltages but
no currents, because the application of superposition requires that current
74
3-1
75
(a) u = 0
60a
u
(6; u = 20
>6
>[<
<|
cot
60a-|
\\
V
\
ri
(c) u = 40
>
(e) u = 80
(f) u = 100
(g) u = 120"
Fig. 4. Effect of overlap angle u on the number of valves conducting simultaneously.
sources be opened while other sources are acting. We obtain, for the instan
taneous currents and voltages in Figure 1 or 5,
h = h = E = h = H = *3
(3a)
(36)
va = vP = ea = Em cos (cot + 60)

vb eb = Em cos (cot 60)
vc:=vn==ec== Em cs (cot 180)
(3c)
(3d)
(3c)
h=
V3 Em cos (cot + 30)
VX = v2 = 0
3 = eba =
y4 = y5 =
= eCb =
V3
vd=
V3
sin cot
V.3
vn
Fig. 5. Bridge converter with valves 1 and 2 conducting.
(d) u = 60
vd=vp-v = ea-ec = eac =
12
rui
(3/)
(3g)
(36)
cos (cu? 150)
cos (cot + 150)
(3i)
Equations for these other intervals may be found from Eqs. (3) by appropriate
changes of subscripts and phase angles.
It was shown in Chapter 2 that an uncontrolled valve ignites (begins to
conduct) as soon as the voltage across it becomes .positive. If valves 1 and 2
have been conducting, valve 3 ignites as soon as eb becomes greater than
ea; that is, at point R, Figure 3. This instant is taken as cot = 0. At the same
instant, under present assumptions, valve 1is extinguished (ceases to conduct).
Valves having control grids can be made to delay ignition but not to advance
it. The delay angle is denoted by a and corresponds to a time delay of
ajco sec. If delayed this long, valve 3 ignites when cot a ; valve 4, when
cot = ot + 60; valve 5, when cot = a + 120; and so on. The delay a cannot
exceed 180; for example, point T, 180 after R in Figure 3, marks the end of
the range over which valve 3 can ignite, for only in this range is eb greater
than ea.
The ignition delay affects both the direct voltage and the ac displacement
factor, as will now be shown.
The instantaneous direct voltage vd across the bridge on the valve side
of the dc reactor is composed of 60 arcs of the alternating line-to-line voltages,
as shown by the vertical lines of the shaded area A0 in Figure 3 for the period
in which valves 1 and conduct with a = 0, or as shown in a different way
in Figure 6a.
The average direct voltage Vd is found by integrating the instantaneous
voltages over such a period or any 60 period. For a = 0 (no delay) and with
cot called 6, it is
3 C
=Vd0=-A0
71
71
~
(3j)
The condition of this circuit endures for only one-sixth cycle (60) at a time
and is followed by one in which valves 2 and 3 conduct ; then valves 3 and 4,
then 4 and 5, then 5 and 6, then 6 and 1,and so back to 1and 2. See Figure 4a.
71
eacd9=-71
J -/3
sin (6 +
30)T
m 2 sin 30 =
/-
J 60
V3 i?m cos (6 + 30) d6
J 60
3 Em
71
[Sin 30 - sin (-30)]
= 1.65 Em
(4)
3-1
1
where Em is the crest value of line-to-neutral alternating voltage. In terms

of rms line-to-neutral and line-to-line voltages (ELN and ELL, respectively),
- 3E,.n
2.34Etn
Vdo =
LN =
'LN =
71
Instant when valve 3
ignites in an
uncontrolled rectifier
77
1.35Er.r.
E
h/lEr.r.
= 1-35
Ell -
(5)
LL
71
Vd0 is usually called the ideal no-load direct
voltage.
For delay angle a, both limits of integration are simply increased by a, as
shown in Figure 6b to g, giving
I.
cos (0 + 30) dB = Vd0 sin (B +30)
VA = V.dO
a 60
Fd0[sin (a -I- 30) sin (a 30)]

= Vd0(2 sin 30) cos a = Vd0 cos a
(b) a = 30'
(c) a
60'
(d) a = 90'
wt
(e) a = 120'
(f) a = 150'
(6)
Thus we see that one effect of delayed ignition is to reduce the average direct
voltage by the factor cos a. Inasmuch as a can range from 0 to almost 180,
cos a can range from 1 to 1, and Vd can range from Vd0 to VdQ. Since the
current Id cannot reverse because of the unidirectional property of the valves,
negative voltage Vd in conjunction with positive current Id represents reversed
power flow; that is, conversion from dc power to ac power instead of from
ac to dc. In other words, it represents inversion as opposed to rectification.
Obviously a converter cannot invert unless it has grid control.
Figure 6 shows the waves of direct voltage on the valve side of the dc
reactor for various delay angles. It is noteworthy that the ripple and hence
the harmonics in the direct voltage increase with delay up to 90 and then
decrease from 90 to 180.
With losses in the converter neglected, the ac power must equal the dc
power; that is,
3ELNILl cos <f) = VdId = IdVdo cos a
(7)
IL1 is the rms value of the fundamental-frequency component of

alternating line current. The line current has the wave shape shown in Figure
la, consisting of positive and negative rectangular pulses of height Id and
width 2tc/3 rad. This shape is independent of a as long as there is no overlap.
By Fourier analysis the crest value of the fundamental component of this
wave is
1+60
2 f+n/3
2
V2/tl=- | Id cos B' dB' = zh sin B'
where
-n/3
(g) a = 180'
]+60
71
-60
= - Id[sin 60 - sin (-60)] = - Id(2

71
Fig. 6. Instantaneous direct voltage (shown by heavy line) of bridge converter with
ignition delay angle a but no overlap.
60
2V3 L =1.11/,
sin 60)
71
(8)
78
3-1
(a) a = 0
Ea
tibial
(b) a m 30
(o) a = 60
Ea
(d) a. = 90
CONTROL BUT NO OVERLAP
Ial
Ea
(e) a = 120
hi
Ea
I\+
a *\
i
(f) a = 150 /
la1
Fig. 7. Relation between ignition delay and phase displacement.
Inverter
Rectifier
Inverter with
forced commutation
Rectifier with
forced commutation
The rms value of the fundamental current is
iLl =
id
= o.mid
n
Substitution into Eq. (7) of IL1 = 0.780/d and Vd0

cos 0 = cos a
79
in which cos <j> is the displacement factor or vector power factor and <p is the
angle by which the fundamental line current lags the line-to-neutral source
voltage.
From Eqs. (7) and (9) we see that the converter operates as a sort of trans
former having a fixed current ratio but a voltage ratio that varies with the
ignition delay imposed by grid control.
With no ignition delay, the rectangular pulses of alternating line current
are centered on the half sine waves of line-to-neutral voltage, as shown in
Figure 7a, and the fundamental sine wave of current is in phase with that
voltage wave. Ignition delay a shifts the current wave and its fundamental
component by angle = a, as shown in Figure lb, c, d, e, and f Thus the
converter rectifier or inverter draws reactive power Q from the ac system.
See the vector power diagram in Figure 8. The rectifier is said to take lagging
current from the ac system, and the inverter is said either to take lagging
current or to deliver leading current to the ac system. The signs of active and
reactive power in a rectifier are like those in an induction motor, and, in an
inverter, they are like those in an induction generator. The analogy between
the inverter and the induction generator may be extended a step further: both
work best in parallel with synchronous machines, but both may be made to
work with shunt capacitors instead.
A converter having valves with normal control circuitry can work only in
the upper half of the complex power plane (Figure 8), because ignition can
be delayed but not advanced from its uncontrolled position. Imagine a con
verter in which the valves are replaced by synchronously-controlled switches,
the closing times of which can be shifted in either the lagging or the leading
direction. By operating with the switches closing on negative instead of
Ea
Ea
ANALYSIS WITH GRID
(?)
= 2.34LN yields
(10)
Fig. 8. Power vectors of converter.
80
3-2
positive a, the converter could be made to deliver instead of to consume

reactive power. In other words, such a converter could operate in any quadrant
in Figure 8.
3-2 ANALYSIS WITH GRID CONTROL AND WITH OVERLAP LESS
THAN 60
Vb
Vi=
Instantaneous Currents and Voltages
j
j
4?)-
At end (ft)/ = a + u 5) :
anc*
*3 = 0
(H)
and
i3 = Id
(12)
81
io
i>6
"n
Because the ac source, especially the transformer, has inductance, the

currents in it can vary only at a finite rate, and therefore the transfer of
current from one phase to another requires a finite time, called the commuta
tion time or overlap time, ulco, where u is the overlap angle. In normal
operation it is less than 60: typical full-load values are from 20 to 25. As
illustrated in Figure Ab or c, during commutation three values conduct simul
taneously, but between commutations only two valves conduct. Since a new
commutation begins every 60 and lasts for angle u,the angular interval when
two valves conduct is 60 u. The sequence of conducting valves is 12, 123,
23, 234, 34, 345, 45, 456, 56, 561, 61, 612. If u = 60, as shown in Figure Ad,
a new commutation begins at the same moment that an old one ends, so that
some set of three valves is always conducting. This is a special limiting case,
as is also that for u = 0. The valve sequence is 123, 234, 345, 456, 561, 612.
If 60 < u < 120, an abnormal mode of operation occurs in which alternately
three and four valves conduct. There are intervals of double overlap, in which
commutations are occurring in both half bridges simultaneously, with two
valves of each conducting; and between such intervals are other intervals in
which commutation is in progress in only one half bridge and thus three
valves are conducting in the bridge. See Figure Ae and /. The upper limit of
this mode occurs with u = 120 and four valves always conducting (Figure
At beginning (ft)/ = a) :
60
<?-TO CP
*b
But let us return to an analysis of normal operation with 0 < w < 60 and
with alternately two and three valves conducting. The interval in which valves
1 and 2 conduct was analyzed in Section 3-1, with the results given by
Eqs. (3). This interval ends at cot - a, when valve 3 ignites. In the next
interval, the effective circuit is that in Figure 9, with valves 1, 2, and 3 con
ducting. During this interval the direct current is transferred from valve 1 to
valve 3. Hence
OVERLAP LESS THAN
Fig. 9. Bridge converter with valves 1, 2, and 3 conducting.
The extinction angle is 5 = a + u. Around loop N31N,

eh -
die,
di,
r
r
= L ~ ~ L ~
dt
The emf in this loop, known as the commutating emf, is see Figure 2 and
Eq. (2b):
~ ea =
(13)
dt
Since
\/3 Em sin
(14)
ft)/
h = Id ~~ *35
then
dh _ q _ dh
dt
dt
and Eq. (14) becomes

/
Em sin co/ = 2LC ~

Division by 2Lc and definite integration with respect to /, with the lower limit
being set by Eq. (11) and a running upper limit, give
V3 En
|
|
2Z,
where
sin co/ dt =
/s2(cos a cos cot) = i3
Is2- 2coLr
du
Id
(15)
(16)
Equation (15) shows that z'3, the current of the incoming valve during com
mutation, consists of a constant (dc) term and a sinusoidal term (see Figure
82
3-2
10). The sinusoidal term lags the commutating voltage by 90, as it should in
a purely inductive circuit, and has a crest value hi which is that of the current
in a line-to-line short circuit on the ac source. The constant term, which
serves to make i3 = 0 at the beginning of commutation, depends on a, and
for a = 0 it shifts the sine wave upward by its crest value. The current /
of the outgoing valve has a sine term of the same amplitude as that of i3 but
of opposite phase, and its constant term serves to make / = Idat the beginning.
Currents and i3 would reach highpeak values if the short circuit remained,
as infrequently happens because of an arcback (Chapter 6). Normally, how
ever, the short circuit is removed when the current in. the outgoing valve
reaches zero, for normal valve action prevents it from reversing.
During commutation the line-to-line voltage of the short-circuited phase is
90'
7s2 (cos a - cos cot)
OVERLAP LESS THAN
"1"
va = vb =
~2
C0S
&c
/T7\
for the two phases of the source have equal impedances jcoLc and act as a
voltage divider.
During the overlap interval of valves 1 and 3 (a < cot < a + u), the instan
taneous values of all currents and voltages are given by the following
equations :
ia it = Id Is2(cos a cos cot )

ib = /3 = Is2(cos a cos cot)
(18zz)
h=-h=-h
(18c)
*4 = *5 =*6 = 0
va = vb = vp 0.5ec = 0.5Em cos cot
(18tO
vc = vn = ec= -Em cos cot

v,d = vp - v = 1.5Em cos cot
(18/)
v4 = v5 = v6 = vd =
83
zero, and the two line-to-neutral voltages are equal to each other and to the
average of the corresponding open-circuit voltages:
Ui = v2 = v3 = 0
cot
60
1.5Em cos cot
(18e)
(18g)
(18/r)
(18i)
Equations (18) can be adapted to other overlap or three-valve periods (valves

234, 345, 456, 561, 612) by appropriate changes of subscripts and phase
angles. For the nonoverlap or two-valve periods (12, 23, 34, 45, 56, 61),
Eqs. (3) hold, with appropriate changes. From the resulting equations (Tables
1 and 2), graphs of the instantaneous currents and voltages are drawn
(Figure 11).
Average Direct Current and Voltage
Id ~ Is2 (cos a - cos
At the end of the commutating period, Eqs. (12) hold, and substitution of
them into Eq. (15) yields
Id = Is2{cos a cos 5)
Fig. 10. Currents it and i2 during commutation of valve 1 to valve 3 as arcs of offset
sinusoidal waves of line-to-line short-circuit current of amplitude Is2, with (3 lagging 90
behind the commutation voltage eba.
(19)
This gives the direct current in terms of the ignition and extinction angles.
A similar equation will now be drived for voltage drop AVd due to overlap.
In Figure 12 the effect of overlap is to subtract an area A from the area A0
Table 1. Instantaneous Currents in Three-phase Bridge Converter with Overlap Less than 60'
(Alternately Two and Three Valves Conducting)
Valves
Con
ducting
1,2 cos a
2, 3, 4
3,4
3,4, 5
4, 5
4, 5, 6
5,6
I,2 cos 6
/s2 COS a
/s2 cos(0 60)
Id
Id
Id Is2 cos
+/s2cos(0 60)
0
Id
Id IS2 cos a
/s2cos(0 + 60)
Id
Id
Id I,2 cos a
Is2 COS 0
I,2 COS a
Id
0
0
+I,2 cos(0 60)

Is2 COS a.
6, 1,2
+1,2 0
1,2 COS 0
Is2 cos a+
-Id
1,2 cos
Id
Id
Is2 COS a
Id
+ IsZIdcos 0
Id 1,2 COS a
Id
Id
/s2cos(0 60)
Id
-h
Id
Id+h2 COS a
/s2cos(0 60)
0
Id
$2 cos a
Id~ I
/s2Cos(0 + 6O)
I,2 cos a+
1,2 cos(0-i-6O)
- Id + 1,2 COS a
+ 1,2 COS 0
1,2 COS a
1,2 COS 0
-Id
I,2 cos a
Id
Id I&2 COS a
1,2 cos (060)
60)
+1,2 cos(0
Id
-Id
Jd + /s2C0S
/s2cos(0+6O)
si cos a
h I
+I,2 cos(0
Is2 cos(0 + 60)
-Id
1,2 COS a
Id~Is2 COS
1,2 cos(0 60)
+1,2 cos(0+6O)
-h
COS
Id
2, 3
6, 1
I'd COS
+/s2 cos 0
1,2,3
5, 6, 1
1,2
+ 60)
1,2 cosa+
/s2cos(0+6O)
Table 2. Instantaneous Voltages in Three-phase Bridge Converter with Overlap Less than 60
(Alternately Two and Three Valves Conducting)
Valves
wt=6
conducting
V4-
Vs
V6
Va
vb
Gba
Gca
Gca
Gcb
Go
Gb
l.5ec
i.5ec
&ab
Gca
Gcb
-l.5eb
Gcb
6 be
Vi
Vz
1,2
1,2,3
vP
Vn
Gc
Ga
Gc
Gac
\.5ec -eJ2 -Gc/2
Gc
-ed2
Gc
1.5ec
Gcb
Gbc
Vc
Vd
8 60
/y
<JL
8
2,3
a
Ga
Gb
ec
Gb
Gc
-Gbl2
Gb
ebj2
Gb
es/2
Gab
Ga
Gb
Gc
Gb
Ga
Cba
l.5ea
ea
Ga
1.5ea
G ab
Ga
Gca
+ 60
'2,3,4
-1.5e l.5eb
8+60
3,4
Gab
1.5eb
a -1-120
3,4,5
l.5ea
1.5ea
Gac
Gac
S -!- 120
4,5
a
+ 180
8 + 180
a
+ 240
8 + 240
4, 5,6
1.5ec -\.5ec l.5ec
Gb
-GcS2 -ec/2
Gac
Gbc
Gbc
Gba
Ga
eb
5, 6, 1
1.5e&
1.56
l.5eb
-Gbl2
Gb
6,1
Gbc
Gba
Gba
Gca
Ga
Gb
6,1,2
5,6
a +300
8 + 300
-Gail ~eJ2 ~eJ2
-1.5ea
l.5ea
1.5ea
Ga
Gc
Gc
ea
Gc
Gc
-eJ2
Gc
Gc
Gb
Gcb
Gb
1.5eb
Gb
Gab
-Gb/2 -ebj2
Gc
-Gail ea/2
Ga
Ga
-eaj2
1.5ec
1.5ea
3-2
OVERLAP LESS THAN
60
87
(a) JO
Nli
>
ea + e&
ec
(at
Fig. 12. For derivation of voltage drop caused by overlap.
every sixth of a cycle (7t/3 rad). As already shown, area

larly, area A = AVd n(3: '
"cb
*ab
*bc
eb)dQ= \"eJde
U. - e" +
2
A=
*ba
r u
sin G dQ =
V3 Em (cos a
3
A
AKd = 71
1.5e
bb
8S
OO
to 10
oo
CJ CM
Fig, 11. Instantaneous currents and voltages of three-phase bridge converter with
a = 15 and u= 15: (a) valve currents; (6) line currents; (c) alternating line-to-neutral
voltages; id) direct positive and negative voltages with respect to neutral point of ac source;
(e) direct voltage between poles; (/) voltage across valve 1. Voltages across other valves
have the same wave form but are displaced 60 apart; 4/ means valve 4 ignites; 2e, valve
2 extinguishes; etc.
A0 = Vd0 %\3. Simi
-25
[-cs0J
cos 5)
.
= --(cos a cos 5)
2n
(cos a cos (5)
(20)
Without overlap, the direct voltage was Vd0 cos aEq. (6). With overlap,
it is
Fj0(cos a + cos 8)
Va = VAn cos a - AVa =
(21)
Comparison of Eqs. (19) and (20) shows that the voltage drop is directly
proportional to the current:
Vdo
2Is2
(22)
88
3-2 OVERLAP LESS THAN 60
Hence
Vd = Vd0 I cos a - 21
(23)
sZ
By use of Eqs. (4) and (16), this becomes

(24)
Vd = Vdo cos a - RcId

where
3
- coLc = - Xe = 6fLc
Rc = 71
71
ohms
(25)
Rc is called the equivalent commutating resistance. Although it is the ratio of
drop of direct voltage to direct current, it consumes no power. It could be

likened to the drop in terminal voltage of a dc generator per ampere of current
due to a differential series field winding.
The equivalent circuit of the bridge rectifier, operating at constant alternat
ing voltage and constant ignition angle, is given in Figure 13. The direct
voltages and current in this circuit are average values without ripple.
89
for given values in volts and amperes, respectively. However, the ratio
VJIs2, which is the base dc resistance, is independent of the alternating
voltage, and so is the per-unit value of commutating resistance R'c.
Second System. A more usual base direct current than
of course, is the
rated or full-load current Id. The per-unit current on this new base is Id =
IJIdn. For reasons that will appear later, it is desirable to adopt at the same
time a new base direct voltage Vd0n, which is the value of Vd0 under normal
full-load rectifier operation with terminal voltage Vd at its normal value
Vd. Vd0n must be distinguished' from Vd0, for only one of them can remain
constant while the direct current varies. The per-unit direct voltage on this
new base is Vd = VdjVd0n.
We shall show that in this system there is a simple relation between the
per-unit commutating resistance and the per-unit commutating reactance
based on the rated voltage and current of the converter transformer. These
transformer ratings must be expressed in terms of the base voltage and current
adopted for the dc circuit. The rated line-to-neutral rms valve-side transformer
voltage is, by Eq. (5),
1
Vd0
&LNn =
Rc = 6fL
(27)
vW
Vdo cos a
Fig. 13. Equivalent circuit of bridge rectifier.
Voltage Drop in Terms of Per-unit Quantities
First System. Let Vi0 be chosen as the base voltage and Is2 as the base
current of a per-unit system. Division of Eq. (23) by Vd0 yields
(26)
' V'd = cos a - \Id
where
The rated (full-load) alternating current (rms value including harmonics) is

computed, for simplicity, from the rectangular pulses of height Id amperes
and width 2rc/3 rad that would exist at no overlap. This value is
/ = 4,72/3 = 0.8164,
It is correct within a few percent at full-load overlap.* The base ac impe

dance is then
,b
Eln
4/1 6
Vd0
Id
(Z9)
and the per-unit commutating reactance is
V'd = VdjVd0 is the per-unit direct voltage, and l'd = Idjls2 is the per-unit
direct current on the chosen bases. On the same bases the per-unit commutat
ing resistance is R'c =
It may be noted that in this per-unit system the base quantities Vd0 and
Is2 both vary in direct proportion to the alternating voltage referred to the
valve side of the transformers, which voltage is proportional to that on the
other side of the transformers and to the turns ratio, which is adjustable.
Hence a change in voltage affects the per-unit values of voltage and current
(28)
X,.
{7tRf\f6Vd0\
2R,
*=zrlrAJ-i=2*'
(30>
Hence
Y"
r"=t'
See Appendix A for more accurate values with overlap.
(31)
90
3-2
OVERLAP LESS THAN
60
91
that is, the per-unit commutating resistance on the new base is half the per-unit
commutating reactance based on the transformer rating. Equation (23) becomes
which is exact only if u = 0 but which is true with a maximum error of 4.3%
at u = 60 and only 1.1% for u 30 (the normal operating range). It follows
that the displacement factor is
cos a -
, cos a + cos 8
-cos (p -
Vd = Vd0
or, dividing by Vd0 and putting
V'J =
I'd
Vd0n
(32)
Vd0/Vd0n = V'JQ,
V-cosa-l'd'
(40)
By the use of Eqs. (21) and (24), it may be written also as
(33)*
cos (f>
cos a
' dO
(41)
' dO
By the substitution of Eq. (40) into Eq. (34), we get
Relations between AC and DC Quantities

Simple Approximate Relationships. These relationships hold between the ac
quantities at the point where the voltage waves are sinusoidal (between the
emfs and inductances in Figure 1) and the dc quantities. From Eqs. (4) and
(20), we obtain the following relationship between the alternating and direct
voltages, which is valid for u < 60 :
36(cs + cos
(34)
With losses neglected, active ac power equals dc power:
P.=P*
(35)
Pa = 3ELNIL1 cos $
(36)
Pd=VdId
(37)
where
Vd s
Eln cos 0
The concept of the converter's having a fixed current ratio IJI and an
adjustable voltage ratio VJELN is still valid. Now, however, the current ratio
varies a few percent with load, and the ac displacement factor depends on
loadEq. (41) in addition to ignition delay angle.
Equations (35), (39), and (42) give approximate relations, not involving
the converter angles a and <5, for power, current, and voltage, respectively.
These three equations, in terms of per-unit quantities on any power base
equal on both sides and on any pair of current bases proportioned as in
Eq. (39), become
(43)
(44)
and
Equating the right-hand sides of Eqs. (36) and (37), substituting Eq. (34)
for Vd, and solving for the active alternating current, we get
\/cos a + cos
'cos*=(VX
2 j
(38)
V'd E'a cos cj>
(45)
Reactive power on the ac side may be found from
Q = Pa tan <{>
(46)
where <j) is given by Eq. (40) or (41). Of course, there is no reactive power
on the dc side.
Now comes the approximation,
/I]S76Wmo
(42)
(39)
Vd0n , equations similar to

* Perhaps because of neglect of the distinction between Vd0 and
1) have appeared in converter
this except with the factor Vd0 omitted (assumed equal to

literature but are not correct unless Vd0 is kept constant at the particular value Vd0n . If,
for example, a rectifier is operated at constant terminal voltage Vd or V"d but with Vd0 (or
V/0) increasing with Ia, the factor Vd0 should be retained.
Effect of Transformer Ratio. In practice, a bank of transformers is con

nected between the ac source and the bridge-connected valves, and the ratio
of these transformers is adjustable under load by means of tap-changers. The
ac quantities of most interest are those on the network side of the trans
formers.
Let T be the ratio of line-to-line voltages on the valve side to those on the
network side at no load and at any tap, and let T be the value of T correspond
ing to rated voltages on both sides. Also let T' = TjT be the per-unit ratio
92
3-3
of the transformers, that is, the factor by which the ratio is off its nominal
value Tn.
Then Eq. (43) is unchanged, but Eqs. (44) and (45) become
()
V'd T'E'a cos (j>
(48)
in which the ac quantities are referred to the network side.
Exact Relations A harmonic analysis of the current wave (Appendix B)

shows that its fundamental-frequency inphase and quadrature lagging com
ponents (with respect to ELN) are
cos a + cos 8
IL1 cos 0 = IL10 ---
....
(49)
ANALYSIS WITH OVERLAP GREATER THAN
93
60
circuit, or low alternating voltage. Therefore, the present section may be

skipped by readers interested only in the normal operation.
Inspection of Figure 4e or / shows that the overlap interval u, during which
current is transferred, say, from valve 1 to valve 3 in one half bridge, begins
and ends with subintervals, each of angular span u 60, in which four
valves are conducting, there being simultaneously with commutation from
VI to V3, a commutation in the other half bridge, either from V6 to V2 or
from V2 to V4. Between the beginning and ending subintervals is a middle
subinterval of angular span 120 u in which no other commutation occurs
and only three valves are conducting. When four valves are conducting
(Figure 14), they constitute a three-phase short circuit on the ac source and
(ia) (va)
(1)
(k)
<s>
and
. ,
hi sin -jhi sin <j>
and the scalar fundamental current is
hi =
cos 4>f
+ (hi sin <P)2]112
(52)
The reactive power is
Q = 3ELN(ILi sin <f>)
= P tan
<f>
(53)
where
cos 2a cos 2o
2u + sin 2a
tan <p = -
sin 25
Fig. 14. Bridge converter with valves 1, 2, 3, and 4 (or 6, 1, 2, and 3) conducting.
a pole-to-pole short circuit on the dc terminals. When three valves are con
ducting (Figure 9), they constitute a line-to-line short circuit on the ac source,
as already considered in Section 3-2. In the four-valve subintervals, the direct
voltage is zero; in the three-valve intervals, it is an arc of a sine wave of
amplitude 1.5Em (the altitude of the voltage triangle in Figure 2).
-tt
u csc u csc(a + <5) cot(a + 5)
Currents during Commutation

(54)
3-3 ANALYSIS WITH OVERLAP GREATER THAN 60
The current i3 in valve 3 will be traced through the three subintervals that
constitute the interval of angular span u in which i3 rises from 0 to Id. During
this same interval ix decreases from Id to 0 and is always equal to Id i3.
Operation of the bridge converter with overlap angle in the range between
60 and 120 is abnormal, being encountered only under overload, dc short
First Subinterval. The first subinterval extends from cot = a to cot =5 60.
Valves 1, 2, 3, and 6 are conducting, and the upper one of each pair of
quantities in Figure 14 should be used.
94
3-3
,c-et-2Lcf-Le|=
(SO
I'X
dia 2ea -eb-ec
(57)
3Lc
By addition to the numerator of
ea + eb + ec=0
(58)
Eq. (57) simplifies to
Lc
cos (cot + 60) ~~-=
dt
dt
(59)
Similar equations hold for currents in the other two phases. Of course, these
results could be obtained directly by considering a three-phase short circuit

on the positive-sequence circuit of the ac source.
By the integration of Eq. (59) we get
coLr
+ 60) d{cot)
cos {cot
cos (cot 30)]
(60)
_ 2
's3 ~r~ /- *s2

Lc V3
(b 1)
being the crest value of the alternating component of current in a three-phase

short circuit. Equation (60) satisfies the initial condition that i3 = 0 when
a.
At the end of the first subinterval cot = 5
60 and
h = 7s3[cos (gc 30) cos (5 90)]
(62)
Second Subinterval. The second subinterval extends from cot = <5 60 to

+ 60. Valves 1, 2, and 3 are conducting. The circuit in Figure 9 holds.
cot = a
cos {cot 90) d{cot)
2coL
<5 _6Q'
fx
/s3[cos {cot - lW)$Ler =
fx
a/3 Ecos {d - 60) - cos cot ]
(64)
Tcos
(a 30) cos (5 90)
cos (<S 60)
cos cot
(65)
By vector addition,
-cos (5 - 90)
cos {d
- 60) = icos {5 + 30)
(66)
h = 7js3
jcs
(67)
(a 30) + \ cos (5 + 30)
cos cot
At the end of the subinterval, cot = a + 60 ; hence
where
cot
<"'
Hence
-/,3[cosM-30)rw'
= /s3[cos (a 30)
V"3 E,
to which must be added the initial value, equal to the final value in the pre
ceding subinterval and given by Eq. (62). The sum is
rat
<3 =
(63)
Integration gives
If they are solved for dijdt, the result is
Lc
95
f =zf=4rcos(ffl'-90O)
(55>
dt
60
There is a line-to-line short circuit on phases a and b of the source through

valves 1 and 3. Hence
The mesh equations may be written
dt
i3 =
cos
(a 30)
+ \ cos (5 + 30)
cos (a +
60)J
Vector combination of the first and third terms results in
u-ijy/icos tan-1
[f
+ icos
0)j
(68)
Third Subinterval. The third subinterval extends from cot = a + 60 to

1, 2, 3, and 4 are conducting. The circuit in Figure 14 applies
again, but with the lower symbol of each pair. The alternating components
of the phase currents are given by the same equations as those for the first
cot = 5. Valves
96
3-3
subinterval except that valve 3 is now in series with phase b, so that the ac
component is 90 behind Eb . The dc component is different from what it was
in the first subinterval. The total current in valve 3 is
i3 = ib = Is3 cos (cot 150) 4-
(69)
73
where 73 is a constant of integration determined from the fact that, when

cot = a + 60, i3 has the value given by Eq. (68):
/s3cosa tan1-y|+
j'3 - /s3j-~cosa tan-1
-cos(<5 4- 30) = /s3 cos (a 90) 4- 73
(a - 30) + icos (S + 30)]

cos (5 + 30) - cos (cot + 30)]
h = fs3[i cos (a - 30) + i
(70)
(71)
At the end of the subinterval and whole interval, cot = 6 and i3 = Id. This
fact enables us to obtain the direct current as a function of the ignition and
extinction angles.
jd = /s3[ cos (a - 30) + icos (S + 30) - cos (3 + 30)]
30) + icos (5 4- 30)
cos cot
(cos a' cos 5')
(72)
+ 30. For u > 60, Eq. (72) replaces Eq. (19),
30) + \ cos (5 4- 30)
(75)
(76)
Vl = v2 = v3 = 0
(77)
v4 = vs=v6 = vd
Va
II
Vc
= v =
II
\.5ec
1.5Em cos cot
0.5ec = 0.5Em cos cot
ec = Em cos cot
oo
(79)
(80)
h =Id- IAicos (a - 30)+i cos (3 4- 30) - cos (cot 4- 30)]
(81)
By analogy with ix in the first subinterval, with cot decreased by 60,
i2
ic = Id 7s3[cos (a 30) cos (cot 90)]
(82)
From Eq. (71),
/3 = ib = 7s3[cos (a 30) 4- \ cos (5 + 30) cos (cot 4- 30)] (83)

(84)
(85)
(87)
Subsequent Subintervals. The equations may be obtained from those for

the same number of conducting valves by appropriate changes of subscripts
and phase angles. The results are shown in Tables 3 and 4. The wave forms
are plotted in Figure 15 for a 45 and u = 75.
Average Direct Current and Voltage
Second Subinterval (Valves 1, 2, 3 Conducting)
7s3cos(a
h = is = h = 0
= v2 = v 3 = v4 = v5 = v6 = va = vb = vc = vp = vn = vd = 0
Instantaneous Currents and Voltages
i2 ic = Id
cos (a
ia = ix i4 Id 7s3[fcos (a 30) 4- cos (5 + 30) cos (cot 30)] (86)
V3
where a' = a 30 and 3' = 3

which is for u < 60.
ix = ia = Jd
II
i*4 = 7s3[cos (a 30) cos (cot 90)]

is - h = 0
- 30) - cos (3 + 30)]
-L Is2[cos (a - 30) - cos (3 + 30)]

V3
97
By analogy with i3 in the first subinterval, with cot decreased by 60,
On combination of the last two terms, this simplifies to
h = K3[ cos (a
ii
60
Third Subinterval (Valves 1, 2, 3, 4 Conducting)
- cos (a 90) + \ cos (5 + 30)
= Is3[i cos
*3
cos cor
(73)
(74)
Average current is given by Eq. (72).

Average voltage is found by averaging the instantaneous voltage over a
60 interval comprising a subinterval in which it is zero and one in which it
is an arc of a sinusoid of amplitude 1.57Tm. Let the latter subinterval be the
second one used in the discussion of commutation. In it the instantaneous
Table 3. Instantaneous Currents in Three-phase Bridge Converter with Overlap Greater than 60
(Alternately Three and Four Valves Conducting)
Valves
!l3
conducting
ii
ii
iz
is
is
ic
ib
la
y'
8-60
a
+ 60
1,2,3
Id A*
1,2, 3,4
Id -A
-Id
C-h
ld-B L -C
~D
-E
Id
-Id
h-G
Id
to
8
a
+ 120
+ 60
+ 180
,8
+ 120
+ 240
8 + 180
a
D-h
Id -D
2, 3, 4, 5
Id -E Id -F
3,4,5
3,4,5,6
h -H Id -K
K-h
h- H-K
4, 5,6
Id -L
Id
L-h
-L
Id
4, 5, 6,1
.0
Id M L-N
M+N-h
h-N
5,6,1
.0
h-P
Id
-Id
h-P
5,6,1,2
Id Q
Id-R
R Id
h- Q-R
Id-F
+ F-Id
+300
6,1,2
6, 1, 2, 3
2, 3,4
Id
Id-U
S
T
*See Table 3A for definition of symbols A, B,
h-S
Id-T
-3
Table 3A
A = 7s3[cos (a - 30) cos (8 30) - (V3/2) cos 0]
B 7s3[ cos (a 30) i
cos (8 30) cos (0 30)]
C /s3[cos (a 30) cos (0 90)]
D = 7s3[cos (a - 30) + cos (8 30) - (Vl/2) cos (0 - 60]
E = 7s3[ cos (a 30) + cos (8 30) cos (0 30)]
F 7s3[cos (a 30) cos (0 150)]
G = 7s3[cos (a - 30) +* cos (8 30) - (VJ/2) cos (0 - 120)]
H 7j3[ cos (a 30) +% cos (8 30) cos (0 90)]
K = 7s3[cos (a 30) cos (0 210)]
L = 7,3[cos (a - 30) +icos (8 + 30) - (Vl12) cos (0 - 180)]
M= 7,3[ cos (a - 30) +h cos (8 30) - cos (0 - 150)]
N=Is3[cos(a -30) - cos (0 - 270)]
P = 7s3[cos (a - 30) J cos (8 30) - (V3/2) cos (0 - 240)]
Q = Jssti cos (a - 30) icos (8 30) -cos (0 -210))
R = 7s3[cos (a 30) cos (0 330)]
+
+
+
+
+
+
S 7,3[cos(a - 30) + cos (8 + 30) - (V3/2) cos (0 - 300)]

T=
cos (a - 30)
cos (8 + 30) - cos (0 - 270)]
U= 7,3[cos (a 30) cos (0 30)]
-S
. .. U.
cq
T+U-h
-T
3-3 ANALYSIS WITH OVERLAP GREATER THAN 60

in
iH
13. lb
in
101
ii
v->
(a)
'
<a
'
(b)
11
1
fS3I
in
>n
<w
!Q
*1
is
(c)
J=
"jo-
Cu w
g $
HH
.S >
Q>
(A
hS
f1
eba
If
1.5e,
1.5ec 1.5<?2, /
Nr l'5ea 1.5ec
1.5eft
1.5ca
1.5e,
<
in_in_|n__
_-
(d)
1
H
in
in
in
$ .S
4-
1
g
o
o
3
!lII
CO
N-
<N
CO
CO
<N
CS
<N
oo
+a
CO
*t
'S-
>
+
to
<n
in
ro*-
co*
n
0
CO
SO
+
GO
oo
1-"
+a
GO
+a
it
<N
-T
cN
"
n"
in
Tf"
m*
0
oo
ii
CO
+a
GO
BibMsca da F'H* PoHUcnica

.:0
CO
-T
o
*
0
CN
GO
CO
co
Fig. 15. Instantaneous currents and voltages of three-phase bridge converter with
and u = 75 : (a) valve and phase currents; (b) alternating line-to-neutral voltage,
phase a; (c) direct positive and negative voltages with respect to neutral point of ac source;
{d) direct voltage between poles (broken line) and voltage across valve 2 (solid line).
a = 45
102
3-4
COMPLETE
CHARACTERISTICS OF RECTIFIER
103
voltage is, by Eq. (78), vd = \.SEm cos 6. The average voltage over 60 (rt/3 rad)
is.
/*a+60
Vd = -71
J <5-60
_]c5 60
/3
= -y- dotcos (a 30)
~|a+60"
qp
- 90)
1.5m cos e dd = cos(6
271
n/3
+ cos (5 + 30)]
Fd0(cos a' + cos <$')
(88)
with a' and S' as in Eq. (72). Equation (88) for u > 60 is analogous to Eq. (21)
for u < 60.
Direct Voltage as a Function of Direct Current for Constant Ignition Angle
By elimination of co S between Eqs. (72) and (88), we get
=V3coS(a-30)-
"dO
(89)
*s2
I.
Fig. 16. For explaining that the minimum ignition delay angle is 30 if the overlap angle
is 60 or greater.
In the first per-unit system, this becomes
y'd =
where
V'd = VdjVd0
and
Vd
cos (
30)
1.5/rf
(,
(90)
l'd = Idjls2. In the second system,
V3 V'do cos (a - 30) - 1.5 I"d
(91)
where V"d = Vd{Vd0n, Vd0 = KJ0/Fd0n, I"d =IJId, and X"c is the per-unit
commutating reactance based on the rating of the converter transformer.
Comparison of Eq. (90) with Eq. (26) and of Eq. (91) with Eq. (33) shows
that the equivalent commutating resistance for u > 60 is three times that for
u < 60.
Minimum Ignition Angle
If the overlap is greater than 60, ignition of a valve in one half bridge
must begin while the preceding commutation is still under way in the other
half bridge (Figure 4e), putting a line-to-line short circuit on the ac source.
For example, valve 3 must fire while valves 2 and 6 are still commutating and
putting a short circuit on phases b and c. The anode of valve 3 is at potential
vb = (fc + O/2 = -eJ2} and its cathode is at potential va = ea. See Figure 16.
Valve 3 cannot ignite until its anode potential becomes a little greater than
its cathode potential, that is, until vb > va, which, for balanced alternating
voltages, occurs first at point B, where cot = 30. If it were not for the short
circuit, vb would be eb, and ignition would occur at point A, where cot = 0.
Hence the minimum ignition delay angle is 30. This delay occurs spon
taneously even with no grid control.
The same phenomenon may be seen in the voltage v2 across valve 2 in
Figure 15d. This voltage becomes positive at cot = 30, which for valve 2
corresponds to a = 30.
.
3-4 COMPLETE CHARACTERISTICS OF RECTIFIER
A three-phase bridge rectifier with no grid control operates successively in

three different modes as the direct current is increased from zero to the shortcircuit current /s3:
First mode
Alternately two and three valves conducting.
a = 0.
u increases from 0 to 60.
I'd increases from 0 to 0.500.
Vd decreases linearly from 1 to 0.750. :
104
Second mode
Three valves conducting.
a increases from 0 to 30.
u = 60.
I'd increases from 0.500 to 0.866.
V'd decreases on a curve from 0.750 to 0.433.
Third mode
Alternately three and four valves conducting.
a = 30.
I'd increases from 0.866 to 1.155.
V'd decreases linearly from 0.433 to 0.
A rectifier with grid control set for a = a0, where a0 is between 0 and 30,
has three modes similar to those of the uncontrolled rectifier :
First mode
Alternately two and three valves conducting.
a = a0.
Second mode
Three valves conducting.
<x increases from a0 to 30.
u = 60.
Third mode
Alternately three and four valves conducting.
oc = 30.
For a 30, the second mode disappears, and transition is made from the
first mode directly to the third, with a = a0.
The spontaneous increase of a in the second mode is sometimes called
auto phase control The incoming valve cannot ignite until the polarity of the
voltage across it is reversed by the completion of commutation in the other
half bridge. The situation is similar to that already described in Section 3-3,
page 102, for u > 60.
The direct current and voltage at transition points between modes can be
calculated by means of the equations already given for Id and Vd as functions
of a and 5. The relations will be clarified by study of the first chart described
in Chapter 4.
Normally the rectifier operates in the first mode. The other modes occur
3-5
INVERSION
105
from overloads, dc short circuits, or low alternating voltage. Such abnormal

operation may be studied conveniently in the laboratory by using an exag
gerated value of commutating reactance.
3-5 INVERSION
General
Because the valves conduct in only one direction, the current in a converter
cannot be reversed, and power reversal can be obtained only by the reversal
of the average direct voltage Vd. The voltage then opposes the current, as in
a dc motor, and is called a countervoltage.
In Section 3-1 it was shown that, if there is no overlap, Vd reverses at
a 90 ; that is, rectification occurs for 0 < a < 90 and inversion for
90 < a < 180. Since in a practical case there is always some overlap, the
value of a at which inversion begins depends on the overlap, or on the
current, and may be found from Eq. (21) or (24). From Eq. (21), the transi
tional value is
(92)
which is always less than 90.

Moreover, <5 ought to be less than n by at least the angle y0 corresponding
to the time required for deionization of the arc, which is 1 to 8; hence
a ought to be less than it u. Figure 17 shows the voltage eba causing com
mutation of current from valve 1 to valve 3. If the current has not been
completely shifted to valve 3, with margin for deionization of valve 1, before
cat = 71, when eba reverses, the current thereafter is transferred from valve,
3 back to valve 1. This occurrence is called a commutation failure and is
discussed in Chapter 6.
The commutation voltage for hv dc inverters is furnished by synchronous
machines. If the ac system receiving power from the dc line has no generators,
or none running, a synchronous condenser is usedin Gotland, for example.
Some small single-phase and polyphase inverters use tuned LC circuits
instead.
Notation for Ignition and Extinction Angles

In rectifier theory, the ignition angle a was defined as the angle by which
ignition is delayed from the instant (cot 0 for valve 3) at which the com
mutating voltage (eba for valve 3) is zero and increasing. Similarly, the
106
3-5
inversion
107
cos p and cos <5 =
cos y, with these results:
h = Ucos y - cos p)
(96)
_ Fd0(cos y + cos p)
(97)
l'd = cos y cos p
(98)
__ cos y + cos p
(99)
sign of Vd and putting cos a =
F,=
In the first per-unit system,
cot
1/r>
v*
'
In the second per-unit system,

Inverter
Rectifier
X'cI'd
,
=o(cos y-cos/3)
(100)
P0(cos y + cos P)
2
(101)
For constant ignition advance angle P, Eqs. (24), (26), and (33) become,
respectively,
Vd = Vd0 cos p + RJd

V'd = cos p + \l'd
Fig. 17. Showing relations among angles used in converter theory and why the curvature
of the front of a current pulse of an inverter differs from that of a rectifier.
extinction angle 8 is measured by the delay from that same instant (cot = 0 for
valve 1). Although angles defined in the same way and having values between
90 and 180 could be used in inverter theory, commoner practice is to define
ignition angle p and extinction angle y by their advance with respect to the
instant (cot 180 for ignition of valve 3 and extinction of valve 1) when the
commutation voltage is zero and decreasing. See Figure 17. The relation
among the several inverter angles is as follows :
P = n-ot
(93)
y =n
(94)
u =8
a = 13 y
(95)
Equations for Average Direct Current and Voltage in Terms of /? and y

Inverter voltage, considered negative in the general converter equations, is
usually taken as positive when written specifically for an inverter. General
equations (19) and (21) are changed to inverter equations by changing the
V'd
Vd0 cos p + \X"C Id
(102)
(103)
(104)
Because inverters are commonly controlled so as to operate at constant

extinction advance angle y (see Section 5-7), it is useful to have the relations
between Vd and Id for this condition. Elimination of cos p from Eqs. (96)
and (97) gives
Vd = Vdo cos y-RJd
(105)
(The corresponding equations in per-unit quantities are like (103) and (104)
but with reversed sign of the last term.) Under this condition the equivalent
commutation resistance is Rc and is negative. It may be likened to the
voltage drop per unit of current due to the series field of a differentially com
pounded dc motor, forced to run at constant speed by driving an induction
generator.
Equivalent circuits of the inverter are shown in Figure 18.

The foregoing equations are for overlap less than 60. For greater overlap,
the equations given in Section 3-3 are valid. If desired, these may be modified
to change the sign convention for~Vd and to use P and y instead of a and 8.
108
Id-
.
5
Vdo COS B
T-
(a)
lA/
W9
VdO COS 7
(b)
Fig. 18. Equivalent circuits of inverter.
Wave Forms
Wave forms of instantaneous voltages and currents are shown in Figure 19

for inverter operation with y = 15 and u 15, hence with ft = 30.
The sinusoidal arcs of current during commutation have, for inverter
operation, opposite curvature from those in rectifier operation. See Figure 17.
If both current waves have the same overlap and if a for one equals y for the
other, the shape of one wave is like that of the other turned end-for-end
about a vertical axis.
The voltage across an inverter valve is compared with the voltage across a
rectifier valve in Figure 20. Again ur = u-t and ar = yh with subscript r denoting
a rectifier and ian inverter. The wave form of the inverter is like that of the
rectifier rotated 180 in the plane of the paper. The following similarities and
dissimilarities are noteworthy:
1. The average voltage across a rectifier valve is negative and is called

inverse voltage', the average voltage across an inverter valve is positive; hence
the necessity for reliable grid control of inverter valves is apparent.
2. In both modes of operation, the voltage across the valve is positive just
before conduction begins, but in the rectifier it is positive for a shorter while
than in the inverter, approaching zero as a approaches zero.
3. In both modes of operation, the voltage across the valve is negative
immediately after extinction of the arc, but in the inverter it is negative for a
much shorter while (the commutation margin) than in the rectifier.

4. Abrupt changes in voltage across the valve occur at ignition and ex
tinction. For a, = yt and equal alternating voltages, the voltage jump at
ignition is greater in the inverter than in the rectifier, but as to the voltage
jump at extinction, the opposite is true.
Formulas for the major voltage jumps are as follows :
Rectifier
Inverter
Ignition
V~3 Em sin a
V 3 Em. sin ft
(106)
Extinction
V3 Em sin 8
3 Em sin y
(107)
I I
[
<&c1 1
O
lO <>
<Q
1 1
oo
'
co
co r~1
CM CM
~H
l l
CO
"S-
CM-CM
II
Fig. 19. Wave forms of currents and voltages of three-phase bridge inverter with a = 150,
u 15, and y = 15: (a) valve currents; (b) line currents; (c) line-to-neutral
alternating voltages; (d) positive and negative direct voltages with respect to neutral point
of ac source; (e) voltage across valve 2; (/) direct voltage between poles.
jS = 30,
aJ
TT
tTr?
( t.. 7.
Ct-ts*.
110
3-5
111
INVERSION
Voltage
Current
(a)
*60
-60'
(a) First range
Voltage
60s-
-60'
(b) Second range
(c) Third range
Fig. 21. Relations among inverter angles right after extinction of a valve: (a) /3 = 45,
u = 30, y = =15; (b) jS = 75, = 45, y=30, =15; (c) jS = 97.5, =52.5,
y = 45, =15. A 1.5Em sin(a>? -f 30); B = V 3Em sinatf; C l-Sism
(lot - 30).
(b)
Fig. 20. Wave forms of voltage across a valve in (a) inverter and (b) rectifier: 1. inverse
period; 2. blocking period; 3. conducting period.
There are also four minor voltage jumps per cycle, two of which are half as
great as the jump at ignition, and the other two, half as great as the jump at
extinction. These occur at moments when other valves are being ignited or
extinguished.
Commutation Margin
A distinction is made between extinction advance angle y and commutation
margin angle . The two are often confused, because they are equal under
normal operating conditions.

Extinction advance angle is the time angle between the end of conduction
and the reversal of the sign of the sinusoidal commutation voltage of the
source. It is given by
y = 71
Commutation margin angle is the time angle between the end of conduction
and the reversal of the sign of the nonsinusoidal voltage across the outgoing
valve.
The voltage across an inverter valve (Figures 20a and 21a) has a positive
dent D because of the succeeding commutation. This dent normally occurs

after the sinusoidal voltage on which it is superposed has become positive.
Under this condition = y. With increasing current or decreasing commuta
tion voltage, however, the width of this dent, which is the overlap angle w,
increases; and with increased overlap or earlier ignition, the dent encroaches
on the period in which the valve voltage would otherwise be negative (Figure
216) and thus makes < y. After the front of the dent becomes entirely
negative, further advance of the dent (Figure 21c) does not decrease the
commutation margin further.
The three situations loosely described above in terms of the advance of the
dent can be specified rigorously as different ranges of ignition advance angle
/?. In each range the commutation margin angle follows a different law.
First Range (Figure 21a). The front of the dent is entirely above the
horizontal axis.
< 60c
=y
(108a)
(1086)
Second Range (Figure 216). The front of the dent is partly above and
partly below the horizontal axis.
60 <
< 90c
= 60 u = y (fi~ 60)
(109a)
(1096)
112
3-6
Third Range (Figure 21c). The front of the dent is entirely below the
horizontal axis.
P > 90
C = y-30
(110a)
(110b)
These laws can be verified by a study of Figure 21 inthe light of the follow
ing considerations. The ignition and extinction advance angles are measured
back from the instant of reversal of the commutating voltage. For the com
mutation that results in the extinction of the arc in the valve across which the
voltage is graphed in the figure, this instant is marked 0, and the extinction
angle y for this valve is indicated; however, the ignition of the incoming
valve of the same commutation is indistinguishable in the graph. The ignition
and extinction in the next commutation (occurring in the other half bridge)
occur 60 later, at the front and tail of the dent D. Hence its front is 60
after (or /? 60 before) the reference point O, and its tail is 60 y after the
reference point. The sine wave forming the upper limit of the dent leads the
main wave by 30.
;
\
anodes. Either scheme requires a current divider for forcing equal division of
the current among the several parallel paths. Otherwise, the first anode to
begin conduction would limit the voltage of the remaining anodes to the arc
drop, which is insufficient to ignite arcs to those anodes.
A precaution usually taken with multianode valves is to delay ignition by
a minimum angle of about 5 to 7, so that there is a definite and adequate
voltage from each anode to the cathode just before ignition. This increases the
probability that all anodes ignite simultaneously as soon as the control grids
are made positive. This can be done with little effect on the average direct
voltage, which varies as cos a ; the voltage across the valve just before
ignition varies as sin a; for example, with valves having peak inverse voltages
of 100 kV, by changing a from 0 to 7, the firing voltage is changed from the
bare minimum (about 2 kV) to 100 sin 7 = 12.3 kV, and the no-load direct
voltage is decreased from 95.5 kV to 95.5 cos 7 = 94.8 kV, or only by
j
j
The current divider, shown in Figure 22 for a four-anode valve, consists of

four current transformers of ratio 1 : 4. It also increases the certainty of all
anodes firing. If one anode fires first, the high rate of change of current in the
primary windings of the transformers for the other anodes induces a high
enough voltage intheir open-circuited secondary windings to strike arcs almost
immediately on those anodes also.
Current
The current and voltage required for high-power long-distance transmission

are beyond the ratings of single valves and must therefore be raised by parallel
or series connections.
The current can be increased by using the following:
1.
2.
3.
4.
transformers/ratio 1:4
\JL>
rwr
imn
VW
iJVW
Valves in parallel in each bridge arm

Two or more anodes in parallel for each tank and cathode
Bridges in parallel
Some combination of these means
Anodes
The voltage can be increased by connecting:
1. Valves in series in each bridge arm

2. Bridges in series
3. Both
Anodes in parallel have an advantage over valves in parallel in that the

former scheme reduces the duplication of tanks, cathodes, ignition and
excitation anodes, etc., and saves space. Voltage grading and control grids
are repeated on each anode assembly. ASEA valves have two, four, or six
113
0.7%.
I
3-6 SERIES AND PARALLEL ARRANGEMENTS OF VALVES, ANODES,
OR BRIDGES
"
SERIES AND PARALLEL ARRANGEMENTS
Cathode
Fig. 22. Current divider for four anodes in parallel.
114
3-7
The transformers and the resistors across their secondary windings main
tain almost equal currents in the several anodes. At first glance, the use of
transformers of direct current appears strange. It must be remembered,
however, that the current in each valve (except the bypass valve, to be dis
cussed in Section 6-2, page 199) is a train of pulses having both ac and dc
components. Transformation of the ac components, together with the clamp
ing effect of the unilateral conduction of each anode, suffices to divide the
dc components also.
The resistors further aid in maintaining equal current division by increasing
the voltage on the anodes carrying less than their share of the total current
and decreasing the voltage on the anodes carrying more than their share. On
the assumption that each secondary winding in Figure 22 always carries onefourth of the primary current, if each anode likewise carries one-fourth, there
are no currents in the resistors. If the anodes carry unequal currents, the
surplus or deficit of each anode current with respect to the normal current
flows through the resistor in the same or opposite direction, respectively,
introducing equalizing voltages into the anode circuits.
By making the resistors nonlinear they perform the additional function of
limiting the voltage across each transformer winding. This voltage is adequate
in amount and duration for ignition at the minimum delay angle. The
product of voltage and duration equals the change of core flux and is limited
by saturation. If the voltage were not limited, it would be much higher and
shorter at large ignition delays (near 90), and the transformer insulation
would have to withstand the higher voltage. The nonlinear resistors limit the
voltage at large delays, giving a pulse of magnitude and duration not too
different from that at minimum ignition angle and avoiding the need for
excessive insulation.*
The need for bridges in parallel has not yet developed in hv dc transmission.
Two bridges could be connected through an interphase transformer in the
manner shown in Figure 12 in Chapter 2 for parallel connection of two half
bridges. The two bridges would probably be fed through transformers con
nected for a 30 phase difference between one bridge and the other so as to
give 12-pulse operation of the converter.
Valves in series as opposed to bridges in series have the advantages of
fewer and larger transformers or transformer windings. However, they have
the following disadvantages:
No reduction in harmonics
More valve auxiliaries at more different insulation levels with respect to
ground
Uneven voltage distribution across the valves as a result of stray capacitances
* See Ref. 20, pp. 192-193.
MULTIBRIDGE CONVERTERS
115
With respect to misoperations of valves, which are discussed in Chapter 6,

series connections greatly decrease the probability of arcbacks and firethroughs but increase the probability of misfires.'
The unequal division of voltage across valves in series requires further
explanation. Each valve, because of the size of its tank and the number of its
auxiliaries at tank potential,has a much greater capacitance to ground than to
the adjacent valves in a series chain. The result, as is well known in the case
of strings of insulators, is to give more voltage across the elements remote
from ground than across those near ground. In the case of valves, the voltage
of concern is that during the nonconducting period. Capacitive voltage
dividers can be used to achieve equal voltage division, but they contribute
more energy to the parasitic high-frequency oscillations that occur at the
abrupt changes in voltage (see Figure 20). The addition of resistors to the
voltage dividers damps such oscillations but gives a small additional loss.
The predominant present practice is to use single valves with multiple
anodes and to connect bridges in series. Several manufacturers, however, are
trying to develop single-anode valves of high rated current.
3-7 MULTIBRIDGE CONVERTERS
In the present state of the art, two or more bridges in series on the dc side
needed for obtaining as high a direct voltage as required for
are usually
economical transmission.
The extension of converter theory to converters having B bridges in series
on the dc side requires consideration of (a) harmonics, (b) relations between
ac and dc quantities, and (c) the effect of mutual commutating reactance on
inverter operation.
Harmonics
As a rule, multibridge converters have an even number of bridges arranged

in pairs, one bridge of each pair being supplied with three-phase voltages
displaced by 30 from those supplied to the other bridge. As already illus
trated in Chapter 2, this arrangement gives a 12-pulse converter instead of a
6-pulse one, and certain harmonics (fifth, seventh, seventeenth, nineteenth, ...
on the ac side and sixth, eighteenth, ... on the dc side) theoretically are
eliminated and practically are greatly reduced. The two sets of three-phase
voltages with 30 phase displacement can be obtained from two banks of
transformers, one connected YY and the other, YA, or from one threewinding bank connected YYA.
116
3-7
Higher pulse numbers can be obtained in converters having more than two
bridges. For example, an 18-pulse 3-bridge converter and a 24-pulse 4-bridge
converter are possible. The transformer connections required, however, are
more complex than those for 12-pulse converters, and most hv dc engineers
think that it is more practical to build a 12-pulse converter provided with
additional filter capability than to reduce harmonics by use of a pulse number
higher than 12.
Harmonics are discussed further in Chapter 8.
Pr+JQr
I,
Pi +JQi
1\BrTr
cos a
IL1 = 0.780Sr/d
amperes
(111)
and
Vd=l35BTELLco$4>
volts
(112)
Still, of course,
Pa = Pd
watts
(113)
The equations in per-unit quantities eqs. (47), (48), (43) are unchanged
but must be interpreted with respect to new bases that are for the whole
converter instead of for one bridge.
The equivalent commutating resistance in ohms of a -bridge converter is
B times the value for one bridge, but the per-unit value of this resistance is the
same for B bridges as for one bridge and gives the same per-unit voltage drop
PJdIVdo
P-
L ~Rci
Kl
Rcr
\AAoyyV /*//>oT>
h>
cos y
<3
BiTi ' 1
Ini
VrfOr
cos a
XX
Vdr
Vdi
Vjoi
T
COS
X
Ac -
In hv dc multibridge converters, the bridges are in series on the dc side

and in parallel on the ac side. Therefore, for a given direct current and
alternating voltage, the alternating current and the direct voltage are multiplied
not only by the transformer ratio T but also by the number of bridges B, and
so is the power on both sides. Accordingly, Eqs. (39) and (42) for ordinary
units of measure become
-o
- Rectifier Modification of the Relations between AC and DC Quantities
117
X
-Line-
inverter-
Dc-
Ac-
Fig. 23. Equivalent circuit for steady state in two-terminal dc link.
neither leakage impedance nor exciting admittance: the effect of leakage

reactance in producing drop of direct voltage is accounted for by the equiva
lent commutating resistance. The dc line is represented by its resistance only,
capacitance and inductance being neglected.
Effect of Mutual Reactance on Inverter Operation

Overlap Less than 30. Consider a two-bridge 12-pulse converter. An equi
valent positive-sequence circuit of the ac network viewed from the two sets
of valve-side transformer terminals is the three-branched star in Figure 24. All
xs-x
<>Br. 1
Modified Equivalent Circuit
Now seems to be a good time to modify the equivalent circuits of the

rectifier and the inverter (Figures 13 and 18) so as to include the transformer
ratio and the number of bridges and to combine these circuits with that of the
dc line, thus forming an equivalent circuit of an entire two-terminal dc link
in the steady state. Figure 23 is such a circuit.
Subscripts r and isignify rectifier and inverter, respectively, and a and d
signify ac and dc. The transformer symbols represent ideal transformers with
Fig. 24. Equivalent star circuit of the ac circuit between two converter bridges (Br. 1,
Br. 2) and a common voltage source E.
its branches are predominantly inductive and are assumed to be entirely so.
Tr is an ideal transformer having a 30 phase shift. If the two bridges are fed
by separate transformer banks having equal voltage ratios and equal leakage
reactances, the two branches of the equivalent circuit adjacent to the bridges
are equal, and a single equivalent emf E is adequate. The third branch has a
reactance Xm due to the reactance of the ac system beyond the transformers.
It is the effect of this branch which we investigate.
118
3-7
In a 12-pulse converter supplied with balanced three-phase voltages, a

commutation begins every twelfth of a cycle (30), alternately in one bridge
and the other. The commutations cause line-to-line short circuits at the ac
terminals of the respective bridges. Because of the mutual reactance, a com
mutation in one bridge causes some distortion of the voltages of the other
bridge.
In Figure 25a equilateral triangle abc is a phasor diagram of the source
Bridge 2
119
represented by triangle a"b"c, whose vertex a" lies between a and a! and
vertex b" lies between b and b'. The ratio of vertical lengths in the phasor
diagram depends on the ratio of reactances thus:
bb
X
/
= ""
bb'
ab"
b'b"
ab
bb'
aa
x=/c
aa!
aa
aa
Bridge 1
(H4)
X. X,
= 1 -k
(115)
At the terminals of bridge 2 the voltages V would be the same as those at M

if both transformers were connected alike. The effect of the actual transformer
connections is to distort triangle abc in Figure 25b to a'b'c'. In both parts a
and b of the figure the effect of the commutation is to shrink all vertical
distances proportionally while horizontal distances are unchanged.
A commutation short-circuiting phase be of bridge 2 shrinks the hori
zontal components of all voltage vectors, as shown in parts c and d in Figure
25, while leaving vertical components unchanged. Commutations on other
phases shrink the components of the vectors parallel to other axes, there being
in all six such axes spaced 30 apart, as are successive commutations. The
next six commutations of one cycle use the same six axes again.
As a result of the mutual reactance, commutations in one bridge cause
dents in the voltage waves of the other bridge in addition to the dents caused
by its own commutations. The voltage wave of most concern is that of
voltage across a valve of an inverter (Figure 26). Particular attention is called
(d)
b' b
Fig. 25. Distortion of triangles of line-to-line voltages on bridges 1 and 2 due to com
mutations (a), (b) on phase ab of bridge 1, (c), (d) on phase be of bridge 2.
//
M / / //
( / ///
cot
emfs, which are assumed to be constant and balanced. When no commutation
is occurring, the alternating voltages VB1 across bridge 1 are the same as those
of the source because constant current produces no inductive drop. The
voltages VB2 across bridge 2 are likewise balanced, but because of the different
transformer connection they are advanced 30 with respect to the voltages of
bridge 1; they are shown as a, b, c in Figure 25b.
Now assume a commutation to occur in bridge 1, placing a short circuit on
phases a and b and collapsing the voltage ab to zero. The voltages on bridge 1
are then a'b'c in Figure 2 5a. At star point M of the equivalent circuit, the
voltages are deformed to a lesser extent than at the bridge 1 terminals, being
Fig. 26. Voltage across valve of 12-pulse two-bridge inverter, showing additional dents
D' and D" due to. common reactance, y 15, u = 15, /S = 30.
to the additional dent D' that occurs 30 sooner after the extinction of the
arc than does dent D, whose effect on commutation margin was discussed in
Section 3-5, page 110.

The effect of dent D' is shown in Figure 27. If /? < 30, as shown in Figure
21a, then D' does' not reduce the commutation margin; that is = y. If,
120
3-7
>
at
hlM
Fig. 27. Decrease of commutation margin angle
of inverter due to additional dent D'.
however, /? > 30, as in Figure 21b, then D' encroaches on the commutation
margin and < y. More specifically, = 30 u, and y = (3 u. The
difference y continues to increase with increasing /? until the vertical front
of D' all becomes negative, after which y remains constant until the whole
dent has passed below the horizontal axis.
The decrease of commutation margin due to dent D' is objectionable
because, by requiring greater extinction and ignition angles, it decreases the
power capability of the inverter, increases its demand for reactive power, and
increases the voltage jumps on the valves. Moreover, the effects of this dent
occur at values of /? normally reached at less than full load.
Overlap Greater than 30. If overlap angle u < 30, only one commutation
121
occurs at a time (with five valves conducting in two bridges), and there are
intervals between commutations (with four valves conducting) when the
voltages are undistorted. In the foregoing, such values of overlap were
assumed.
If u = 30, only one commutation occurs at a time, but as soon as one is
completed, another one begins. All the time five valves are conducting.
If 30 < u < 60, there are intervals in which a commutation in one bridge
begins before completion of the next earlier commutation in the other bridge.
This condition, in which six valves are conducting, may be called double
overlap. The distortion of voltages during double overlap is worse than that
during single overlap, although not so bad as during the double overlap be
tween the halves of the same bridge (with u > 60), which places a threephase short circuit on the valve-side transformer terminals, as described in
Section 3-3.
In the present case there are at the same time two line-to-line short circuits
on different transformer banks (or different sets of windings of the same
bank). These short circuits are on terminals at which the open-circuit voltages
differ in phase by 30. The triangle of line-to-line voltages of each bridge
collapses to a straight line, which is an altitude of the equilateral opencircuit triangle. If Xm = 0, hence k = 0 Eq. (114)the voltage line (vector)
of one bridge is 30 from that of the other. As k increases, these vectors
shorten, and their phase difference decreases. At k = 0.42 the magnitude has
decreased from 1.50isAr to 1.32ELN, and the phase difference, from 30 to!7.
At k = 1, the two line-to-line short circuits become equivalent to a threephase short circuit, and all voltages on the valve side of the transformers
vanish. The circuit analysis for finding these and other voltages is straight
forward but laborious and is not presented here.
Double overlap in a 12-pulse converter with mutual reactance between
bridges has several objectionable effects, including a more rapid drop in
direct voltage and the likelihood of commutation failure. Double overlap is
not likely to occur in normal operation, for the overlap at rated voltage and
current generally is less than 30, say, 20 to 25; however, it can occur at
small overloads or at small undervoltages.
For overlaps between 30 and 60 two circuit conditions occur alternately.
In one, a commutation is in progress in one bridge; and, in the other, com
mutations are in progress in both bridges, as just described. The number of
conducting valves in the two bridges is alternately five and six.
With u = 60, there are always commutations in both bridges at the same
time, and six valves are conducting.
With u between 60 and 90, there are alternately six and seven valves con
ducting. When seven are conducting, there is double overlap in one bridge
and single overlap in the other.
122
With u between 90 and 120, alternately seven and eight valves conduct.
When eight conduct, there are double overlaps in both bridges.
Complete analysis of all these conditions is lengthy and not of much
practical interest. Such analyses have been made by Shekhtman13 and
Freris.24
Elimination of Mutual Reactance
The objectionable effects of mutual reactance in a 12-pulse two-bridge
converter, especially the effect in decreasing the commutation margin of the
inverter, must be eliminated; and this implies the elimination or compensa
tion of the mutual reactance itself. There are two methods by which this can
be done.
Filters. The mentioned objectionable effects are due to the influence of
commutations occurring in one bridge on the wave shapes of the alternating
voltages appearing on the other bridge. The wave shapes are changed in six
short arcs per cycle, each lasting less than one-sixth of a cycle. If the differ
ences between voltage wave shapes of such converters with and without
mutual reactance were analyzed into a Fourier series, it would be found that
the differences consisted principally of certain harmonic frequencies.
Harmonic filters (Chapter 8) are necessary in most converters and are
usually installed on the network side of the transformers. The effect of the
filters is to make the ac bus voltage substantially sinusoidal. The commutating
inductance, which can be defined as the inductance between the valves and
the nearest point in the ac network where the voltages are substantially
sinusoidal, is then the leakage inductance of the converter transformers. If
separate transformer banks are provided for each bridge, there is no mutual
commutating reactance between bridges. The system reactance beyond the
bus is still common to the several bridges as regards fundamental-frequency
voltage drop. However, as currents of higher harmonic frequencies are
effectively short-circuited by the filters and do not enter the ac network
except in amounts that are negligible for the present discussion, commutation
in one bridge does not alter the voltage wave forms in other bridges.
The fundamental-frequency voltage drop in the ac network is calculated by
ordinary phasor methods, not by converter theory. Since the filter appears as
a shunt capacitance at fundamental frequency, the "drop" may actually be
a rise if the reactive current furnished by the filter exceeds that consumed by
the converter. The ideal no-load direct voltage Vd0 is proportional to the
alternating voltage of the bus. The drop (or rise) of direct terminal voltage
with respect to Vd0 is calculated by converter theory : Vd = Vd0 cos a RJd,
or Vd0 cos y - RcId . For a given alternating voltage at the converter, this
drop is independent of the power factor.
problems
123
Compensation of MutualReactance. Another method,8 used at the Gotland

inverter station, is to connect a set of center-tapped reactors as shown for
one phase in Figure 28. When commutation occurs in one bridge, the drop in
voltage in that bridge caused thereby, induces, through the mutual reactance
between halves of the reactor winding, a voltage rise in the other bridge that
compensates for the drop in Xm. For exact compensation, the mutual
reactance X'm between halves of the reactor should be equal to the system
reactance Xm . Since the self-reactances Xa are proportional to the squares of
their number of turns and the mutual reactance Xm to the product of the
number of turns of the halves, if the same flux links all turns, Xa s Xm ,
each being one-fourth of the reactance of the entire reactor. The net mutual
reactance vecomes Xm X'm, and the net self-reactance (commutating re
actance) becomes Xm + Xa + XT , which is greater by Xa than the selfreactance without the reactor. This increase may be offset by reducing the
leakage reactance XT of each converter transformer. Since the system re
actance Xm may vary, depending on the lines, transformers, and generators in
service, it is advisable to make X'm equal to the greatest value of Xm . With
smaller values of Xm , there is some excess of compensation, and the sign of the
additional voltage dents is reversed, which is not harmful. This condition,
which corresponds to a negative value of k, has been analyzed by Freris.24
Y Y
XT
-H
xm
xa
Valve
XL
groups
(bridges)
Reactor-""
XT
Y A
Fig. 28. Tapped reactor for compensating the coupling between valve groups due to
power-system reactance Xm. X, is self-reactance of each half of reactor winding and Xm,
mutual reactance between the halves; XT is leakage reactance of each transformer bank.
PROBLEMS
1. Draw curves of instantaneous voltage across a valve as functions of

wt for w .0 and a = 0 to 180 in increments of 30 (seven curves
altogether).
124
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
ANALYSIS of the bridge converter
Draw the wave shape of current in one phase of delta-connected trans

former secondary windings feeding a bridge rectifier circuit with negli
gible overlap, and calculate the effective value of the current.
What orders of harmonics are present in the direct voltage of a six-pulse
converter? In the alternating current?
Find the amplitude of the Ml harmonic of the direct voltage of a sixpulse converter with ignition angle a and no overlap. Express it as a
fraction of Vd0 .
Find the amplitude of the Mi harmonic of the alternating current of a
six-pulse converter with ignition angle a and no overlap. Express it in
terms of the direct current.
What is the rms value of all the harmonics combined, excluding the
fundamental, of the alternating current of a converter in terms of the
direct current if the latter is free from ripple and if the overlap angle is
very small ?
Draw curves of wave forms of valve current and voltage of a threephase two-way rectifier with a = 15, u = 60.
Draw curves of current through a valve and voltage across the same
valve as functions of cot for a = 0 and u = 30.
Do Problem 8 with a = 15 and u = 30.
Do Problem 8 with a = 30 and u 30.
Do Problem 8 with = 120 and u = 15.
Draw curves of one cycle of instantaneous line-to-neutral and line-toline voltages at the terminals of a Y-connected secondary winding of the
converter transformer feeding one bridge, with a = 30 and u = 10.
Same as Problem 16 except a = 15 and u = 30.

Same as Problem 16 except a 30 and u = 30.
Same as Problem 16 excpet a = 45 and u = 15.
Same as Problem 16 except a = 30 and u = 60.
21. Compute the rms value of the line-to-neutral voltage on the ac side of
the bridge (Figure 11c).
22. Draw the voltage and current wave forms of a six-phase, one-way
diametrical rectifier for a = 15, u - 30.
problems
125
23. Derive equations for the average direct voltage and current of a sixphase, one-way diametrical rectifier as functions of the ac source voltage,
commutating reactance, ignition delay angle, and extinction angle.
Confine to overlap less than 60.
24. Show that, in a three-phase bridge rectifier operating with no delay and
with 60 overlap, the direct current is half of the crest value of the lineto-line short-circuit current on the valve side of the transformer bank.
25. Find the maximum permissible commutating reactance, in per unit
based on transformer rating, to allow commutation to be completed in
an overlap angle of 60 at rated direct voltage and 1.5 times rated direct
current.
26. Find the ratio of the kVA rating of the transformer bank for a threephase, two-way converter to the rated dc power in kW if a = 15 and
X" = 0.15 pu based on the transformer's rating,
27. Prove that the two expressions for tan< in Eq. (54) are equivalent.
28. Draw curves as specified in Problem 30 but for a = 30, u = 75
(double overlap).
29. Plot pu direct current l'd versus pu direct voltage V'd from open circuit
to short circuit (a) for a = 0, (b) for a = 30, (c) for a = 60.
30. Draw curves of (a) instantaneous voltage of each dc pole of the bridge
with respect to transformer neutral, (b) the six valve currents, (c) one
transformer secondary current, Y connection, and (d) voltage across
one valve for a = 15, u = 60 (auto phase control).
3 1 Find the values of V'd,Id, and Pd of an uncontrolled rectifier at the point
of maximum power.
32. Plot a graph of reactive power versus active power on the ac side of an
inverter operating at a constant extinction angle y = 15. Use per-unit
values with base power Vd0Is2 . Terminate the curve at u = 60.
33. Plot y versus ft for constant = 15.

34. Draw curves as specified in Problem 30 except for y = 35, ft = 115
(inverter with double overlap).
35. Show that, if there are stray capacitances Cx between each pair of valves
and C2 from each valve to ground, the division of voltage between
valves is unequal. Derive expressions for the fraction of the total voltage
across each of two, three, and four valves in series, and compute the
numerical values for C2 = 3 Cx. Find numerical values of the capaci
tances of a voltage divider that equalizes the voltages across four valves
to within 5%.
126
36. In the current divider in Figure 22, find the voltage across the non
conducting anodes with respect to the common cathode (a) when the
whole current flows in one anode and (b) when it is equally divided
among two or (c) three anodes.
37. Draw curves of one cycle of instantaneous line-to-neutral voltage of one
phase and of instantaneous line-to-line voltage of one phase at the
terminals of a Y-connected secondary winding of the converter trans
former feeding one bridge of a two-bridge 12-pulse converter with
coupling coefficient k = 0.422 between bridges. Take a = 30 and
u 10.
38. In a certain ac power system under light-load conditions, the reactive
power produced by the shunt capacitance of the hv ac transmission
lines exceeds that consumed by the series inductances of the lines and
transformers and leads to undesirably high voltage at and near a
rectifier station. The situation is further aggravated when the rectifier
also is operating at light load, because then the reactive power consumed
by the rectifier is less than that produced by its ac harmonic filters. One
method proposed for improving these conditions is to make the rectifier
consume additional reactive power by controlling it to operate at larger
ignition delay angle than its normal value of 15 but subject to the
limitation that neither the peak inverse voltage of the valves nor the
greatest voltage jump across a valve should exceed their respective
values at rectifier operation at rated alternating voltage, rated direct
current and voltage, and normal ignition delay. Under these conditions
the overlap angle is 25. Compute and plot the additional reactive
power consumption of the converter (in per unit of rated power) as a
function of per-unit direct current.
BIBLIOGRAPHY
1. "The Current and Voltage Conditions in Large Rectifiers" (in German), by
W. Dallenbach and E. Gerecke, Archiv fur Elektrotechnik, Vol. 14, pp. 171-248, 1924.
2. Principles of Mercury-Arc Rectifiers and Their Circuits, by D. C. Prince and F. B.
Vogdes, McGraw-Hill, New York, 1927.
3. " The Current and Voltage Relations in the Graetz Three-Phase Rectifier Circuit " (in
German), by K. Maier, E.u.M., Vol. 53, pp. 577-581, Dec. 8, 1935.
4. "Three-Phase Rectifier Circuits," by A. J. Maslin, Electronics, Vol. 9, pp. 28-31,

December 1936. Comment by Dallenbach on p. 80 of Ref. 12.
5. " The Theory of the Control Problem of H.V.D.C. Transmission with Rectifiers and
Inverters in Bridge Circuit," by F. Buseman, Technical Report, reference Z/T74,
B.E. & A.I.R.A. London, Mar. 2, 1948.
BIBLIOGRAPHY
127
6. The Transmission of Power by High Voltage Direct Current (in German), by Karl
Baudisch, Springer, Berlin, 1950, pp. 116-122.
7. " Extended Regulation Curves for 6-Phase Double-Way and Double-Wye Rectifiers,"
by I. K. Dortort, A.l.E.E. Trans., Vol. 72, Part I, pp. 192-198, May 1953. Disc., pp.
198-202.
8. " The Operation of Several Phase-Displaced Inverters on the Same Receiving Net
work," by E. Uhlmann, Direct Current, Vol. 1, pp. 106-110, June 1953. Additional
dent in voltage across valve due to common reactance decreases the commutation
margin.
9. "Influence of A-C. Reactance on Voltage Regulation of 6-Phase Rectifiers," by
R. L. Witzke, J. V. Kresser, and J. K. Dillard, A.I.E.E. Trans., Vol. 72, Part I, pp.
244-252, July 1953. Disc., p. 252.
10. " The Relations of Currents and Voltages in the Rectifier Connected in the Graetz
Three-Phase Bridge" (in Italian), by Walter Dallenbach, VElettrotecnica, Vol. 44,
No. 3, pp. 133-143, March 1957.
11. Recommendations for Mercury-arc Converters, International Electrotechnical Com
mission, Geneva, Switzerland, Publication No. 84, 1st ed., 1957. Definitions, symbols.
12. "The Current and Voltage Conditions in the Graetz Three-Phase Rectifier Bridge
Circuit," by Walter Dallenbach, Direct Current, Vol. 4, pp. 72-80, December 1958.
13. "Operating Modes and External Characteristics of a Twelve-Pulse Cascade-Bridge
Converter Circuit" (in Russian), by M. G. Shekhtman, N.I.I.P.T., Vol. 5, pp. 23-63,
1960.
14. "Bridge Rectifier and Inverter Parameters," Chapter 3 of High Voltage Direct
Current Power Transmission, by Colin Adamson and N. G. Hingorani, Garraway,
London, 1960.
15. " Current and Voltage Conditions from No Load to Short Circuit in Three-Phase
Bridge Circuits," by F. Hoelters, Direct Current, Vol. 5, pp. 112-121, 132, March 1961
16. "The Universal Characteristic of the Three-Phase Bridge Converter," by L. L. Freris,
Direct Current, Vol. 6, pp. 198-201, October 1961.
17. " Reactance Drop in Mercury-arc Power Rectifiers," by O. E. Mainer, Direct Current
Vol. 7, pp. 182-184, July 1962.
18. Schaltungslehre der Stromrichtertechnik (Converter Circuit Theory, in German), by
Th. Wasserrab, Springer, 1962, 466 pp. Especially pp. 241-256.
19. "An Analysis of the Three-Phase Bridge Converter," by L. L. Freris, Direct. Current,
Vol. 8, pp. 6-11, 19, January 1963. Discussion by J. R. G. Schofield, p. 136, May.
20. High Voltage Direct Current Converters andSystems, edited by B. J. Cory, MacDonafd,
London, 1965, Chapter 4, "Analysis of Bridge Convertor Operation," by L. L. Freris.
Also part of Chapter 8.
21. "Analysis of a Hybrid Bridge Rectifier," by L. L. Freris, Direct Current, Vol. 11, pp.
22-33, February 1966. Bridge with three, controlled valves and three uncontrolled
diodes.
22. " Voltage Regulation in 3-Phase Bridge Rectifier Connection from No Load to Short
Circuit," by J. A. Budek and A. H. Marchant, Direct Current, Vol. 11, pp. 38-41,
February 1966. Experimental check.
23. " Series Connection of Mercury Arc Valves," by G. E. Gardner, J. H. Holliday, and
D. J. Johnson, I.E.E. Conference PublicationNo. 22, High Voltage D.C. Transmission,
Manchester, Sept. 19-23, 1966, Part 1, Paper No. 43, pp. 211-214.
I
j
'1
\
j
j
128
24. "Effects of Interaction among Groups in a Multi-group A.C.-D.C. Convertor," by

L. L. Freris, l.E.E. Proc., Vol. 114, No. 7, pp. 965-973, July 1967.
25. "Single-phase Bridge Converter with a New Control and a Corresponding Three
phase Converter: Part I, Operational Features," by N. G. Hingorani, Direct Current
(new series), Vol. 1, No. 1, pp. 25-30, April 1969; "Part II, Method of Control,"
by N. G. Hingorani and Philip Chadwick, ibid., pp. 31-35. Disc., No. 3, p. 124,
February 1970.
26. "The Performance of a Convertor with Anode Reactance," by L. L. Freris, I.E.E.E.
Trans, on P.A. & S., Vol. 88, pp. 1537-1544, October 1969.
Converter Charts
Two different charts are described, both of which show relations among direct
voltage Vd and current Id, the three converter angles a, <5, and u, and the ac
quantities, including active power P, reactive power Q, apparent power S,
displacement factor cos (j>, and alternating current Ia .
The first of these charts has rectangular coordinates x, y proportional to
direct current and voltage Id and Vd . The second one has rectangular coordin
ates proportional to active and reactive powers P and Q. Both charts are
based on the assumptions made in Chapter 3, and both use, for generality,
the first system of dimensionless or per-unit variables described there.
Rectifier and inverter operation are represented by different regions of the
charts.
4-1 CHART 1 WITH RECTANGULAR COORDINATES OF DIRECT
CURRENT AND VOLTAGE
The horizontalcoordinate x is chosen as the ratio of the direct current to the

crest value of symmetrical alternating current in a line-to-line short circuit:
2coLJd
It can range from zero to 2/ 3, its value during a short circuit on the dc
terminals. It cannot reverse.
The vertical coordinate y is the ratio of the direct voltage under general
conditions to its value with no load and no delay of ignition:
v_
v*
v30
"
n_
r'io 3,/3E
(2)
Note that both coordinates vary inversely as the alternating voltage

referred to the valve side of the transformers.
The upper half plane, with positive Id and Vd, represents rectification. The
lower half plane, with positive Id and negative Vd, represents inversion.
129
130
CONVERTER CHARTS
4-1
CHART
131
Overlap Angle Less than 60 (0 u 60)
+ 1.0
Loci of Constant a and 8. Coordinates a and y are related to the ignition
delay angle a and extinction angle 8, as shown in Eqs. (19) and (21) in Chapter
3; thus
cos 8
(3)
2y = cos a + cos <5
(4)
x cos cc
Vd
whence, by addition and subtraction,
+ 2y = 2 cos a
x + 2y = 2 cos 8
Rectification
+0.5
Rectification
!
0.5
1.0
Rearrangement gives
-0,5
y = cos a x
(5)
.y = cos <5
(6)
+4*
Equation (5) represents the locus of constant ignition angle a in the xy plane.
It is a straight line of slope , intercepting the y axis at cos a and the x axis
at 2 cos a. There is a family of such parallel lines, one for each value of a
(Figure 1). Similarly, Eq. (6) gives the locus of constant extinction angle 5 in
the same plane. It is a straight line of slope +, y intercept cos <5, and x
intercept 2 cos 8. Again, there is a family of such lines, all parallel (Figure
2). The values of a and 8 may range from 0 to 180. Since p = n a, the loci
of constant a are also loci of constant p, although the numerical values of a
and p differ from one another on the same locus. Similar remarks apply to y
and 5. For inverter operation, ft and y are commonly used instead of a and 8.
Inversion
nversion
-1.0
-1.0
Fig. 1. Loci of constant ignition

angle a in the /i, Vi plane for
" < 60.
Substitution of Eqs. (3), (4), (7), and (10) into Eqs. (8) and (9) yields
u
o =o
2y
2 cos - cos 2
2
, u
. X- sin
~ sin
x = 2
2
2
Loci of Constant u. The overlap angle u is given by

u=8 a
(7)
The simplest way to draw a locus of constant u is to note that it is a curve

passing through the points of intersection of constant a lines and constant 8
lines for which a and 8 differ by a constant angle (Figure 3). The range of u
on the part of the chart now under discussion is from 0 to 60. The locus of
u = 0 is the vertical axis. The loci for other values of u appear elliptical.
A proof that they are indeed ellipses follows.
We begin with the trigonometic identities:
+ a) cos -J()5 a)
cos 8 cos a = 2 sin (<3 + a) sin -],-(<5 a)
8 +a X
cos 5 + cos a = 2 cos \{8
Let
(8)
(9)
(10)
Fig. 2. Loci of constant extinction

angle S in the I'&, V& plane for
u < 60.
Solution for cos 2/2 and sin 2/2 gives

2
2
cos - =
and
cos u}2
2 sin uj2
sin - =
""
(11)
Substitution of these expressions into the identity
2
sur
x
+ cos2 - = 1
(12)
gives
r+/ >
-*
2 sin uj2J
\cosw/2
=1
(13)
132
CONVERTER CHARTS
4-1
CHART
133
Overlap Angle Greater than 60 (60 u 120)
For this range of overlap we have from Eqs. (72) and (88) in Chapter 3,
instead of Eqs. (3) and (4), the following:
Rectification
yjdx = cos (a 30) cos (8 + 30)
(16)
2 \
= \y = cos (a - 30) + cos (5 + 30)
(17)
V3
By introduction of the following new variables,
y]3x = x'
=/
we
a -30 = a'
5 + 30 = 8'
obtain equations having the form of Eqs. (3) and (4):
x' = cos a' cos 8'
2/
= cos a'
(18)
+ cos 5'
(19)
Let us also introduce, by analogy with Eq. (7),

u' = 5' - a' = (8 + 30) - (a - 30) = 8 - a + 60 = u + 60
-0,5
(20)
Derivations like those of Eqs. (5), (6), and (13) yield
/ = cos a' - \x'

/ = cos 8' + \x'
/ / Y
Y \Z7fi)
2 sin u'f2J + \cos u'f2
Inversion
(21)
(22)
=1
(23)
The new (primed) variables are now replaced by the old (unprimed) ones.
Equations (21) and (22) become
-1.0
Fig. 3. Loci of constant overlap angle in the I'a, Vi plane for u < 60.
3
yj3 cos (a 30) -x
y=
V 3 cos (8 + 30) + -x
This has the form

=i
(14)
which is the standard equation of an ellipse with center at the origin, hori
zontal half axis a, and vertical half axis b. In the present case,
a = 2 sin -
b ~ cos -
(15)
(24)
(25)
These represent straight lines, as before, but with different intercepts and
with slopes three times as great. Figure 4 shows the lines for a = 30 in both
ranges of u. The lines for a = 30 and 8 = 150 are shown in Figure 3. They,
with the elliptical arc for u = 60, form the boundaries of the region of
134
CONVERTER CHARTS
0.5
1.732
-\
0.4
y
0.3
0.433
s=
60
0.2
u = 120'
1.0 1.155
0.1
Fig. 4. Broken line for a 30 formed by segments of two straight lines having the
following properties:
--intercept
x-intercept
Range of u
0.9
Slope
1.0
1.155
(120)
0 < < 60
Vi 1.732
v'3/2 = 0.866
60 <u< 120
2/V3
V3 = 1.732
1.155
-1/2
-3/2
double overlap, shown in more detail in Figure 5. This region has a small
area and is characterized by high current and low voltage.
The loci of constant u in this region are found by replacing the primed
variables in Eq. (23) with the respective unprimed ones :
1 +[-7=
] =1
[12 sin~3x
(u + 60)/2J
LV 3 cos (u + 60)/2j
-
0.2
(26)
This has the form of Eq. (14) and represents an ellipse with semiaxes
2
. + 60
a = -7=. sm -
V3
and
w 60
,,
b = \j/-3 cos +-
2
(27)
-0.5
For u = 60, which is the boundary between two regions of the chart where
different equations are applicable, Eqs. (24), (25), and (27), for u 60, give
the same results as do Eqs. (5), (6), and (13) respectively, for u g 60.
Fig. 5. Enlargement of part CDE of chart in Figure 3, u > 60.
136
CONVERTER CHARTS
4-1
Loci of Constant AC Quantities (on the Network Side)
137
+ 1.0
Relations Between AC and DC Quantities. The relations between per-unit

ac and dc quantities are given by Eqs. (43), (47), and (48) in Chapter 3,
namely,
For power :
P'a = P'd = P ' = V'dl'd = xy
For current:
VaT'l'd = T'x
(29)
v'd
E'a cos < = =
(30)
'
For inphase voltage :
CHART
Rectification
(28)
actual turn's ratio

converter transformer, both numerator
nominal turns ratio
and denominator being ratios of valve-winding turns to network-winding
where T
of the
is ::-
turns.
By Eq. (41) in Chapter 3, however, the displacement factor is
cos <f>Vd = y
* = jyss'=707T'
(31)
Hence the per-unit alternating voltage on the network side of the transformer
is
(32)
It is independent of the direct current and voltage and, therefore, does not
appear on the chart.
Loci The following loci, all shown in the chart in Figure 6, are independent
of transformer ratio :
Approximate * Loci of Constant Displacement Factor (cos <j>) are horizontal

lines at ordinate
y = cos
(33)
They coincide with the loci of constant direct voltage V'd.
Approximate Loci of Constant Apparent Power S'
S'
EX=y.T'x = x = Ii
(34)
given by
(35)
and coincide with the loci of constant direct current l'd .
* The
Fig. 6. Loci of constant active power P', reactive power Q', apparent power S', and
displacement factor cos <f>.
Loci of Constant Active Power P'. These loci are rectangular hyperbolas
Hence the loci are vertical lines at abscissa
* = S"
Inversion
distinction between exact and approximate loci is discussed in connection with

Chart 2 in Section 4-2.
xy = P'
or
P'
(36)
Those in the first quadrant are for positive power (rectification), and those in
the fourth quadrant are for negative power (inversion). The locus for P ' = 0
consists of the y axis and the positive x axis.
138
4-2
CONVERTER CHARTS
Approximate Loci of Constant Reactive Power
q'
x = -y===
(37)
r = S' =
or
These curves are of the fourth degree and are unnamed. All values of
positive, except on the vertical axis, where Q' = 0.
The following loci depend on the transformer ratio :
Q' are
Approximate Loci of Constant Alternating Current I'a. These loci are vertical
lines at abscissa
X
(39)
=~
At nominal ratio ( T ' 1) they coincide with the loci of constant S' of the
same numerical value. For other values of T' they are shifted; the corre
sponding loci do not appear in the charts.
Complete Chart
Figures 3 and 6 could be superposed to form a single chart. The result,

however, might be confusing unless it were drawn in two or more colors.
Loci of Constant Converter Angles
The loci of constant ignition angle a, constant extinction angle <5, and
constant overlap angle u are best derived from the exact equations for perunit vector power in terms of these angles. From Appendix B the equation
for 0 u 60 is
Loci
(40)
Ordinate, per-unit reactive power :
(4i)
rb
Pb = Vd0Is2 =
9E
of Constant
+ 30) +j2u]
Overlap Angle u. By putting <5
/2a, and rearranging Eq. (44), we obtain
Together they form the complex power plane, P' + jQ'.
(42)
=a
(45)
+ u, factoring out
(46)
Let us first examine graphically the factor (1 /2m). As shown in Figure 8,

the vectors 1 and /2m are the two equal sides of an isosceles triangle, whose
remaining side has length 2 sin u and, considered as the vector difference of the
equal sides, has direction (90 m) referred to the positive real axis. Thus
1 2u
(44)
120, it is
P'+jQ'=j?,I+ l-0/2
4 -
Abscissa, per-unit active power :
where
~/2S + j2u)
P ' + JQ' = [/2(a - 30) - /2{&
The same per-unit system is used with this chart as with Chart 1. The
rectangular coordinates are the following :
y=Q' =
(43)
These loci of constant values of the foregoing coordinates are shown in

Figure 7. Those of constant (j> are labeled with values of the displacement
factor cos<.
A third set of coordinates, curvilinear ones, are the loci of constant
ignition, extinction, and overlap angles, shown in Figure 9 and discussed
below.
For 60
4-2 CHART 2 WITH RECTANGULAR COORDINATES OF ACTIVE

AND REACTIVE POWER
Angle <j), measured counterclockwise from the positive x axis representing the
phase angle by which the fundamental alternating current l[ lags behind the
alternating source voltage E.
P ' + JQ' = i(/2
x = P' =
139
Radius, per-unit apparent power :
//-r\ 2~|V2
Superposed is a set of polar coordinates:
Q'
Q =J(S ')2 - (P ')2 = V*2 - (xy)2 = Xy/l - y2

or
chart
sin m
u 90
=2 1-
(47)
With constant u, the only variable factor in Eq. (46) is /2a. With increasing a,
140
4-2
CONVERTER CHARTS
CHART
141
Rectification
inversion
(2m
sin u
2u
Reference
-12u
Fig. 8. To show that 1 /2k = 2 sin u/ u
90,
locus of constant u is the arc of a circle of radius (sin u)j2 and center at
0 +juj2 with central angle extending from u 90 to 270 u.
For u > 60 Eq. (45) is used instead of (44) and with a similar rearrange
ment. The loci of constant u are still circles, and their centers are still at
0 +ju/2. The radius of each is given by \ sin (u + 60), and the valid arc
extends from central angle u - 30 at a = 30 to 210 ~u at a = 150 u.
The loci for u 0 and u = 120 are single points, the former at the origin
and the latter at 0 +yVt/3. .
Loci of Constant Ignition and Extinction Angles, Let us consider first the
locus of a = 0. For a = 0, Eq. (46) becomes
-0.5
+0.5
Active power (per unit) P'

Fig. 7. Loci of constant apparent power and of constant displacement factor on Chart 2.
These are also approximate loci of constant direct current and of constant direct voltage,
respectively.
the vector represented by the second term of this equation rotates counter
clockwise through angle 2a, and its tip traces a circle of radius (sin u)J2. The
valid arc of this circle extends through 360 2u, from u 90 at a = 0 to
270 u at a = 180 u, and is symmetrical with respect to the imaginary
axis. The first term in Eq. (46) represents a vector locating the center of the
circle on the imaginary axis at distance w/2 rad above the origin. Thus, the
P'+jQ'=i+j~i/2u
(48)
This equation, as u increases, represents the curve generated by a point on the

circumference of a circle of radius rolling up the right-hand side of the
imaginary axis, that is, a cycloid. A cusp of the cycloid occurs at the origin
when u = 0. The next cusp would occur at 0 +jn[2 when u = 180. The only
valid arc for the three-phase bridge converter, however, corresponds to that
for u from 0 to 60.
The loci for other constant values of a are cycloids of the same size, shape,
and orientation as that for a = 0 but shifted both horizontally and vertically
so that the cusp that was at the origin for a = 0 follows a cycloid like that
for a = 0 but rotated a half turn around the origin. Every cycloid of the family
142
CONVERTER CHARTS
passes through the origin, which is the point corresponding to u = 0 on all of
them.
The most practical way of constructing these loci seems to be the following.
First draw the circles of constant u, and on them mark the point for a = 0 and
the equally spaced points for other values of a, using constarit increments of
a. The loci of constant a may then be drawn by connecting, by a smooth
curve, points for the same value of a but different values of u.
Each such point, representing certain values of a and u, also represents a
value of <5 = a + u that may be marked on each point. The loci of constant 5
are then drawn by connecting all points for the same value of 8. The pairs of
loci for a
and 8 = n (or y = at) are mirror images with respect to the
imaginary axis. Hence the loci of constant <5 are also arcs of cycloids.
4-2 CHART 2
This method of constructing the loci of constant a

and those of constant 8
is applicable to u > 60 as well as to u < 60.
All these loci are shown, in. Figure 9. A part of the
chart is drawn to a
larger scale in Figure 10.
0.4
0.3
cy
1.0
H,
Inversion
Rectification
0>
Q.
|0.2
O
"N
it
S 0,5
= 30'
o.i
20'
Active power (per unit) F
0.3
0.4
Fig. 10. Pan of Chan 2, enlarged.
Loci of Constant Direct Current and Voltage

These may be plotted from either approximate or exact
O (u = 0)
0.5
Fig. 9. Converter Chart 2 loci of constant converter angles in P', Q' plane.
+ 0.5
equations.
Loci Basedon Approximate Relations. Per-unit
direct current l'd is approxi
mately equal to per-unit fundamental apparent power
S and per-unit direct
voltage Vd is approximately equal to displacement
factor cos </>. Hence the
loci of constant direct current based on an approximate
relation, are, like the
loci of constant alternating current and
constant apparent power, circular
t44
4-2
CONVERTER CHARTS
arcs having radius l'd and their center at the origin. These loci can be drawn
with a compass. The loci of constant direct voltage, based on an approximate
relation, are straight lines through the origin, having slope tan (j>, where
(j) = cos-1 V'd. These lines can be drawn with a straightedge.
The two sets of loci together constitute a polar-coordinate grid (Figure 7).
Loci Based on Exact Relations. Loci of this form are related to those of
constant a and constant <5 by the following equations :
l'd = cos a cos 5

2V'd = cos a + cos 5
(49)
(50)
chart
whence
cos oc = Vd -j-
\l'd
(51)
cos 5 = V'd-\I'd
Ignition angle a
or extinction angle y
90>
Fig, 11. Converter Chart 2loci of constant direct voltage

P', Q' plane.
20
Vi and direct current I'd in
(52)
The exact loci of constant l'd and constant V'd can be constructed as follows :
Values of l'd and V'd are selected, and corresponding values of a and 6 are
calculated by Eqs. (51) and (52), and u by 8 a. Values of P ' + jQ' are then
calculated by Eq. (44) or (45), depending on whether u is less than or greater
than 60. Finally, lines are drawn connecting points of equal l'd and other
lines connecting points of equal V'd . The former are approximately, but not
exactly, concentric circles; the latter are approximately, but not exactly,
straight, radial lines. See Figure 11.
Rectification
Inversion
145
Overlap angle u, deg
Fig. 12. Error in reactive power calculated by the approximate method.
146
BIBLIOGRAPHY
CONVERTER CHARTS
147
Difference Between Approximate and Exact Loci. The point representing
a given pair (Id,Vd) in the P ' + jQ' plane according to the approximate
relations has the same abscissa (P' V'dl'd) as the point representing the same
pair according to exact relations. The approximate ordinate Q\ however,
given by
Qsfd-Jl-(V'df
(53)
is greater than the exact one by an amount that varies from 0 to nearly 8%
and is greater along the outer boundaries of the chart (a = 0" and 5 = 180)
than on the vertical axis. In the normal operating range, with a = 15 or
y = 16 and u between 0 and 25, the error in Q' does not exceed 3 per cent.
See Figure 12 for further information.
BIBLIOGRAPHY
1. "The Theory of the Control Problem of H.V.D.C. Transmission with Rectifiers and
Inverters in Bridge Circuit," by F. Busemann, Publication Z/T74, B.E.&A.I.R.A.,
Mar. 2, 1948, Figures 15, 19.
Hingorani, Garraway, London, 1960, Chapter 3.
3. "A Chart Showing the Relations between Electrical Quantities on the A-C and D-C
Sides of a Converter," by E. W. Kimbark, I.E.E.E. Trans on P.A.&S., Vol. 82, pp.
1050-1054, December 1963. Also in I.E.E.E. Publication S-155, D-C Transmission,
June 1963, pp. 34-43, and in Direct Current, Vol. 8, pp. 166-169, June 1963. Chart 2
with different scales than herein.
4. Schaltungslehre der Stromrichtertechnik (Converter Circuit Theory, in German), by
Th. Wasserab, Springer, Berlin/Gottingen/Heidelberg, 1962, pp. 337-340.
4-3 RELATIONS BETWEEN THE TWO CHARTS
Every point on Chart 1 represents specific values of all the variables

(I'd, Vd, a, 5, u, I'JT', P', Q', S', and cos0). Likewise on Chart 2. Further
more, for every point on Chart 1 there is a corresponding point on Chart 2
such that the set of values denoted by the point on Chart 1 is identical to the
set denoted by the point on Chart 2. Hence we can imagine that every point
on one chart could be moved to the position of the corresponding point on the
other chart by a distortion similar to the unequal stretching of a sheet of rubber.
The origin on Chart 2, however, does not transform to a single point on Chart
1 but, instead, becomes the whole vertical axis. The transformation is not
conformal.
PROBLEMS
1. By use of a converter chart, find the maximum power output of a bridge

rectifier operated at constant ac input voltage and transformer tap. Also
find the corresponding values of l'd, V'd, a, and w.
2. Derive the equation of the loci of constant u in rectangular coordinates
(P',Q13. Show that P' =(cos2 a cos2 <5)/2 =(sin2 <5 sin2 a)/2.
4. Design and draw a converter chart on a grid of equiangular triangular
coordinates representing the three converter angles a, 5, and u. Demarcate
the valid area. Draw the loci of constant direct current and voltage, J'd
and Vd.
:(
:1
1,
5-1
5
Control
\
I
This chapter discusses the control of dc transmission under normal con
ditions. Grid control of the converter valves, however,is used not only for the
control of normal transmission but also for clearing faults in the converters
and on the transmission line. These aspects of protection are discussed in
Chapters 6 and 7.
|
I
Critical Grid Voltage. The critical grid voltage is that which separates
grid voltages which permit ignition from those which block ignition. The
critical grid voltage varies oppositely from the anode voltage and is roughly
proportional to it (Figure 1), but the exact value varies somewhat with
temperature and vapor pressure in the valve. In addition, it depends on the
geometry of the valve.
|
Let us begin the discussion by recalling the conditions under which the
conduction of valves can be initiated and terminated or, in other words,
under which the arc can be ignited and extinguished.
In mercury-arc valves, the control grid can delay ignition but, after con
duction has begun, cannot stop it. The same limitation applies to the control
of thyristors by a gate signal. There are other kinds of valves, such as vacuum
(or "hard") electronic tubes and transistors, in which both ignition and
extinction can be controlled. Some devices that can conduct in either direction,
such as mechanical switches and Marx arc switches, also can control both
ignition and extinction. None of these devices, however, is commonly used
in high-voltage high-power converters; therefore, we confine our attention
to the kinds of values that can control ignition only and, in particular, to the
mercury-arc valve.
Such a valve begins to conduct if the following two conditions are satisfied:
(a) if the anode voltage, with respect to the cathode, is positive and (b) if the
control-grid voltage, also with respect to the cathode, is more positive than
the critical value. For brevity we express the second condition by merely
saying that the grid is positive.
The cessation of conduction of such a valve depends primarily on the
external circuit. The current must become zero through action of that
148
circuit. It would be better to say that, if the valve were not a valve but a plain
conductor, the current through it would reverse. Since, however, the valve is
a valve, the current through it cannot reverse but becomes and remains nil.
Instead, the voltage across the valve becomes negative. Next, this voltage
must remain negative long enough for the adequate deionization of the arc
(requiring 50 to 400 fis in a hv mercury-arc valve, or 1 to 8 of a 50- or 60-Hz
wave). After deionization, negative grid voltage blocks reignition, even if the
anode later becomes positive.
Practically, in a converter, the circuit condition that causes extinction is
the firing of the next valve of the set and the completion of commutation to
that valve.
As we saw in.Chapter 3, the need of adequate time for deionization is more
critical in an inverter than in a rectifier, for the anode remains negative for a
much shorter time after extinction (see Figure 20 in Chapter 3).
Form of Grid Pulse
Valve Characteristics
149
|
j
5-1 GRID CONTROL
GRID CONTROL
I.
I
I
|
I
Front of Pulse. For assuring ignition, the grid voltage should rise from a
value well below the minimum control value to one well above the maximum
critical value. For accurate timing of ignition (independent of variations of
critical voltage), the grid voltage should rise very rapidlyideally, instantly.
Fig. 1. Anode voltage
Va and corresponding critical grid voltage Vgc during a cycle.
150
5-1
CONTROL
Duration of Pulse. Because a valve, once ignited, normally remains con

ducting until completion of commutation to the next valve of the set, a short
grid pulse suffices theoretically. Two practical circumstances, however,
indicate the superiority of a long pulse. The first is the possibility of premature
arc extinction, known as quenching (see Section 6-8). If quenching occurs, the
grid pulse ought to be present to permit immediate reignition. The second
practical point is that the starting up of transmission should be facilitated.
This requires a continuous series circuit through two valves in each bridge
of both rectifier and inverter. The least number of valves in series per pole is
four in a six-pulse system or eight in a twelve-pulse system. Though all these
valves are initially nonconducting, they would, if in the steady state, be in
various phases of their conduction cycle; hence short pulses at the normal
instants of ignition would not be effective in starting transmission. Either
special short starting pulses or normal long pulses are required. The long
pulses meet the need both of starting transmission and of immediate reignition after arc quenching, and, therefore, long pulses are used in practice.
Each grid pulse should last until the beginning of the grid pulse of the next
valve in the commutating set but no longer. Duration at least that long assures
that at least one valve in each set can conduct so that, through the several sets
in series, there is always a complete path. If the pulse lasts much longer, it
may lead to failure of commutation in an inverter (see Section 6-5).
The grid pulses coincide with the pulses of anode current that would exist
in the same valves if there were no overlap. Under balanced conditions, the
length of each grid pulse is one-third cycle (120). Under unbalanced or
transient conditions, the grid pulses may all be longer or shorter than onethird cycle, or some may be longer and others shorter.
Ideal Shape. The ideal grid pulse then is rectangular, having a flat top and
vertical front and tail (Figure 2a).
cot
(a)
(b)
Fig. 2. Wave forms of grid pulses: (a) ideal; (b) practical.
GRID CONTROL
151
Practical Shape. It is difficult in practice and unnecessary to maintain a

flat top, and some droop may occur. It is sufficient that the top remain above
the critical grid voltage for the maximum pulse duration that may be re
quired (200 or 0.55 cycle should be adequate). Vertical fronts and tails are
impossible; however, the front should be made as steep as possible in the
normal range of critical voltage. In practice, the front lasts about 100 ps (2),
and the variation of firing time due to limited steepness is about 25 ps (0.5).
Some overshoot is permissible at the ends of the front and tail. A typical
pulse form is shown in Figure 2b.
Current and Voltage. Typical values are 1 or 2 A and several hundred volts.
Grid Bias and Blocking
A constant negative bias voltage of a few hundred volts is applied to the

control grids of all valves continuously. When a valve is supposed to con
duct, a positive pulse is added to this negative bias. The height of the grid
pulses is more than enough to overcome the bias.
If,,for clearing a fault in a converter bridge, it becomes necessary to block
all the main valves of the bridge, the transmission of positive pulses to all
their grids is interrupted, letting the negative bias remain on the grids.
Transmission of Grid Bias and Pulses to Grids

The bias and pulses are applied in series between cathode and grid. Diffi
culties arise in that the control signals and auxiliary power must originate at
or near ground potential, although most of the cathodes are at different high
direct voltages with respect to ground (see Figure 3).
Ac power is transmitted through insulating transformers to each valve or
each group of valves having a common cathode potential. This power is used
to supply grid bias through an auxiliary rectifier. It is fed also to various valve
auxiliaries (to be described in Volume 2, Chapter 14).
Grid pulses are usually formed and amplified at potentials near ground.
The amplified pulse is fed to the grid through a well-insulated pulse trans
former. Pulse transformers were developed for pulsing magnetrons in
microwave radar transmitters during World War II, and the principles of
their design have been thoroughly studied and published.47
Another method used for transmitting grid pulses is to pulse a beam of
light or infrared rays27 that impinges on photocells at cathode potential. The
received pulses are amplified there, the amplifiers being powered through the
aforementioned insulating transformers.
152
CONTROL
Vdoi
-=-
Vd()i cos a
Vdl
Vd2
5-3
POWER REVERSAL
Vd02 COS
-=-
153
Vd02
X
Rectifier -
-Line-
Inverter
Fig. 4. Equivalent circuit of dc transmission valid for average currents and voltages in
the steady state.
resistances are fixed, the current is proportional to the difference of the two
internal voltages and is controlled by controlling these voltages. The direct
voltage at any designated point of the line, as well as the current, can be
controlled by controlling the two internal voltages; for example, if the line is
uniform and if the two commutating resistances are equal, the voltage at the
midpoint of the line is the average of the internal voltages. The direct voltage
at any other point of the line is a weighted average of the internal voltages.
More generally, any two independent quantities, for example, power and
voltage, could be controlled by the two internal voltages.
Each internal voltage can be controlled by either of two different methods :
grid control or control of the alternating voltage. The internal voltage of the
rectifier is written in Figure 4 and Eq. (1) as Vd01 cos a. Grid control, delaying
the ignition angle a (time ocjco), reduces the internal voltage from the ideal
no-load voltage VdQ1 by the factor cos oc. (It will be remembered that the
voltage drop due to overlap is represented by the voltage across the com
mutating resistance Rcl.) The ideal no-load voltage Vd01 is directly propor
tional to the alternating voltagesee Eq. (5) in Chapter 3. The alternating
voltage in some exceptional cases could be controlled by generator excitation,
but it is usually controlled by tap changing on the converter transformers.
Grid control is rapid (1 to 10 ms), but tap changing is slow (5 to 6 sec per
step). Both these means of voltage control are applied cooperatively at each
terminal. Grid control is used initially for rapid action and is followed by
tap changing for restoring certain quantities (ignition angle in the rectifier or
voltage in the inverter) to their normal values.
Fig. 3. Voltages to ground of cathodes of valves in bipolar 12-pulse converter with

a = 0 and a = 0.
5-2 BASIC MEANS OF CONTROL
The current in a dc line operating in the steady state is given by Ohm's law
as the difference in its terminal voltages divided by its resistance. By incorporatingthe equivalent circuits ofthe converters ("given in Figures 13 and 18 in
Chapter 3) with that of the line, Ohm's law may be extended to the internal
voltages, embracing the commutation drops as shown in Figure 4. The current
Id in the line is then given by
r
_ PrfOl cos a Vd02 cos (/? or y)
(i)
In this equation cos fi is used in the numerator and + Rc2 in the denominator
if the inverter is operated with constant ignition angle (3; cos y and Rc2 are
used if the extinction angle y is constant. For present purposes, the former
mode of operation is assumed, because it is the ignition angle [3 that can be
directly controlled; the extinction angle y is controlled indirectly through
controlling (3 to values computed from the direct current Id, the commutating
voltage, and the desired extinction angle, as discussed in Section 5-10.
Direct current Id, then, depends on the voltage dropnumerator of
Eq. (1) divided by the total resistance (denominator). Since in practice the
5-3 POWER REVERSAL
It should be noted that in Eq. (1), Id and hence the difference of internal
voltages are always positive because of the unilateral conduction ofthe valves
(symbolized by the diodes in Figure 4). If it is desired to reverse the direction
154
5-4
CONTROL
of power transmission, the polarity of the direct voltages at both ends of the
line must be reversed while maintaining the sign of their algebraic difference.
Station 2 then becomes the rectifier and station 1 the inverter. The ter
minal voltage of the rectifier is always greater in absolute value than that
of the inverter, although it is lesser algebraically in the event of negative
voltage.
In some hv dc cable lines it is desired not to reverse polarity of conductor
voltages (see Volume 2, Chapter 11, for further discussion). It then becomes
necessary to have reversing switches between each converter and the line.
A method of reversing current, sometimes used in low-voltage dc industrial
practice, by having two converters in parallel, one for each direction of
current, is prohibitively expensive for hv dc transmission.
Further discussion of power reversal by voltage reversal will be found in
LIMITATIONS OP MANUAL CONTROL
155
-Inverter
-Rectifier
0.5
1.5
2.0
2.5
IdRdn
Normal
Low
5-4 LIMITATIONS OF MANUAL CONTROL
Alternating
voltage
Lower
The direct voltage at either end of a transmission line can vary in a sudden,
unexpected, and undesired manner because of short circuits or other dis
turbances on the ac systems' or because of faults in the converters. It is then
the responsibility of the rapid grid control to maintain or restore the desired
conditions on the dc line as far as possible with the available range of control.
This use of rapid, automatic control is more important than the effecting of
a rapid change of conditions on the dc line.
The need for rapid, automatic control will be shown by first supposing that
each converter has only manual control of the ignition angle. Viewing the
transmission line from its midpoint, draw a graph of the characteristic curve
of voltage at this point versus current for the converter and half line on each
side (Figure 5). Both characteristics are straight lines. For the left-hand
(rectifier) side, the intercept on the vertical axis is the internal voltage of
the rectifier, Vd0l cos a, and the slope is (Rc i + RJ2). For the right-hand
(inverter) side, the intercept is the internal voltage of the inverter, VdQ2 cos fi,
and the slope is +(Rc2 + RJ2). Since, at a given place and time, there can
be only one value of voltage and one value of current and since both these
values must conform to the characteristics of both converters, they must be
given by the coordinates of the intersection of the two characteristic lines. Let
the intercepts of the two lines be adjusted so that their intersection lies at
rated current and voltage (Jdn ,Vdn) at point N. The slopes of the lines have
been drawn on the assumption that at rated current the voltage drop due to
Fig. 5. Changes of direct current due to decreases of alternating voltage at either terminal
of the dc line when both converters are operating at constant ignition angles.
commutation is 8% of rated voltage and the line drop is

8
+ 9/2 = 12.5%.
9% of the
same.
Now suppose that the alternating voltage at the inverter drops by 12.5% of
Vd . Then the inverter characteristic drops to that represented by the first
parallel broken line below the solid line, and its intersection with the rectifier
characteristic moves to point A, corresponding to 1.5/d. A further 12.5%
voltage drop increases the current to 2Idn (point B). If the alternating voltage
at the rectifier drops by 12.5% of Vdn while that of the inverter is normal, the
rectifier characteristic drops to the first parallel broken line below the solid
line, the intersection moves to point C, and the current drops to 0.5Idn . An
additional 12.5% drop of alternating voltage lowers the characteristic to the
second parallel broken line and moves the intersection to point D: the current
becomes nil.
In this example, a dip in alternating voltage produces a percentage change
in direct current of four times the percentage change of voltage. Such large
fluctuations of current cannot be tolerated. The high overcurrents are
especially undesirable, for they may lead to arcbacks of the rectifier (Section
6-3), commutation failures of the inverter (Section 6-5), and damage to the
valves. The need for rapid control of current is indicated.
156
CONTROL
5-5 CONSTANT CURRENT VERSUS CONSTANT VOLTAGE
Two obvious alternative ways of operating a dc transmission system while

permitting control of transmitted power are the following:
1. Current held constant while voltage varies as the power does
2. Voltage held nearly constant while current varies as the power does
The same two methods could be used for ac transmission and distribution.
In the constant-current system, the various loads and one or more power
sources are connected in series ; a load or source is turned off by bypassing it
after bringing its emf to zero if it has one. This method has been widely used
for street lighting circuits, and was used on several of the earlier dc trans
mission projects. In the constant-voltage system, the various loads and sources
are connected in parallel; a load or source is taken out of service by opening
the respective branch. This system is in general use in ac transmission and
distribution systems and dc distribution systems.
In a hv dc transmission system having only two terminals (none to date
have had more), the distinction between series and parallel connection of the
two converters (rectifier and inverter) disappears. The choice between con
stant current and constant voltage must be made on other grounds. The chief
5-6.
DESIRED FEATURES OF CONTROL
157
voltage-dependent losses are always much less than the current-dependent

losses.
Thus, consideration of losses favors the constant-voltage system, but
limitation of current favors the constant-current system. In the past there was
some uncertainty as to the choice, with the earlier dc projects using constant
current. Now, however, it is possible by the use of automatic control to
combine the best features of each.
5-6 DESIRED FEATURES OF CONTROL
The following features are desirable:
1. Limitation of the maximum current so as to avoid damage to valves

and other current-carrying devices
2. Limitation of the fluctuation of current due to the fluctuation of
alternating voltage
3. Keeping the power factor as high as possible
4. Prevention of commutation failures of the inverter
5. Prevention of arcback of the rectifier valves
6. Inmultianode valves, providing a sufficient anode voltage before ignition
differences concern the following :
occurs
1. The limitation of variation of current caused by variation of alternating

voltage or by faults on the dc line or in the converters
2. The energy losses and efficiency
7. Keeping the voltage at the sending end of the line constant at its rated
value insofar as possible in order to minimize losses for a given power
8. Controlling the power delivered or, in some cases, the frequency at one
end.
It is well known that short-circuit currents on a constant-voltage ac system

can be very great although limited by system impedance, of which reactance
is the major component. On a constant-voltage dc system, fault currents
could conceivably be much greater, being limited only by circuit resistance.
On a constant-current system, however, short-circuit currents are ideally
limited to the value of the load current and in practice to about twice rated
current. Accidental open circuits could give rise to high voltages, but in
practice open-circuit faults are much rarer than short-circuit faults.
In a constant-voltage system the I2R loss in the conductors is proportional
to the square of the power transmitted ; in a constant-current system, this loss
always has its full-load value. If the system transmits less than rated power
some of the time, as is true in nearly all transmission systems, the daily or
annual energy loss is muchless in a constant-voltage system than in a constantcurrent system. The opposite is true of those losses depending on voltage,
such as corona loss and losses in insulation leakage. In practice, however, the
Many of these features have already been discussed, but others need some
discussion before the scheme actually used is described so that it may better
be appreciated.
There are four reasons for keeping the power factor high; two concerning
the converter itself and the other two concerning the ac system to which it is
connected. The first reason is to keep the rated power of the converters as
high as possible for given current and voltage ratings of valves and trans
formers. (The voltage rating of the transformer and the voltage across the
valves when not conducting are both proportional to Vd0 , but the power is
proportional to Vd .) The second reason is to reduce the stresses on the valves
and damping circuits. The third reason is to minimize the required current
rating and the copper losses in the ac lines to the converter. The fourth reason
is to minimize voltage drops at the ac terminals of the converter as its loading
increases. The last two reasons apply to any large ac loads.
The power factor can be raised by adding shunt capacitors, and if this is
158
CONTROL
5-7 ACTUAL CONTROL CHARACTERISTICS
done, the disadvantages become the cost of the capacitors and switching them
as the load on the converter varies.
The power factor of the converter itself is
cos <f>
Vdo
y[cos a + cos (a + )]
(2)
159
Inverter
(const 7)'
voltage
for a rectifier or
cos </> = i-[cos y
+ cos (y + w)]
(3)
for an inverter. For a reasonably high power factor, a or y should be made as

small as possible. In a rectifier, this is easy: we can make a = 0, for which
cos a = 1.(For multianode valves a should be about 5, giving cos a = 0.996.)*
In an inverter it is more difficult. In order to avoid a commutation failure,
commutation must be completed before the commutating voltage reverses
at y = 0; hence y must be greater than zero by some margin. We cannot
control y directly but instead must control the ignition advance angle
(3 = y + u in accordance with the value of overlap u expected in view of the
prevailing direct current and commutating voltage. Because of some in
accuracy in the computation of /? and a possibility of changes in direct current
and alternating voltage even after commutation has begun, sufficient com
mutation margin above the minimum angle required for deionization of the
mercury arc must be allowed. The easy and safe way would be to choose a
large value of /?. This way, however, lowers the power factor and raises the
stresses on the valves. A better way is to compute the firing angle required to
obtain a constant extinction angle y. How this is done is described later
(Section 5-10). It is called constant-extinction angle (C.E.A .) control.f
5-7 ACTUAL CONTROL CHARACTERISTICS
Individual Characteristics of Rectifier and Inverter

These are plotted in rectangular coordinates of direct current Id and direct
voltage Vd at some common point, say, at the sending end of the dc line. Let
the rectifier be equipped with a constant-current regulator. If this regulator
functions ideally, the rectifier characteristic is a vertical line (AB in Figure 6).
In practice it is not quite vertical but does have a high negative slope. Let the
* The reason is given in Section 3-6.
f ASEA uses the term consecutive control.
.Rectifier
(const Id)
Fig. 6. Characteristics of control schemefirst step of explanation.
inverter be equipped with a C.E.A. regulator. Then the inverter characteristic

is a line given by
K = Vd02 cos y + (Rt - Rcl)Id
(4)
On the assumption that the commutation resistance Rc2 is somewhat greater

than the line resistance Rl} this line (CD in Figure 6) has a small negative
slope.
Since at the common point there can be only one voltage and one current,
their values must be given by the coordinates of the intersection of the
rectifier and inverter characteristics (E).
Both characteristic lines can be shifted. The rectifier characteristic can be
shifted horizontally by adjusting the current command(order or setting), which
is one of the inputs to the current regulator, the other being the measured
current. If the measured current is less than the current command, the
regulator advances the firing time (decreases the delay angle a), thus raising the
rectifier internal voltage in proportion to cos a and thus raising Id see
Eq. (1). If the opposite is true, it increases a, thus decreasing cos a and Id. The
inverter characteristic can be raised or lowered by means of the tap changer
on the transformer at the inverter station, which varies the alternating voltage
on the valve side. As soon as the tap changer is moved each step, the C.E.A.
regulator very quickly restores the desired value of y. The internal direct
voltage at the inverter is changed in proportion to the alternating voltage,
since cos y is constant, and this tends to change the direct current, which,
however, is quickly restored to the set value by the current regulator at the
160
5-7
CONTROL
rectifier station. The dc reactors at both ends of the line tend to prevent rapid
changes of current, thus easing the duty of the current regulator.
It may be said with fair accuracy that the rectifier controls the direct current
and that the inverter controls the direct voltage. The statement would be
exact only if the rectifier characteristic were truly vertical and if the inverter
characteristic were truly horizontal. Otherwise, each control affects both
current and voltage, although it affects one of them but slightly and the other
greatly.
If the inverter voltage is raised, as described, the rectifier voltage must be
raised by an equal amount in order to keep the current constant. This can be
done quickly only by the current regulator, which increases cos a by electronic
control. This can continue only until a = 0 and cos a 1. The valve cannot be
ignited earlier, and so it is spared the vain effort of trying to increase cos a
by making a negative. The rectifier voltage can be increased further only by
changing taps on the rectifier transformer. If the inverter voltage is being
increased by its tap changer, the rectifier voltage can be increased just as
rapidly by its, assuming that both tap changers are of similar design. The
rapid electronic current regulator need only be able to raise the rectifier
voltage by as much as, or a little more than, it is raised by one step of the
inverter tap changer.
In practice, the rectifier tap changer is automatically controlled so as to
bring a into the range between 10 and 20. These values represent a com
promise between (a) keeping the power factor high, which requires small a,
and (b) having a margin for quick increases in rectifier voltage, which
requires great a.
In the foregoing we have considered the behavior of the control system for
slow changes of voltage. Equally important is its behavior for rapid changes
of voltage, especially for rapid decreases of voltage due to short circuits on the
ac systems, to electromechanical oscillations ("swings") between ac genera
tors, or collapse of the voltage of one valve group (bridge). These changes
may occur at either end or both ends simultaneously, usually in unequal
degrees.
Consider first a reduction of inverter voltage. The inverter characteristic is
shifted downward (from CD to FG in Figure 6). The new operating point is
H. The line then operates at reduced voltage but at substantially the same
current as before; the power is reduced in proportion to the voltage. The dip
in voltage may be of short duration, in which case operating conditions are
restored to substantially the initial ones. If the low alternating voltage is
maintained, the inverter tap changer raises the direct voltage until either it
becomes normal or the limit of the tap-changer range is reached.
Now consider a decrease of alternating voltage at the rectifier. If cos a were
constant, the direct voltage at the rectifier would decrease proportionally.
ACTUAL CONTROL CHARACTERISTICS
161
The rapid current regulator, however, trying to maintain constant current,

raises the direct rectifier voltage either to its initialvalue or until a 0 (or 5).
If the voltage dip is more than a few percent, the minimum a limit is reached
first. The rectifier characteristic really consists of two line segments : one of
constant minimum ignition angle a0 and one of constant current, as shown in
Figure 7 by ABH. The inverter characteristic is assumed to be CD, as before.
Rectifier
(a = an)
Normal voltage
Inverter
/(r=7)
Reducedj
voltage
Inverter
Rectifier
(const/,)
(const Ia)
Fig. 7. Actual characteristics of control scheme.
Now a big dip in rectifier voltage shifts the rectifier characteristic down to
A'B'H, which does not intersect the inverter characteristic. Consequently,
the current and power drop to zero after a short delay due to the dc reactors.
In order to avoid such a.great change of current and power caused by a
moderate dip in alternating voltage (exaggerated in Figure 7), the inverter is
also equipped with a fast current regulator, but it is set at a lower current than
the rectifier's regulator. The inverter characteristic is now DFG, consisting
of two segments, one of C.E.A.(y = y) as before, and one of constant current.
It intersects the new rectifier characteristic at L. It may now be said that the
inverter is controlling the direct current, and the rectifier, the direct voltage
an interchange of functions from those pertaining to normal voltage con
ditions. (More generally, the station having the lower value of VdQ1 cos a0
RclId or Vd02 cos y (Rc2 Rt)(Id AId) controls the voltage and the other
station the current.)
The difference between the current command of the rectifier and that of the
inverter is called the current margin and is denoted by A Id. It is generally 15%
of the rated current, although it could be made smaller. It must be great
162
5-7
CONTROL
enough so that the two steep constant-current lines do not cross each other
in spite of errors of current measurement.
During a large dip in rectifier voltage, the power is reduced not only in
proportion to the voltage but also because of the reduction in current A Id.
Thus a voltage dip at the rectifier end reduces the power more than does an
equal dip at the inverter end. This, however, is much better than having the
power suddenly become zero.
Moreover, if the dip in power is objectionable, a supplementary current
control can be added, which, whenever the current command exceeds the
measured current by approximately the magnitude of the current margin,
automatically increases the current command first to the rectifier and a little
later to the inverter by the amount of the current margin, and thus restores
the current to its correct value in a few tenths of a second. Later, if the
measured current exceeds the new current command by a similar amount, that
command is reduced by that amount first at the inverter and a little later at
the rectifier.
Under the abnormal condition just considered (low rectifier voltage with
current controlled by the inverter), the rectifier current regulator sees that the
current is too low and tries to raise it by raising the rectifier voltage by de
creasing the ignition delay. It is unable to do so, however, either because the
delay is already zero or because the minimum a control overrides the current
control. The inverter, in order to control the current, is operating at a greater
extinction angle than the minimum specified value. This occurs because the
current regulator fires the inverter valves before the C.E.A. control has a
chance to do so.
Under normal voltage conditions or low inverter voltage, the inverter
current regulator sees that the current is too high according to its own setting.
It tries to lower the current by raising the inverter voltage. To do so, it must
decrease the extinction angle y by decreasing the ignition advance angle /?.
It cannot do so because the C.E.A. regulator is already igniting the valves
before the current regulator would do so.
Suppose that, when a change of current is desired, the new current com
mands are to be set manually in response, perhaps, to telephoned orders. The
setting at one station must be changed before that at the other station. If the
current is to be increased, the current setting is raised first at the rectifier and
second at the inverter. But if the current is to be decreased, the current setting
should be lowered first at the inverter and second at the rectifier. In either
case, the current margin is first increased and later decreased to its normal
value. In this way there is no danger of accidentally changing the sign of the
current margin and thus suddenly reversing the power. If the change of
current is to be made automatically, the order of events is the same as
described, but the changes can be accomplished more quickly.
ACTUAL CONTROL CHARACTERISTICS
163
Combined Characteristic of Rectifier and Inverter

In many dc transmission links each converter must function sometimes as
a rectifier and at other times as an inverter. At times both converters are
called on to work as inverters in order to deenergize the line rapidly. There

fore each converter is given a combined characteristic, as shown in Figure 8,
Converter !
Converter 2
Vd
/
Converter I
/'Converter 2
If
U
1
1I
if1
C.EA
I I C.C. |
l
>1
>1
1
i
c.c.
1'
Hn
JL
J 1
C.E.A.
Converter 1
Converter 2
(b)
Fig. 8. Control characteristics permitting reversal of power flow: C.I.A., constant

ignition angle; C.C., constant current; C.E.A., constant extinction angle.
consisting of three linear portions: C.I.A., C.C., and C.E.A.

With the characteristics shown by solid lines, power is transmitted from
converter 1 to converter 2. If the characteristics are changed to those shown
by the broken lines, the direction of transmission is reversed by the reversal
of direct voltage with no change in direct current. Both stations are given the
same current command, but, at the station designated as inverter, a signal
164
5.-9
CONTROL
representing the current margin is subtracted from that current command,

giving a smaller net current command. When it is desired to reverse the
direction of power, the margin signal must be transferred to the station that
becomes the inverter station.
During the reversal of power and voltage the shunt capacitance of the line
must be first discharged and then recharged with the opposite polarity. This
process implies a greater current at the end of the line initially the inverter
than at the end initially the rectifier. The difference of terminal currents can
not exceed the current margin. Hence the shortest time of voltage reversal is
1-
A/d
seconds
(5)
where C is the line capacitance, A Vd the algebraic change of direct voltage,

and Ald the current margin.
The current margin signal corresponds to the horizontal separation A Id of
the constant-current characteristics of the two converters along the hori
zontal axis or between the corners Px and P2 in Figure 8. Because of the slope
of these C.C. characteristics, the actual separation between them varies with
Vd ,being least at the normal working point. The margin signal must be great
enough to maintain a positive margin there in spite of errors in the current
measurement and regulation. Operation at the intersection of the two steep
C.C. characteristics, with both current regulators operating, would be erratic.
The necessity for the slope of the C.C. characteristics is explained in
Section 5-11.
5-8 CONSTANT- MINIMUM-IGNITION-ANGLE CONTROL
The next step in our study of the control of a converter is to examine in

more detail how each of the three straight-line segments of the combined
characteristic (Figure 8) can be obtained.
1. Constant minimum ignition angle

2. Constant current
3. Constant minimum extinction angle
If the constant minimum ignition delay angle a0 is to be zero, no special

provision need be made for it because zero is inherently the minimum
possible delay.
If, however, the use of multianode valves requires a greater minimum
delay for example, a0 = 5 controlis required. The following method could
be used. The voltage across each valve is measured, and if it is less than a
165
specified valuefor example, 3 Vm sin 5the constant-current control is

prevented from igniting the valve. Since the purpose of the delay is to ensure
a certain voltage across the valve before igniting it, the method is logical,
although it allows some variation in a to opposite changes in magnitude Vm
of alternating voltage.
In practice, the voltages across the valves would not be used, but rather the
secondary voltages of a control transformer. In order to meet the requirements
of other control circuits, these voltages must be sinusoidal (free of notches
caused by commutation). The primary windings of the control transformer
must be connected to the network side of the main converter transformer in
order to obtain sinusoidal control voltages, and both transformers must be
similarly connectedfor example, in YAso that the control voltages are in
phase with the commutating voltages.
-J
TC
CONSTANT-CURRENT CONTROL
5-9 CONSTANT-CURRENT CONTROL
The second segment of the converter characteristic (Figure 8) is one of

constant current.
Constant-current control involves the following:

1. Measurement of the direct current Id
2. Comparison of Id with the set value Ids (also called reference value,
current order, or current command)
3. Amplification of the difference Ida Id, called the error
4. Application of the output signal of the amplifier to a phase-shift circuit
that alters the ignition angle a of the valves inthe proper direction for reducing
the error
If the measured current in a rectifier is less than the set current, a must be
decreased in order to increase cos a and thus raise the internal voltage of the
rectifier Vd0 cos a. The difference between the internal voltages of the rectifier
and the inverter is thereby increased, and the direct current is increased pro
portionally see Eq. (1). If the measured current exceeds the set current, a
must be increased instead of decreased, and all the quantities mentioned
above are changed in the opposite sens,e.
In the inverter, if the measured current is too low, the internal voltage must
be decreased instead of being increased as in the rectifier in order to increase
the difference of internal voltages. This refers, however, to the absolute value
of the inverter voltage. If we consider the inverter voltage to be negative,
which is usual if the same converter sometimes rectifies and at other times
inverts, the algebraic value of inverter voltage must be increased, as in a recti
fier; and to accomplish this, a must be decreased, as in a rectifier. The curve
166
5-10
CONTROL
CONSTAN T EXTINCTION-ANGLE CONTROL
of cos a versus a in the range from 0 to it is monotonic (Figure 9) ; anywhere

in this range, a decrease of a increases the algebraic internal voltage Vd0 cos
a. This means that the same constant-current controller can be used on a
given converter without change of connections during both rectification and in
version. (In practice, however, the same current setting is transmitted to both
terminals of a dc line, and the current margin is subtracted from the current
setting of the inverter; that is, the error signal for the inverter's current
regulator is s = Ids AId Id.) This prevents the current regulators at both
terminals from functioning simultaneously.
Con.
PS
Dc line
Tr;>
Fig. 10. Schematic diagram of constant-current regulator:.Ids, current command signal;

A Id, current margin signal; Id> line current and signal; R1} input resistor; R2, feedback
resistor; C, feedback capacitor; A, high-gain amplifier; PS, phase-shift circuit; Con.,
converter; Vd, direct voltage of converter; Ld, dc reactor; Tr, dc current transformer;
2, summer; e, error signal.
+1
The slope of the "constant-current" segment in the Vd, Id plane in the

steady state may be found by putting into Eq. (7) s = 0, v = Vd + <RcId, and
s = Ids
Id and then taking the derivative
dVj = -(K+Rc)
dL
-l
Fig. 9. Curve of cos a versus a.
The current regulator (Figure 10) is a simple kind of feedback amplifier

characterized by a gain and a time constant. Its differential equation is
v
where
167
= Kt
+ T
dt
(6)
v = instantaneous value of Vd0 cos a

T = R2C = time constant
K = gain of amplifier and phase-shift circuit
s = error signal
Ts
+1
Usually K>RC. In order to obtain a true constant-current characteristic,

represented by a vertical line, the gain K of this kind of regulator would have
to be infinite. As shown in Section 5-11, however, the gain cannot be made
too great without producing instability.
Only one current regulator per pole per terminal is required : it can control
all the valves in the several bridges. In bipolar schemes, the current regulators
of the two poles are normally given equal settings, so that the neutral current
is small if both poles are in operation. Where ground-return current is
objectionable, the neutral current can put an additional signal into the regu
lators in the proper sense to better equalize the currents of the two poles.
5-10 CONSTANT-EXTINCTION-ANGLE CONTROL
The corresponding transfer function is

v
(8)
(7)
where s is the variable of the Laplace transform. In practice, the transfer

function may be more complicated than that of Eq. (7).
The third segment of the combined control characteristic is one of constant

extinction angle (C.E.A.). The necessity for C.E.A. control of inverters has
been shown already, but it remains to be explained how such control can be
accomplished.
Each inverter valve must be ignited at such a time that extinction occurs
at a later time, which, however, must be earlier by an adequate margin than
168
CONTROL
5-10
the time when the commutation voltage reverses. Furthermore, after conduc
tion ceases, the voltage across the outgoing valve must be negative and must
remain so long enough for adequate deionization of the arc path. Let both
time t and the corresponding time angle cot be measured from the instant when
the commutation voltage of the valve in question first turns positive. The
instantaneous value of the commutation voltage is then
eba=/3Emsmmt
(9)
Commutation can begin after cot = 0 and must be completed before cot = n.
Let it be required that, under normal conditions, commutation be completed
at cot = n y, where y is the normal extinction advance angle. For accom
plishing this, commutation must begin at an ignition angle cot = a = n P,
which depends on the commutation voltage (crest value 3 Em), the direct
current Id to be commutated, the commutation inductance Lc , and the
desired yn . For the present, all four of these are assumed to remain constant
during a particular commutation, but the voltage and current can vary from
one commutation to the next.
The relation among the five quantities (a, y, Em, Id, and Lc) was derived
in Chapter 3. It depends on the fact that the time integral of the commutation
and the required firing angle ft is given by
- V3 Em cos cott
1,2
ebadt = y/3Em
Jn
-v/s E-(cos
=CO
C0/2
>f3Em
V3S,m sin a>t
(12)
(13)
cos ut
90 lag
circuit
Polyphase
1_
rectifier
En
-Id
/""N
dld
dt
Differentiator
Pot.
-V3E,m COS yn
O
cos yn
Comparator
-N
-<o
dt
Summer
and trigger
-2XM
JL
2XC
(10)
COS COfi)
volts
The required ignition time tx can be found by a real-time analog computer

functioning in accordance with Eq. (12) and shown in Figure 11. Each term
i-
*j3 Em cos yn - 2 XcId
cos 0 = cos yn - I'd
Pot.
-coscof-!'2
,
.
f2sin
cot dt = *j3Em&
Jn
Ja
or
voltage is equal to the change of magnetic flux linkages produced. The latter
is 2LcId. The former is
169
CONSTANT EXTINCTION-ANGLE CONTROL
valve
Pulse
forming
circuit
coq = a = ignition angle
where
cot2 = 7i
Fig. 11. Schematic circuit of analog computer for C.E.A. control. Pot., potentiometer.
The signal dhldt is explained in connection with Figure 15.
yn = normal extinction angle
Then the integral becomes
J3 E[cos (n
co
y) cos cotj
/s E
-( cos yn
=-
cos cor,) volt-seconds
CO
Equating the two quantities, we get
of the equation is represented on a greatly reduced scale. One input signal is

Id in the main converter circuit.
This can be realized by using the voltage across a shunt through which Jd or
a known fraction of Id passes. Another input voltage J3 Em cos y is a
direct voltage that is proportional both to the desired cos yn (a constant) and
to the crest value of the commutation voltage. This signal may be derived
from a direct voltage obtained by full-wave rectification of the single-phase
commutation voltage for the valve in question. These two inputs are summed
algebraically, and the sum is compared with the instantaneous value of the
third input, which is an alternating voltage with a crest value proportional to
a voltage proportional to the direct current
-v/3 m(cos y + cos coq) = 2coLcId ~2XJd
volts
(11)
170
5-10 CONSTANT-EXTINCTrON-ANGLE CONTROL
CONTROL
that of the commutation voltage and a phase 90 behind that voltage. When
the two quantities become equal, a pulse is generated that initiates the grid
pulse to ignite the valve in question. The quantities involved in the analog
comparison are shown in Figure 12.
The wave of commutation voltage is obtained from one phase of a bank
of control transformers (potential transformers) connected to the ac bus. The
control voltage should be proportional to the open-circuit ac voltage on the
valve side of the main converter transformers, but it cannot be obtained from
the valve side because of the objectionable notches in the voltage wave caused
by commutations. The ac bus voltage has a good wave form if adequate ac
harmonic filters are provided. It does not remain proportional to the valveside open-circuit voltage, however, when the tap changers are operated.
A solution to this dilemma, if the taps are on the winding on the network side,
is to have the control transformers connected to a fixed tap on that winding.19
171
R<
fi,
fisut
El
JR
fiut
-o
Eu
-o-i
Eb
+ V3m
(b)
Fig. 13. Phase-shift circuits.
-jR =
jwC
-vVA
- V3 Em cos ait
Fig. 12. Wave forms of voltages in C.E.A. analog computer.
The negative cosine voltage wave required for C.E.A. control can be derived
from the sine wave of commutation voltage in either of two ways. One way is
to use an RC or RL phase-shift circuit adjusted for an output lagging the
input by 90 (Figure 13). The other way is to integrate the commutation
voltage.19,36 The analog integrator (Figure 14) yields the negative of the
time integral of the input voltage.
For the correct operation of the C.E.A. control under unbalanced alternat
ing voltages, such as may occur during short circuits on the ac system, each
valve should have its own analog computer. The negative cosine wave is
derived from the sine wave of commutation voltage for the valve in question
fin
fi,
Fig. 14. Integrator; A is a high-gain amplifier.
taken from the secondary winding of the control transformer. The proper
phases are shown in the following table:
Outgoing valve
(to be extinguished)
Commutation voltage
Incoming valve
(to be ignited)
ebo
3
eca
6c b
6 ab
ebc
172
CONTROL
5-10 CONSTANT-EXTINCTION-ANGLE CONTROL
173
It would be easy to obtain the negative cosine waves from the polyphase
supply, but with unbalanced voltages these waves would be incorrect in both
magnitude and phase.
The signal y/3 Em cos y should likewise be taken from a rectifier con
nected to the phase listed above. From the two waves derived from the same
phase, a three-phase voltage can be obtained for feeding a six-pulse rectifier,
whose output can be better filtered than that of a single-phase two-pulse
rectifier. This signal from one phase can be used for the computers of the two
valves of a pair.
The signal proportional to 7dcan be used for all the valves of the converter.
Up to now we have assumed that Em and 7d were constant during a single
commutation, although they might vary between successive commutations
of the same valve. Now let us briefly consider the effect of their varying while
a commutation is in progress. Increased direct current requires more voltagetime integral for commutation, and decreased commutation voltage provides
less voltage-time integral between the same angular limits. Obviously, after
commutation has begun, its time of beginning cannot be altered by anything
that might happen afterward. Increased direct current, or decreased com
mutation voltage, or both together, require more time (and time angle) for
completion of commutation, and if enough time is not available for completion
and deionization before the sign of the commutation voltage reverses, the
direct current that began to be shifted from one valve to another is shifted
back to the valve it came from. This result is called a commutation failure and
is discussed further in Section 6-5.
Decreased alternating voltage, caused, perhaps, by a short circuit on the
ac network, lowers the direct voltage of the inverter proportionally. Con
sequently the direct current through the inverter increases at a rate deter
mined principally by the inductance of the dc reactor. The chief purpose of
this reactor is to decrease the probability of commutation failure of the
inverter caused by a disturbance occurring too late for making the necessary
correction in firing angle. For further discussion see Section 7-2, page 236.
Some commutation failures can be prevented by the introduction of a
supplementary input signal proportional to the rate of change of direct
current, dlfdt. This can be obtained in several ways, two of which are shown
in Figure 15. The method thus modified, however, is still ineffective unless the
increase of current occurs long enough before the time when commutation
otherwise would begin. This control should be arranged so that, although
increasing current advances the firing angle, decreasing current does not
retard it. The diodes in Figure 15 accomplish this.
The usual value of y is 16. Only 1 to 8 of this is required for deioniza
tion, the larger angles applying to the higher-power valves.30 The remainder
is margin for variations in direct current and in commutating voltage occur
ring too late to permit adequate correction of /?.
W\r
eR = Rid
Ux
-T
eL~L
P
Id
dld
dt
(a)
"H
eout
(b)
Fig. 15. Circuits for obtaining a voltage signal proportional to dljdt: (a) RL
circuit(b) RC circuit.
The foregoing discussion has neglected the effect, described in Section 3-5,
page 110, of the dent in the wave of voltage across the outgoing valve due to
the subsequent commutation. At large values of l'd, which is proportional to
Id/Em in Eq. (12), this dent decreases the commutation margin for a given
extinction angle. Hence, in order to prevent commutation failures at high
direct currents or low alternating voltages or both, the extinction advance
angle y must be greater than its normal value y , and the ignition advance
angle must be made accordingly greater. Curves of required (3 and y as
functions of l'd for constant commutation margin angle ( = 15 are shown in
Figure 16.
The C.E.A. control system should be modified so that at high values of it
Id
functions to increase y and to attain approximately constant commutation
margin .
It may be noted that the C.E.A. control as usually performed and as
described above is not a feedback control system as is the constant-current
control. A method of control based on the measurement of the extinction
angle, comparison of the measured angle with the desired angle, and altera
tion of the ignition angle in the direction to decrease the error has been
proposed.34 It has the obvious shortcoming that, after sudden changes of
circuit conditions, the correction is made too late. The standard method
described is superior in that it can prevent most commutation failures that
would occur without it. Another proposed methodis to superpose a correction
based on measurement upon the standard method.19
Isochronous Control. In both C.C. and C.E.A. control it is likely that,

because of unbalances or harmonic distortion of the ac voltages used for
commutation, the six valves of a bridge are not ignited at exactly equal
intervals of time. Then the result, explained in Section 8-3, is that the converter
generates uncharacteristic harmonics that are undesirable. The uncharacteris
tic harmonics may be almost completely suppressed by the adoption of a
174
5-11
CONTROL
STABILITY OF CONTROL
175
One method of increasing the damping of oscillations is to add damper

circuits either across the dc reactor (Figure 17a) or across the end of the dc
line (Figure 17Z>). Such circuits are discussed in Section 7-5.
The control system for normal operation of the dc line, comprising
constant-current control of the rectifier and C.E.A. control of the inverter, if
properly designed and adjusted, can provide suitable damping of line oscilla
tions and can do so more economically than the above-mentioned damper
circuits. If, however, the controls are improperly designed or adjusted, they
can cause oscillations to grow. This condition is called instability.* '
135
120
105
Dc reactor
Line
Line
45'
15'
L3rd a-L*4thJ
2nd
1st range
0.5
1.0
Per-unit direct current Id
/3
Fig. 16. Ignition advance angle and extinction advance angle y versus per-unit direct
current I'd in an inverter operated at constant commutation margin angle 15. Broken
line shows j3 for y 15.
modified form of control37,38 in which the valves are ignited at equal time
intervals and the ignition angles of all valves are retarded or advanced
equally so as to obtain the desired current, in C.C. control, or the desired
minimum extinction angle in the valve for which that angle is least, in C.E.A.
control.
5-11 STABILITY OF CONTROL
The dc line, together with the dc reactors at each end, constitutes a lightly
damped system that can be set into oscillation by various disturbances, such
as line-to-ground short circuits, converter faults, and improperly controlled
energization of the line. The natural frequencies of these oscillations usually
lie in the range of 10 to 100 Hz. Overvoltages on the line insulation may
result from improper energization or from converter faults that impress a
component of power-frequency alternating voltage on the dc line. It is
desirable that the oscillations be rapidly damped so as to limit the overshoot
of voltage to a moderate value.
(b)
Fig. 17. Circuits for damping line oscillations: (a) in parallel with dc reactor; (b) in
shunt with dc line.
Following is an approximate analysis of the stability of a dc link com

prising the line, the dc reactors, the rectifier under constant-current control,
and the inverter under C.E.A. control.4
It is assumed at the outset that both of these controls are instantaneous.
Each converter may then be represented by a direct emf in series with a re
sistance. The rectifier on C.C. control displays a high positive resistance
K+ Rcl, which depends principally on the gain of the C.C. regulator. The
inverter on C.E.A. control displays a low negative resistance, Rc2. See
Figure 18.
The transmission line may be represented, to a first approximation, by its
equivalent or nominal T circuit. This circuit is shown in Figure 19a together
with circuits representing the dc reactors and the converters. All these
circuits are then combined to form a single T as in Figure 196.
* The term used here as in the theory of feedback control systems has a somewhat different
meaning from that of the terms "power system stability" or "synchronous stability."
5- 11
176
STABILITY OF CONTROL
177
CONTROL
The characteristic equation of the circuit is found by setting the determinant

of the impedances equal to zero :
Vd
1 Rectifier (C.C. control)
Zi Zm
-Zm z2
\
Inverter
=0
(16)
or
(C.E.A. control)
zLz2-zl =0
(17)
Substitution of the values of the three impedances gives

Id
Rectifier
K + i?cl
Ldi
Dc reactor
Dc transmission line
Dc reactor
Rdi
Li/2
Li/Z
Rll2
Ril2
Rdi
R2 + L2s + )
Rx +Lxs -+
Fig. 18. Characteristics of converters in Vd , Id plane.

Inverter
-Ra
=0
After performing the indicated multiplications, canceling the terms (1/Cs)z,

and clearing of fractions, we have
LxL2Cs3 + (LXR2 + L2Rx)Cs2 + (RXR2C + Lx -1- Ls +

Ei\
il
il
(14b)
L2RxCs2 + (./?ii?2l + Ls + (J?j -f- R2) = 0
Ri
It
(b)
Fig. 19. Equivalent circuit of dc link for analysis of stability of control (a) before com
bination of line with terminal equipment; (6) after combination.
The equations of this circuit are
Z1I1 ZJ2 = Ex
ZmIx + Z2I2 = e2
y+
where
Zx
= Rx
+ Lxs + Cs
1
(15a)
Z2 = R2 + L2s + Cs
(15b)
z = -L
Cs
(15c)
s = a +jco = complex frequency
(19)
(14a)
(a)
Li
+ R2) = 0
This is a cubic equation in s, and its roots are values of s that characterize the
transient phenomena. A cubic equation with real coefficients has three roots,
at least one of which is real. The other two may be real or a conjugate com
plex pair. From our knowledge of the nature of the circuit, we know that
the latter alternative is more probable. The corresponding solution in the
time domain, found from the inverse Laplace transform of Eq. (19), would
have a damped direct current and a damped oscillatory current. If R2 is.
negative, one or both of these terms may have negative damping; it is more
likely that only the oscillatory term has it.
If numerical values of the coefficients were given, the cubic polynomial
could be factored into a linear one and a quadratic. Factoring of the algebraic
cubic is very cumbersome. So let us make another simplifying assumption.
Since we know that ordinarily
> Lxs, let us put Lx = 0. Equation (19) then
simplifies to the following quadratic:
Li
Ri
(18)
=0
+ i442
l2rx c
(20)
which corresponds to the simplified circuit in Figure 20. Since Rx p R2 , we

R2
-Wv
Ri
(15d)
L2 RXC
z.C
Fig. 20. Simplification of Figure 196 by putting Ex
E2 0 and Li = 0.
178
CONTROL
. 5-12
may further simplify Equation (20) by putting R2 0 in the last term, which
then becomes 1/L2 C. Equation (20) has the form
or
s2 + 2os + (co2 + a2) = 0
(21)
s2 + 2(o)ns + co2 = 0
(22)
The undamped natural frequency is
(23)
and the damping coefficient is
i/r2
(24)
It is the average of two reciprocal time constants, the first of which, containing
R2, may be negative. For positive damping a > 0; hence, if R2 < 0,
(Figure 19a) and since usually K > Rcl

there
is
maximum
value of K for which the system
a
<0,
if
+ Rdl + RJ2, R2
is stable. In practice it is desirable to have considerable positive damping,
but somewhat less than critical damping (( = 1). Assuming that = 0.7,
which corresponds to 5% overshoot in response to a step input, we get
Since
Rt = K+ Rcl + Rdi + RJ2
R -.
L% _
\aJl2c- cr2
(26)
The foregoing theory gives pessimistic results if dc resistances are used,

because some of the power losses are higher at the frequency of hunting than
at zero frequency.
Temporary operation with the inverter on constant-current control and the
rectifier on constant a control is more stable than that described above,
because then both converters have positive-resistance characteristics.
Another form of instability or hunting can occur in the rectifier, even if the
inverter operates at constant ignition angle and thus has a positive resistance,
because of the intermittent operation of the current control.5 This control
operates only at the firing of each valve ; that is, six times per cycle in a threephase bridge and not in between. With high gain and short time constant,
this control may overcorrect deviations of the current from the set value. If
the current is too low, the firing of the next valve is advanced. If the correc
tion is made too rapidly, the current becomes too high before the next valve
TAP-CHANGER CONTROL
179
is fired; accordingly, the firing1 of that valve is delayed. This correction is

likewise made rapidly, resulting in the current's becoming too low before the
next valve fires, and the firing of that valve is advanced. This process, con
tinued, produces an oscillation of firing angle and direct current and voltage
at a frequency three times that of the ac line and also produces even harmonics
in the alternating current (see Section 8-3, page 318).
Such oscillations are prevented by making the time constant of the current
regulator long, or the gain small, so that overcorrection cannot occur, or else
by isochronous control.
It is desirable that the steady-state gain of the current regulator be made so
high that the constant-current portions of the control characteristics shown
in Figure 8 are almost vertical, that is, so that the current changes by only
a few percent for a voltage change equal to rated voltage. If this gain is high,
a long time constant (about 5 sec) must be used to make the control stable.
For small changes of current, corrections are still made rapidly, because the
speed of correction depends on the ratio KjT, gain divided by time constant
Eq. (6).
5-12 TAP-CHANGER CONTROL
The rectifier tap changer is controlled so that if a becomes less than 10, it
raises the direct voltage by raising the transformer ratio T; and if a becomes
greater than 20, it lowers the direct voltage by lowering T. As a result, a lies
between 10 and 20 in the steady state unless the tap changer has reached one
of its limits.
The inverter tap changer is controlled so that the direct voltage at some
designated point of the line, preferably the sending end, is close to its rated
value. More precisely stated, the tap changer raises the direct voltage by
raising T if that voltage is lower than the desired value by more than a specified
amount and lowers the direct voltage by lowering T if the voltage is higher
than the desired value by a specified amount.
The direct voltage at the point of the fine where it is to be held constant is
computed by adding to the direct voltage measured at the inverter station the
RtId drop in the dc line, obtained to scale by sending the current output of the
dc current transformer through a resistor representing the line resistance to
scale. This arrangement may be called a line-drop compensator. It is subject
to some error attributable to changes in the fine resistance with ambient
temperature and current.
In order to avoid hunting of the tap changer, the dead band must be wider
than the size of the tap step. Otherwise, the voltage might be too high on one
180
5-13
CONTROL
tap step and too low on the adjacent step, with the result that the control
would force the tap changer to oscillate between the two adjacent steps.
In the rectifier, the voltage band between a = 10 and 20 is Kd0(cos 10
- cos 20) = 0.045 Vd0 , or 4. 5 % of Vd0 An allowance of 2 in measure
ment of the angles reduces the voltage band to KJ0(cos 12 - cos 18) =
0.027Fd0 = 2.7% of Vd0 . Hence the tap step must be less than 2.7%, say, 1.3
to 2.0%. In other words, the dead band is approximately 1.3 to 2 tap steps.
The inverter transformers have the same tap step as the rectifier trans
former, especially because most dc lines are intended to transmit in either
direction. With due allowance for errors in voltage measurement and in linedrop compensation, the voltage band should be at least +1.3 to 2.0%,
depending on the size of tap step.
5-13 POWER CONTROL AND CURRENT LIMITS
A common requirement is that a dc line deliver a scheduled value of

power. Although constant-current control comes near to meeting this re
quirement if the direct voltage remains reasonably constant, more accurate
power control can be obtained by automatically varying the direct current so
as to compensate for changes in direct voltage caused by variation of line
resistance, variation of alternating voltage beyond that which can be cor
rected by tap changers or grid control, and by outages of one or more valve
181
POWER CONTROL AND CURRENT LIMITS
It is necessary to put several limits on the current control (Figure 21).
1. Maximum Current Limit. The purpose of this limit, which might be

from 1.0 to 1.2 times rated current, is to avoid thermal damage, especially to
the valves.
2. Minimum Current Limit. The purpose of this limit, which is usually 0.1
of rated current, is to avoid operation with very short overlap, which in
creases the probability of arcbacks (see Section 7-4, page 266) or with dis
continuous current, which leads to overvoltages (see Section 7-2, page 240
and 7-8, page 286).
3. Voltage- Dependent Current Limit. The same power can be obtained
with a high voltage and low current or with a low voltage and high current
(points A and B or C and Din Figure 21). Operation at the latter condition is
objectionable not only because of higher power losses and higher reactivepower demand but also because of voltage instability. Such operation is
likely to occur on starting transmission, while the direct voltage is being
raised slowly. It is prevented by the use of the limit represented by the sloping
line OE in Figure 21. The slope of this line must exceed the (negative) slope
of the constant a lines, which depends on both the commutating resistance of
the converter and the impedance of the ac system feeding the converter.
Upper voltage limit-
groups.
Power control could be accomplished in either of two ways. The first and
most obvious way would be to measure the power on either the ac or dc
circuit, to compare it with a power command, and to use the difference (error
signal) to advance or retard the ignition of the valves. In order to retain the
advantages of current control, however, including its speed and the inclusion
of current limits (to be discussed later), the power control does not replace
current control but supplements it. The power command and the measured
direct voltage are fed into an analog divider, whose output signal represents
the direct current Id = PjVd. This becomes the current command, which is
one of the input signals to the current control.
The voltage measurements at both terminals must be corrected by current
compensation to a common point so that equal power commands at both
terminals result in equal current commands. The current margin signal is still
subtracted from the current command of the inverter.
Increase and decrease of power command, if made manually, should be
executed at the two stations in the order already described for current
control.
0.25 Vd,
Loci o
Ida
/max
Fig. 21. Voltage and current limits (bounding shaded area).
Id
*
182
5-15 MULT1TERMINAL LINES
CONTROL
Sf=kAf
183
gf -kAf
5-14 FREQUENCY CONTROL
The frequency of an ac power system is controlled by controlling the power

input to the prime movers of selected generating units. If the frequency is
low, the input to these units is temporarily increased, and the excess of
mechanical power input over electrical power output goes into increase of
kinetic energy: hence, increase of speed and frequency. If the frequency is
high, the input to the prime movers is temporarily decreased, and the kinetic
energy is drawn on to supply the required electrical output.
If the power rating of a dc line is comparable with or greater than the rating
of the running generators in the ac system to which the line is connected, the
line terminal should share in the frequency regulation or even perform it
unaided.
Two examples of frequency control by dc line terminals come to mind: one
in which the frequency of the receiving system is controlled and another in
which the frequency of the sending system is controlled.
The island of Gotland is supplied with electric power principally by the
hv dc cable from the Swedish mainland. The small steam-electric generating
plant on the island is shut down much of the time, leaving the synchronous
condenser at the inverter station as the main synchronous machine. The
frequency of the ac system on the island is controlled by regulating the power
delivered to it from the dc line.
The Sardinian scheme ordinarily exports power to the Italian mainland.
The frequency of the Sardinian ac system is regulated partly or wholly by the
amount of power on the dc line.
In both of these examples, the frequency of the smaller (island) ac system
is controlled by the dc line, and this is true for either direction of power flow,
although the prevailing directions are as already mentioned.
A frequency discriminator circuit is used at the terminal where frequency
is to be controlled. Its output signal is proportional to the deviation of fre
quency from its normal value and is used as an error signal to advance or
retard the ignition of the valves. Low frequency .should advance the ignition
of the valves of the converter in the controlled network if it is inverting, thus
increasing the received power, but should retard the ignition if the converter
is rectifying, thus decreasing the sent power. Hence a reversing switch is
required on the output of the discriminator. Figure 22 shows the error signal
in each case.
The same error signal from the discriminator is transmitted to the distant
terminal. At the rectifier it is normally used without change, but at the inverter
a margin signal is subtracted from it, so that, although the frequency is
correct, it appears too low regardless of whether the frequency-controlled
(a)
(b)
Fig. 22. Frequency error signal
versus frequency /: (a) frequency-controlled network
exporting; (b) frequency-controlled network importing.
network is importing or exporting power. In either case, the inverter advances

its ignition angle in a vain attempt to raise the frequency and, as a result,
reaches the minimum allowable extinction angle and thus determines the
direct voltage. In the event, however, of low voltage at the rectifier, limiting
the line voltage, the inverter does take over the frequency control, although
there is then a small frequency error due to the margin signal. The frequency
control is analogous to the current control previously described : the converter
with lower voltage determines the direct voltage of the line, and the one with
higher voltage (as explained in Section 5-7) determines the frequency.
Both converters are provided with current limits, which can over-ride the
frequency error signal, and with stabilizing signals proportional to rate of
change of current (dljdt).
5-15 MULTITERMINAL LINES
All the dc lines built heretofore have been essentially two-terminal lines,
having one rectifier station and one inverter station. Nevertheless, multiterminal lines have been proposed, and control schemes have been developed
for them. One of these is a constant-voltage scheme with all converters connected in parallel. Another is a constant-current scheme with all converters
connected in series.
There are schemes for example, Kingsnorth in which a bipolar line
forms a trunk from which monopolar branches go to different terminals
(Figure 23).49 Each pole is controlled independently, and the neutral con
ductor of the trunk carries the difference of the currents of the two poles.
The two truly multiterminal schemes to be described may be either bipolar
or monopolar. If bipolar, each pole may serve some different stations. The
schemes are described as monopolar.
?
j j
;
!I
j
wm
184
5-15
CONTROL
Inverter 1
Rectifier 1
Rectifier 1
Vdi
\/
Rectifier 2
Inverter 3
Rectifier 2
h
?dl
Fig. 23. Bipolar line with monopolar branches.
Inverter 4
Vds
Vd2
Inverter 2
185
MULTITERMINAL LINES
"T
Vat
[*-13
vd
AI
I
Id2
-Ir
H'
h
-hi
-Id2
(a)
Vdi
Constant-voltage Parallel Scheme48
A line having four terminals two rectifiers and two invertersis assumed
for illustration (Figure 24). The control scheme is merely an extension of that
already described for a constant-voltage two-terminal line. Current com
mands Ix , I2, h > ft > having an algebraic sum equal to the current margin
A I, are sent from a control station to the respective converter stations.
Rectifier currents are taken as positive; inverter currents, as negative. The
station having the lowest ceiling voltage, that is, the lowest Vd0 cos oc0 or
cos y , controls the line voltage. This station is normally one of the inverters
operating at constant extinction angle. The other three stations operate on
constant current and at voltages lower than their respective ceiling voltages.
The current at the voltage-controlling station is the algebraic difference of
the current command and the current margin. (In the inverter this is an
arithmetical sum.)
vd
t*
tfe
I-h
y
L 1
Fig. 24. Four-terminal dc line with converters in parallel.
These normal relationships are illustrated in Figure 25a. Here a separate

diagram is drawn for each of the four stations instead of combining the
characteristics on one diagram as was done for the two-terminal line (Figure
7). Another difference is that inverter currents are shown as negative. As
before, a common direct voltage Vd is the link between the diagrams for the
1
*-
h p*
i
hi
Vdi
VdZ
vd2
Id.2
i4 -
-JdS
-hi
(b)
Fig. 25. Control diagrams of four-terminal dc lines. Line voltage is controlled (a) by
inverter station 3; (b) by rectifier station 2.
several stations,* but the relationship among the four currentsthat their
sum is zerois less apparent.
If the ceiling direct voltage at one of the stations operating on constant
current fell (say, because of a drop in alternating voltage or because of a
converter fault), that station would become the voltage-controlling station
and its current would decrease by the current margin. Figure 25b shows the
current-voltage relationships with rectifier station 2 controlling the voltage.
The current of the former voltage-controlling inverter station 3 rises (alge
braically) by the current margin. The currents of stations 1 and 4 are un
changed.
If any of the converters is to be changed from a rectifier to an inverter or
vice versa, its current must be brought to zero, a reversing switch must be
operated there, and, finally, the current must be increased to the desired
value. The voltage polarity of the line could not be reversed unless it were
desired to reverse the sign of power at every terminal at the same time.
* A single-valued voltage-current characteristic for each station would be obtained strictly

only if the four stations were connected in star to a common point and not as shown in
Figure 24.
186
5- 16
CONTROL
Constant-current Series Scheme

One disadvantage of this scheme (full I2R loss even at light load) was
already mentioned. This disadvantage can be somewhat alleviated by de
creasing the current instead of the voltages in periods when all stations are at
light load. There is another disadvantage, however, arising from characteris
tics of the converters and from the possibility that sometimes some converters
may be lightly loaded and others heavily loaded. The lightly loaded converters
operate at low voltage (well below rated voltage), which can be obtained
either by a low transformer tap or by operating at a large control angle
(a or y). The first method requires a wide tap range, which increases the cost
of the converter transformer. The second method subjects the valves to large
voltage jumps, which increase the probability of converter faults, increase the
losses in the valve damping circuits (Section 7-3), and increase the consump
tion of reactive power.
A method of overcoming both of these objectionable features has been
proposed.50 Each converter has an even number of bridges. Assume, for
simplicity, two bridges. The voltage of each bridge can be adjusted from a
maximum of one polarity to a maximum of the opposite polarity. If both
bridges have maximum voltages of the same polarity, the converter is working
at full load either rectifying or inverting. If both have equal voltages of
opposite polarity, the converter voltage and power throughput are zero. The
problem is to cover the whole range of voltage across the pair of bridges
while keeping the voltages of the individual bridges as high (in absolute
value) as possible. Figure 26 shows a manner in which the net voltage of two
bridges can be varied continuously over the entire range while each bridge
works in the range of 50 to 100% of its full voltage except at points at which
1.0
MEASURING DEVICES
187
the voltages of both bridges are swiftly changed without changing the net
voltage of the pair.
Additional ideas on control of multiterminal lines are set forth in Refs. 51
and 54.
5-16 MEASURING DEVICES
In the control and protection of hv dc converters and lines, devices are re

quired for obtaining signals, usually direct voltages, proportional to the
current and voltage of the main dc circuits and to various other quantities,
such as power, frequency deviation, and ignition delay. Some of the more
important such devices are discussed in this section.
Direct Current
Measurements of direct current are required for the current regulator and
one of the inputs to the constant-extinction angle control of the inverter. For
both these purposes very fast response is necessary.
A direct current proportional to but less than the direct current in the main
circuit can be obtained, in a circuit isolated from the main circuit, by means
of a dc current transformer .58-64 This device consists of two saturable reactors
(transductors), each having two windings, an ac supply usually fed through
a small step-down transformer, and a rectifier employing several small diodes.
The reactor cores have a sharp saturation point and a very low mmf Fm
for saturation (Figure 27). One winding (the primary one) carries the direct
Magnetic flux <j>
Bridge 1
-Fm
0,5
MMF
s, -0.5
Bridge 2
-1.0
-2.0
-Fs
-1.5
1.0
0.5
-0.5
0
-1.0
Net voltage of converter/rated voltage per bridge
-F
1.5
Fig. 26. Control of direct voltage of two-bridge converter on series system.
2.0
Fig. 27. Idealized magnetization curve of a high-permeability core for a dc current
transformer.
188
5-16
CONTROL
current ld to be measured. If this current is great, the primary winding is not

really a winding but merely a straight bar or cable passing through the center
of the window of the core an arrangement commonly used also in ac
current transformers for high primary currents. Such a primary is considered
to have one turn. The secondary winding has many turns, and the current
ratio is inversely as the turns ratio. The secondary windings are excited by
alternating current.
In the common form of dc current transformer, the two secondary windings
are connected in series opposition and are in series with the ac source and a
full-wave rectifier (Figure 28). The primary current Id exerts a mmf Fp NpId
Primary bar
MEASURING DEVICES
189
The flux< in each reactor core is the ordinate of the respective point on the
magnetization curve. If both reactors are saturated, the fluxes in both cores
are almost equal. If one is saturated and the other unsaturated (points R and
S), the fluxes are unequal.
The relation between these fluxes and the applied alternating voltage e will
now be developed. If the resistance of the secondary circuit, including second
ary windings of transformer and reactors, rectifier, and load, is negligible, the
applied voltage is consumed in the inductive drop almost all of which is in the
two reactors :
The negative sign arises from the opposite connection of the two secondary
windings. The time integral of this voltage is
->
source
edt = Ns<f> = Ns(<j>A-<j>B)
(29)
It is the difference of secondary flux linkages of the two reactors. If the

integration begins when is = 0 and if both reactors are alike, as assumed, the
constants of integration for both cores are equal and disappear in the differ
ence. Suppose that the voltage of the ac source is sinusoidal:
Saturable
reactors
Load
Rectifier
e = Em cos cot
Then the flux linkage of the secondary circuit is likewise sinusoidal and lags
90 behind the voltage (Figure 29a), because
'
E
= Ns<f> = e dt = sin cot
o
Fig. 28.. Circuit diagram of conventional series type of dc current transformer.
that, in the absence of a secondary current, drives the cores of both reactors
far into saturation, producing flux $ and mmf F represented by point P in
Figure 27. The secondary current ts has a mmf Fs NJS that is added to Fp in
reactor A and subtracted from it in reactor B. Thus the total mmfs are
(21a)
FA = Fp + Fs = NpId+Nsis
(27b)
and
Fb = Fp ~ Fs = NpId Nsis
The points on the magnetization curve representing these conditions are
always equidistant horizontally from P, as, for example, are M and N.
Points M and N are for a very small secondary current, and both reactor
cores are still saturated. A larger secondary current drives these points farther
apart (points Q or S and R), causing reactor B to become unsaturated.
(30)
(31)
At every instant the net flux linkage of the two reactors must conform to the
value on this sine curve. The crest value Em of the applied voltage must be
chosen so that the crest value of the flux linkage EJco is somewhat less than
the height of the hysteresis loop, making it impossible for the two reactors to
be saturated in opposite directions. With no primary current, both are un
saturated throughout the cycle. With primary current having mmf greater
than Fm , one reactor or the other is unsaturated during most of the cycle
(points R and S). Only for short time intervals near the zeros of the fluxlinkage wave (near crests or valleys of the voltage wave) are both reactors
saturated (points Mand N). Hence during most of the cycle the mmf Fs of the
secondary current is is very nearly equal to the mmf Fp of the primary current
Id, differing by Fm, depending on whether |0| is increasing or decreasing.
While \<fi\ is increasing the operating point of one reactor is moving down
ward on the left-hand branch of the magnetization curve and Fs = Fp + Fm ;
190
5-16
CONTROL
MEASURING DEVICES
191
Hingorani64 has devised and described a new form of dc current trans

former in which the output current is free from notches and which uses only
2x
(a)
saturable reactors.
Ripples in the direct current are transmitted to the secondary circuit through
the unsaturated reactor acting as an ac current transformer unless special
means are used to suppress the ripple.
Another feature of this current transformer is noteworthy: the output
current is independent of the direction of Id. This is not necessarily a short
coming in applications in which the primary current is always in the same
direction, as it is in a converter.
two
Direct Voltage
(c)
Fig. 29. Wave forms in the dc current transformer of Figure 28.
while \<p\ is decreasing, the operating point is moving upward on the righthand branch of the curve and Fs= Fp~ Fm. If Fm is very small, this differ
ence is unimportant, and, in any event, the average of these two quantities is
the desired value Fp .
The wave shape of the secondary current is is shown in Figure 29b. This
current is rectified to form the output current i0 shown in Figure 29c. The
output current faithfully follows the magnitude of primary or input current
except for two minor features :
1. Narrow notches in small time intervals when both reactors are unsatur
ated (near zeros of iJ/s or peaks and valleys of e)
2. Differences of +Im occurring before and Im after a step at crest
values of \\j/s\ (zeros of e\ Im being the magnetizing current corresponding to
FmThe notches are the chief source of error, for Im is very small if suitable
material is used in the cores. The notches are undesirable in applications in
which the output signal must operate fast-acting devices.
The notches can be obviated by the use of two or three devices like that
described supplied with alternating voltage from different phases of a poly
phase source and with their output terminals paralleled. The final output
current is at every instant equal to that of the unit having the greatest
current.
Measurements of direct voltage are required for the control of the inverter
tap changers, for one of the inputs to a power controller, for current limitation
at low voltage, and for indication of abnormal voltages.
At the very high voltages used in dc transmission, measurement of the line
voltage is difficult and expensive. A resistance voltage divider is enclosed in
an oil-filled porcelain insulator. Its total resistance must be high for limiting
the power converted to heat to an amount that can be dissipated without
undue complication and expense. As a result, only a fraction of a watt is
available at ground potential, and an amplifier is required between the voltage
divider and the several instruments and controllers using a voltage signal.
The combination of the high resistance of the divider and its unavoidable
stray capacitance results in some time delay in the response of the output
voltage to changes in the input voltage. The time constant is of the order of
50 ms.
Power
Dc power can be measured by forming the product of the measured direct
current and voltage. A Hall effect device can be used for this purpose.. Its
output voltage is very low and must be amplified.
Ac power measurements can be used to give a reasonably accurate measure
ment of dc power, because the losses in the converter are small.
For reasons stated in Section 5-13 the direct measurement of power is
seldom used in the control of the power of a converter.
Frequency Deviation
In frequency control, a direct voltage proportional to frequency deviation
is required. The circuit used for this purpose is called afrequency discriminator.
It has a parallel LC circuit tuned to the normal frequency. The quadrature
192
PROBLEMS
CONTROL
current through such a circuit is very nearly proportional to the frequency

deviation for deviations in the range of interest. It leads the voltage for
positive deviation (frequency above normal) and lags behind it for negative
deviation (frequency below normal). This current must be compared with
another alternating current or voltage in quadrature with the voltage im
pressed on the tuned circuit.
Ignition Delay Angle
Control of the tap changer of the rectifier transformer requires measure

ment of the ignition delay angle a of the rectifier for determining whether
this angle is
1. Less than 10: direct voltage should be lowered.
2. Between 10 and 20: no action.
3. Greater than 20: direct voltage should be raised.
This angle can be measured in a number of ways. One way would be a
measurement of the time between the zero point on the sine wave of cornmutating voltage and the front of the grid pulse, which would then be com
pared with the times 10/360/and 20/360/. Another way would be to measure
the maximum instantaneous positive voltage across the valve, normally the
voltage just before it ignites, with 3_Em sin 10 and 3 Em sin 20 obtained
from a voltage divider fed from
Em sin cot, rectified. Of course, these
low
of voltage.
level
a
at
comparisons would be made
Problems 4 to 7 pertain to the circuits known as monocyclic squares and

shown in Figure 30. The three-phase form was used in the initial phase of the
Mechanicville-Schenectady experimental dc transmission to convert constant
alternating voltage to constant alternating current and vice versa.
4. {a) Prove that at the frequency for which Xc and XL are in resonance, the
smgle-phase monocyclic square with negligible losses, loaded with a
variable impedance, delivers a current to that impedance which is pro
portional to the input voltage or delivers a voltage which is proportional
to the input current. (b) What is the ratio of voltage to current ? (c) What
is the reactive-power input for a given reactive-power output? (d) Com
pare the power-factor angles of the input and output circuits.
5. Repeat Problem 4 for the three-phase monocyclic square.
6. Study the three-phase monocyclic square with variable frequency.
(a) How much does the ratio of output current to input voltage vary?
(b) How does the reactive power consumed by the circuit vary?
7. Which phase sequence on the input side of the three-phase monocyclic
square having a rectifier on the output side consumes the least reactive
power from the constant-voltage network ?
V
\/3
Input
PROBLEMS
Output
Input
I. Show that Eq. (12) can be written in the form

- Vd0 cos cotx - Vd0 cos y - 2RcId
2. Show that, if the direct current is changing slowly during commutation,
the following relation is valid:
(cos y -cos /?) =Rc(Idy +1

where I4fiis the value of Id at the beginning of commutation {cot = n - /?)
and Idy is the value at the end of commutation (cot = n- y).
3. How much inductance Ld must the dc reactor have in order barely to
prevent a commutation failure in an inverter operating at 100 kV and
1.0 kA if the 60-Hz voltage falls to half of its normal value immediately
after commutation has begun? The commutation resistance is 8 Q,
y = 16, and the deionization angle is 1. Assume that the direct voltage
on the line side of the reactor remains constant during commutation.
- Vd0
(a)
CV
AC
v CC
(b)
AC
CC DC
Line
Rectifier
CC AC,
CV
AC
Inverter
(c)
Fig. 30. Monocyclic squares: (a) single-phase; (b) three-phase; (c) their use in constantcurrent dc transmission.
194
CONTROL
f
BIBLIOGRAPHY
1. "The Theory of the Control Problem of H.V.D.C. Transmission with Rectifiers and
Inverters in Bridge Circuit," by F. Busemann, Technical Report, Ref. Z/T74,
B.E. & A.I.R.A. , Mar. 2, 1948.
2. "The Stability of Frequency of A.C. Systems Connected by a D.C. Transmission
Line " by F. Busemann, Technical Report, Ref. Z/T75, B.E. & A.I.R.A., Mar. 5, 1948.
3.
"Grid Control, Regulation, Switching High-Voltage Converters On and Off," Chap

ter VI (pp. 52-62) of Power Transmission by Direct Current, by Ya. M. Chervonenkis,
Academy of Sciences of the U.S.S.R., 1948, translated from Russian by the Israel
Program for Scientific Translations, Jerusalem, 1963.
4. "Hunting of Compounded Inverters for H.V. D.C. Transmission," by F. Busemann,

B.E. & A.I.R.A. (E.R.A.) Report, Ref. B/T 105, Feb. 21, 1951, 17 pp. By tests on a
model, the theory in Ref. 1 (Report Z/T74) was confirmed, except that the losses on
the model were higher than expected.
5. "Hunting of Rectifiers with Marked Compounding," by F. Busemann, B.E. &
A.I.R.A., Report B/T104, Feb. 14, 1951.
6. "Practical Reactive Power Requirements of H.V. D.C. Inverters with Natural
Commutation," by F. Busemann, Direct Current , Vol. 1, pp. 35-38, September 1952.
Compounding of rectifier and inverter reduces reactive power required in normal
operation in order to withstand reduction of alternating voltage due to remote fault
without commutation failure.
7. "The Gotland D.C. Link: The Grid Control and Regulation Equipment," by Harry
Forsell, Direct Current, vol. 2, pp. 109-114 and 166-170, June and December 1955.
8. "Analysis of the Stability of the Control Systems of the Kashira-Moscow Direct
Current Transmission Line," by A. M. Reider, Direct Current, Vol. 3, pp. 227-240,
December 1957.
9. "Protection of D.C. Transmission," by B. S. Melik-Sarkisov, N.I.I.P.T., Vol 2,
pp. 84-95, 1957. In Russian.
10. "Compounding Device of the Invertor of the D-C Transmission Kashira-Moscow,"
by A. M. Reider, N.I.I.P.T., Vol. 2, pp. 96-111, 1957. In Russian.
11. "The Stability of a D.C. Interconnection Regulating System When Using the New
Valveless Current Regulator," by E. A. Berlin, Direct Current, Vol. 3, pp. 264-266,
March 1958.
12. "Control Systems," by U. Lamm, report to C.I.G.R.E. Study Committee No. 10,
June 1959, published in Direct Current, Vol. 4, pp. 162-163, September 1959.
Hingorani, Garraway, London, 1960, Chapters 4 and 5 (pp. 49-84).
14. " Control Systems," by U. Lamm, Appendix VI to C.I.G.R.E., 1960, Report No. 417,
"Report on the Work of Study Committee No. 10, D.C. Transmission at E.H.V.,"
pp. 29-33.
15. "Semiconductor Circuitry and Arrangements for the Protection and Control of
Bridge-Connected Current Convenors," by Colin Adamson and N. G. Hingorani,
Direct Current, Vol. 5, pp. 52-64, September 1960.
BIBLIOGRAPHY
195
16. "Control of Power Flow in a D.C. Link," by L. L. Freris and B. J. Cory, Direct
Current, Vol. 5, pp. 72-77, December I960.
17. "Corrector for the Compounding Device," by G. V. Smirnov, N.I.I.P.T., Vol. 6,
pp. 149-154, 1960. In Russian.
18. "The Scope of Semiconductor Devices in the Control and Protection of H.V.D.C
Systems," by N. G. Hingorani, Direct Current, Vol. 6, pp. 180-181, September 1961.
19. "Control of H.V.D.C. Converters," by Colin Adamson and N. G. Hingorani,

Direct Current, Vol. 7, pp. 148-155, June 1962.
20. "Steadiness of Systems for Regulating D.C. Power Transmission Improved by
Altering the Principle on Which the Invertor is Compounded," by E. M. Berlin,
Direct Current, Vol. 7, pp. 319-325, December 1962.
21. " Compound Current Regulator for the D.C. Transmission Volgograd-Donbass," by
E. M. Berlin and M. G. Zavarina, N.I.I.P.T., Vol. 9, pp. 86-107, 1962. In Russian.
22. "The Steady State Stability of a D.C. Transmission Link When Operating into a
Receiving Power Supply System of Comparable Capacity," by V. M. Kvyatkovskii
and N. M. Mel'gunov, Direct Current, Vol. 8, pp. 107-115, April 1963.
23. "A Note on the Reverse-Parallel Operation of Rectifier Bridges," by K.D. Srivastava
and R. M. Davis, Direct Current, Vol. 8, pp. 202-203, July 1963.
24. "Operation and Control of H.V.D.C. Transmission," by P. G. Engstrom, I.E.E.E.
Trans, on P.A. & S., Vol. 83, pp. 71-77, January 1964. Disc., p. 77.
25. "D.C. Link and A. C. Networks Combined Power Control Requirements," by
Francesco Iliceto, Direct Current, Vol. 9, pp. 37-39, 60, May 1964. Load-frequency
control.
[
j
26. High Voltage Direct Current Convertors and Systems, edited by B. J. Cory, Macdonald,
London, 1965. Chapter 5, "Control of H.V. D.C. Systems," by L.L.Freris,pp. 73-93;
Chapter 8, " Some Design Aspects of the Cross-Channel Power Link," by L. A. Harris,
pp. 185-187, 199-211.
27. "Infra-red Radiation Pulses for the Control of a Converter," by R. Feinberg, M. E
Roberts, and J. Armstrong, Direct Current, Vol. 11, p. 72, May 1966. Pulses trans
mitted by semiconductor crystal lamps to photodiodes.
28. " System Stability and Control-Protection System of A.C.-D.C. Link," by T. Yamada,
S. Fujii, and T. Horigome, C.I.G.R.E., Report 325, 1966.
29. " Principles of Control for H.V.D.C. Transmission," by J. D. Ainsworth and C. J. G.
Martin, I.E.E. Conference Publication No. 22, High Voltage D.C. Transmission,
30. Discussion by Uno Lamm, I.E.E. Conference Publication No. 22, High Voltage D.C.
Transmission, Manchester, Sept. 19-23, 1966, Part 2, p. 74.
31. " Operation of the Control and Protection System of the Volgograd-Donbass Link,"
by K. Gusakovsky, A. Posse, and A. Reider, Contribution No. 95, I.E.E. Con
ference Publication No. 22, High Voltage D.C. Transmission, Manchester, Sept. 19-23,
1966, Part 2, pp. 130-135.
32. "EHV-DC Simulator," by J. E. Hudson, E. M. Hunter, and D. D. Wilson, l.E.E.E.
Trans, on P.A. & S., Vol. 85, pp. 1101-1107, November 1966.
33. " Simplified Study of the Stability of Regulation of a D.C. Link," by J. Clad6 and
A. Lacoste, Direct Current, Vol. 12, pp. 9-12, 14-15, 18-22, February 1967.
196
CONTROL
34. "A Method to Detect the Deionization Margin Angle and to Prevent the Commuta
tion Failure of an Inverter for D.C. Transmission," by T. Machida and Y. Yoshida,
I.E.E.E. Trans, on P.A. & S., Vol. 86, pp. 259-262, March 1967.
35. "A Method of Automatic Frequency Ratio Control by a D.C. System," by T.
Machida, Y. Yoshida, and H. Nakamura, I.E.E.E. Trans, on P.A. & S., Vol. 86, pp.
263-267, March 1967.
36. "A New Constant Extinction Angle Control for AC/DC/AC Static Convertors," by
Narain G. Hingorani and Philip Chadwick, I.E.E.E. Trans, on P.A. & S., Vol. 87, pp.
866-872, March 1968. Disc., p. 872. (I.E.E.E. Paper 31 TP 67-497.)
37. " The Phase-Locked Oscillator A New Control System for Controlled Static Con
verters," by J. D. Ainsworth, I.E.E.E. Trans, on P.A. & S., Vol. 87, pp. 859-865,
March 1968. Disc., p. 865.
38. "Present and Future Status of HVDC Transmission," by H. M&rtensson and
B. Skoglund, I.E.E.E. Publication 68 C 57, E.H.V. Conference held in Montreal,
1968. Contains a brief discussion of equiangular control.
39. " Untersuchung von Schutz- und Regelungsproblemen in der Modellanlage Rheinau "
(Investigations of Control and Protection Problems in the Rheinau Model Plant, in
German), by Reiner Foerst, Gerhard Heyner, Karl-Werner Kanngiesser, and Michael
Becker, ETZ, edition A, Vol. 89, No. 9 pp. 213-218, Apr. 26, 1968. English translation
in HGt)-Schriften-reihe, pp. 47-52, published by the Arbeitsgemeinschaft HGt).
40. "Der Loschwinkel-Regelkreis einer HGt) als Abstatsystem " (The Extinction-Angle
Regulating Circuit of a HVDC Transmission Link as a Sampled-Data System, in
German), by Reiner Foerst, ETZ, edition A, Vol 89, No 16, pp. 381-385, Aug. 2,
1968. Based on dissertation done at Darmstadt Technical College, 1967.
41. "Single-Phase Bridge Converter with a New Control and a Corresponding ThreePhase Converter Part II, Method of Control," by N. G. Hingorani and Philip
Chadwick, Direct Current (new series), Vol. 1, No. 1, pp. 31-35, April 1969.
42. " Control of the Extinction Angle of the Inverter of a High-voltage d.c. Transmission
System," by Winfried Muttelsee and York Rogowski, Direct Current, Vol. 1 (new
series), No. 2, pp. 49-53, August 1969. Feedback control of extinction angle using
low loop gain in steady operation but high loop gain when the extinction angle is
too small, as during a rapid increase of direct current.
43. "A Refined HVDC Control System," by Ake Ekstrom and GoteLiss, I.E.E.E. Trans.
on P.A. & S., Vol. 89, No. 5/6, pp. 723-730, May/June, 1970. Disc., pp. 730-732.
Provides equally spaced ignition pulses.
44. "Stability Analysis of the HVDC Transmission Control System Using Theoretically
Calculated Nyquist Diagrams," by Kjell Eriksson, Gote Liss, and Erik V. Persson,
I.E.E.E. Trans. onP.A.&S., Vol. 89, No. 5/6, pp, 733-739, May/June, 1970. Disc.,
pp. 793-740.
45. "Stability Analysis of the DC Power System," by Dennis P. Carroll and Paul C.
Krause, I.E.E.E. Trans, on P.A. & S., Vol. 89, No. 6, pp. 1112-1118, July/August,
1970. Disc., pp. 1118-1119.
46. "Control Methods for Improving the Reactive Power Characteristics of HVDC
Links," by Karl W. Kanngiesser and H. P. Lips, I.E.E.E. Trans, on P.A. & S., Vol. 89,
No. 6, pp. 1120-1125, July/August, 1970.
Pulse Transformers
47.
Part III of Pulse Generators, Vol. 5 of M.I.T. Radiation Laboratory Series, by N.

Glasoe and J. Lebacqz, originally published by McGraw-Hill, New York, 1948;
reprinted by Boston Technical Lithographers, Lexington, Mass., 1964.
BIBLIOGRAPHY
197
I
i
Control of Multlterminal Lines
48. "Some Aspects of Tapping of H.V.D.C. Transmission Systems," by Uno Lamm,

Erich Uhlmann, and Per Danfors, Direct Current, Vol. 8, pp. 124-129, May 1963.
Editorial comment in Vol. 8, pp. 87, 120, April 1963.
49. "The Place of H.V.D.C. in the Power Transmission Field," by G. D. Breuer and
E. M. Hunter, I.E.E.E. Publication S-155, D-C Transmission, June 1963, pp. 5-18.
Disc., pp. 222-223.
50. "Series Connection of Converter Stations in an H.V.D.C. Transmission," by John
Reeve and J. Arrillaga, Direct Current, Vol. 10, pp. 72-74, 76-78, May 1965.
51. "Possibility of H.V.D.C. Transmission with More than Two Stations," Section
5.3, pp. 84-91, of Ref. 26.
52. " Universal Control-Protection System for Multiterminal D-C Power Transmission,"
by T. Yamada, H. Kondo, T. Horigome, and T. Sekine, I.E.E.E. Paper No. 31
PP 67-57, Winter 1966.
53. "Simulator Studies of Multi-terminal H.V. D.C. Systems," by H. M&rtensson and
T. Adielson, I.E.E. Conference Publication No. 22, H.V. D.C. Transmission, Man
chester, Sept. 19-23, 1966, Part 1, pp. 134-145. Three-terminal system controlled
sccording to the principles of Ref. 48. Oscillograms of voltage and currents during
blocking and deblocking of valve groups, etc.
54. " Multi-terminal D.C. Transmission," by K. W. Huddart and W. G. Watson, I.E.E.
Conference Publication No. 22, H.V. D.C. Transmission, Manchester, 1966, Part 1,
Paper 17, pp. 94-98.
55. " Multiterminal Operation of HVDC Converter Stations," by Reiner Foerst, Gerhard
Heyner, Karl Werner Kanngiesser, and Hermann Waldmann, I.E.E.E. Trans, on
P.A. & S., Vol. 88, pp. 1042-1050, July 1969. Disc., pp. 1050-1052.
56. " Model Tests on Three Terminal HVDC Systems," by R. Foerst and G. Heyner,
C.I.G.R.E., report 14-02, 1970, 10 pp.
Measuring Devices, Including DC Current Transformers
57.
58.
59.
60.
61.
ffo.
63.
64.
"Measurements in High-Voltage D.C. Power-Transmission Installations," Chapter

VIII (pp. 73-78) of Power Transmission by Direct Current, Ya. M. Chervonenkis,
1948 (see Ref. 3).
"A Direct Current Instrument Transformer," by I. A. Zaitsev and A. V. Kalyaev,
Direct Current, Vol. 2, pp. 19-21, June 1954. From Elektrichestvo, 1954, No. 2.
"Applications of the Transductor in Rectifier Technique," by L. F. Borg and P. G.
Engstrom, Direct Current, Vol. 2, pp. 58-65, December 1954.
"The Choice of Parameters and Design of a D.C. Instrument Transformer," by A. M.
Ryvkin, translated by B. F. Evans, Direct Current, Vol. 2, pp. 97-100, March 1955.
Magnetic Amplifiers, by H. F. Storm, Wiley, New York, 1955, pp. 167-169, 407-408.
Magnetic-amplifier Circuits, by William A. Geyger, McGraw-Hill, New York, 2nd
ed., 1957, pp. 44-46.
"New Transductor-Type D.C. Transformer Particularly Applicable to H.V. D.C.
Systems," by Colin Adamson, Proc. I.E.E., Vol. 110, pp. 739-750, April 1963. Paper
No. 4101P.
"An Accurate Transductor-Type D.C. Measuring Transformer," by N. G. Hingorani,
Direct Current, Vol. 9, pp. 53, 56-60, May 1964.
I
i
j
j
I
|
6-2
6
Misoperation of Converters
Control, the subject of Chapter 5, and protection, discussed in this chapter

and the next, are closely related in hv dc systems, because grid control of the
valves in the converters is used for both purposes. The valves are used
instead of circuit breakers for deenergizing the dc line when a short circuit
occurs on it and for reenergizing the line after the short circuit disappears.
Faults occur sometimes in the valves themselves. Some of these faults are
self-clearing and most others are cleared by operation of other valves. Only
for a few extreme faults is it necessary to open an ac circuit breaker or to take
a converter bridge out of service for more than a second.
6-1 MALFUNCTIONS OF MERCURY-ARC VALVES
Following are the definitions of the commonest malfunctions of valves or

their accessories. The words in parentheses are alternate names for some of
these malfunctions:
Arcback (backfire) Conduction in the reverse direction
Arcthrough (fire-through, shoot-through) Conduction during a scheduled
blocking period
Quenching (arc quenching, arc chopping) Premature extinction of the arc
during a scheduled conducting period
MisfireFailure to ignite in spite of positive grid and anode voltages
BYPASS VALVES
199
In Figure 20 in Chapter 3, showing the voltage across a valve while rec

tifying or inverting, the cycle is divided into three periods: (1) inverse, (2)
blocking, and (3) conducting. Fromthe foregoing definitions of malfunctions,
it is clear that arcback can occur only in the inverse period; arcthrough, only
in the blocking period; quenching, only in the conducting period; misfire,
only at the beginning of the scheduled conducting period ; and commutation
failure, only right after the end of the scheduled conducting period and
extending into the scheduled blocking period.
A rectifier valve has a longer inverse period and a higher inverse voltage
than the same valve during inversion ; hence arcbacks are more frequent and
more severe during rectification than during inversion. An inverter valve has a
longer blocking period and a higher forward voltage than the same valve
during rectification; hence arcthroughs are more frequent and more severe
during inversion than during rectification. Commutation failure occurs during
inversion only.
6-2 BYPASS VALVES
General
Most of the valve faults that are not self-clearing with the valve in service
are cleared by relieving the valve from current for a fraction of a second. This
is the purpose of the bypass valve.
Normally each converter bridge has a bypass valve across its dc terminals,
as shown in Figure 1. Since the bridge already has six main valves, the bypass
30 Ac
/10
/10
These expressions originated in connection with mercury-arc valves. Perhaps

they will be extended to thyristors, or perhaps a new vocabulary will come
into being.
A malfunction of an inverter that may not be a fault of the valves is the
following:
failure (breakthrough) Failure to complete commutation

before the commutating emf reverses
Commutation
198
Fig. 1. Addition of bypass valve (7) and switch (8) and disconnecting switches (9 and 10) to
a three-phase converter bridge.
200
MISOPERATION OF CONVERTERS
6-2
valve is sometimes called the seventh valve. It is usually a valve of different

design from the main valves, and, in addition, some of its accessories are
different, as described below under "Special Auxiliaries," page 205.
The control grids of the bypass valves are normally blocked. When a bridge
is to be bypassed, its bypass valve is unblocked, and the main valves are
simultaneously blocked by discontinuing the transmission of positive pulses
to their grids. The direct current then shifts from the main valves to the bypass
valve, not instantly but in a few milliseconds by a process like normal
commutation between two main valves. Later, by simultaneously unblocking
the main valves and blocking the bypass valve, the direct current is trans
ferred back to the main valves. Commutation of current to and from the
bypass valve and the conditions under which it can occur are analyzed later
under "Transfer of Current."
If a bridge is to be removed from service for replacement of a defective
valve or for other maintenance work, the direct current is first transferred
to the bypass valve, after which the bypass switch (8 in Figure 1) is closed and
takes over the direct current. Finally, the ac and dc disconnecting switches
(9 and 10) are opened, isolating the bridge. For putting the bridge back into
service, these switching operations are performed in reverse order.
A monopolar dc link or each pole of a bipolar link has at least two bridges
in series (one in the rectifier and one in the inverter) and usually more than
two. Suppose that there are four bridges, two at each terminal (Figure 2). Let
us consider the effect of bypassing one bridge on the operation of the remain
ing bridges. First assume an inverter bridge to be bypassed. The remaining
inverter bridge continues to operate at normal extinction angle, and the total
countervoltage of the inverter is approximately halved. The direct current
tends to increase, but the dc reactor and the constant-current control at the
rectifier hold it nearly at the ordered value, to do which the ignition angle of
Rectifier
Positive dc line
BYPASS VALVES
201
both rectifier bridges is increased. Next, assume that a rectifier bridge instead
of an inverter bridge is bypassed. The rectiffer emf is almost halvedit would
be exactly halved if the remaining rectifier bridge did not decrease the ignition
delay angle to its minimum valueand, therefore, the current decreases.
The constant-current control of the inverter, however, increases the ignition
advance angle so as to prevent the current from decreasing by more than the
current margin.
Because operation at large control angles a or y increases the stresses on

valves and on valve-damper circuits and increases the consumption of reactive
power, such operation for more than a few minutes is undesirable. If a bridge
must be bypassed for a long time, a bridge on the same pole at the other
terminal should be bypassed also, and bypass switches should be closed in
parallel with each unblocked bypass valve.
Transfer of Current
Transfer to the Bypass Valve. This transfer is analyzed now as it

takes place when the bridge is operating normally as a rectifier. Assume that
the direct current in the line is kept constant by the action of the dc reactor
and the constant-current control. Assume, too, that at instant C, Figure 3,
while valves 1 and 2 are conducting, the grids of all the main valves are given
r-j
(a)
180*
60
-60
Inverter
AY
YA
(b)
YY
YY
CM
1,2,7
Fig. 2. Bypass valves B on converters with two bridges per pole.
Fig. 3. Commutation from a normally operating rectifier bridge to the bypass valve:
(a) voltages; (b) currents.
202
6-2
a blocking signal (that is, the positive grid pulses are turned off) and that
valve 7 (the bypass valve) is unblocked; that is, its grid is made positive. At
this moment the anode voltage of valve 7, v7 = vdvn vp, is negative
and it cannot ignite. At instant D valve 3 would normally ignite but cannot
for lack of a grid pulse. Valves 1 and 2 continue to conduct, and the direct
voltage, vd = vp vn , declines. At instant E, the direct voltage of the bridge
becomes zero and starts to reverse. Immediately the bypass valve ignites.
Thereupon the effective circuit (consisting of the conducting branches in
Figure 1) becomes that in Figure 4. Its left-hand mesh has a line-to-line emf
BYPASS VALVES
203
In an inverter, the direct voltage across the bypass valve is normally

positive, permitting that valve to ignite as soon as it is unblocked. The direct
line voltage immediately becomes zero, and commutation begins from the
conducting main valves to the bypass valve. This commutation has an ignition
angle that varies somewhat with the instant of unblocking; however, while
any two given valves say, valves 1 and 2 are conducting, the correspond
ing commutation emf (in this case, eca) is near its positive crest; therefore,
commutation is rapid. In Figure 5 AE is the range of cot in which both valves
1 and 2 are normally conducting; the corresponding arc of commutating
voltage is crosshatched.
Normally
positive for rectifier
negative for inverter
et
= Vh
Fig. 4. Equivalent circuit for commutation from regular valves 1 and 2 to bypass valve 7 of
a rectifier or inverter bridge.
eca , inductance 2Lc , and three valves in series. The bypass current i7 and the
commutating voltage are both zero at the moment when valve 7 fires (a 0
for the bypass valve). Taking this moment as t = 0, the voltage is
eCa
3 Em sin 03 1
(1)
and the bypass current, lagging 90, is
i7 =/s2(l - cosrof)
where
IS2 =
(2)
V3 EJ2coLc . The currents in the main valves are

h = h =h~ h =Id~ 4>(1 - cos cot)
(3)
Commutation ends when it = i2 0 and i7 = Id at cot = u7. The overlap angle

is given by
BCD
cos u7 = 1
*s2
(4)
After commutation is completed, only the right-hand mesh of the circuit in

Figure 4 is conducting.
Fig. 5. Commutation from normally operating inverter bridge to bypass valve: (a)
voltages; (b) currents.
204
6-2 BYPASS VALVES
The circuit diagram in Fig. 4 is still applicable, but with the difference of
polarity of direct voltage shown thereon.
Transferfrom the Bypass Valve. When a rectifier bridge is unblocked, with
its ignition delay angle at the normal value (around 15), its voltage builds up
rapidly with normal polarity. The voltage across the bypass valve under these
conditions is negative, as already noted; and current is transferred from this
valve after which it must cease conducting. Preferably the bridge voltage is
raised more slowly (by appropriate control of the ignition delay angle) for
limiting the overshoot of voltage due to the lightly damped natural mode of
the line and its reactive terminations.
In an inverter, on the other hand, the normal voltage is opposite to that of
the same converter when rectifying and would constitute a positive voltage
across the bypass valve. That valve would therefore continue to conduct even
if its grid had a blocking (negative) voltage, for the grid can exert no control
on a valve that is already conducting. The bridge, therefore, works into a dc
short circuit. Its current output is not limited by constant-current control.
Instead, because the current for this control is measured on the line side of
the bypass valve, it tries to raise the bridge voltage in an attempt to reduce the
line current, which now exceeds the inverter current setting by even more than
the current margin.
When the transfer of current from the bypass valve to the main valves is
required, the control should not only block the bypass valve and unblock the
main valves but also should reduce the ignition angle /? enough to make the
inverter bridge temporarily work as a rectifier. We have already noted that
the polarity of voltage of a rectifier is that required to transfer the line current
from the bypass valve to the main valves. After the completion of this com
mutation and adequate deionization of the arc in the bypass valve, the
decreased) so that the bridge
ignition of the main valves may be delayed
voltage builds up with proper polarity for inverter operation. The bypass
valve is now prevented from reigniting, despite its positive anode voltage, by
its negative grid voltage.
On examining these phenomena in more detail, we note that, with ft = 90,
the instantaneous open-circuit voltage of the bridge, which is still actually
short-circuited by the bypass valve, is positive about half of the time and
negative the other half (see Figure 6d in Chapter 3). This alternating emf,
in series with the relatively small commutating inductance 2Lc, can in a small
fraction of a cycle produce a short-circuit current in the amount and direc
tion to force the net current in the bypass valve to become zero. (The current
supplied from the bridge does not pass through the dc reactor.) Thus, the
value of fi required to force extinction in the bypass valve may be somewhat
less than 90 and, thus, be in the zone considered, from the standpoint of
205
average voltage and current, to be inversion rather than rectification; that is,
the required rectification occurs in about one-twelfth cycle (30) or less.
Special Auxiliaries
The bypass valve requires a different kind of current divider and a different
kind of grid-pulse supply from those used with the other bridge valves. The
reason is that the bypass valve is required to carry current continuously for a
longer time than the other valves normally do, and, also, its grid voltage must
be positive for a longer time.
The current divider for a multianode bypass valve differs from that for
other multianode valves (Figure 22 in Chapter 3) only in having equal resis
tors in series with each anode.4 A current divider like those of the main valves
gives nearly simultaneous ignition of the several anodes and equal current
division during pulses as short as those normally used (one-third cycle).
However, it cannot maintain good current division for times up to several
seconds. The added resistors do this. The power loss in the resistors would be
intolerable for continuous operation but is tolerable for a few seconds. If a
bridge must be bypassed for a longer time, a switch is closed in parallel with
the bypass valve, relieving that valve and its resistors of current. The voltage
drop in the resistors at rated current is of the same order of magnitude as that
in the arc.
The grid pulses for the regular valves are usually formed at ground poten
tial and are transmitted to the grids through pulse transformers. Such trans
formers cannot transform direct current and, therefore, work well only with
short pulses. Bypass valves do not always ignite immediately after they are
unblocked. Moreover, a positive grid voltage maintained for the duration of
desired conduction is a safeguard in the event of arc quenching.
A method of furnishing grid signals to the bypass valves is to connect a
free-running square-wave generator to the primary winding of the pulse
transformer. The secondary voltage of that transformer is rectified, and the
resulting direct voltage is applied to the control grid.5
Because bypass valves are sometimes ignited at a low anode-cathode
voltage, it is less certain that all the anodes of a multianode bypass valve will
ignite.
Use of Main Valves for Bypassing
Proposals have been made for the elimination of the bypass valve and
using, instead, a pair of the main valves on the same phase of the ac supply
for bypassing the bridge.6,7 This proposal has been adopted for thyristors,
206
6-3
which do not arc back, but the bypass valve is still required for clearing
arcbacks in hv high-power mercury-arc converters. Even on such converters,
however, it is proposed to use main valves as backup in the event of failure of
the regular bypass valve to ignite.
6-3 ARCBACK
Each of the malfunctions listed in Section 6-1 is discussed as to its causes,

its consequences, and the protection against it. The wave shapes of currents
and voltages in the converter during each kind of malfunction are analyzed.
Causes
Arcbacks are the commonest and best known as well as the severest kind of
malfunction of rectifier valves but are less frequent in inverters. Since arcback
is reverse conduction, it can occur only when there is inverse voltage across a
valve. In rectification each valve is exposed to inverse voltage during approxi
mately two-thirds of each cycle but to forward voltage for a much shorter time
and with a lower crest value. In inversion the opposite is true (see Figure 20
in Chapter 3).
Arcbacks are a random phenomenon. We know some of the factors that
tend to increase the average frequency of occurrence but do not know how
entirely to eliminate them. An average occurrence of one or two arcbacks per
valve per month is considered satisfactory.
Among the factors that tend to increase the occurrence of arcbacks are the
following:
1. High peak inverse voltage
2. High voltage jumps (Section 3-5, page 108), especially of the jump at
arc extinction
3. High rate of change of current at the end of conduction
4. Overcurrent
5. Condensation of mercury vapor on anodes
6. Impurity of materials in anodes and grids
7. High rate of increase of inverse voltage
Most of these factors can be controlled. Factors 1 and 2 can be reduced by
decrease of rated valve voltage; 3 and 4, by decrease of rated current. These
measures, however, reduce the power handled per valve and, hence, raise
ARCBACK
207
the cost of the converter per unit of power. Factors 2 and 3 can be improved
by the use of small converter angles (a, /?, y, S); but these angles must be
increased to large values temporarily in such control operations as starting
up, maintaining constant current during dips of alternating voltage, or in
causing the transfer of line current from the bypass valve to the main valves
of an inverter. These operations do increase the incidence of arcbacks.
Factor 5 is minimized by maintaining the anodes at a higher temperature than
the cathodes. Factor 7 is minimized by the use of RC damper circuits in
parallel with each valve (discussed in Section 7-3). Factor 3 can be made to
occur less frequently by not allowing operation with very small overlap (see
Sections 5-13, page 181, 7-2, page 240, and 7-4, page 266).
i
I
Consequences
As shown in the subsequent analyses, arcbacks result in line-to-line short

circuits and, sometimes, in three-phase short circuits. These short circuits
subject the transformers and valves to much greater current than does normal
operation. The transformer windings must be firmly braced to withstand more
numerous short circuits than those to which ordinary power transformers
are subjected. The high valve currents increase the rate of erosion of anodes
and, hence, lead to the need for more frequent inspection and maintenance of
valves. Such currents also make immediate repetition of the arcback almost
certain unless the valve is relieved of current for half a second or so, allowing
it to return to normal temperature and to become so well deionized that it can
again withstand anode voltage.
Two kinds of arcbacks in rectifiers are analyzed. The first analysis is of
successful blocking and bypassing, thereby limiting the crest value and the
duration of the arcback current. The second analysis is of an uncontrolled
rectifier, with greater crest value and duration of the arcback current than in
the first case. The transformer and valves should be designed to withstand
uncontrolled arcbacks without permanent damage.
|
j
|j
5;
Analysis of Arcback with Blocking and Bypassing
An arcback is assumed to occur in valve 1 very soon after it has finished

commutating with valve 3. All the main valves not conducting at this time
are assumed to be blocked successfully, and, at the same time, the bypass
valve is assumed to be unblocked. The sequence of events is described in a
"narrative" in which reference is made to the circuit diagrams in Figure 6
and the curves in Figure 7.
j
I
m
6-3 ARCBACK
209
Narrative
A.
B.
C.
id
IV
3V
Vp
'P
(a)
Fig. 6. Equivalent circuits for study of arcback of valve 1 with blocking and bypassing:
(a) first period, with valves 1, 2, and 3 conducting, and second period, with valves 1, 2, 3,
and 7 conducting; (b) third period, with valves 1, 3, and 7 conducting.
Crest
+ 18.8 kA
cot = 0. ea = eb Valves 1 and 2 are conducting.

cot = a. Valve 3 ignites and begins to commutate with valve 1.
cot =a + u. Commutation from valve 1 to valve 3 is completed, and an
inverse voltage step immediately appears across valve 1, which has just
stopped conducting. Assume that valve 1 arcs back, renewing the short
circuit on phases a and b of the transformer; vp , which had momentarily
jumped by eb, now reverts to the average of ea and eb . v continues equal
to ec . Currents it and i3 continue on the same offset cosine waves that
they were on during normal commutation. Valves 1, 2, and 3 are con
ducting; see the equivalent circuit in Figure 6a. Blocking of further grid
pulses is assumed to occur promptly at the latest, before instant E.
D. cot = 60 + a. Valve 4 would normally ignite now. However, even if its
grid were pulsed now, it could not ignite until instant E (at intersection
of curves ec and eJ2), because its anode-cathode voltage is still
negative. Hence, from Cto Eis. the time available for blocking.
E. cot = 90. Direct voltage vd becomes zero and tends to reverse. Bypass
valve 7 ignites and begins to commutate with valve 2. Second period
begins, in which valves 1, 2, 3, and 7 are conducting; see Figure 6a.
They put a three-phase short circuit on the transformer. The bypass
valve short-circuits the dc terminals. All valve voltages, as well as volt
ages of ac and dc terminals, are now zero.
F. End of commutation. Current in valve 2 becomes zero and remains so.
Valve 7 current is ld. The third period begins, in which valves 1, 3 and 7
are conducting; see Figure 6b. Valve 7 keeps the dc terminals shortcircuited. Valves 1 and 3 constitute a line-to-line short circuit on the
transformer.
cot = 180. The commutating emf reverses. Currents in valves 1 and 3
reach their maxima and begin to decrease.
H. Currents in valves 1 and 3 reach zero. If valve 3 extinguishes, the original
arcback is finished. Henceforth only the bypass valve conducts. See also
Section 6-10, page 231, for the case in which valve 3 arcs back.
G.
i7 (bypass)
ojt, deg
ii (in valve
arcing back)
-10
Equations for Instantaneous Voltages and Currents
1stperiod, B to E, valves 1, 2 and 3 conducting (Figure 6a):

-18.8 kA
~20i
D E
Instants: ABC
Period:
Valves
1st
1,3 k- 1,2,3
conducting:
F
2nd
S,2, 3, 7
Fig. 7. Valve currents during arcback controlled by blocking and bypassing.
*-b = - -ec
Va = Vb = Vp = -
(5)
vc = vn = ec
(6)
= v2 = v3 = 0
(7)
210
6-3
h ~ h = h + 7s2(cos cot h = -ic = h

i3 = ib Is2(cos a cos a>f)
z4 = i5 = = i-t 0
cos a)
(9)
All voltages are zero.

cos (cot
~h~ hi
cos a + la
30) 0.5 V3
Analysis of Uncontrolled Arcback
(13)
An arcback is assumed to occur on an uncontrolled rectifier with no bypass

valve or with a blocked one. The phase-to-phase short circuit on the trans
former develops into a three-phase short circuit, yielding a greater crest
current in the valve which arcs back than that which occurs if all the non
conducting valves are blocked. The uncontrolled arcback is described below
with reference to the circuit diagrams in Figures 8 to 16 and to the wave
forms in Figure 17 (page 217). The numerical example assumes Is2 = 10.0 kA,
Id = 1.8 kA, and a. = 0.
(14)
Narrative
(10)
(11)
(12)
2nd period, E to F, valves 1, 2, 3, and 7 conducting (Figure 6a) :
A.
2
= ic
= /s3[cos (cot + 90)
+ 1] + 7d
(15)
\l3
3 =
ib = /J cos(ffl? - 150) - 0.5 + cos a
(16)
i5 = i6 = 0
(17)
h = [cos (cot + 90)
+ 1]
(18)
T *-b
Cc
= --
(19)
(20)
Vl=v3=v4 = v6 =v7 = vd=0
(21)
=iec
(22)
ii = z3 = ia =
ib = 7S2[cos cot cos a cos co/2]
/i3[cos (cot2 150) 0.5]
(23)
where t2 corresponds to point F.

h~ h=h
h=h=0
(24)
h = Id
Numerical
(25)
e,
Lc
ij
-H-
B. cot = u. End of normal commutation. Valves 2 and 3 are conducting (cir

cuit 2, Figure 9), and the short circuit disappears. Inverse voltage sud
denly appears across valve 1. Assume that valve 1 arcs back. The short
circuit is restored. The circuit in Figure 8, with valves 1, 2, and 3 con
ducting, is again valid. it and i3 continue to follow displaced cosine
curves having their maximum and minimum, respectively, at instant A.
-& ea
va
eb
Vb
ui 114
N'Qy
wId
f5=6
Example
Figure 7 is plotted for
Q-
n"
Fig. 8. Equivalent circuit 1: valves 1, 2, and 3 conducting.
vc = ec
~v2 = v5
Commutation begins from valve 1 to valve 3. Valves

(see circuit 1, Figure 8). There is a short circuit
and
conducting
3
are
1, 2,
on phases a and b. vp follows the average of ea and eb .
cot = 0. eaeb.
V4 = US = l>6
3rdperiod, Fto H, valves 1, 3, and 7 conducting (Figure 6b):
va = vb = vn = vp=
Is2 = 10.0 kA
and a = 15. The crest values
and i3
211
(8)
V4 = v5 = v6 = V-, = -vd = \.5ec
h = ia
ARCBACK
(hence /s3 = 11.6 kA), Id= 1.11 kA,

are 18.8 kA. Point His at cot = 331.
-&
-w-
Vn
Fig. 9. Equivalent circuit 2: valves 2 and 3 conducting.
212
c.
6-3
M1S0PERATI0N OF CONVERTERS
wt = 60. Valve 4 normally would ignite at this time but cannot because
ARCBACK
213
on phases a and c through valves 1 and 5. Positive voltage appears sud

denly across valve 6, causing it to ignite. Valves 1, 2, 5, and 6 are now
conducting (see circuit 5, Figure 12), reestablishing a three-phase short
circuit. Commutation begins from valve 2 to valve 6.
the voltage across it is still negative by reason of the short circuit on

phases a and b.
D. cot = 90. The voltage across valves 4, 5 and 6 passes through zero,
becoming positive. Any one of these three valves could ignite, bringing
the voltage across them to zero and preventing the other two from
igniting. Assume that only valve 5 ignites.* Then valves 1, 2, 3, and 5 are
conducting (see circuit 3, Figure 10). There is a three-phase short cir
cuit through valves 1, 3 and 5 and a dc short circuit through valves
2 and 5. Currents ia, ib , and ic have ac components constituting a
positive-sequence set, each current lagging the corresponding emf by
90. These currents also have dc components (offsets) determined by
initial conditions. See Eqs. (27), (29), and (30) below.
-Q-nsw*" w
N
Id
-M-
Ic
L,.
la
Fig. 12. Equivalent circuit 5 : valves 1, 2, 5, and 6 conducting.
1b
H. cot = 271.5. Commutation is completed. Valves 1, 5, and 6 are con

ducting (see circuit 6, Figure 13). There again is a short circuit on phases
a andc.
-Hic
v"
HFig. 10. Equivalent circuit 3 : valves 1, 2, 3, and 5 conducting.
Lc
vc 5
eb
J.
2
ev. eb = Vb
Id
-wVZ
v$
* See "Effect of alternate assumptions" on page 217.
"b
w-
I.
Id
V2 = V$ = U4
ot = 150. z3 reaches its crest as eb reverses.

F. (ot 210. it reaches its negative crest as ea reverses.
G. cot = 261.5. z3 becomes and remains zero, after which valves 1, 2, and
5 are conducting (see circuit 4, Figure 11). There is now a short circuit
E.
la
330. Voltage across valves 2, 3, and 4 passes through zero and

becomes positive. Any one of them may ignite : assume that valve 3 does
so.* Valves 1, 3, 5, and 6 are conducting (see circuit 7, Figure 14),
reestablishing a three-phase short circuit.
(ot = 397.
i5 becomes zero and remains so. Valves 1, 3 and 6 are con
ducting (see circuit 8, Figure 15). Positive voltage immediately appears
across valve 2, and it ignites. Valves 1, 2, 3, and 6 are now conducting
(see circuit 9, Figure 16), forming three-phase and dc short circuits.
Valve 6 begins to commutate to valve 2.
(ot
* See "Effect of alternate assumptions" ori page 217.
214
6-3
ARCBACK
215
Equations
Algebraic evaluation of the constants of integration is cumbersome with so

many different successive circuits; therefore these constants are denoted by
K's, and their numerical values are given for the example.
1st period, A to D, valves 1, 2, and 3 conducting (Figure 8):
Same as Eqs. (5) to (12) for the previous case, with cos a = 1.
2ndperiod, D to G, valves 1, 2, 3, and 5 conducting (Figure 10):
Fig. 14. Equivalent circuit 7: valves 1, 3, 5, and 6 conducting.
(26)
*"i = 4 = 4a cos (cot - 30) +Kt = 11.6 cos (cot - 30) - 13.9 kA
(27)
(28)
1*2 = 4= 1.8 kA
i3 = ib = /s3 cos (cot - 150) + Kz = 11.6 cos (cot - 150) + 4.2 kA (29)
i5 =ic + 4 -4 [cos (cot + 90) + I] = 11.6(sinco/~ I) kA
(30)
4 = *6 = 0
(31)
V5 =
- 1)2
3rdperiod, G to H, valves 1, 2, 5, and 6 conducting (Figure 12):
(32)
4 = 4 = 4 cos (cot - 30) + K3 = 11.6 cos (cot - 30) - 14.0 kA

see Eq. (37)
*2 = 4 - 4
*3 = 4 = 0
see Eq. (33)
4 =4-4
4 = 4s cos (cot + 30) + K4 = 11.6 cos (cot + 30) - 4.2 kA
(33)
(34)
(35)
(36)
(37)
4th period, H to I, valves 1, 5, and 6 conducting (Figure 13):
ea + ec
va = vc = vp = j
Fig. 16. Equivalent circuit 9: valves 1, 2, 3, and 6 conducting.
K. co7 = 409.3. Commutation is completed. Valves 1, 2, and 3 are con

ducting (circuit 1, Figure 8). There is a short circuit on phases a and b.
L. cot = 450. The voltage across valves 4, 5, and 6 becomes positive, and
one of them ignites.*
* See "Effect of alternate assumptions" on page 217.
(38)
v = eb
(39)
*h = v5 = v6 = 0
(40)
-vd=%eb
4 = 4 = 42 cos (cot - 60) + K5 = 10.0 cos (cot - 60) - 10.9 kA
(41)
% =
'
eb
v2 =v3 =v4 =
(42)
7\
2 16
6-3
(43)
h = h = *4 == 0
Is2 cos (cot + 120) + K6 = 10.0 cos (at + 120) + 12.7 kA

H = ~h = h= 1.8 kA
*5
(44)
Valves:
+25
(45)
+20
(46)
+ 10
217
1,2,3,6
1,2,5,6
Instant:
ARCBACK
GiH
1,2,3
1,2,3,5
1,5,6
1,3,5,6
300'
360'
1,2,3
5th period, / to /, valves 1, 3, 5, and 6 conducting (Figure 14):
ix = ia = isZ cos (at - 30) + K7 = 11.6 cos (at ~ 30)
1.4 kA
h = 7s3[cos (at 150) + 1] = 11.6 [cos (co?
(47)
(48)
i2 = i4 = 0
150) + 1] kA
23.0 kA
15.8 kA
120'
180'
240'
420'
(49)
i5 = ic = /s3 cos (at + 90) + K8 = 11.6 cos (at + 90) + 0.6 kA
(50)
*6 I
d = 1.8 kA
(51)
h = h ~ Id
($2)
-25.5 kA
-20
6th period, J to K, valves 1, 2, 3, and 6 conducting (Figure 16):
+E,
(53)
ix = ia = js3 cos (at 30) + K9 = 11.6 cos (at - 30) 16.8 kA

i2 = ic = /s3 cos (at 90) + K10 = 11.6 cos (at 90) 7.0 kA
(54)
h~ I
d~ h
(5(5)
U = i5=0
i6 =Id- i2
(57)
(55)
360'
180'
360'
-E,
O>
(58)
1th period, Kto L, valves 1, 2, and 3 conducting (Figure 8):
180'
-E,
Use Eqs. (5) to (8), 10), and (12) from the previous case.
(59)
it = ia = ls2 cos at + Ktl - 10.0 cos at 12.4 kA
(60)
i3 == ib = Is2 cos at + K12 = 10-0 cos at + 14.2 kA
Crest Values of Currents. In the first cycle the crest values in the numerical
example are f = 25.5 kA,/3 = 15.8 kA, and i5 = 23.0 kA. Those for and
i5 are higher than the crest currents with successful blocking. The greatest
crest current is that of the valve which arcs back. The fact that |z A| and
i3 are greater at L (at = 450) than at D (at = 90) indicates that still higher
crest values may be reached in the second cycle, and, perhaps, in succeeding
cycles. However, decay of the dc components due to resistance, which was
neglected, could alter this inference.
Fig. 17. Arcback in uncontrolled rectifier without bypassing: (a) valve currents; (b)
alternating emf's and voltages of dc bridge terminals with respect to ac neutral point N;
(c) valve voltages.
Effect of Alternate Assumptions. At instants D, /, and L, any one of three

valves could ignite, and a certain one was assumed to do so. If another one
were chosen at D, the course of events and the crest values of currents would
be slightly different. Regardless of the choice at D, however, the same valves
(1, 5, and 6) would be conducting from Hto I. Also, regardless of the choice
218
6-3
ARCBACK
219
at /, the same valves (1, 2, and 3) would be conducting from Kto L. Detailed
analysis of the other choices is left as an exercise (Problems 3 and 4).
Table 1 gives these crest values in terms of Is2 , Isl , and E/X. For the numeri
cal example, with hz = 10.0 kA, these crest values are 20.0 and 27.3 kA,
respectively. E[X =
is the effective value of three-phase short-circuit
Maximum Crest Currents
current.
Study of the foregoing analyses of arcbacks shows that the crest currents
depend somewhat on Id and a, being greatest for Id = 0 and a = 0, the condi
tions for a fully offset cosine wave. Figure 18 illustrates these conditions for
two kinds of arcback.
For the controlled arcback the crest current in valves 1 and 3 in the first
cycle is clearly
(61)
ti=2Is2
/s3/>/2
Table 1. Maximum Crest Currents in Arcbacks

Type of Arcback
f//.3
f//
iKEIX)
Controlled
Uncontrolled
1.73
2.37
2.00
2.73
2.45
3.35
and for the uncontrolled arcback it is
J1=7i2 + 1.57s3=7i2(1 + V3)
(62)
An inverter valve can arc back only in the small time interval in which the
voltage across it is negative, that is, with cot between 180 y and 180. The in
verse current is of short duration, and its crest value is small, perhaps even
less than rated current. Because the valve suffers very little excess heating in
this case and because arcing back is a random phenomenon, there is little
likelihood of the valve's arcing back again after its next commutation or of a
consequent arcback. Inverter arcbacks are cleared without need of blocking
ait
270
180
90
Arcback in Inverters
_1L
-1
and bypassing.
The phenomena after an arcback are similar to those after a commutation
failure (Section 6-5).
-2
(a)
Detection and Cure of Arcbacks

210'
90" 120'
Is3[cos (ait
300'
30) 1.366]
(b)
Fig. 18. Computation of crest arcback current in valve 1: (a) with blocking and bypassing;
(b) with no blocking nor bypassing.
The most obvious way of detecting an arcback is by reverse current in a

valve. A magnetic core linked with the valve current is magnetized to satura
tion by the normal, forward current. A big reverse current caused by an
arcback magnetizes the core to saturation in the opposite direction, and the
change of flux in the core induces a pulse of emf in a secondary winding.
The drawback of using such a device to initiate blocking and bypassing
of a bridge is that one is required for each valve and that the signal thus
derived must be transmitted to ground from a point at a very high voltage
with respect to ground. Such a signal, however, is suitable for operating an
arcback counter for each valve. The counters can be at valve potential and
can be read from a safe distance.
Another way of detecting arcbacks is to compare instantaneously the
highest rectified alternating current of the three phases on the valve side of
the transformer (ac terminals of the bridge) with the direct current. In normal
operation the two currents are equal, but during an arcback the rectified
220
6-4
alternating current greatly exceeds the direct current. The difference can be
used to give the blocking signal. The alternating currents may be taken from
the secondary windings of current transformers in the three transformer
bushings or valve-hall entrance bushings, while the direct current is taken
from a dc current transformer at or near ground potential. The secondary
alternating currents and rectifiers are required for each bridge and so are
fewer than the number of valves. The same direct current can be used for all
the bridges of a pole.
As will appear later (Section 6-9), the same comparison of direct current
and rectified alternating current can be used to indicate various inverter
DC Short Circuit with Blocking and Bypassing
|
|
i
Reduction of Arcbacks by Series Connection of Valves
If the probability of arcback in one valve were independent of the arcing

back of another valve, the series connection of two valves in each arm of the
bridge circuit would greatly decrease the probability of both valves' arcing
back at the same time and thus short-circuiting the ac supply. An arcback
in one valve of a series pair, however, puts the entire inverse voltage on the
other valve instead of dividing it equally between the two valves. The good
valve is thus subjected to twice normal voltage, and such voltage may make it
more likely to arc back. If each valve were capable of withstanding the full
voltage all the time, the frequency of their both arcing back would be very
small, but the aggregate rating of all the valves in a bridge would be twice that
of the usual bridge with one valve per arm. Perhaps a compromise should be
adopted in which the normal voltage rating of each valve in a series pair is
between one-half and one times the inverse voltage across the pair.
The use of two valves in series in each bridge arm is common, perhaps
even standard, in Russia but has not yet been adopted to any extent elsewhere.
6-4 SHORT CIRCUIT ON A RECTIFIER
A short circuit on the line side of the dc reactor is a line fault and is con
sidered in Section 7-6. A short circuit on the valve side of the dc reactor is a
converter fault and is considered here. It is assumed to be a fault on the dc
terminals of one bridge. It is similar in many respects to an arcback, and the
protection against it is the same as that against an arcback, that is, blocking of
221
the main valves and unblocking of the bypass valve whenever the rectified
alternating currents of a bridge greatly exceed the direct current.
faults.
When an arcback is detected, the main bridge valves should be blocked,

and the bypass valve should be unblocked as soon as possible. The bridge
should remain bypassed until the faulty valve is capable of withstanding
normal inverse voltage; then the bridge may be returned to service. The usual
times of bypassing range from 0.5 to 1 sec.
SHORT CIRCUIT ON A RECTIFIER
I
?
1
j
j
Assume that the fault occurs while valves 2 and 3 are conducting and that
the remaining main valves are blocked before valve 4 is due to fire. There is
a phase-to-phase short circuit on phases b and c through valves 2 and 3 and
the fault. This continues for about half a cycle, during which the valve cur
rents reach a crest value of approximately hi and then decline to zero. In the
absence of arcbacks in both valves, which is very improbable, these currents
cannot reverse, and so the valves extinguish and are blocked. The bypass
valve cannot ignite until the fault current reverses, which occurs a few degrees
before the currents in valves 2 and 3 become zero, and can do so only if the
fault is an arcing one. Before then the polarity of voltage across the bypass
valve, due to the fault arc, is negative.
A dc short circuit is bound to occur, if it does at all, when the phase-tophase emf associated with the two conducting valves is near its maximum
value, and, consequently, the dc offset of the sinusoidal short-circuit current
is small. An arcback, on the other hand, occurring soon after the end of
commutation when the commutating voltage is small, is likely to produce an
almost fully offset sine wave of current.
Another difference between an arcback and a dc short circuit is that in an
arcback one of the two valves through which the short circuit occurs is
conducting backward; in a dc short circuit both valves are conducting
forward.
In multianode valves, still another difference is that the dc short-circuit
current is shared by all the anodes.
DC Short Circuit on Uncontrolled Rectifier
In the absence of any blocking, a three-phase short circuit immediately

develops ; for example, if the dc short circuit occurs while valves 2 and 3 are
conducting, a positive voltage immediately appears across valve 4, igniting
that valve. The three phases are now connected through valves 2, 3, and 4
and the dc short circuit. Each valve carries the alternating current of one
phase, lagging the respective phase emf by 90. Each phase current has, in
addition to the alternating component, a direct component (or offset) depen
dent on conditions when the dc short circuit occurs. The maximum crest
current, caused by full offset, is 2/s3 and is not so great as that possible
during an arcback on an uncontrolled rectifier.
As long as a three-phase short circuit endures, the voltage across every
valve, conducting or nonconducting, is limited to the arc voltage and is
T*
\
222
6-5
insufficient to ignite any of the nonconducting valves. When the current in any
valve decreases to zero, however, and is extinguished, a positive voltage
appears instantly across the other valve on the same phase, igniting it; for
example, when valve 2 extinguishes, valve 5 ignites, both of these valves being
connected to phase c. Thus the three-phase short circuit is maintained except
during extremely brief time intervals when there is a phase-to-phase short
circuit.
The dc components decay but usually too slowly to change the crest
current significantly during the first cycle.
COMMUTATION FAILURE
223
as 15 and the ignition angle as 40. For simplicity in computation, the failure
is assumed to result from late firing of valve 3, with direct current and alter
nating voltage remaining unchanged, although in practice changes in the last
two quantities are likely to be involved either as causes or effects. Is2 is
assumed to be 10 kA and Idto be 2 kA.
-L
(a)
Maximum Crest Currents in DC Short Circuits
These are given below in the same manner as the arcback currents in
Table 1. They are less than the corresponding arcback currents.
Table 2. Crest Currents in DC Short Circuits
Type
Controlled
Uncontrolled
///.3
Hhz
0.87
1.00
(loo)' 2.31
mix)
(b)
2.83
6-5 COMMUTATION FAILURE
(c)
180
20
Causes
The commonest misoperation of an inverter is a failure of commutation.

A true commutation failure is not due to any misoperation of a valve but
to conditions in the ac or dc circuits outside of the bridge in which the failure
occurs or to inadequate control of the time of ignition. Because of increased
direct current, low alternating voltage (caused perhaps by an ac short cir
cuit), late ignition, or a combination of these, commutation is not completed
before the alternating commutation emf reverses. Thereafter the direct cur
rent is shifted back from the incoming valve to the valve that was expected
to go out. It is shown below that nearly all inverter valve faults lead to
results similar to those caused by a commutation failure.
Instant
DE G
M N
Valves
conducting
Fig. 19. Wave forms of voltages and currents with failure of commutation of an inverter :
(a) phase emf's and dc pole voltages with respect to neutral of ac source; (b) direct voltage,
and (c) valve currents. Valve 1 fails to commutate to valve 3.
Narrative
Analysis
A failure of commutation from valve 1 to valve 3 is analyzed with refer

ence to the wave forms in Figure 19 and to the various simplified diagrams of
an inverter bridge mentioned hereafter. The normal extinction angle is taken
B 105
C 145
Normal commutation begins from valve 6 to valve 2.

Commutation is completed. Valves 1 and 2 are still conducting.
Valve 3 fires 5 late, and commutation begins from valve 1 to
valve 3.
.
224
6-5 COMMUTATION FAILURE
The commutating voltage eb ea reverses before commutation is

completed, denoting a failure of commutation. Reverse transfer
of current begins.
E 200 Before reverse commutation from valve 3 to valve 1 is completed,
valve 4 ignites, and commutation begins from valve 2 to valve 4.
Valves 1, 2, 3, and 4 are conducting; this is a condition of double
overlap similar to that discussed in Section 3-3 and shown in the
circuit in Figure 14 in Chapter 3. Both the ac and the dc terminals
of the bridge are short-circuited; hence all voltages are zero.
F 207.4 Reverse commutation from valve 3 to valve 1 is completed.
Valves 1, 2, and 4 are conducting (see Figure 20). Commutation
from valve 2 to valve 4 is still under way. The dc terminals are still
short-circuited through valves 1 and 4. There is a short circuit on
ac terminals a and c, but that on terminal b has ended.
225
D 180
Fig. 22. Equivalent circuit when valves 1, 4, and 6 are conducting.
J 345
K 360c
L 380
M 440
Fig. 20. Equivalent circuit when valves 1, 2, and 4 are conducting.
G 214.8 Valve 2 extinguishes, completing commutation to valve 4. Valves

1 and 4 are conducting, short-circuiting the dc terminals. Now,
however, there are no alternating currents. See Figure 21.
ve
ec
<S)
va
ea
Vb
eb
O-
Fig. 21. Equivalent circuit when valves 1 and 4 are conducting.
H 260
/
320
Valve 5 is pulsed, but it cannot ignite because the voltage across it

is negative.
Valve 6 ignites. Commutation from valve 4 to valve 6 begins.
Valves 1, 4, and 6 are conducting (Figure 22). The dc terminals
N 465
P 500
Q 525
are still short-circuited through valves 1 and 4, and ac terminals

a and b are short-circuited through valves 4 and 6.
Valve 4 extinguishes, completing commutation to valve 6 and
removing the dc short circuit. The direct voltage becomes positive

for a short while, which is abnormal for inverter operation.
Valves 1 and 6 still conduct: this is a normal combination but
occurs at an abnormal part of the cycle because valve 1 is not due
to fire until instant L.
The direct voltage reverses, becoming negative, that is, having
normal polarity for inversion.
Valve 1 is pulsed but is already conducting.
Valve 2 ignites, and normal commutation from valve 6 to valve 2
begins.
Valve 6 extinguishes, completing commutation.
Valve 3 ignites, and commutation from valve 1 to valve 3 begins.
Such commutation failed in the previous cycle, but the error in
ignition time that caused failure is assumed to have been corrected.
Valve 1 extinguishes, completing commutation. The inverter is
now operating normally.
Symptoms
From Eto / (a period of 145) the dc terminals are short-circuited, and the
direct voltage of the bridge is zero; for 205 after E it is below normal.
During this time, the direct current, which in the foregoing analysis was
assumed constant, would really increase at a rate determined mainly by the
inductance of the dc reactor. From G to I
(105), there is no current in any
of the ac terminals. Thus, there is a time interval in which the direct current
is greater than the rectified alternating current, which is a condition opposite
to that during arcback of a rectifier valve.
226
6-6
Cure
After the occurrence of a commutation failure, the succeeding commuta
tion is initiated earlier by the C.E.A. control system (Section 5-10) and is
usually successful. If the failure is caused by low alternating voltage, the
reappearance of normal alternating voltage helps prevent further failures. In
the event of persistent commutation failures, the bridge in which they occur
should be blocked and bypassed.
Double Commutation Failure
227
is advanced sufficiently by the C.E.A. control, or double overlap, as shown in

Figure 19c, E to F, hastens that commutation.
If a double commutation failure occurs, it is self-curing, as is a single
commutation failure.
Consequent commutation failures
These are discussed in Section 7-2, page 236.
6-6 ARCTHROUGH
This is the failure of two successive commutations in the same cycle.

If the unsuccessful commutation from valve 1 to valve 3 and back to valve 1
is completed before valve 4 fires, and if the conditions that caused the first
failure persist, a second failure can occur in the commutation from valve 2
to valve 4, causing the direct current to be transferred back to valve 2.
Valves 1 and 2 are now left conducting until the time in the next cycle when
they would normally be conducting again. By this time, the conditions
that caused commutation failure may no longer exist, and normal com
mutations from 1 to 3 and from 2 to 4, and so on, may take place. During
the time that only valves 1 and 2 are conducting, the alternating voltages
of terminals a and c care connected to dc terminals p and n, respectively.
Thus an alternating voltage is impressed on the dc line. The voltage at the
dc terminals of the bridge in question is reversed for approximately onehalf cycle between instants Hand L in Figure 23. During this time, the inver
ter bridge works as a rectifier and aids the rectifier at the other end of the line
in circulating current in the line. Therefore, the current increases more
during a double commutation failure than during a single one.
Fortunately double commutation failures seldom occur. Usually, after one
commutation failure, either the firing angle initiating the next commutation
240
ARCTHROUGH
300
360
420
480
540 600
Fig. 23. Direct voltage during double commutation failure.
Causes
Arcthrough is conduction during a scheduled blocking period. It can be
caused by failure of the negative grid bias, by a defect in the grid circuit, by
the too early occurrence of a positive grid pulse, or by a sufficiently great
positive transient overvoltage on the grid or anode.

Arcthrough in a Rectifier
This arcthrough merely reduces the ignition delay angle from its normal
value (typically 15) to a smaller value or zero. Its effect on the wave shapes
of current and voltage is small. A persistent arcthrough in one valve, however,
introduces a dc component into the transformer currents, because this valve
conducts longer than does the other valve of the pair connected to the same
ac terminal, and the preceding valve conducts for a shorter time than its
paired valve. For the same reason, the harmonic components of both the
alternating currents and the direct voltage are altered.
Arcthrough in an Inverter
Arcthrough in an inverter produces similar effects to those produced by

commutation failure. Suppose that valve 3 arcs through at point C (Figure 24)
because of failure of blocking when there is a high positive voltage across
this valve. At this time valves 1 and 6 are already conducting. Valves 3 and 6
form a short circuit across the dc terminals of the bridge, making the direct
voltage across the bridge drop to zero and barring the direct current from
the ac terminals. After commutation from valve 6 to valve 2 is completed at
F, the short circuit is removed and direct voltage is restored, although it is
initially in the positive direction, that is, abnormal polarity for an inverter.
The direct voltage builds up to its normal negative value at I, and normal
operation is temporarily restored. If the cause of failure of blocking of valve
228
6-8
D F
STU V
MN
J
I
WX
QUENCHING
229
When valve 3 ignites, a dc short circuit is established through valves 3 and 6,

making the direct voltage vanish. The short circuit is removed on completion
of normal commutation from valve 3 to valve 5, but usually it reappears in
every cycle.
(a)
Misfire in Rectifier
-U
Vd
60'
240
360
420
480
540
(b)
Fig. 24. Arcthrough of valve 3 of inverter: (a) valve currents; (6) direct voltage. At C
valve 3 arcs through while valve 6 is conducting; at Mit arcs through again. Valve 6 ignites
at P and extinguishes at F and T.
3 has not been removed, however, that valve arcs through again at M as soon
as the voltage across it becomes positive. When valve 6 ignites at P, the short
circuit of the dc terminals is reestablished and persists longer than it did
previously, until valve 6 extinguishes again at T, after which the direct voltage
is reestablished. The short circuit reoccurs once per cycle until the cause of
arcthrough is removed or until the bridge is bypassed.
Protection
The phenomena caused by misfire of a valve in a rectifier are similar to

those in an inverter except that (a) the short circuit exists for a shorter part
of each cycle, typically for 75 instead of 135, and (b) there is a small jump of
direct voltage at the beginning of the short circuit instead of at the end of it
and there is a very great jump at the end of the short circuit.
Protection
See Section 6-9, page 231.
6-8 QUENCHING
Causes
Quenching of the arc during conduction is a result of insufficient ionization.

One cause of this is too low a temperature of the valve tank.
See Section 6-9, page 231.
Prevention
6-7 MISFIRE
Causes
Misfire is the failure of a valve to ignite. It may be caused by failure of a
grid pulse or by extinction of the cathode spot by failure of the excitation
system or by low or reversed net cathode current.
Analysis of Misfire in Inverter
Suppose that valve 2 fails to ignite. Then obviously commutation from valve
6 to valve 2 cannot take place, and valve 6 continues to conduct. The con
sequences are much the same as if the commutation had begun but failed.
The tank temperature is automatically controlled to the right value.

Effects
Quenching of the arc soon after conduction starts has almost the same
effects as failure to begin conduction (misfire). These include a short circuit
of the dc terminals of the bridge for a short while, with the consequent col
lapse of direct voltage and interruption of the alternating current. In addition,
like current chopping in any inductive circuit, quenching may give a high
voltage across the break, that is, across the valve, as well as across lumped
inductances. See Section 7-8, "Protection against Current Chopping,"
page 286.
230
6-10 CONSEQUENTIAL FAULTS IN RECTIFIER
23 1
Persistence
6-9 GENERALIZATION OF INVERTER FAULTS AND CERTAIN

RECTIFIER FAULTS
Commutation failures and arcbacks in inverters may be expected to appear

only in one cycle and to be self-clearing. Arcthrough, misfire, and quenching,
depending on whether the cause is temporary or permanent, may clear
themselves in one cycle or persist for several or many cycles.
Similarities
All faults in inverters result in a short circuit of the dc terminals of a bridge

for a certain time interval, accompanied by the collapse of the direct voltage
of the bridge and, during most of this time interval, the absence of alter
nating current in all phases. This occurs, typically, for an interval of 135
in one cycle or in any number of successive cycles. These faults include :
Misfire
Arcback
Arcthrough
Quenching
Single commutation failure
Figure 25 shows the wave forms of direct voltage of the affected bridge for
each of these faults. The wave forms differ only in minor details, some of
which are caused by the calculations having been made for different values of
direct current. Part of the wave form for misfire, drawn in broken lines, has
fewer notches than the corresponding parts of the other three curves. The
first cycle of arcthrough (Figure 24b) may differ from the subsequent cycles,
depending on the instant of the cycle when arcthrough occurs.
The direct voltage during double commutation failure (Figure 23) has a
substantially different wave form from those of the other inverter faults in
that the direct voltage reverses for about half a cycle. It, too, however, has an
interval (50 in the example) of dc short circuit, and during this interval not
only is the direct voltage zero but also the maximum absolute value of phase
current is less than the direct current.
Misfire or arc quenching in a rectifier exhibits the same characteristics as
does a dc short circuit, but these are of shorter duration than are most inverter
faults.
60" intervals
co t
3,4
3,4
Fig. 25. Direct voltage during inverter faults: (1) arcback, /d = 0.18; (2) single commuta
tion failure, Is 0.2; (3) arcthrough, I'd = 0.1 ; (4) misfire or quenching, Is = 0.1.
Detection of Inverter Faults and Protection against Them

Two methods of detection suggest themselves: (a) vanishing of direct
voltage and (b) maximum rectified current of all phases being less than the
direct current. The latter condition is the opposite of that noted in rectifier
arcback, where the alternating currents of two phases become much greater
than the direct current. The method of detection usually employed compares
the alternating and direct currents. A time-delay relay, called the persistent
arcthrough relay, is actuated by an excess of direct current over alternating
current and causes unblocking of the bypass valve and blocking of the main
valves of the affected bridge.
6-10 CONSEQUENTIAL FAULTS IN RECTIFIER
Cause
During an arcback at least two valves carry excessive current the valve
that arcs back and carries reverse current and another valve that carries
forward current. Let us call the latter valve the partner valve. These two
valves constitute a phase-to-phase short circuit on the transformer. The short
circuit cannot be extinguished by the bypass valve alone.
The arcing-back valve is out of control and must be relieved of current as
soon as possible, so that it may recover. The partner valve, in spite of having
been subjected to excessive current, can usually function normally; that is, it
can cease conducting when its current falls to zero and it can withstand there
after a voltage of either polarity. The approximately equal sharing of the
overcurrent by the several anodes of the partner valve favors its normal
behavior, in contrast to the other valve, where the arcback is usually con
fined to one anode that carries the whole short-circuit current. Normal
behavior of the partner valve ends the short circuit.
Occasionally, however, the overheated partner valve misoperates and fails
to interrupt the short circuit. Its misoperation, being a consequence of the
arcback in the other valve, is called a consequentialfault.
232
PROBLEMS
Such faults seldom, if ever, occurred in the earlier dc schemes, Gotland

and Channel Crossing, but they became a more serious problem in the New
Zealand, Konti-Skan, and Sakuma schemes, with their higher-power valves.
Intensive research was undertaken for improving the performance of
valves.9,10
Types
The currents during two kinds of consequential faults in valve 3, following
an arcback in valve 1, are shown in Figure 26. Until time H, when the cur
rents in both valves become zero simultaneously, these currents are identical
to those of the simple arcback previously described (Section 6-3, page 207,
and Figure 7).
ABC
E F
\60
H IJ
'ii
Protection
CAB's are more serious than simple arcbacks, because they cannot be
cleared by grid control. They must be cleared by a slower method; hence the
valves involved become hotter and may, therefore, require more frequent
overhaul. The reduction of dc power lasts longer. In order to clear a con
sequential fault, the transformer of the affected bridge must be deenergized
by an ac circuit breaker. Reclosure should be delayed about 40 sec. Not only
is this much slower than grid control, but also it causes a greater reduction in
transmitted power than that of the affected bridge unless a circuit breaker is
provided for each bridge. A circuit breaker per bridge is more expensive
than a circuit breaker per pole and might not be required if consequential
faults were very infrequent. If the power rating per bridge is great, however,
a circuit breaker per bridge might be justified on other grounds.
//
u
-J-
h
120
LM
CAT-/
77
233
110 f
\ /AB
//
\
i
240
PROBLEMS
//CAB
l\
300/ 3 10
//
V\
C>N420
1
480
\\
\\
5 40
i
600
660//
wt, deg
\\
//
//
\\
\\
\\
720
//
//
Fig. 26. Consequential arcback (CAB) or consequential arcthrough (CAT) of rectifier

valve 3 after arcback (AB) of valve 1.
The first kind of fault is properly called a consequential arcthrough (CAT).

The currents remain zero until point / at cot 360 (Figure 26), when the
voltage across valve 3 first becomes positive, whereupon this valve arcs
through and carries forward current. At the same time valve 1 arcs back
again and carries reverse current. These current waves are fully offset and
rise to crests of almost 2/s2 if the effect of resistance is small. Thus the second
crests at K may be higher than the first ones at G.
In the second kind of fault the consequential arcback (CAB)valve 3
arcs back almost as soon as its voltage becomes negative, and valve 1 arcs
through at the same time. Thus both valves continue to conduct. In the
absence of decay of the dc component, the second crests are equal to the
first ones.
The difference between these two kinds of fault is not important, and some
writers call both of them CAB's, Most CAB's occur within a very short time
(0.3 ms) of point H or J, but they can occur at any time within several cycles.
"Analyze," as used in the following problem statements, means to find

the pertinent equivalent circuits and the corresponding equations of con
verter voltages and currents required for drawing accurate graphs of the
following quantities as functions of cot : vp,vni vd, and the valve currents. The
time covered should be from 1 to 1.5 cycles, beginning with one normal
commutation before abnormal operation commences.
1. Analyze commutation from a normally operating rectifier bridge to the
bypass valve if the latter is unblocked just after commutation has begun
from valve 1 to valve 3. Assume Id = 2 kA, Is2 = 10 kA, a = 15.
Blocking and unblocking occur at cot = 0.
2. Analyze in detail the transfer of current from the bypass valve of an
inverter bridge to the main valves.
3. Analyze an uncontrolled bridge converter after an arcback of valve 1 on
the assumption that valve 4 ignites at D and valve 2 at I.
4. Like Problem 3 except that valve 6 ignites at D and valve 4 at I.
5. Find the maximum current during an arcback occurring in an inverter
valve immediately after extinction at y = 15. Is2 = 10.0 kA,
Id= 1.8 kA.
6. Analyze commutation from a normally operating inverter bridge to the
bypass valve if the latter is unblocked and the main valves are blocked
just after commutation has begun from valve 1 to valve 3.
7. Analyze the performance of a rectifier bridge after the main valves are
blocked without unblocking the bypass valve.
234
8. Analyze arcback in an inverter. Assume that valve 1 arcs back at the end
of its commutation with valve 3 at y = 15 and that there is no blocking.
Id = 1.8 kA, Is2 = 10.0 kA.
9. Analyze a short circuit on the dc terminals of a bridge of an uncontrolled
rectifier with no bypass valve under the following assumptions: The
short circuit occurs at cot = 35. Id = 1.0 kA, Is2 = 10.0 kA.
10. Same as Problem 9 except that the rectifier has grid control with a 15
and has a bypass valve. Blocking of the main valves and unblocking
of the bypass valve occur at cot = 45.
11. Analyze a double commutation failure in an inverter that normally
operates with y = 15,u = 15, and, hence, = 30. Is2 = 10.0 kA. The
failure occurs because valves 3 and 4 fire 5 late, that is, with fi = 25.
12. Analyze arcthrough of valve 3 of an inverter at cot = 75 while valves 1
and 2 are conducting. Id= 1.0 kA, Is2 10.0 kA, y = 15.
13. Analyze misfire of valve 2 of an inverter for which Id = 1.0 kA,
Is2 = 10.0 kA, and y = 15.
14. Analyze misfire of valve 4 of a rectifier operating with Id= 1.0 kA,
/s2 = 10.0 kA, and oc = 15.
7
Protection
7-1 GENERAL
Protection of ac systems is usually divided into two principal categories :

1. Overvoltage protection, accomplished by means of overhead shield
wires, protective gaps, and lightning arresters
2. Overcurrent protection, accomplished by fuses, circuit breakers, pro
tective relays, and current-limiting reactors
Protection of hv dc systems may be divided into the following four categories :

BIBLIOGRAPHY
1. "Protection of H.V. D.C. Systems," Chapter 6 of High Voltage Direct Current Power
Transmission, by Colin Adamson and N. G. Hingorani, Garraway, London, 1960.
2. "The Protection of H.V. D.C. Convertors and Systems," by C. Adamson, Chapter 6
of High Voltage Direct Current Convertors and Systems, edited by B. J. Cory, Macdonald, London, 1965.
3. " Some Design Aspects of the Cross-Channel Power Link," by L.A. Harris, Chapter 8
4.
5.
6.
7.
8.
9.
10.
of Cory (Ref. 2).

Harris (Ref. 3), Section 8.7, pp. 192-193.
Harris (Ref. 3), Section 8.12.7, pp. 206-207.
"Bypass Operation in Bridge Convertors for High-Voltage D.C. Transmission," by
John J. Vithayathil, Proc. I.E.E., Vol. 112, pp. 359-365, February 1965.
" Sequential Arrangements for the Elimination of By-Pass Valves in H.V.D.C. Con
verters," by M. Morales, Direct Current, Vol. 10, pp. 164-167, 176, November 1965.
"A Digital Protection Scheme for HVDC Convertors," by J. Reeve, I.E.E. Con
ference Publication No. 22, High Voltage D.C. Transmission, Manchester, Sept. 19-23,
1966, Part 1, Paper No. 62, pp. 304-305. Outlines a method of detecting valve faults
and of distinguishing the kind of fault and the valves involved.
"Consequent Arc Backs and Consequent Arc Throughs inHV Mercury-Arc Valves,"
by Sten S. B. Berneryd, Birger Funke, and Thomas S. Robinson, I.E.E.E. Trans, on
P.A.&S., Vol. 89, No. 5/6, pp. 740-746, May/June 1970. Disc., pp. 746-747.
"Overcurrent Surge Behaviour of HVDC Mercury-Arc Valves and Improvement of
Their Consequential Arc-Back Performance," by Harry Gibson, ibid., pp. 747-752.
Disc., pp. 752-754.
1. Overvoltage protection, accomplished similarly to that of an ac system,

although with some differences
2. Overcurrent protection, accomplished by control of the valves
3. Damper circuits
4. Dc reactors
Except as to overvoltage protection, protection of hv dc systems is entirely

different from protection of hv ac systems.
The categories of dc protection are considered in reverse order.
7-2 DC REACTORS
These reactors, usually having inductances of 0.4 to 1.0 H, are connected

in series with each pole of each converter station. They serve the following
purposes :
1. To prevent consequent commutation failures in the inverter by limiting

the rate of increase of direct current during commutation in one bridge when
the direct voltage of another bridge collapses
2. To decrease the incidence of commutation failures in the inverter during
dips in alternating voltage
235
236
7-2
PROTECTION
3. To decrease harmonic voltages and currents in the dc line

4. To smooth the ripple in the direct current sufficiently to prevent the
current from becoming discontinuous or almost so at light loads
5. To limit current in the bypass valves due to discharge of the shunt
capacitances of the dc line and terminal equipment in the event that all the
bypass valves of one pole are fired simultaneously
6. To limit the crest current in the rectifier due to a short circuit on the
dc line
The inductance of such a reactor is usually determined from considerations
1 and 4. If possible, the inductance should be chosen also so as to prevent
resonance of the dc circuit at the power frequency.
'
Prevention of Consequent Commutation Failures
In Chapter 6 it was shown that nearly all inverter faults lead to the collapse
of the direct voltage of a bridge. This voltage remains at zero for about
three-eighths of a cycle, which is several times greater than the time available
for commutation. The countervoltage of the affected inverter pole falls to
(b 1)fb times its normal value if there are b bridges in series; for example,
if there are two bridges per pole, the voltage of the affected pole drops to
half of its normal value. Consequently, the direct current through the inverter
rises, tending toward a new steady value such that the new RIdrop in the line
is the difference between the rectifier voltage and the reduced inverter voltage.
The rise of current prolongs commutation and, if excessive, causes commu
tation failure, which, in turn, leads to an additional large decrease of inverter
voltage, which further increases the rate of rise of direct current. Thus the
failure is likely to spread to all the bridges of the affected pole, decreasing
the power transmitted by the dc line to half of its prefault value.
It is good practice to limit the magnitude of such a disturbance to the dc
and ac systems by confining it to one bridge. To do so, the rise in direct
current during a commutation in one bridge, caused by a collapse of direct
voltage in another bridge, must be limited to a value that does not cause
commutation failure.
The factors governing the amount and rate of rise of direct current are,
in addition to the dc reactor, the surge impedance of the dc line, lumped
capacitances at the inverter station, and current regulators.
The surge impedance Zs limits the increase of direct current A Id drawn by
the inverter from the line due to a decrease of inverter voltage AVd to
A Id = A VdjZs. This limitation holds for the time of round-trip wave travel on
the line. On a short line the reflected wave returns so soon, and on a cable
the surge impedance is so low, that the rise of current is not adequately limited.
DC REACTORS
237
Even the surge impedance of a long overhead line would not limit the
discharge of lumped shunt capacitances, such as the surge capacitor see
" Protection against Lightning," page 285 through the inverter.
The current regulator at the inverter only prevents a decrease of direct
current in excess of the current margin. The current regulator at the rectifier
has too long a time constant and, even if it were instantaneous, would be
powerless to prevent discharge of shunt capacitance of the line and of the
inverter station through the inverter.
Practically speaking, then, a dc reactor is always required, and its induetance is the principal factor limiting the rate of rise of direct current. The
magnitude of rise during commutation depends on the overlap time.
In calculations of the inductance required for sufficiently limiting the rise
of direct current after the collapse of the direct voltage of a bridge so that a
commutation failure does not ensue, the following assumptions are made :
1. The direct voltage on the line side of the dc reactor remains constant
initial value. This assumption may be slightly pessimistic: its corollary
is that only the dc reactor limits the rise of direct current.
2. The alternating voltage at the inverter remains constant. A sudden drop
in this voltage would make commutation failure more likely, but it is improb
able that such a drop would coincide with misfire or arcthrough of a valve.
3. The tap changer does not move.
4. Initially the direct current has its rated value Idn. This requires greater
overlap than would a smaller value of Id and thus gives more time in which
Id can increase.
5. The collapse of direct voltage occurs simultaneously with the beginning
of commutation. This is the worst time for it to occur. If it occurred earlier,
the C.E.A. control could advance the ignition angle some. If it occurred later,
there would be less time left in which the direct current could increase before
the completion of commutation.
6. Because of error in the C.E.A. control, ignition is assumed to be 1
I
j
j
j
at its
late.
Let the normal extinction angle be yn - 16, and let the minimum extinc
tion angle to be permitted as a result of collapse of direct voltage be
= 8.*
The normal ignition advance angle ft is calculated from Eq. (96) in
Chapter 3 :
/
cos ft, = cos y--~
(1)
52
* The deionization time of high-voltage, high-current mercury-arc valves corresponds to
an angle of 5 to 8.
'
jl
238
PROTECTION
7-2
and 1 is subtracted from it. Next, the greatest permissible current at the
end of commutation, that which reduces y to ym, is computed from a valid
modification of Eq. (1) in which Idn is replaced by the arithmetic mean of
the values of Id at the beginning and end of commutation, hp and /dv, respec
tively. The modified equation is :
(2)
= /,2(cosy-cos/?)
10.0 (cos 8 - cos 37.6) -2 x 1.80

= 20(0.990 - 0.792) - 3.60
= 20 x 0.198 - 3.60 = 3.96 - 3.60 = 0.36 kA (20%)
AId = 2x
By Eq. (6),
37.6- 8.0
ym,
hp = hn,
and
hy = hn + &h-
A Id = 2/s3[cos ym - cos (fin - 1)] - 2Idn
=L37X1"
S6C
By Eq. (5),
Then
(3)
Since the current rises linearly at the rate
Example
dld Ald AVd

dt
At
Ld
(4)
AVsAt
La=AL
(5)
then
where
200 x 1.37 x
= 036
10-3
= -76 H Ans-
Find the smallest fraction of normal value to which the balanced alter
nating voltage at the inverter in Example 1 can fall suddenly without causing
a commutation failure. Assume that the fall of voltage occurs, as in
Example 1, immediately after the beginning of a commutation at fi = 37.6
with Id = 1.8 kA. Assume also that a commutation failure occurs if y < 1.0.
SOLUTION
1 - y*
At fin-
360/
Example
29.6
W60=600
y
239
By Eq. (3),
Let us put
fi=fin-\,
DC REACTORS
(6)
Find the inductance of the dc reactor required to prevent consequent

commutation failure in the inverter described below:
2
Number of bridges per pole
Rated voltage per bridge
200 kV
Rated current
1.80 kA
10.0 kA
7s2
60 Hz
Frequency
SOLUTION
By Eq. (1),
cos fin = cos 16 - 1.8/10.0 = 0.961 - 0.180 = 0.781

= 38.6
1
fin = 37.6
Commutation is made more difficult not only by increase A Idof the current
to be commutated but also by a decrease in the commutating voltage, which
is proportional to Vd0. Initially
'
dO ~
2Va
cos y
+ cos fi
2 x 400
cos 16 + cos 38.6C
800
= 459 kV
0.961 + 0.781
Let the alternating voltage fall from E to kE. Simultaneously, Vd0 falls to
kVd0 and Is2 falls to kls2. The direct terminal voltage falls from =400kVto
E2S
v'-kv
yd~KVdO
= k 459
= k 459
y'
+ cos F
2
cos 1.0 + cos37.6c
1.000 +0.792
---= k 411
7-2
240
PROTECTION
The drop in Vd is
AVd = Vd - V'd = 400 - k 411 kV

jj
The time that commutation lasts is
360
\f)
J = 1.70xl0- sec
360 x 60
The increase in direct current during this time is
W = (400 -.411) 1.70 x 10 =

0.761
Ld
Another expression for A Id is
A Id = 1kls2 {cos f - cos/?') - 21d0
A/id
':
= 2k 10.0 (1.000 - 0.792) - 2 x 1.80

= 20Ar x 0.208 - 3.60 = 4.16k ~ 3.60
We may equate the two expressions for A Id and solve for k\

4.16k - 3.60 = 0.893 - 0.919k
5.08k = 4.49
k = 0.885 Ans.
As a check, let us calculate A Idby putting this value of k into both expressions :
AId = 4.16 x 0.885 - 3.60 = 3.68 - 3.60 = 0.08 kA = 80 A

A 7d = 0.893 - 0.919 x 0.885 = 0.893 - 0.813 = 0.080 kA {Check)
The increase in direct current is only 4.5%, and the decrease in alternating
voltage is only 11.5%.
Ripple in Direct Current
This ripple concerns us in two ways. First, it may induce noise in telephone
circuits that parallel the dc line. This is discussed in Chapter 8. Second,
since the magnitude of the ripple is almost independent of the average value
of direct current, the ripple sets the lowest average current at which the
current is continuous. Below this, the current becomes discontinuous or
intermittent. In twelve-pulse operation, the current is then interrupted twelve
times per cycle. Such operation of a high-power converter is objectionable
on account of the high overvoltages {L dijdt) induced in the windings of the
converter transformer and the dc reactor by the high rate of change of
current at the instants of interruption. These overvoltages are related to
those caused by arc quenching.
Actually, operation, even with continuous current, is objectionable if the
DC REACTORS
241
overlap is too small. Two reasons may be given. The first is that the two
jumps in the direct voltage at the beginning and end of commutation merge
to form one jump twice as great. Consequently, the overshoot of direct
voltage caused by a high-frequency oscillation is twice as great and may flash
over the protective gap placed across the dc terminals of each bridge. See
Section 7-8, page 286. The second reason is that the current oscillation pro
duced in a valve that is beginning to conduct (see "Arcback," page 266) is
capacitively coupled into the valve that is ending conduction. If the overlap
is short, the oscillation does not die out before the latter valve ends conduc
tion and may produce a high rate of change of current at the end of conduc
tion. This is a condition conducive to arcback and should be avoided, because
it tends to increase the erosion of the anodes.
The minimum current limit is introduced into the control system of the
converter for this purpose. It should be set at about twice the critical current
below which there is no overlap.
Calculation of Ripple. Let the following assumptions be made for a study
of the effect of the dc reactor on the ripple in the direct current:
1.
2.
3.
4.
Six-pulse operation of one bridge

Constant ripple-free direct voltage on the line side of the dc reactor
No overlap
Periodic variation of the direct voltage on the valve side of the dc
reactor, as assumed in Chapter 3, consisting of arcs of sine waves of powersystem frequency
5. Arbitrary ignition angle a
6. No resistance in dc reactor
Six-pulse operation obviously gives a greater ripple than, twelve-pulse or
mixed operation with the same direct voltage, and such operation occurs in
twelve-pulse converters when one bridge is out of service. Assumption 2
depends on good filtering of the direct voltage by the dc reactor itself, with
or without tuned harmonic filters (Chapter 8). At light loadsthose for
which there is a possibility of intermittent currentthe overlap is small and
has a very small effect on the wave shape of direct current.
As a result of assumptions 2 and 6, the equivalent circuit to be analyzed
is that in Figure 1, whence
did =
dt
(7)
Therefore
1
"~L.
(P. - Vt) dt
(8)
242
PROTECTION
7-2
Ld
Id
w
VdVd
Vd
DC REACTORS
243
0.2
0
Vd
(a)
-0.2
Fig. 1. Circuit for analysis of relation between current in and voltage across the dc
reactor.
0.8
In a six-pulse converter with no overlap, the voltage vd on the valve side of

the reactor has the wave shape shown in Figure 6 in Chapter 3, consisting
of 60 arcs of sine waves having amplitudes equal to the crest line-to-line
alternating voltages, which is, by Eq. (4) in Chapter 3,
Max
0.6
I
0.4
ld
Em=V,
(9)
dO
0.2
The equation of the first such arc after the origin of time is
vd =
-j3Em cos (cot 30)
for a cot a + 60
In the absence of overlap, the voltage on the line side of the reactor is
Fig. 2. Ripple in direct (a) voltage and (b) current for = 30.
By vector addition of the three terms in the brackets, this becomes
The voltage across the reactor in the time range specified is
fi
Vd = Vd0\
- cos (cot
30) cos a
(12)
[sin (cot
30) sin (a 30)]
(cot a) cos a
(13)
and is plotted in Figure 26 for a =30. Maxima and minima of id occur

when vd Vd = 0. When v d Vd > 0, id is increasing; when vd Vd< 0, id
is decreasing. The wave shape of id is a scallop with the cusps downward.
The zero of id has been taken as its value at cot = a. This is arbitrary: a
constant of integration may be added.
The average value of the ripple current is
V+n/3
L=
dO
id d(cot) = coL,
Id = f (0.0931 sin a)
coLd
This is plotted in Figure 2a for a = 30. Since Vd is the average value of vd,
the average voltage across Ld is zero.
The instantaneous direct current id is found by substitution of Eq. (12) into
Eq. (8) and performing the indicated integration over the range of cot from
atoa + 60. The result is
h=
<x>t
(11)
Vd = Vdo cos cc
vd
75
(10)
71
71
sin a - sin (a 30) - cos
(14)
(15)
For allvalues of a except 0 and 180,the function - Vd has discontinuities

at the junctions between arcs at cot = a, a + 60, a + 120, etc. For most
values of a (10.1 < a < 169.9) a zero of vd Vd lies on this discontinuity,
and minimum id occurs at the same time. In this range of a, Eq. (15) gives
the difference between the average and the minimum values of direct current.
Hence it also gives the least average value for which the direct current is
continuous or uninterrupted. For a = 0, it is clear that neither of the zeros of
vd Vd in a 60 arc falls at cot = a, and the value of id at cot = a is not the
minimum. It can be shown that if a = 10.1 or 169.9, a zero of vd Vd
lies at one end of the discontinuity. These are the limiting values of a for
which Eq. (15) can be used to find the minimum continuous direct current.
The ripple at values of a near 0 or 180 is of little practical interest, because
both normal and abnormal operation occur at values of a that produce
greater ripple.
For twelve-pulse operation the numerical coefficient in Eq. (15) is replaced
by 0.023.
"/
244
7-2
PROTECTION
Example
Find the inductance of the dc reactor required to keep the direct current
continuous when the converter in Example 1 is operating with one bridge
per pole at
5% of rated current and a = 90.
SOLUTION
DC REACTOR
245
reactor and the total shunt capacitance of the station, and the dc line is
represented by one of the approximations shown in Figure 3. A long over
head line may be represented by a Thevenin equivalent circuit for transients
comprising the open-circuit voltage and surge impedance (Figure 3a). A
cable may be represented by a nominal T or n circuit with series resistance
and charged shunt capacitance; the n circuit is shown in Figure 3b, where
the rectifier is represented by a constant-current source.
Operation with one bridge per pole is six-pulse.

The difference between average and minimum direct current is given by
Eq. (15), whence
V<iq x 0.0931 sin a
Ld =
(Oldm
Vd0 at light load may be taken equal to the no-load direct voltage of one
bridge, 200 kV.
sin a = sin 90 = 1
co 2nf = 27160 = 377 rad/sec
idm for barely continuous
Dc hne
Terminal
current is the minimum average direct current,
given as 0.05 x 1.8 kA = 0.09 kA.
200 x 0.0931 xl
L'-iTr
Dc cable
Termina
A4nTr
= 0'49H
R,
-vW-
The inductance chosen in Example 1 is ample, giving minimum continuous

current at 3.2% of rated current.
i,r
Ld -LCi
' 2
2
Zg Cg
_L
\Ld
Discharge of Line through Bypass Valves

Figure 2 in Chapter 6 shows how the bypass valves are connected. Those
at the inverter are of such polarity that the line can discharge through them.
Assume that the dc line is charged to direct voltage Ed by the rectifier while
all inverter valves are blocked. Then assume that all the inverter bypass
valves on one pole are unblocked as a result of accidental loss of the grid
bias. The line discharges through the bypass valves.

An accurate calculation of the discharge current as a function of time would
require representation of the line as a circuit of uniformly distributed para
meters as well as representation of all the lumped parameters at the inverter
terminal not only those of the dc reactor but also those of the surge cap
acitor and dc filters, if used. Such a calculation might best be performed by
use of a dc model or simulator (Volume 2, Chapter 18).
An approximate calculation, however, can be based on a simplified circuit
in which the inverter station is represented by the inductance of the dc
(c)
(d)
Fig. 3. Approximate circuits for calculation of discharge of (a) overhead line and (b)
cable through bypass valves BP. Simplified circuits for transient component of discharge
current :(c) overhead line, {d) cable. Ch line capacitance; C station capacitance; Ed, line
voltage; Ld, dc reactor; Ri, line resistance; Zs, surge impedance.
In either case, the current through the bypass valves consists of a steadystate (dc) component and a transient component.
For the circuit in Figure 3a, the dc component during the first round-trip
wave-travel time is EdjZs. The transient current, found by analysis of the
simplified circuit in Figure 3c, is a damped sinusoid. The current through
the bypass valves is :

=
z.il~s
cos cot
coL,
CO
]}
1 sin cut
(16)
246
7-3
PROTECTION
where
-zfa
co =
(I7a)
Vtojl a2
(17c)
SOLUTION
The discharge current as a function of time is calculated from Eq. (16)

and is plotted in Figure 4. The crest currents are found to be the following:
Ld (H)
Crest Current (kA)
0.1
0.3
0.5
1.61
1.11
1.013
0.3 H/
/ 0.5
' '
"""
-/
/ / /
/ / / /
/ / / /x
/ / /
_ _ -- _
'
/ // /
\ // /
I// /
Example 4
An 800-mi 300-kV overhead line having surge impedance of 300 H has

an inverter station with two 150-kV 1.2-kA bridges in series, a station shunt
capacitance of 2 piF, and a dc reactor of inductance Ld. Find the crest dis
charge current through the bypass valves for Ld = 0.1, 0.3, and 0.5 H.
\\
_ ___ _
---
V
\\
/ o.l H
/
/
//
For the circuit in Figure 3b, the steady-state current is Id. The transient
current consists of a dc term with exponential decay and a damped sinusoidal
term.
The effect of the dc reactor on the crest value of discharge current may be
seen from the following example.
I
\
~~
(17b)
247
VOLTAGE OSCILLATIONS AND VALVE DAMPERS
'/
11
III!
Time t, ms
Fig. 4. Discharge current of overhead dc line as affected by the inductance of the dc
reactor. The broken lines are the envelopes of iL for Ld = 0.1 H.
7-3 VOLTAGE OSCILLATIONS AND VALVE DAMPERS
In order to avoid too frequent arcbacks, one must limit the rate of rise of
inverse voltage and the peak inverse voltage across mercury-arc valves to
values depending on the valve design. Damping circuits are used for satisfy
ing these limitations rather than derating the valve. Such circuits are required
also across thyristors for avoiding their breaking down on inverse voltage
exceeding the rated value.
The crest current does not exceed the rated current of the bypass valves
(1.2 kA) except fot Ld = 0.1 H.
The round-trip wave-travel time is 2ljv = 2 x 1300km/(300 km/ms) =
8.7 ms. The crest currents occur earlier (1 to 4 ms); hence the line represen
tation is valid.
The effect of the dc reactor on crest discharge current can be explained

by reference to Eqs. (16) and (17) and to Figure 4. The curves lie within
converging envelopes (shown in broken lines) having initial ordinates which
depend only secondarily on Ld and which decrease exponentially by a damp
ing factor a independent of Ld. The frequency of the current, however,
decreases as Ld increases; hence the crest value occurs later, when the
envelopes have converged more.
Wave Form of Voltage across Valve

The theoretical wave shape across a rectifier valve without damping cir
cuits or stray capacitance is shown in Figure 20b in Chapter 3 and again in
Figure 5 in the present chapter. Of special importance for the present dis
cussion is the vertical discontinuity or sudden voltage jump,
Vje =
y/3Em sin 5
(18)
immediately after extinction of the arc.

In reality, the small stray capacitance that always exists across a valve
causes the rate of rise of inverse voltage to be finite but higher than the
248
PROTECTION
7-3
VOLTAGE OSCILLATIONS AND VALVB DAMPERS
249
1.5 E,
kSk
-VJ2Fig. 5. Theoretical voltage across rectifier valve, showing voltage jumps. Drawn for a
= u = 15: Vji =\3Em sin a is ignition voltage jump; Vjc = \/3 Em sin 8 is extinction
voltage jump. The smaller jumps (F}</2 and VJc/2) occur at ignition and extinction of
other valves.
permissible rate. Also, in conjunction with the leakage inductance of the

transformer, it causes an oscillation of voltage typically in the frequency
range 10 to 20 kHz and overshoot approaching 100%.
A satisfactorily low rate of rise of voltage can be obtained by connecting
a capacitor across each valve. The frequency is then typically 1 to 2 kHz.
The overshoot, however, is still too great, for the inherent losses in the
oscillatory circuit are small.
The overshoot can be reduced by connecting a resistor in series with each
capacitor. Typical values of overshoot are 10 to 25%. The series RC com
bination is called a valve damper. Figure 6 shows a converter bridge with
valve dampers. The bypass valve needs no damper of its own, being ade
quately served by those of the six main valves.

The dampers at the Lydd terminal of the Cross Channel dc link have
R = 3.5 kfl, C 25 nF. The calculated overshoot, with these dampers and
commutating inductance of 56 mH per phase, is 22%.
The RC dampers serve the additional purpose of improving the transient
voltage division between two or more bridges in series.
Calculation of Wave Form of Voltage across Valve with Dampers We con

sider the voltage across valve 1 immediately after the end of commutation
with valve 3. The circuit shown in Figure 6, in which stray capacitance is
neglected, can be simplified to that shown in Figure 7 by the following con
siderations :
1. During the time considered, valves 2 and 3 are conducting and thus
short-circuiting their own dampers. These dampers and valves are replaced
by solid connections.
Fig. 6. Connection of RC damper circuits across each valve of a converter bridge.

Valves 2 and 3 are conducting. Valve 1 has just finished conducting.
2D
Valve
Switch
Active network
Passive network
Fig. 7. Simplification of the converter circuit of Figure 6 for finding the transient voltage
across valve 1 immediately after the end of its conducting period. Block D represents an
RC damper circuit; 2D represents two such circuits in parallel.
5
i
i
j;
2. The four nonconducting valves are replaced by open circuits.

3. The dampers across valves 5 and 6 are in parallel and may be combined
to one of twice the admittance.
4. With respect to terminals a and b (across valve 1), the given circuit is,
by Norton's theorem, equivalent to one in which the internal sources are
j
j
j
j
'
250
7-3
PROTECTION
replaced by an external current source connected between those terminals.

The internal sources are the alternating voltage sources and a constantcurrent source representing the direct current. These are replaced by short
circuits and an open circuit, respectively. The external current source is that
which gives the same short-circuit current as the original circuit. This is a
sinusoidal curve of power frequency equal to the current through valve 1
during commutation, continued after the end of commutation as a negative
(reverse) current equal to that which would occur in an immediate arcback.
When the switch representing the valve is closed, all the current from this
source goes through the switch, and there is no voltage across it. When the
switch is opened at t = 0, the current of the source goes through the passive
network and develops a voltage across it equal to the recovery voltage across
the valve.
Further simplifications are made by the following considerations:
bo
2D
2 S
IOC
(b)
(a)
Fig. 9. Circuit equivalent to that of Figure 8 with respect to terminals a and b.
5 = 1/35 = admittance of inductance 3Lc
6. The three Y-connected inductances in Figure 7 are replaced by an

equivalent A, giving the circuit in Figure 8. This eliminates node N.
D=
Cs
= admittance of one damper
RCs+ 1
(21b)
Figure 9b, differing from 9a only in the detailing of the equivalent dampers
1 and 2, makes the complex natural frequencies apparent. They are
bo
sus3 = -<Tt jCOj_
(22a)
(22b)
$2* 4 =
2D
(21a)
and
(19)
3 L,
X:2C
where
5. During the short time that is of interest, say, one cycle of the natural
frequency, it is sufficiently accurate to represent the source current as a ramp,
h-t
251
where
a*
31,
and
j2
i
JL = 0.fal
(23a,b)
are the damping factors;
co, =
Fig. 8. Circuit equivalent to the passive network of Figure 7 with respect to terminals
a,b, c after eliminating node N.
7. Now it can be shown that, with respect to terminals a, b, the circuit in

Figure 9a is equivalent to that in Figure 8. The equivalence can be proved by
writing expressions for the impedance of each circuit. That in Figure 9a is
Z*(s) =
65 +
10Z> + 25 + 2D
85+ 12D
(25 + 2D)(6B + 10D)
(20)
"'~V3Z+
and
co, =
"2~~j5Lrc
= 0.775co,
(24a,b)
are the undamped natural frequencies; and

CO 1 =
VcoJi - <7?
and
co2 =
y/a>i2
(25a,b)
are the actual damped natural frequencies.

The equation for the voltage across the passive circuit as a function of time
can be found by taking the inverse Laplacian transform of the equation
V1(4 = ZJs)I1(s)
(26)
252
7-3 VOLTAGE OSCILLATIONS AND VALVE DAMPERS
PROTECTION
where Zab is given by Eqs. (20) and (21), and IL is the Laplacian transform of
Eq. (19), which is
cr,
= 0.915
/,=
cox
(27)
Eq. (28) becomes
e'CTl'cos coxt
i(0 =
sin
+ 4 s ~2tf cos (x>i t

~
Example
Hz
1=1810
2n
= 1640 Hz
f1= 1M?5
2n
= 0.613
COt
The voltage is
253
since,*) 1
o>2
(28)
Calculate the wave form of voltage across a valve and its associated damper
at the Lydd terminal of the Cross Channel link. The damper parameters are
R =3500 Q
C = 25 nF
The leakage inductance per phase of the converter transformers (Y connec

tion) calculated at /= 50 Hz is
Lc = 56 mH
Vl(t) = 100[0.750e-10-42f (cos 11.380* + 0.915 sin 11,380*)

+ 0.250s ~6250t (cos 10,190* + 0.613 sin 10,190*) - 1] kV
A curve plotted from the foregoing equation is plotted in Figure 10. The
calculated overshoot is 22.25%. If the high-frequency resistance of the trans
former windings had been taken into account, the calculated overshoot would
have been somewhat less. If stray capacitances across the valves had been
taken into account, the curve would have been shifted to the right a little
and joined to the origin through a short arc.11 The inclusion of stray capaci
tance in the analysis complicates the mathematics but does not appreciably
modify the rate of rise and crest value of the reverse recovery voltage across
the valve.
In order to put Eq. (28) into a more general form, introduce the dimensionless variables
Assume the extinction voltage jump to be
*'
Vj = 100 kV
SOLUTION
0.1
Time after extinction, ms

0.2
0.3
0.4
-Jlcc
0.5
By Eqs. (23),
1=
<j
3500
= 10,420 sec-1
6 x 56 x lO"3
= 0.6a i = 6250 sec-1
-50
By Eqs. (24),
25 x 10~9)"1/2 = 15,440 rad/sec

=11,980 rad/sec
= (3 x 0.056 x
o)2
=0.775o
-100
By Eqs. (25),
Overshoot
<ox
V(15,440)2 - (10,420)2 = 11,380 rad/sec
V(ll,980)2 - (6,250)2 = 10,190 rad/sec
-130
Fig. 10. Inverse recovery voltage across rectifier valve.
(29)
254
7-3
PROTECTION
from which
255
SOLUTION
cr't'
m=t.
at
alt=,
f'V 3 (<r')~2
t'
v' = 0.15e~r/3 { cos [?'
----
/
= = ,*'
V3'
and
t'
V5'
3!
sin co'.t
103/25 x 10-9\1/2
(56
lO-3/
'
-=
--
, 3.5
From Figure 11, the overshoot is 22%.
t'5-(a')2
+ 0.25e"<,r''5cos2
Valve Damper Loss
sin j/ 1
(30)
The form of the curve of v' versus (' depends only on the parameter a'. A
graph of overshoot as a function of a' is given in Figure 11.
Example
The power lost in the valve damper circuits is important both (1) because
of its effect on the overall efficiency of the converter station and (2) because
it determines the heat to be dissipated from the damper resistors, and hence
affects their size, cooling requirement, and cost.
Reasonably accurate formulas for the loss have been presented by Ainsworth11 and are as follows :
Pd = Phf + Ptf
where
Phf = 1.75fiVLL2 (C + Cs) (sin2 a + sin2 5)

pj= 2nfx2VLL2C2R [2.46 +0.875 (sin 2a +3 sin 2d - 2u)
Find the voltage overshoot in Example 5 by use of the curve in Figure 11.
(31)
fx
+0.433 (cos 2a +cos 23)]
(32)
(33)
= fundamental frequency, Hz
VLL = rms line-to-line secondary voltage, V
C = damper capacitance, F
Cs = stray capacitance, F
R = damper resistance, Q
a = ignition delay angle
6 = extinction delay angle
u overlap angle, rad
Pd = power loss in damper for one valve, W
Phf = high-frequency component of Pd
Pis = low-frequency component of Pd
100
75
50
The derivation of these formulas is now outlined. It is assumed, as a close

approximation, that the current in the damper resistor is that obtained by
25
Generalized damping coefficient <r'

Fig. 11. Overshoot of inverse recovery voltage across valve as a function of generalized
damping coefficient: ct' = (R/2)v/ (C/Lc).
applying the theoretical voltage illustrated in Figure 5 directly to the RC

circuit, neglecting the transient voltage drop in the commutating reactance.
One cycle of the wave form of the theoretical voltage consists of eight arcs
of fundamental-frequency sine waves, separated by vertical discontinuities or
jumps one such arc (for the period of conduction of the valve) being of
zero amplitude. The current resulting from this voltage wave in each of the
eight intervals corresponding to the separate sinusoidal arcs has two com
ponents : (a) a transient component it resulting from the preceding voltage
256
7-3
PROTECTION
jump and (b) a sinusoidal component is resulting from the sinusoidal arc of
voltage.
The transient component of current actually has the form of a rapidly
damped high-frequency oscillation, as discussed in Section 7-3, page 247,
but, according to the assumption made at the beginning of this derivation,
it would be a nonoscillatory exponential wave of short time constant RC.
Under either assumption, this component vanishes long before the occur
rence of the next voltage jump.
The sinusoidal component of current is an arc of fundamental frequency
leading the corresponding voltage wave by nearly 90, since at this frequency
the reactance of the capacitor predominates.
The instantaneous power consumed by the resistor is
p = Ri2 = R(jt
+ Q2 = R{j2 + 2itis + if)
watts
(34)
In the last member of this equation, the first term is the high-frequency loss,
and the sum of the second and third terms is arbitrarily called the lowfrequency loss, although the second term depends on both components of
current. It is obvious that the first and third terms are always positive but
that the second term may have either sign. The expression must be integrated
with respect to time over the eight parts of one cycle; and the result must then
be multiplied by /j in order to find the energy dissipated per second, which is
the power loss. The energy entering the capacitor over a whole cycle is zero.
High-Frequency Loss. It is not difficult to show that a voltage jump Vj

causes a loss of energy of CVfjl joules in the resistor (Problem 4). This result
is the same as the energy stored in a capacitance C charged to voltage Vj. If
the capacitor were discharged through the resistor, this energy would be dis
sipated in the resistor. The energy dissipated is given by the same expression
regardless of whether the capacitor is wholly or partially charged or discharged
as a result of the voltage jump.
The voltage across a valve (Figure 5) has eight jumps per cycle, the largest
of which are the ignition jump F/ =
sin a and the extinction jump
sin 5. In addition, there are six smaller jumps due to commu
VJe =
tations in other valves; three of these are equal to VjJ2 and three are equal
to VJej2. The energy loss per cycle is, therefore,
Whf
= 0.5CZV]
0.5C(\/2FLL)2(sin2 a + sin2 <5)[1 + 3(0.5)2]
jumps.
Example
Find the damper loss per valve and per bridge of the converter described
below for both rectifier and inverter operation at full load. Each converter
is rated at 1.8 kA, + 400 kV, 1440 MW, and has three bridges per pole. The
resistance of the bipolar line is Ri = 18.95 U per pole. The sending-end
voltage is kept at rated value. The equivalent commutating resistance is
7.65 Q per bridge. The rectifier normally operates at a = 15 and the inverter
at y 16. The fundamental frequency is 60 Hz. The damper circuits have
R = 1200 Q, C = 0.100 /iF. Neglect stray capacitance.
SOLUTION
Receiving-end direct voltage per pole = 400 18.95 x 1.8

= 400.0 -34.1 =365.9 kV
Receiving-end voltage per bridge = 365.9/3 = 122.0 kV
Sending-end voltage per bridge = 400.0/3 = 133.3 kV
At the sending end,
Vd0 cos <x = Vd + RJd = 133.3 + 7.65

= 133.3
x 1.8
+ 13.8 = 147.1 kV per bridge
=ra==152-5kVperbridge
Vll=1~M1.35
LL
U3.3 kV
= 0.966
cos 5 = cos a. --
Vd0
8 = 38.2
--
TFFs152.5
-66
0.181 = 0.785
u = 5 - a = 38.2 - 15.0 = 23.2 = 0.405 rad
joules (35)
which, multiplied by fu gives the high-frequency power loss as in Eq. (32).

The high-frequency loss is by far the larger part of the damper loss, and
therefore, if only a rough approximation to the total damper loss is required,
257
the low-frequency loss may be ignored. The high-frequency loss is inde

pendent of the resistance and is directly proportional to the fundamental
frequency, to the capacitance, and to the sum of the squares of the voltage
*J2VLL
\J2VLL
At the receiving end,
Vdo cos y
= Vd+RJd = 122.0 + 7.65 x 1.8

= 122.0
+ 13.8 = 135.8 kV per bridge
258
7-3
PROTECTION
135.8
135.8
=
= 141-3 kV per bridge
Vdo 6
cos y pTnTi
0.961
= 2940(2.46
259
+ 0.875[sin (-80) + 3 sin (-32) - 0.842]
+ 0.433[cos (-80) + cos (-32)]}
+ 0.875[ 0.985 + 3(- 0.530) - 0.842]

+ 0.433(0.170 + 0.848)}
= 2940[2.46 - 0.875(0.985 + 1.590 + 0.842) + 0.433 x 1.018]
= 2940(2.46 - 0.875 x 3.416 + 0.441)
= 2940(2.46 - 2.99 + 0.44) = 2940(-0.09)
= 2940(2.46
Vll =
cos /? = cos y
104'! kv
2RcId
2x13.8
-
= 0.961 = 0.961 - 0.195 = 0.766
v do
141.3
P = 40.0
u=)3-y = 24.0 = 0.421 rad

damper loss in rectification.
Pd = 55.9 0.3 = 55.6 kW per valve
By Eqs. (32), (33), and (31)
Phf= 1.75/i VIl(C + Cs) (sin2 a + sin2 5)
= 1.75 x 60(113.3 x 103)2 10-7(sin2 15.0

= 1.348 x 10s[(0.259)2 + (0.618)2]
= 1.348 x 105(0.067 + 0.382 -0.449)
= 6.06 x 104 W = 60.6 kW per valve
+ sin2 38.2)
+ 0.875(sin 2a + 3 sin 2<S - 2u)

Plf
+ 0.433(cos 2a + cos 2<5)]
= 2ti(60)2(113.3 x 103)2(10~7)2 x 1200(2.46 + 0.875 (sin 30.0
+ 3 sin 76.4 - 0.810) + 0.433(cos 30.0 + cos 76.4)
= 3480[2.46 +0.875 (0.500 + 3 x 0.972 - 0.810)+ 0.433(0.866 0.235)]
+
= 3480(2.46 + 0.875 x 2.606 + 0.433 x 1.101)
= 3480(2.46 + 2.28 + 0.48) = 3480 x 5.22 = 18,200 W
= 18.2 kW per valve
Pd = 60.6 + 18.2 = 78.8 kW per valve
Pd 6 x 78.8 = 473 kW per bridge.
The high-frequency loss is 77% of the damper loss.
DAMPER LOSS IN INVERSION
= 1.138 x
= 1.138 x
103)210~7[sin2 (180 - 40)

105[(0.644)2 + (0.276)2]
105(0.415 + 0.076 = 0.491)
+ sin2(180 - 16)]
104 W = 55.9 kW per valve

Plf = 2tt(60)2(104. 1 x 103)2(10-7)21200(2.46 + 0.875[sin 2(180 - 40)
= 5.59
pd = 6 x 55.6 = 334 kW per bridge
The high-frequency loss is slightly greater than the whole damper loss.
The damper loss in rectificationis473/240,000 = 1.97 x 10 ~3 pu = 0.20% of
the rectifier power. In inversion it is 334/220,000 =1.52 x 10~3 pu = 0.15%
of the inverter power.
= 2nf\VtL2C2R[2A6
Phf = 1.75 x 60(104.1 x
-264 W = -0.26 kW per valve
+ 3 sin 2(180 -16) -2x0.421] +0.433 [cos 2(180 -40)
+ cos 2(180- 16)]}
If valves were developed that could withstand a more rapid rise of inverse
voltage and still have an acceptably infrequent occurrence of arcback,
smaller capacitances could be used in the dampers, producing a correspond
ingly smaller power loss in the damper resistors.
The damper loss during inversion, although normally less than that during
rectification, becomes greater under the abnormal condition of low direct
voltage together with the inability of the tap changer to lower the alternating
voltage proportionally because of having reached its limit. The required
ratio of direct voltage to alternating voltage is then obtained by advancing
the ignition angle of the valves, leading to bigger voltage jumps and to higher
damper losses. The inverter is then operating on its constant-current charac
teristic instead of on minimum extinction angle. One cause of low direct
voltage at the inverter is the bypassing of a bridge at the rectifier. If this
condition is sustained, the corresponding bridge of the inverter should be
bypassed also. Even so, the direct voltage per remaining bridge on the same
pole is, at rated current, lower than its normal value, because the line voltage
drop is divided among fewer bridges. The inverter tap range should be
designed to permit continuous operation of the inverter at rated current
without the damper loss's exceeding its value for rectification at rated current
and voltage, although the extinction advance angle may be somewhat greater
than its normal value.
Even if the tap range were adequate for allowing continuous inversion at
normal extinction angle and highest alternating voltage, both rectifier and
inverter would operate at abnormally large ignition angles for short times,
such as during start-up or blocking or unblocking of a bridge. The damper
260
PROTECTION
resistors, which are usually oil-cooled, have a long enough thermal time con
stant to endure such brief overstressing without overheating. The damper
capacitors, however, cannot withstand even such a brief application of a
considerable overvoltage without damage by ionization of the dielectric.

The capacitors should be chosen to withstand, without ionization, the crest
overvoltage caused by the greatest voltage jumps to which they may be
subjected, which is during rectification with ignition angle almost 90.
7-4 CURRENT OSCILLATIONS AND ANODE DAMPERS
If ignition of a valve is delayed, a postive voltage builds up across it which

collapses when the valve is ignited. Any stray capacitance across the valve is
charged to this voltage and discharges through the valve as soon as the latter
ignites. Because of inductance, either stray or lumped, in the discharge circuit,
the discharge is oscillatory ; and because of low resistance in this circuit, the
oscillations are lightly damped. Because of the multiplicity of stray capaci
tances and both stray and lumped inductances, there are oscillations of a
multiplicity of frequencies. Generally speaking, however, they can be grouped
into two different bands, which may be called medium and high frequencies.
The medium-frequency band is around 20 to 60 kHz, the high-frequency
band 0.5 to 10 MHz.16
These current oscillations have several detrimental effects:
1. Extinction of the cathode spot in the incoming valve, with consequent
misfire
2. Increased likelihood of arcback of the outgoing valve if the overlap is
short
3. Radio interference
All these effects depend on the magnitude of the ignition voltage jump and are
therefore greater during temporary operation at abnormally large converter
angles than during normal operation.
Radio interference is discussed in Vol. 2, Chapter 16. The other two effects
are discussed in what immediately follows.
Extinction of Cathode Spot
The mercury-arc valves used in hv dc converters are the kind known as

excitrons, in which the mercury vapor is kept ionized during nonconducting
periods by means of arcs from one or more excitation anodes to the mercurypool cathode. The arcs terminate in a cathode spot, which is the source of
electrons, the movement of which through the vapor constitutes the valve
7-4
CURRENT OSCILLATIONS AND ANODE DAMPERS
261
current. If the net cathode current (the sum of the currents from the main
anodes and the excitation anodes) becomes zero, the spot becomes unstable
and could be extinguished. Then the valve cannot conduct again until the
spot is reestablished by the ignition electrode. Extinction would occur if,
during an oscillation, the main current goes negative by the amount of the
excitation current.
Because one-way conduction is the distinguishing characteristic of a valve,
one might doubt that the main anode current could ever reverse except when
an arcback occurs. A negative current of brief duration, however, can flow
because of the appreciable time required for deionization of the space between
the main anode and the control grid. This time is ample for a half cycle of
a high-frequency oscillation, although in the same time the power-frequency
current could change but little.
Oscillations of the current of the incoming valve at the beginning of com
mutation can be calculated reasonably well by use of the equivalent circuits
in Figure 12. In circuit a, the stray capacitance C is initially charged to
La ,
~\SLr
(a)
(b)
Fig. 12. Basic circuit of current oscillation in incoming valve: (a) stray capacitance and
inductance only; (b) anode reactor and resistor added.
voltage V}i. When the valve is ignited, the capacitance discharges through
the stray inductance L and the valve. A high-frequency oscillation occurs, as
illustrated by curve a in Figure 13, in which the negative main current
greatly exceeds the excitation current, causing spot extinction at point P.
Spot extinction is avoided, and radio interference is decreased, by the
addition of the circuit components shown in Figure 12b:
1. A reactor La having an inductance of 1.0 to 2.5 mH connected in series

with the valve
2. A resistor JR, typically of 2 to 7 kft,* in parallel with the reactor
* The resistor across the anode reactor of the bypass valve is only 200 to 400 Q.
262
7-4
PROTECTION
160
Example
263
The following data pertain to a certain converter;

Commutating voltage, Ec= 113 kV, rms phase to phase.
Stray capacitance across valve, C = 100 pF.
Stray inductance of valve, L = 100 H.
Excitation current, 50 A.
Commutating inductance, 2Lc = 42.2 mH, phase to phase.
Stray resistance is assumed to be negligible.
> -50
03
>
(A)
-160
<
C
OJ
0)
1. Find the undamped natural frequency and amplitude of the oscillation

of current in the incoming valve at the beginning of commutation under the
most severe condition.
2. If the amplitude of the current oscillation exceeds the excitation current,
find the least inductance of an anode reactor required to limit the reverse
current to 80% of the excitation current. What is the new undamped
frequency?
3. Find the resistance to be connected in parallel with the anode reactor
in order to damp the oscillation by a factor of 0.5 per half cycle.
Time, ms
_>
SOLUTION
-25
1. The voltage jump at ignition is
Vji =
-50
(B)
Fig. 13. Current of incoming valve at beginning of commutation: (a) without anode
reactor; (6) with anode reactor; (c) with anode reactor and resistor; (d) power-frequency
component of valve current; (e) negative of excitation current.
The reactor lowers both the frequency and the amplitude of the current
oscillation, as illustrated by curve b in Fig. 13. The resistor increases the
damping of the oscillation and thus further decreases the reverse main current,
as illustrated by curve c in Figure 135. Such a moderate reversal of the main
current, having crest value less than the excitation current, is harmless.
The reactor is connected on the anode side of the valve and is therefore
called an anode reactor. The combination of reactor and resistor may well
be called an anode damper. If the damper were connected to the cathode side,
it would be less effective, because it would be bypassed by a greater stray
capacitance than that on the anode side. In some converters, dampers are
connected on both sides of the valve.21
\/2Ell sin a
and is greatest at a = 90, or sin a = 1. In the present case it is

y..=
V2 x
113 x 1 = 160 kV
The undamped natural frequency is
litLC
=
1 _
x 10-10
2tiV10-4
1.6 x 106 Hz =1.6 MHz
The amplitude of the voltage oscillation is

= 160 kV
The amplitude of the accompanying current oscillation is
264
PROTECTION
7-4
where
265
3. In one half cycle, cot = cot = n rad, and

R = J~ =
= 0.5
whence
Then
m~
exp( at) = exp( icot) = exp(
= 1000n
tt = In 2 = 0.693
160,000
1000
and
= 0.22
C=
Also
The undamped current oscillation, given by
it = 160 sin (107r)
amperes
and superposed on the power-frequency current, approximated as a ramp,
---
104
Vnt = -160 x = 3.8
2T,
2 x 21.1 x lO"3
106r
2.
R=
amperes
is plotted in Figure 13 as curve a. The current has a crest value equal to 3.2
times the excitation current and results in extinction of the cathode spot
and cessation of anode current.
80% of the excitation current is
160,000
= -J = ! = 4000
40
which is four times as great as in part a. Since R0 = \Zl{C and C is un

changed, L must be increased by a factor of 42 = 16 and becomes
L 16 x 10~4H = 1.6 mH
The value of the stray inductance, which is included in this value, should be
subtracted to find the inductance of the anode reactor, which is then
4 = 1.6-0.1 =
1.5 mH
The new natural frequency is one-quarter as great as before, or
MHz
1.6
f= -
--
9ioo Q = 9.1 kCl
0.22
The natural frequency, decreased by damping becomes
f=fn Vl-C2 = fn Vl - (0.221)2 = 0.98\fn

= 0.98
x 400 kHz = 390 kHz
A;==il=18A
The value of R0 required to limit the crest current to 40 A is
Vn
4
2R
With stray inductance neglected, the current has an initial jump,
0.8 x 50 = 40 A
2R
whence
2R\C)
= 0.4 MHz = 400 kHz
The current oscillation,
it = 40 sin (0.25 x 107?)
amperes
superposed on the power-frequency current is plotted in Figure 13 as curve b.
9.1
The general expression for the damped oscillatory current is
it (A sin cot 4- B cos cot) exp( con/)

where
co = 2.44 x
106 rad/sec,
co = 2.50 x 106 rad/sec,
and
C = 0-22
From the initial conditions,
z't = 18 A
and
dit_Vn_
dt
L .
160 xlO3
=1()8A/sec
1.6 x 10-3
'
the constants A and B are evaluated, and we get
it = [45 sin (2.44 x 106t) + 18 cos (2.44 x 106r)] exp(-0.55 x I06t) amperes
The current i= it + is is plotted in Figure 13 as curve c.
The significance of the damping of the high-frequency oscillation will be
more apparent from what follows.
266
7-4
PROTECTION
267
Arcback
Current oscillations occur not only in the current of the incoming valve
but also in that of the outgoing valve of the same commutating pair. The
current oscillation in the outgoing valve is caused by capacitive coupling
between the two valves, represented in its simplest form by the circuit in
Figure 14 for commutation from valve 1 to valve 3. If the direct current is too
small and the voltagejump too great, the amplitude of the high-frequency cur
rent oscillation is great enough to force the current in the outgoing valve to
zero prematurely with a high rate of decrease, as sketched in Figure 15, curve
ix . Extinction of the current in the outgoing valve is immediately followed by
a high-frequency voltage oscillation across that valve, induced by the same
current oscillation in the incoming valve. This voltage oscillation is super
posed on the inverse recovery voltage already discussed in Section 7-3 so that
the initial rate of increase of that voltage is thereby increased.
A high rate of decrease of current in the outgoing valve as that current ap
proaches zero, followed immediately by a high rate of increase of inverse
->
ik 3
Co
<
o
.20.
-10 1
-1
Time, jus
43
11
Fig. 15. Currents in a pair of valves (1 and 3) undergoing commutation with short
overlap (1 or 46 /us): h, current of outgoing valve (valve 1); /3, current of incoming valve
(valve 3); i[ and iz ,theoretical currents of valves 1 and 3, respectively, with high-frequency
oscillations neglected.
voltage across that valve, combines two of the factors listed in Section 6-3 as
conducive to arcback.
The minimum current limit in the control system (Section 5-13) must be
set for an acceptably low incidence of arcback. Not only must the minimum
current setting exceed the initial amplitude of current oscillation in the out
going valve, but also the damping of that oscillation must be adequate to
prevent the current's reaching zero before the normal end of commutation. If
the high-frequency current oscillation is damped out before the end of com
mutation, both the high rate of decrease of current and the high rate of in
crease of inverse voltage are prevented.
The valves may operate momentarily with low current and short overlap
during start-up, fault clearing, etc., but they should not be permitted to work
thus for a long time.
Example 9
Let the following data apply to the circuit in Figure 14:
Eab = 109 kV rms

/= 60 Hz
Fig. 14. Basic circuit for analysis of current oscillations in two commutating valves at
the beginning of commutation: 1. outgoing valve. 2. Conducting noncommutating valve.
3. Incoming valve. C0 , Cm, stray capacitances; La , anode reactor; Lc ,commutating induc
tances ; R, anode resistor ; ea , eb , ec , alternating voltage sources ; Id , direct-current source.
a =15
u= 1
Lc = 55.5 mH
La = 2.5 mH
C0 = 1.0-nF
Cm = 1.0 nF
i? = 7000Q
268
7-4
PROTECTION
Calculate and plot the currents in each of the twp valves during commuta
tion from valve 1 to valve 3. Also calculate the rate of decrease of current
in valve 3 as it becomes zero.
current oscillations and anode dampers
where
CO<X
CO,
SOLUTION
power-frequency and DC phenomena.
= Eab
V2 sin a = 109 -Jl x 0.259 = 40 kv
co =
h s2(cs a cos 8) = Is2(cos 15 cos 16)

= /s2(0.9659 - 0.9613) = 0.0046/s2
~
where
e =
2E
2nfx2(Lc + La)
154 1Q3 _
*
377 x 0.116
Ty
Id
x 3500 = 16.0 A
The power-frequency currents are
4 3500[cos 15 cos (cot + 15)] amperes

i[ 16.0 1*3
The normal overlap time is 1/(60 x 360) = 46.2 x 10~6 sec = 46.2 (is.
damped oscillatory
currents.
n/co2ny - o2y
(41)
(42)
(43)
LyCy
1
(44)
2RCX
(45)
2RC
(46)
L=L
Hence
= 0.0046
(40)
yjLxCx
The direct current is
r52 -
e\
The voltage jump across valve 3
at ignition is
Vj>
The solution for the transient compo
nents of the value currents is
LL<
La+Lc
(47)
CX = C0
(48)
Cy=C0 + 2C
(49)
The derivation of these equations is too long to reproduce here. It depends

on the existence of two natural modes of oscillation, each of which may be
represented by a parallel RLC circuit (Figure 16) and on the initial condi
tions, which are Va = 0, Vb=Vji, iu=> 0, i3t = Vfi/R, diu}dt = 0, and
dhJdt = VjJL,, .
ht = ix + iy
(36)
ht = ix ~ iy
(37)
The valve currents are plotted in Figure 15.
Ly =
where
ix
=[cos +~t(r f ;e)

- cos
- <sin oyj
iy =
sin
exp(~o
Co
C0 + 2C
(3)
(x)
exp( o-ji)
(39)
269
(y)
Fig. 16. Substitute networks of the.circuit of Figure 14.
LaLc
La+Lc
270
7-5
PROTECTION
oscillations of voltage. By the use of appropriate frequency-selective feedback,

damping of power-frequency oscillations is introduced.19
The start-control unit of the control system is designed to increase the
7-5 DC LINE OSCILLATIONS AND LINE DAMPERS

The dc line and its reactive terminations, particularly the dc reactors,
form a lightly damped oscillatory circuit having a natural frequency in the
range of 10 to 60 Hzfor example, 40 Hz on the Cross Channel line. Thr
oscillations may produce overvoltages. If the oscillations of current build up
to such an extent that the current becomes zero at the low point of the cycle,
a spike of overvoltage is produced. This is more likely to happen at light load.
I-
ji
r|
j
i'|
Causes of Oscillation
Oscillations may be caused by impressed alternating or step voltages or by
short circuits. Step voltages and short circuits cause oscillations at the natural
frequency or frequencies of the line. Alternating voltages cause oscillations
of both natural and impressed power frequency. If the natural frequency is
near the impressed frequency, oscillations tend to build up by resonance.
Alternating voltage may be impressed because of the following:
1. Blocking the main valves and misfire of the bypass valves. Then two
terminals of the ac source are connected continuously to the dc line through
the valves that were conducting when blocking took place. The rms ac
1.
2.
3.
4.
Starting up a rectifier
Bypassing or unbypassing a bridge
Recovery from reduced alternating voltage caused by an ac short circuit
Reenergization of the dc line after a short circuit
direct voltage exponentially without overshoot instead of in a sudden step.

This unit, however, does not come into action for all step voltages but prin
cipally for starting up and for unblocking a bridge.
Each bridge is provided with a start-control unit that performs two
functions :
1. On the unblocking of a rectifier bridge, its start-control unit lets the

direct voltage of the bridge rise exponentially to its normal value, with a
long time constant compared with the natural frequency of the dc line, thus
preventing any considerable overshoot of voltage.
2. On the unblocking of an inverter bridge, its starting unit first causes
the bridge .to operate momentarily as a rectifier in order to extinguish the
bypass valve and thereafter lets the direct voltage rise exponentially to its
normal value.
The exponential rise is obtained by an RC circuit inserted in the control path.
i]
voltage is about 0.74FdO.19

2. A persistent fault (commutation failure, misfire, or arcthrough) in an
inverter main valve. The impressed voltage is nonsinusoidal and may be
resolved into sinusoidal components (Fourier series) of which the fundamen
tal is greatest, its rms value being about 0A3VdO.19 Persistent misfire of a
at a =0.
rectifier main valve gives a fundamental voltage of only 0.31
Step voltages may be impressed by the following:
271
DC LINE OSCILLATIONS AND LINE DAMPERS
Damping Circuits. Line oscillations can be damped by connecting resistors

in parallel with either of the main elements of the oscillatory circuit, that is,
the capacitance of the cable or line and the inductance of the dc reactor, as
indicated in Figure 17 by Rx and R2 , respectively. The most effective value
of Rl or R2 is equal to the positive-sequence surge impedance of the line. For
avoiding an inordinately large power loss in Ru because of the high direct
voltage impressed on it, a blocking capacitor must be connected in series
with it, as shown in Figure 18a. This capacitor is very expensive because of
its large capacitance and high voltage rating. A smaller capacitance can be
used if an inductor is also used, as in Figures 18b or c.
*2
|
I
Damping of Oscillations
Overshoot of voltage after impressed step voltage and buildup of voltage

by resonance to impressed alternating voltage can be reduced by damping.
Two methods of damping are available.
|
|
Control Circuitry, The current regulator of a converter tends to hold constant current and thus to restrain oscillations of current that accompany
Ld
Ri
'T~ C
I
f
|
i
i
Fig. 17. Locations

resistors.
and R2 for line damping
(a)
(b)
(c)
Fig. 18. Forms of line damper con

nected across dc line.
272
PROTECTION
A resistor, such as R2 , connected across the dc reactor, incurs a loss from

the harmonic voltages across it. A series RLC circuit, as in Figure 19, may
be designed to reduce losses from converter harmonics while still adequately
damping line oscillations.
-V\A/-1(
7-6
CLEARING LINE FAULTS AND REENERGIZING THE LINE
273
the fault. Thus the voltage profile would shift from curve 1 to curve 2 in
Figure 20. Note that the inverter would have to reverse its voltage and
become a rectifier in order to maintain its current. With both converter
currents maintained, the final steady current in the fault would be the current
margin, which is typically 0.10 to 0.15 times rated current.
w
Ld
Fig. 19. Line damper across dc reactor.
Choice. The overvoltages in most line oscillations can be kept within

acceptable limits, say, 1.7 times rated direct voltage, by a well-designed
control system. Then higher transient overvoltages can occur only in the
event of a very improbable fault in the control system. Such overvoltages may
flash over the insulation of an overhead line. Cable insulation can be pro
tected from puncture by gaps. Flashover of the line or gap constitutes a line
fault. The clearing of line faults is discussed in the following section. Generally,
the expense of line dampers is not justifiable. Perhaps they could be justified
to protect a dc cable from overvoltages due to resonance to impressed alter
nating voltage.
7-6 CLEARING LINE FAULTS AND REENERGIZING THE LINE
Short circuits on a two-terminal dc line are cleared by grid control of the

valves. On a bipolar line each pole has separate and independent faultclearing devices. Some methods of clearing short circuits are described below.
Effect of Current Control
Even if no additional protective devices were used, the current regulators

described in Section 5-9 would limit the short-circuit current by decreasing
the line voltage. The short circuit momentarily causes the rectifier current to
increase (see " Overshoot," page 276) and the inverter current to decrease. In
attempting to decrease the current to the set value, the current regulator of
the rectifier decreases the direct voltage. In attempting to increase the current
to the set value minus the current margin, the current regulator of the inverter
also decreases the direct voltage. If no other control action intervened, the
reduction of direct voltage would continue until each converter developed a
voltage just sufficient to supply the R1drop in the line from the converter to
Fig. 20. Voltage profiles on a dc line: 1. Normal condition. 2. Short circuit at center,
currents maintained. 3. Fault clearing: a = 135 at rectifier; a= 100 at inverter.
Control for Fault Clearing

Although the current control limits the fault current to a small value, it is
not adequate for extinguishing the fault arc. For sure extinction, both the
fault current and the recovery voltage across the fault path should be brought
to zero. Additional control is required to do this.
The best way to bring the line current and voltage to zero rapidly is to
establish terminal voltages of such polarities that both converters invert.

Then both converters drain energy from the dc circuit, where it is stored in
the electric and magnetic fields associated with the shunt capacitance and
274
7-6
PROTECTION
series inductance of the line and of the dc reactors, and deliver the energy
to the ac networks. The difference of terminal voltages required for both
converters to invert is opposite to that for maintaining the direct current.
They produce a gradient, somewhat like that in curve 3 in Figure 20, which
decreases the current and tends to reverse it. Valve action prevents reversal.
Therefore, the line current becomes and remains zero. The fault current does
likewise. Because the faulted conductor was grounded through the fault until
or nearly until the line current became zero, the conductor is left at or near
ground potential so that the recovery voltage across the arc path is negligible
and the arc does not restrike.
In order to establish terminal voltages of the correct polarity for fault
clearing, two steps are taken:
1. The inverter normally has the correct polarity, but in order to prevent
reversal of polarity, as described above on page 272, its ignition delay
angle is given a lower limit of about 100 (fi 80), which allows the inverter
voltage to run down to a low value but not to reverse.
2. The polarity of the rectifier voltage must be reversed so that it inverts.
As soon as the fault is detected, the ignition delay angle is shifted to some
value considerably beyond 90, say, 135.
Fault detectors are ordinarily used to initiate the increase of a, and these
are described shortly.
A method described by Uhlmann5 but not in general use obviates the
need for fault detectors.
Fault Detectors
A short circuit on the dc line is detected at the rectifier station either by

(a) a sudden drop in direct voltage, measured on the line side of the dc
reactor, or (b) sustained low direct voltage. Either of these conditions can
initiate the fault-clearing procedure.
Typical settings of the fault detectors are as follows :
1. The voltage-derivative unit responds after 3 ms if the rate of decrease of
direct voltage exceeds a preset value, say, 0.5 times normal voltage per milli
second, and if the direct voltage goes below 0.75 times normal voltage.
2. The voltage-level unit responds after 50 ms if the direct voltage is below
0.45 times its normal value.
The time delays and voltage settings are intended to provide selectivity
between dc line faults and ac faults or converter faults. Such selectivity is
desirable because, although a dead time is necessary for arc deionization of a
dc line fault, it serves no useful purpose in the event of an ac fault or a converter
'
275
fault but unnecessarily delays the recovery of the rectifier to normal operation.
Low alternating voltage at either converter station is a cause of propor
tionally low direct voltage. The rate of decrease of direct voltage on the line
from this cause, however, is slower than that caused by a dc line fault, because
the discharge of the dc line is delayed by the dc reactors. Of course, the rate
of decrease of dc line voltage is also slow in the event of a dc line fault
through a high resistance ; and this is the reason for providing the voltagelevel unit. The unit may be actuated by ac faults.
One-phase-to-ground faults are by far the most frequent ac faults, but they
cannot reduce the direct voltage by much more than 33%, which is not a
great enough reduction to actuate the voltage-level unit. Multiphase ac faults,
however, may actuate the voltage-level unit unless either the time delay set
ting of the unit is longer than the clearing time of such faults, which is
normally 67 to 100 ms, or the unit is blocked by low ac voltage. If such
blocking is provided, it may be set to last beyond the backup clearing time
of ac faults (200 to 400 ms). It can be actuated directly by low alternating
voltage at the rectifier station and, through a telecommunication channel, by
low alternating voltage at the inverter station.
Converter faults also cause low direct voltage, but, again, the dc reactors
limit the rate of fall of the voltage to a value that does not actuate the voltagederivative unit. Converter faults affecting only one bridge out of two or more
on the same pole do not actuate the voltage-level unit. Faults affecting
several bridges simultaneously may actuate the unit unless some blocking
scheme is provided. Blocking for inverter faults, if provided, would reach the
rectifier station by way of the telecommunication channel used in the event
of low alternating voltage at the inverter. Simultaneous faults on several
inverter bridges would most likely have low alternating voltage as a con
tributing factor, and these would block the voltage-level unit. The clearing
time of converter faults is another factor that determines whether the voltagelevel unit would be actuated.
Reenergization
After an overhead line conductor has been deenergized for approximately

0.2 to 0.5 sec, an arcing fault is generally sufficiently deionized so that the arc
path withstands normal voltage. The automatic fault clearing may, if desired,
be followed by automatic reenergization. In most instances the line is restored
to normal voltage and to the same power as before the fault.* If the fault is
* An advantage of dc transmission is that if a line cannot be reenergized at full voltage

because of wet, dirty insulators, for example it can sometimes be reenergized at a lower
voltage by bypassing some bridges.
276
7-6
PROTECTION
permanent or if the arc restrikes because of inadequate deionization, the fault

is cleared again. If desired, two or three attempts at reenergization may be
made automatically in succession with increasing dead time.
The procedure of clearing short circuits on a dc line and then reenergizing
the line is the counterpart of fault clearing and rapid reclosure on an ac line,
but it differs in using grid control of the valves instead of the opening and
reclosing of circuit breakers. Reenergization of a dc line is done by raising
the direct voltage, not in one sudden big step, as by closing a switch, but
slowly and continuously under control of the starting control unit so that
there is no overshooting of the voltage. Reenergization of an ac line through
a circuit breaker may cause transient overvoltages of two or three times
normal crest voltage unless special means, such as series resistors, are used to
limit the overshoot.
Direct-current lines consisting wholly of cables are not reenergized, as a
rule, because cable faults are nearly always permanent.
277
U
U
Recttfier
ed
Fig. 21. Circuit for calculation of current in a short circuit of a rectifier beyond the dc
reactor.
The circuit diagram is shown in Figure 21. The wave of direct voltage ed
under normal conditions is shown by the upper envelope in Figure 22a.
The fault is assumed to occur when cot = 60 + . Because of the increased
delay in igniting the next valve, ed follows the heavy curve, whose equation is
V 3Em sin cut
(50)
Overshoot of Direct Current
Fault Current. Initially the fault current is limited only by the surge imped
ance Zs of the line to 2VJZS> which is considerably greater than the current
margin. For example, on a dc overhead line rated at 375 kV and 1.8 kA and
having a surge impedance of about 300 Q, the initial fault current is
2 x 375/300 = 2.5 kA, but the current margin is only 0.15 x 1.8 = 0.27 kA.
This initial current results from the discharge of the shunt capacitance of the
line through the series impedance of the line, and the current regulators at
the terminals cannot possibly limit it. In some cases the crest current in the
fault exceeds this initial value.
Rectifier Current. Overshoot of the rectifier current above the set value
depends on (a) the inductance of the dc reactor, (6) the gain and time con
stant of the current regulator, (c) the distance of the fault from the rectifier,
(d) the wave form of the direct voltage, and (e) the instant on this wave at
which the fault occurs.
The general effect of some of these factors is well shown in a paper by
Peterson, Phadke, and Reitan.25'32 They assume conditions that tend to give
the greatest crest current : (a) a six-pulse converter with (b) a = 0, (c) negligible
commutation reactance and hence no overlap, (d) fault at the rectifier end of
the line just beyond the dc reactor, (e) occurring immediately after a commu
tation has started. The current control is idealized by assuming that the fault
is detected immediately and causes the ignition of the next valve to be
delayed 180, thereby reversing the direct voltage on the valve side of the
dc reactor.
n s= 1
120"
n= 2
240'
180
aif
Fig. 22. Rectifier voltage (a) and current (b) on short
360
300
.
420'
circuit beyond dc reactor, a = 0.
278
PROTECTION
7-6
The current is
h = ho + A/d = ho +
where
j*
ed dt
(51)
Ido is the prefault current and Aid is the increase due to the fault :
=
V3
r
a>Ld
sin cot d(cot) = Ib(cos 60 cos cot)
(52)
60
where
h=
(53)
A id is plotted in Figure 22b. It reaches its crest value,

cot= 180:
Aidm when ed = 0 at
Ai'dm = h(cos 60 - cos 180) = 4(0.5 + 1.0) = 1.54
(54)
The foregoing analysis shows that even the best imaginable control of the
direct current cannot prevent a considerable overshoot of that current if a
fault occurs soon after a commutation has occurred. If control action is
delayed longer, the overshoot of current is correspondingly greater.
Suppose now that the first valve to be ignited after the occurrence of the
fault is not delayed at all but that the second valve is delayed by 180. Then
the voltage, after the firing of the first valve, follows the sine curve (labelled
n 1), which lags 60 behind the curve followed in the previous case. The
area under the curve of voltage from the time of the occurrence of the fault
until the voltage becomes zero was Ax + A2'm the previous case, and in the
present case is augmented by A3. An additional delay of one-sixth cycle in
the control action adds area Aa. The relative areas are Ax = A3 = A4 and
A2 = 0.5Ax. The crest currents are 1.5Ib, 2.5Ib, 3.5Ib, etc. In general, for a
control delayed by a time between n and n + 1 sixths of a cycle and then
giving full shift of a, the crest current is
A?'dm = (+ 1.5)4
(55)
The cited paper shows that the effect of normal ignition delay a is to make
= 1 +(
+ 0.5) cos a 0.866 sin a
(56)
which, for a = 15, gives 1.26. It also examines multibridge converters and
the effect of commutating reactance. A 12-pulse 2-bridge converter has the
possibility of faster control action, because the normal firing interval is one-
279
twelfth cycle, and the overshoot with a = 0 and n = 0 does not exceed 1.22Ib,
a value somewhat less than that for a six-pulse converter. The effect of com
mutating reactance on overshoot is shown to be small.
Discussion of the paper brought out that a design criterion is that, in
12-pulse operation, the control system should not allow more than one com
mutation after the occurrence of a short circuit on the dc line : that is, n = 0
or 1. Further, the inductance to be considered is not merely that of the dc
reactor but, in addition, the sum of the commutating reactances of all the
bridges between the faulted pole and ground and the reactance of the ac
system supplying the rectifier.
Field tests have shown crest short-circuit currents of about 1 pu, which,
added to initial load current of 1 pu, gives a total crest current of about 2 pu.
Although this value is well above rated current, it should be emphasized that
it is indeed small compared with the short-circuit currents obtainable in an
ac system.
A later paper38 refines the representation of the constant-current control
and considers short circuits in various locations, including short circuit of one
bridge.
Fault Clearing on a Multiterminal DC Transmission System
The method of fault clearing on a multiterminal system can be similar to

that used on a two-terminal system except that, in the event of a persistent
fault, the faulted section must be switched out. The switching is done with the
fine dead.
Consider as as an example the three-terminal system in Figure 23a or c.
Assume a fault to occur at X. Fault detectors make all converters invert and
so deenergize the dc lines rapidly. After a delay adequate for the deionization
of a temporary fault, an attempt at reenergizing the lines may be made. If
the fault turns out to be persistent, the attempt will be unsuccessful, and all
fines will be deenergized a second time.
Let us assume that in the meantime the location of the fault has been
determined. The exact location need not be known but merely on which line
section it is. Now, the fines being deenergized, an isolating switch is opened
on the faulted fine at switching station D. Thereafter the fine connecting
converters A and B by way of D are reenergized, and power transmission is
resumed. Station C must remain out of service until the faulty fine between
C and D is repaired.
Another layout of a system with three converters and three lines is shown
in Figure 23b. Service at all three stations can be resumed after a faulted line
is switched out, although voltage drops and power losses are greater than
they normally are.
280
7-7
PROTECTION
CIRCUIT BREAKERS
281
switches, lines could be switched into or out of an unfaulted network without

running the voltage down.
Several ideas have been advanced for hv dc switches. 1'3'7,8'13'14 The idea
considered most seriously is to create zeros of current through the contacts
of the switch by the oscillatory discharge of a capacitor through the switch.
The crest currents of the oscillation must be greater than the direct current
to be interrupted.
A schematic diagram of such a circuit breaker is Figure 24a. Normally,
main contacts M and charging contacts A are closed, and capacitor C is
charged to line voltage through high resistance R. Contacts B are open and
(a)
(b)
D
(c)
Fig. 23. Three-terminal dc lines.
7-7 CIRCUIT BREAKERS
The lack of a suitable circuit breaker for hv dc power systems has been
alleged to be a disadvantage of such systems -and, perhaps, really is. The
preceding section has shown that, on a system with two terminals, faults
can be cleared by grid control of the converter valves and that, on a system
of three or four terminals, transitory faults can be cleared by grid control
only and permanent faults by a combination of grid control, fault locators,
and isolating switches. The times required for fault clearing and reenergization of a dc line are not much different from those for an ac line.
Experience with ac systems has shown that they tend to grow into complex
networks. Under present conditions there is little expectation that dc systems
of comparable extent will develop. Dc lines are merely adjuncts to ac
systems. A good hv dc circuit breaker could change this picture.
What is needed is not so much a circuit breaker for interrupting shortcircuit currents, for such currents can be limited by grid control to the mag
nitude of rated currents, but rather a switch capable of interrupting load
currents in circuits at high potential with respect to ground. With such
Fig. 24. High-voltage dc circuit breaker utilizing oscillatory discharge of capacitor C

for producing artificial current zero z : (a) schematic circuit; (6) wave form of current
through main contacts M.
have line voltage across them. When it is desired to interrupt current Id

in the main circuit, the operating mechanism opens contacts A and closes
contacts B, initiating the discharge, and immediately afterward opens main
contacts M, which interrupt the current at a current zero, such as z in Figure
246. Thereafter, contacts B are opened and A are closed.
Another proposed circuit breaker7 is shown in Figure 25. Capacitor C is
initially uncharged. When contacts M open, the main current is diverted to
the capacitor, with the result that the current to be interrupted by Mis small.
00-r
Fig. 25. High-voltage dc circuit breaker with nonlinear resistor.
282
7-8
PROTECTION
The rate of rise of recovery voltage across M is dVJdt = IJC. The non
linear resistor R absorbs energy without greatly adding to the voltage
across M.
Another kind, tested in a model for 20 kV, 10 A, uses the negative-resistance
characteristic of an arc shunted by a capacitor to increase the amplitude of
current oscillation until a zero occurs.14
Still another proposed circuit breaker uses an oil blast through the arc.13,31
hv dc breakers appear to be technically feasible, but their cost may be
excessive.
OVERVOLTAGE PROTECTION
283
per cycle at the end of commutation of each valve by damping circuits

(Section 7-3) rather than by spark gaps.
5. The voltage at which a gap or lightning arrester begins to discharge,
although subject to variation, must always be (a) higher than the highest
normal working voltage and (b) lower than the breakdown voltage of the
insulation being protected. Likewise, the maximum voltage across a lightning
arrester while discharging and the seal-off voltage of the arrester must also be
between the two stated limits.
Insulation problems in dc systems under both normal and abnormal con

ditions are complicated by the simultaneous presence of alternating voltages,
direct voltages, and transient voltages.
7-8 OVERVOLTAGE PROTECTION
External and Internal Overvoltages

Principles
External Overvoltages originate outside the power system. Their principal

source is lightning.
Internal Overvoltages originate in the power system itself. In an ac system
the principal internal overvoltages are switching surges, the most important
kind of which is switching one end of a transmission line onto a live bus
while the other end of the line is open. This can cause transient overvoltages
at the open end approximately two to four times as great as the normal lineto-ground crest voltage. Also, a phase-to-ground short circuit can cause
overvoltages up to twice normal voltage on one of the unfaulted phases.
A dc system has a great variety of sources of internal overvoltages.
.
The following principles are common to both ac and dc systems :

1. Any equipment in which the breakdown of insulation would be expen
sive to repair or would entail a long time for restoration to service must
be protected with certainty against any voltage high enough to puncture
the insulation. Among such equipments are transformers and cables. The
breakdown of air insulation is not so serious except where such insulation is
depended on for safety of personnel; for example, in disconnecting switches
(isolators).
2. Excessive voltages must either be (a) limited by such discharge devices

as lightning arresters or spark gaps or (b) reduced by other means, among
which are shield wires and low tower-footing resistances on overhead lines
and, also, in dc links, by appropriate damping circuits (Sections 7-3 and 7-5)
and by grid control of valves.
3. Next in importance, after, protecting human lives and expensive equip
ment, comes the prevention of frequent, although short, interruptions of
electric power supply. Flashovers of air insulation or of plain spark gaps con
stitute short circuits on the power system, which may cause excessive current
and decreases in voltage and power and which require about 0.5 sec to clear.
On dc circuits the short-circuit current is limited to values that cause little,
if any, damage, but the voltage and power dips are undesirable. Frequent
discharges of lightning arresters may wear or damage the arresters. For the
reasons given, the frequency of fiashover of air insulation or of arcing dis
charge devices must be limited to a reasonable value.
4. Therefore overvoltages that occur frequently must either be withstood
by better insulation or be reduced by other means than spark gaps. A good
example in dc practice is the reduction of voltage overshoots, occurring once
1. A short circuit from one pole to ground on a bipolar line, producing

an overvoltage on the unfaulted pole (see Vol. 2, Chapter 10).
2. Switching surges originating on the ac system, transmitted through the
converters to the dc side. The series connection of valve groups multiplies
the overvoltage by the number of groups.37
3. Arc quenching in valves or discontinuous direct current, causing a high
rate of change of current in transformer windings and in the dc reactor (see
"Protection against Current Chopping," page 286).
4. Persistent misfires or arcthroughs in an inverter valve, impressing alter
nating voltage, in addition to direct voltage, onto the dc line. The alternating
voltage may be magnified by resonance (see Section 7-5).
5. Still worse, the blocking of the main valves of a bridge, together with
failure to unblock the bypass valve. Then the two main valves that were
conducting continue to conduct, thereby connecting two phases of the ac
source to the dc terminals of the bridge.
6. Overshoot of direct voltage of a rectifier due to rapid start or to recovery
of alternating voltage after clearing of an ac short circuit (see Section 7-5).
284
PROTECTION
7. Sudden loss of load on a rectifier (see page 287).

8. Increased overshoot of direct voltage of a bridge at the end of each
commutation due to very short overlap (see page 286).
Means of Seducing Overvoltages
These may be divided into two classes:

1. Those which limit the overvoltage by discharging across a spark gap:
(a) plain gaps, (b) lightning arresters, also called surge diverters
2. Other means of reducing overvoltages at their source: (a) converter
control system, (b) damping circuits, (c) dc reactor, (d) surge capacitors,
(e) overhead shield wires
Some of these means are discussed further.
Plain Spark Gap. The rod gap is simple, cheap, and rugged. It is not
damaged by discharge of great currents. However, it has two disadvantages:
1. Its sparkover voltage varies considerably with rate of rise of voltage,

also called steepness of wave front, and with atmospheric conditions. This
characteristic requires a large spread between normal working voltage and
breakdown voltage of the insulation to be protected.
2. A discharge through a plain gap constitutes a short circuit that can be
extinguished only by removal of the working voltage; for example, by block
ing the converters.
The sparkover voltage of a sphere gap is more nearly constant under good
atmospheric conditions but is influenced by dust and water drops.
Screened rod gaps have been developed that have consistent sparkover
voltages and are not sensitive to dust, etc. The screened rod gap and the
sphere gap, however, like the plain rod gap, short-circuit the power system.
Lightning Arresters have nonlinear resistors in series with precise spark
gaps. The resistors keep the voltage across the arrester within satisfactory
limits for a wide range of discharge currents: the voltage neither falls to zero
nor rises greatly above the sparkover value. The arrester " seals off," that is,
becomes nonconducting, when the voltage across it drops to a certain value
higher than the normal working voltage. Ac lightning arresters, like ac circuit
breakers, depend on the normal current zeros for arc extinction and cannot
extinguish a dc arc. Hence special lightning arresters are required for dc
circuits. Such arresters have been developed recently, and are described on
page 288. Before dc lightning arresters were available, ac arresters were
sometimes used on hv dc circuits to give precise sparkover, and the arc
was extinguished by grid control.
7-8
285
Two limitations of the duty on lightning arresters must be observed:
1. A limit to the energy dissipated by the arrester in a short series of dis

charges, so that the resistor is not damaged
2. A limit to the crest current so that the electrodes of the spark gap are
not excessively eroded, with consequent increase in the sparkover voltage
Protection against Lightning

An overhead dc transmission line is protected against lightning in the same
manner as an overhead ac line. Flashovers of line insulation caused by
lightning seldom do any permanent damage, but they cause short dips in
voltage while the line is automatically deenergized to allow the power-follow
arc to be extinguished and then reenergized. The number of flashovers can
be greatly reduced by shielding the power conductors by placing grounded
conductors above them to intercept most of the lightning strokes. In addition,
the tower-footing resistance must be kept low, so that only a very few strokes
having exceptionally high crest currents raise the tower potential so high that
it "flashes back" to a power conductor.
Because the direct working voltage is significant in comparison with the
tower potential and because the great majority of lightning strokes to trans
mission lines have negative polarity, .most of the flashovers are to the positive
conductor of a bipolar line.

The switchyards of converter stations must be shielded against direct
strokes of lightning. Overvoltages can still come into the station on overhead
ac and dc lines, and the station equipment must be protected against overvoltage waves arriving on the lines. Standard ac lightning arresters are used
on the network side of the converter transformers. These arresters protect
other equipment, such as filters, unless situated too far away, in which case
additional arresters should be placed near the equipment to be protected.
Before suitable dc lightning arresters were available, it was customary to
connect spark gaps between the high-voltage dc conductors and ground near
the line side of the dc reactors for limiting the overvoltages from lightning
surges. If a gap flashed over, it was necessary to deenergize the line, as
described in Section 7-6, for extinguishing the dc power arc. In order to
decrease the frequency of occurrence of flashovers of the gap, a capacitor
(about 1 fiF) was connected from each pole to ground. These surge capacitors
decrease the rate of buildup of overvoltage, and, depending on the capa
citance and the charge in a surge, they limit the crest values of most lightning
surges, so that few of them cause flashovers. The capacitors also decrease the
rate of rise of voltage of the surge.
Electrode lines (Section 9-6) are similarly protected, but with larger surge
capacitors (5 to 10 pF).
286
7-8
PROTECTION
Each dc reactor should be protected by a lightning arrester across it. Since

the normal voltage across such a reactor consists of harmonics but has
negligible dc component, ac arresters are suitable.
Lightning surges arriving at a terminal from an overhead line are barred
by the dc reactors from reaching the converters unless the wave front is so
steep that the reactor appears capacitive or unless the lightning arrester con
nected across the reactor discharges. For protecting the station equipment
in such an event, a spark gap or dc arrester is connected from pole to ground
on the station side of the reactor. In most stations a spark gap here should
have an acceptably low probability of flashover.
Protection against Current Chopping
Arc quenching caused by ion starvation in a mercury-arc valve can result
in high rate of decrease of current (di/dt) in the inductances of the transformer
and the dc reactor, producing a high inductive voltage (L di/dt) across these
inductances and across the valve. The transformer and the dc reactor are
protected against such voltages by ac lightning arresters. The valve is pro
tected by a series of spark gaps associated with the voltage divider (Vol. 2,
Chapter 14). These gaps recover their ability to withstand normal valve
voltage as soon as the valve conducts for a short while.
Operation of a converter with discontinuous (intermittent) current pro
duces overvoltages for the same reason as does arc quenching, but since it
would occur six times per cycle in a six-pulse converter, it cannot be tolerated.
It is therefore prevented by the minimum-current control (Section 5-13).
287
The sparkover voltage of the gaps of the lightning arresters across the dc
terminals of each bridge should be set high enough to prevent sparkover at
zero overlap, although normally this condition is prevented by the minimum
current limit (Section 5-13).
Sudden Loss of Load on Rectifier
Assume a rectifier to be operating at rated current and voltage. If the

decreased to zero or to the minimum current limit more rapidly
than the tap changer could act and if the alternating voltage could remain
constant, the direct voltage would rise by RcIdn, a typical value of which is
9%. However, because of the decreased reactive current drawn by the recti
fier through the inductive reactance of the power system, the alternating
voltage also rises by an amount that depends on this reactance. This rise
might amount to an additional 9% if the short-circuit capability of the ac
network were six times the converter rating, giving a total rise of, say, 18%
in the direct voltages.
The overvoltages are corrected automatically as follows: The powerfactor capacitors are switched off in a fraction of a second, restoring the
alternating voltage to within a few percent of its rated value. The transformer
tap changer lowers the direct voltage to normal in about 1 min.
The flashover voltage of the lightning arresters from dc poles to ground on
both sides of the dc reactors should be substantially higher than 1.2 times
rated voltage. The voltage dividers should also be designed to withstand this
short-time overvoltage.
current were
Overshoot of Direct Voltage of a Bridge Caused by Short Overlap12
The overshoot of the voltage across a valve at the end of its commutation
and the diminution of this overshoot by valve dampers were considered in
Section 7-3.
The voltage across the dc terminals of a bridge, being always equal to the
sum of the voltages across any pair of valves on the same phase, has voltage
jumps with overshoot. In each cycle of the power-system frequency there are
12 jumps in the direct voltage of the bridge, six associated with ignitions and
six with extinctions. The two kinds of jumps generally differ in magnitude,
but as the overlap decreases, the two magnitudes approach equality. At very
short overlaps, the extinction jump occurs before the voltage oscillation due
to the ignition jump has died out, and at exactly zero overlap the two jumps
merge into one jump twice as high. The corresponding overshoot is the same
percentage as that of a single jump but twice as high in absolute value.
Transformer Overvoltage Protection
The converter transformers are usually connected in grounded Y on the

network side. These windings are protected in the manner customary for
power transformers : by an ac lightning arrester connected from each phase
to ground. The valve-side windings of half of the transformer banks are con
nected in ungrounded Y, those of the other half of the banks in delta. The
voltage across every valve-side winding is alternating, but the voltages from
such windings to ground have both alternating and direct components. The
valve-side windings are protected by ac lightning arresters connected phase
to phase and by dc lightning arresters connected phase to ground. Sometimes
an additional dc lightning arrester is connected from the neutral point of
each set of Y-connected valve-side windings to ground.
288
7-8
PROTECTION
289
Protection failure
DC Lightning Arresters23,27
Two special difficulties are encountered in applying lightning arresters to
the protection of the dc part of converter stations against overvoltages :
1. The current through the arrester does not have natural current zeros,
as alternating current does, to aid in resealing the arrester against sustained
direct voltage.
2. The greater part of the apparatus requiring protection and having
direct voltage is buffered by significant lumped inductances from the sources
of overvoltage surges. Overvoltages reaching the converter from the dc line
must pass through the dc reactor; those originating in the ac system, through
the transformers.
With respect to the first difficulty, it should be noted that one kind of modern
lightning arrester used for the protection of ac systems has, in series with
blocks of nonlinear resistance material, spark gaps so designed that they
develop a countervoltage, increasing with time and helping to bring the
arrester current to zero very soon after the subsidence of the applied overvoltage and even before the occurrence of a natural current zero. This same
kind of gap, called an active gap, is used in one kind of dc arrester for
forcing the arrester current to zero. If the overvoltage protection is not to be
jeopardized, however, the current must be brought to zero before the arrester
voltage becomes higher than the desired protective level.
Here the second difficulty appears. The arrester current is supplied through
a reactor (Figure 26), and the current through the reactor is proportional,
not to the voltage across the reactor, but to the time integral of this voltage.
The result is illustrated in Figure 27. The broken line vL there represents the
incoming voltage surge on the line side of the reactor. As long as there is
no changing current in the reactor, there is no voltage across it, and the same
voltage appears across the arrester. When this voltage at time attains a
sufficient value, however, the arrester's gap sparks over. Thereafter, the
voltage va across the arrester, shown by a solid curve, differs from the voltage
vL, and current ia flows through the arrester. The difference between the
L.A.
VL
Fig. 26. Overvoltage surge entering the converter station through an inductance.
Protection level
Normal operating voltage
-<Resealing failure
Fig. 27. Single-stack lightning arrester failing to protect and reseal for a switching surge
entering the protected zone through a reactor. Area I> area II. (From Ref. 27, by per
mission.)
ordinates of the two voltage curves represents the voltage across the reactor,
and the area between the curves is the time integral of this voltage to which
the arrester current is proportional.
The curves in Figure 27 show that the arrester voltage can exceed the
desired protective level before the arrester current is suppressed. Moreover,
they show that, because the arrester voltage reaches a peak and then declines,
it may drop below normal operating voltage before the arrester current
becomes zero, with the result that the arrester fails to reseal and is destroyed
by overheating.
A special dc lightning arrester has been developed,27 having two or more
stacks connected in parallel, each stack containing nonlinear resistors and
active spark gaps. The several stacks carry current in sequence, with results
as illustrated in Figure 28. When the voltage across the first stack to conduct
reaches a certain value a little below the protective level, the second stack is
triggered and begins to conduct. Because this stack has a countervoltage
lower than that of the first stack, all the current is immediately transferred
from the first stack to the second. When the voltage across the second stack
reaches a value approaching the protective level, the current is transferred
either to a third stack or back to the first stack, depending on the number of
stacks provided. (This number is chosen so as to provide the energy-absorbing
capacity required by the particular application.) While one stack is conduct
ing and being heated, the others are being cooled. Thus it is possible to
extinguish the arrester current without the arrester voltage's ever exceeding
the protective level and without overheating, which would otherwise cause
erosion of the gap electrodes and consequent change of their sparkover
voltage.
PROBLEMS
291
Summary of Overvoltage Protection

Vl
The only overvoltages serious enough to flash over gaps or lightning
Protection levei
arresters are these :
Normal
operating
voltage
1. Lightning sprges arriving from ac or dc overhead lines

2. Switching surges arriving from ac lines
3. Alternating voltage of one bridge impressed on the dc line if the line
is resonant to the power frequency or nearly so
4. Chopping or quenching of direct current.
-t
The usual locations of lightning arresters and gaps are shown in Figure 29.
t
Fig. 28. Multiple-stack lightning arrester successfully protecting and resealing against
a switching surge entering the protected zone through a reactor. (From Ref. 27, by
permission.)
(va vL) dt = 0
Dc reactor
/TOTV
Oc line
Converter
transformers
Ac bus
rr
Electrode line
Fig. 29. Locations of lightning arresters and gaps on one pole of a converter station
having two bridges per pole (single-line diagram) : 1. Ac arresters on ac side of transformers.
2. Dc arresters, phase to ground on valve side of transformers. 3. Ac arresters, phase to
phase on valve side of transformers. 4. Dc arrester, neutral to ground on valve side of
transformer. 5. Gaps across valves. 6. Dc arrester, pole to ground. 7. Ac arrester across
dc reactor. 8. Arrester from electrode line to ground. 9. Small inductor. 10. Surge capacitors.
PROBLEMS
1. Analyze the ripple in the direct current during 12-pulse operation, and
verify the numerical coefficient given in the text.
2. Analyze the ripple in the direct current during mixed 6- and 12-pulse
operation (two bridges with YY-connected transformers and one bridge
with YA connection).
3. Analyze the ripple in the direct current during six-pulse operation with
both ignition delay and overlap taken into account.
4. Show that when a capacitance C, initially charged to voltage V0, has its
charge and stored energy modified by a change of voltage + VJ} the
energy dissipated in any series resistance is CVj/2.
5. Show that the wave form of the high-frequency oscillation of voltage
across the bypass valve immediately after a voltage jump across it is the
same as that across a working valve.
6. Investigate the natural frequencies in the wave of voltage across a valve
immediately after ignition of another valve. Does it matter which other
valve ?
7. Assuming that a phase-to-ground short circuit does not change the other
two phase-to-ground voltages, find the reduction in direct voltage of a
rectifier at light load.
8. Find the power loss per valve damper of the inverter in the dc link in
Example 7 on a pole operating with two bridges instead of three at each
end of the line. Assume rated current and, at the rectifier, rated voltage
per bridge. Assume also that the alternating voltage on the valve side of
the inverter transformers is the same as in the example. Does this loss
exceed the damper loss during rectifier operation at rated current and
voltage ?
292
PROTECTION
BIBLIOGRAPHY
15. "Commutation Oscillations in an H.V.D.C. Converter," by R. Feinberg and W. Y.

Chen, l.E.E. Conference Publication no. 22, High Voltage D.C. Transmission, Man
chester, Sept. 19-23, 1966, Part 1, Paper No. 41, pp. 204-206.
16. "Some Special Equipment for D.C. Transmission," by F. G. Goodrich and S. Swensen, ibid., Paper No. 47, pp. 235-240.
17. "Methods of Damping," by L. A. Harris, ibid., Paper No. 51, pp. 255-264.
18. "Impulse Withstand Levels, Clearances, and Creepage Distances for Convertor
Station Layouts," by L. A. Harris and C. W. Mott, ibid., Paper No. 59, pp. 289-292.
19. "Transient Voltage Levels on HVDC Conductor Systems," by J. D. Ainsworth and
J. P. Bowles, ibid., Paper No. 60, pp. 293-296.
20. " Co-ordination of InsulationLevels in HVDC Electrical Installations," by F. H. Last
and A. Stalewski, ibid., Paper No. 61, pp. 297-303.
21. "Some Features of New Zealand's Inter-island HVDC Transmission," by R. J. Fyfe,
M. A. Louden, J. Noble, and D. G. Young, ibid., Paper No. 78, p. 376.
22. "Earth Fault Clearing on an HV DC Transmission Line, with Special Consideration
of the Properties of the DC Arc in Free Air," by A. Kohler, l.E.E.E. Trans, on
P.A. & S., Vol. 86, pp. 298-304, March 1967.
23. " Overvoltage Protection of HVDC Converter Stations," by K. E. Russell, Trans. &
Dist., Vol. 19, No. 9, pp. 91-93, September 1967. Includes brief description of new
G.E. lightning arrester for hv direct current.
BIBLIOGRAPHY
1. "Arc Quenching in Switchgear for Disconnecting High-Power EHV DC Circuits,"
by G. A. Kukekov and P. G. Sorokin, Direct Current, Vol. 1, pp. 213-216, March
1954. Translated from Elektrichestvo, No. 10, pp. 20-24, October 1953.
2. "Protective Means for H.V.D.C. Transmission," by E. Uhlmann, U.S. Patent No.
2,730,667, Jan. 10, 1956.
"Switchgear
for H.V.D.C. Lines," by G. A. Kukekov, P. G. Sorokin, and N. A.
3.
Shipulina, Direct Current, Vol. 4, pp. 123-126, June 1959. From Elektrichestvo,
Vol. 79, pp. 24-27, March 1959.
Hingorani, Garraway, London, 1960, Chapter 6, "Protection of H.V.D.C. Systems,"
pp. 85-114, 184-186.
5. "Clearing of Earth Faults on H.V.D.C. Overhead Lines," by Erich Uhlmann,
Direct Current, Vol. 5, pp. 45-47, 65-66, September 1960. Correction on p. 103,
December 1960.
6. " Lightning Overvoltages on Long D.C. Cables Connected to Overhead Transmission
Lines," by S. Rusck, Elteknik (Stockholm), Vol. 3, No. 7, pp. 107-109, September
1960.
7. "Disconnection of Branch Lines from D.C. Transmission Lines," by N. D. Leshukov
and N. A. Shipulina, Direct Current, Vol. 7, pp. 39-46, 58, February 1962. Reviews
previous work on dc circuit breakers; recommends a certain scheme, and gives cost
estimates.
8. "A New Approach to H.V.D.C. Switchgear: Part I, Theoretical Appraisal of a
Rotating Plasma Machine as a Switch," by J. J. Matthews, Direct Current, Vol. 8,
pp. 184-189, July 1963; "Part II, Experimental Investigations on a Rotating Plasma
Machine," by J. J. Matthews, R. L. May, and H. Taylor, pp. 216-220, August 1963.
(The names of the last two authors were omitted, and a correction was printed in
Vol. 9, p. 25, February 1964.)
9. " Commutation Phenomena in a Static Power Convertor," by R. Feinberg and W. Y.
Chen, l.E.E. Proc., Vol. Ill,pp. 125-134, January 1964. Experimental and theoretical
study of current and voltage oscillations in a three-phase one-way rectifier.
10. "Mercury-Arc Valves for High-Voltage D.C. Transmission," by Uno Lamm, l.E.E.
Proc., Vol. Ill, No. 10, pp. 1747-1753, October 1964. Includes some information on
valve dampers and anode reactors on p. 1752.
11. High Voltage Direct Current Convenors and Systems, edited by B. J. Cory, MacDonald, London, 1965. Chapter 7, "Filters, Damping Circuits and Reactive VoltAmps in H.V. D.C. Convertors," by J. D. Ainsworth, pp. 137-174.
12. "Some Design Aspects of the Cross-Channel Power Link," by L.A.Harris, ibid.,
Chapter 8, esp. pp. 181-195, 205-206, 211-215.
13. "Testing a New Type of Circuit-breaker for H.V.D.C.," by K. Mollenhoff, D.
Morghen, J. Salge, and W. Pucher, Direct Current, Vol. Ill,pp. 3-6, February 1966.
Oil blast, experimental, 645 A, 85 kV.
14. "H.V.D.C. Circuit Breakers Using Oscillating Current Techniques," by Naohei
Yamada and Tung Hai Chin, Direct Current, Vol. 11, pp. 87-89, 92-98, and 107,
August 1966.
|
i'
'1
i
:
|
!
|
f
!
>
i
293
24. "Some Aspects of Insulation Co-ordination and Overvoltage Protection on D.C.

Systems," by L. Csuros, Electra (C.I.G.R.E.), No. 3, pp. 17-23, December 1967.
-25. "Transient Fault Currents in HVDC Rectifiers," by A. G. Phadke, D. K. Reitan,
and H. A. Peterson, l.E.E.E. Paper No. 31 CP 67-441, 1967. See also Ref. 29.
26. "HVDC Transmission: Large DC Infeeds Relative to AC Systems; Considerations of
Higher Voltages; Thyristors," by Aleksa Gavrilovic, l.E.E.E. Publication 68 C-57,
EHV Conference at Montreal, 1968. Section on "Converter Stations for High
Voltages," pp. 2-3.
27. "Present and Future Status ofHVDC Transmission," by H. Martensson and B. Skoglund, l.E.E.E. Publication 68 C-57, EHV Conference at Montreal, 1968. Section on
" Insulationand Overvoltage Protection," especially the part on dc lightning arresters.
28. "Isolationskoordination in Hochspannungs-Gleichstrom-Obertragungs-Anlagen"
(Insulation Co-ordination in hv dc Plants, in German), by Werner Heise, Dusan Povh,
and Ewald Sarbach, ETZ, series A, Vol. 89, No. 9, pp. 204-208, April 26, 1968. English
translation available from Arbeitsgemeinschaft HGO.
-29. "Fundamentals of HVDC Interruption" (in French and English), by W. Pucher,
Electra (C.I.G.R.E.), No. 5, pp. 24-38, June 1968.
30. "Some Aspects of Insulation Co-ordination in H.V.D.C. Systems," by E. Sarbach,
Brown Boveri Rev., Vol. 55, No. 7, pp. 344-351, July 1968.
31. "Circuit Breaker for HVDC Transmission," by D. Kind, E. Marx, K. Moellenhoff,
and J. Salge, C.l.G.R.E. Report 13-08, 1968, 9 pp. Oil-blast type.
"Y:32. "Transients in EHV DC Power Systems: Parti, Rectifier Fault Currents," by Harold
A. Peterson, Arun G. Phadke, and Daniel K. Reitan, I.E.E.E. Trans, on P.A. & S.,
Vol. 88, pp. 981-987, July 1969. Disc., pp. 987-989. Similar to Ref. 25.
33. "Lightning Surges on D.C. Cables Connected to Overhead Lines," by Sune Rusck,
Direct Current (new series), Vol. 1, No. 1, pp. 41-44, 1969.
294
PROTECTION
34. " Overvoltages Appearing on H.V.D.C. Cables Due to Converter Operation," by

3. D. Ainsworth, Arne Johanssen, and K. J. Ralls, Direct Current, Vol. 1 (new series),
No. 2, pp. 83-88, August 1969.
35. " Multiterminal Operation of HVDC Converter Station," by Reiner Foerst, Gerhard
36.
37.
38.
39.
Heyner, Karl Werner Kanngiesser, and Hermann Waldmann, I.E.E.E. Trans, on

P.A. & S., Vol. 88, pp. 1042-1050, July 1969. Disc., pp. 1050-1052. Need for a dc
circuit breaker and tests on one, pp. 1047-1049.
"Overvoltage Protection of HVDC Converter Stations," by K. E. Russell, Trans. &
Dist., Vol. 21, No. 7, pp. 92-95, July 1969.
"Some Considerations on Insulation Level and Testing Procedure of H.V.D.C. Converter Installation," by Y. Kawaguchi, S. Kobayashi, and Y. Sawada, I.E.E.E. Trans.
on P.A. & S., Vol. 89, No. 6, pp. 1134-1148, July/August 1970. Disc., pp. 1149-1151.
Covers capacitive and inductive surge transfer from ac side, surge transfer from dc
side, and impulses from faults in the station.
"Analysis of Transient Short-Circuit Currents in H.V.D.C. Power Systems," by J.
Reeve and S. C. Kapoor, I.E.E.E. Paper 70 TP 533, July 1970.
"A Method of Current Interruption in HVDC Networks by Means of AC CircuitBreakers with Adapted Arc Characteristics and Energy Absorbers," by E. v. Bonin,
B. Koetzold, K. Moller, G. Oberdorfer, J. Schwartz, and H. G. Thiel, C.I.G.R.E.,
Report 13-10, 1970, 15 pp.
|
).
8
Harmonics and Filters
!
\
8-1 SUMMARY
Converters generate harmonic voltages and currents on both ac and dc

sides. A converter of pulse number p generates harmonics principally of
orders
(1)
h=pq
on the dc side and
h=pq 1
(2)
on the ac side, q being any integer. Most hv dc converters have pulse number
6 or 12 and thus produce harmonics of the orders given in Table 1. The
Table 1. Orders of Characteristic Harmonics
Pulse No.
p
6
12
j.
DC Side
AC Side
pq
0, 6, 12, 18, 24, ...

24,...
0, 12,
pq\
1, 5, 7, 11, 13, 17, 19, 23, 25 ..

11,13,
1,
23, 25...
amplitudes of the harmonics decrease with increasing order : the ac harmonic

current of order h is less than Ii/h, where /t is the amplitude of the funda
mental current.
Unless measures are taken to limit the amplitude of the harmonics entering
the ac network and the dc line, some of the following undesirable effects
may occur : overheating of capacitors and generators, instability of the con
verter control, and interference with telecommunication systems, especially
noise on telephone lines. These effects may not be confined to the vicinity
of the converter station but may be propagated over great distances. The
most difficult of these to eliminate is telephone interference.
295
296
HARMONICS AND FILTERS
The principal means of diminishing the harmonic output of converters are

(a) increase of the pulse number and (b) installation of filters. High pulse
numbers have been used in some converters, but it is the general opinion
that for hv dc converters the use of filters is more economical than increase
of the pulse number beyond 12. Filters are nearly always used on the ac side
of converters. Ac filters serve the dual purpose of diminishing ac harmonics
and supplying reactive power at fundamental frequency. On the dc side, the
dc reactor diminishes harmonics, and, in many converters, especially those
connected to dc cables, no additional filtering is required on the dc side.
Dc filters are required, however, on some overhead dc lines.
8-2 CHARACTERISTIC HARMONICS
Definitions and Assumptions
The pulse number of a converter is the number of nonsimultaneous com

mutations per cycle of fundamental alternating voltage.
The order of a harmonic is the ratio of its frequency to the fundamental
(lowest) frequency of a periodic wave. The order of harmonics on the dc side
of a converter, however, is defined with respect to the fundamental frequency
on the ac side.
Characteristic harmonics are those of orders given by Eqs. (1) and (2) in
Section 8-1.
Noncharacteristic harmonics are those of other orders.
Assumptions The following assumptions are made as bases for deriving

the orders, magnitudes, and phases of the characteristic harmonics of a six-
-8-2
CHARACTERISTIC HARMONICS
297
1. The alternating voltage has no harmonics except the first (the funda
mental).
2. The direct current has no harmonics.
There can be higher harmonic currents on the ac side and harmonic
voltages on the dc side, however, and deductions are made concerning these.
Because of assumptions 1, 3, and 4:
3. The overlap angle is the same for every commutation.
4. The ripple of the direct voltage has a period of one-sixth that of the
alternating voltage.
5. Hence the harmonics of the direct voltage are of order 6 and its multiples
12, 18, 24, etc.
6. The alternating currents of the three phases have the same wave shape
but are displaced by one-third cycle in time (120 of the fundamental).
7. The alternating currents have positive and negative parts of the same
shape except that one is inverted; that is, F(9 + 180) = F(9).
8. As a result of deduction 7, there are no even harmonics in the alternating
current.
9. As a result of deduction 6 and the fact that the phase difference for the
Ath harmonic is h times that for the fundamental, the ac harmonics have the
following phase sequences :
Sequence
Orders (A)
Zero (0)
Positive (1)
Negative (2)
0, 3, 6, 9, 12, 15, 18, 21, 24, ... , 3q

I,4, 7, 10, 13, 16, 19, 22, 25, ..., 3q + 1
2, 5, 8, 11, 14, 17, 20, 23, 26, ..., 3q - 1
pulse converter :
Harmonic analysis of the wave shape of the alternating current shows that
1. The alternating voltages are three-phase, sinusoidal, balanced, and of

positive sequence.
2. The direct current is absolutely constant, that is, without ripple. Such
current would be the consequence of having a dc reactor of infinite inductance.
3. The valves are ignited at equal time intervals of one-sixth cycle, that is,
at constant delay angle a measured from the zeros of the respective commutating voltages. By assumption 1, these zeros are equally spaced in time.
4. The commutation inductances are equal in the three phases.
10. No characteristic harmonics of order 3g (triple harmonics) can exist.
Final Conclusions on Orders of Characteristic Harmonics. By deduction 5,

the direct voltage has only harmonics of orders that are multiples of 6, that
is, of orders 6q, where q is an integer.
By deductions 8 and 10, the alternating currents have only odd harmonics
of orders not multiples of 3. Those of orders 6q + 1 have positive sequence,
and those of orders 6q 1 have negative sequence.
Deductions from the Foregoing Assumptions
AC Harmonics at No Overlap
From assumptions 1 and 2 immediately follow deductions 1 and 2, respect

ively :
The wave shapes of alternating voltages and currents conforming to the

assumptions made are shown in Figure 1. The current waves drawn in solid
lines are for any ignition delay angle a but no overlap. The broken curved
298
8-2
299
Fig. 2. Trains of positive and negative rectangular pulses of arbitrary width w.
The general trigonometric form of the Fourier series is

a
F(d) =
L
00
(Ah cos hd + Bh sin hd)

+h=l
(3)
where
A0=~rF(e)d(9)
(4)
71 0
-7c
fin
71
Bh = -71
7A
Fig. 1. Wave forms in a six-pulse bridge: line-to-neutral voltages ea, eb, ec and line
currents ia, ib, ie with YY-connected transformer; also line current IA with AY-connected
transformer.
lines show qualitatively how overlap would modify the fronts and tails of the
current pulses.
Valve Currents and Line Currents on Valve Side. The line-current wave
forms at no overlap are a series of equally spaced rectangular pulses, alternately
positive and negative. Fourier analysis of such a wave shape, for finding the
characteristic alternating-current harmonics in this limiting case, is very
simple; it also serves to illustrate several features of these harmonics. How
ever, let us take an even simpler starting point: the analysis of a train of
positive rectangular pulses of unit height and arbitrary width w radians, that
is, of duration w/co sec (see Figure 2). These pulses might represent the
current through one valve.
F(d) cos hd d9
(5)
F(d) sin hd dd
(6)
Jo
The limits of integration in Eqs. (4), (5), and (6) can be taken more generally
as o and <r + 2n, where o is any angle. A0j2 is the average value of the function
F;Ah and Bh are rectangular components of the hthharmonic.The correspond
ing phasor is
(7)
Ah-jBh=C[&,
where
and
Ch = Ja\ + Bl = crest value

(f>h = tan ~1
Ah
If, in the analysis of the wave shown in Figure 2, the origin of d is taken at
the center of a pulse, F(d) is an " even " function, and Bh = 0 for all h; that is,
the series has only cosine terms. Their amplitudes are found by Eq. (5) thus:
I r
Ah = 71-
J
F(d) cos hddd = -
J -n
1 T . hw
71
.(
hw
T2
C+H-/2cos hd dO
J -wjl
=
2 . hw
(8)
300
8-2
Also
T-i/*_>-
The series is therefore

w
2w
1
2 fw
cos 29
Ft(d) = 1 - + sin- cos 9 + - sin
-
3w
4w
----
cos 30 + - sin cos 40 H

+ 3- sin
2
2
4
(10)
In general, this series has a constant term and cosine terms of every harmonic
frequency. For certain pulse widths, however, certain cosine terms vanish.
This occurs if
hw
2qn
w=-
or
=qn
n\
--4
. w
h sin cos 9
2
.
-2 sin
3w
2
2w
H sin cos
5w
cos 50 H
} (13)
P '
The constant term and all even harmonics have vanished.

Let us now put w = 2nj3 and change the height to Id. For increments of
2 in h, the arguments of the sines increase in increments of 2nj3 rad. For odd
h the sines are all yj3j2, except for triple harmonics, which are zero. The
series is
--
2 v3
=-
7t
Id[/cos 0 -1 cos 50 + -1 cos 70 1 cos 110

5
11
7
\
+
cos 130
cos 170 + ~ cos 190
(14)
This contains only harmonics of orders 6q 1, as predicted earlier. The crest

value of the fundamental-frequency current is
ho =
= ~h= ojmd
Ij
V2 n
(15)
(12)
(16)
.The effective value of the hih harmonic is
ho =
This result can be obtained in at least two ways: (a) by putting the new
function into Eqs. (4), (5), (6) and performing the indicated operations or
(b) by appropriate changes in series 10. These changes are the following:
(1) Shift the pulse n radians; this shifts the fundamental component n rad and
shifts the higher harmonics by hn rad. If h is even, cos (0 hn) = cos 0; but
if h is odd, cos (0 hn) = cos 0. Hence the signs of all odd harmonics are
changed. (2) Invert the pulse. This changes the sign of every term. The net
result is to change the signs of all even-order terms, including the constant
term.
3w
and its effective or rms value is
cos 29
\
1
4w
30 - sin cos 40 +
4
2
/
..., 3q, sin (hw/2) = sin qn = 0. Then the series lacks

harmonic
and
its multiples, called triple harmonics for brevity.
the third
Now, if we consider the negative pulses only, shown by broken lines in
Figure 2, we get
w
4/
cos 0 + sin - cos 30 + - sin

F33 = F,1 + F2 = - ( sin
3
5
2
2
7t\ 2
= =1-103/,
271
so that if h = 3, 6, 9,
2/
F2(9) = - 1
--
Next, let us analyze the train of alternately positive and negative rectangular
pulses. Its Fourier series is
(11)
For example, the pulses of valve current in the three-phase bridge current
have width
W =
301
I-f
(IV)
Series 14 represents the ac line current of phase a on the valve side of the
transformer (Figure 1) if the origin of 0 is taken at the center of the positive
pulse (axis la). The currents ib and ic in the other two phases have the same
wave shape as ia but are displaced 120 (= 2nj3 rad) behind and before ia,
respectively. Their Fourier series, if written for 0 = 0 at axes Ib and Ic, re
spectively, are the same as that for ia written with respect to axis Ia. Likewise,
these series are independent of the ignition delay angle a. If any wave is
shifted forward by an angle (f> , measured for the fundamental period, the hih
harmonic is shifted by h<}> measured for the shorter harmonic period, being
shifted forward if of positive sequence and backward if of negative sequence.
Line Currents on Network Side of Six-pulse Group. If the transformers
are connected YY or AA and have ratios 1: 1, the line currents on the network
302
side have the same wave shape, hence the same harmonics, as those on the
valve side. If, however, the transformers are connected YA or AY, the wave
shape on the network side is different from that on the valve side.
Let the transformers be connected in Y on the valve side and in A on the
network side, and let the ratio of each individual transformer be 1:1. Then
the currents in the delta-connected windings are the same as those in the
corresponding Y-connected windings. Each line current on the delta side is
the difference of two delta currents ; for instance,
ico
P lio
iro
00
If)
len Ico"
T-J"
w
> >
> >
a
Ss
<
T-
*
w
O
co
Line current iA at the bottom of Figure 1 is constructed graphically from the

two waves above it. Let us find its Fourier series with respect to 0 = 0 at the
center of its positive part (axis IA). With respect to this same axis, ib is re
tarded 30 and ic is advanced 30. Table 2 shows the magnitude and phase
of each harmonic component of these three current waves.
If, instead of the ratio of individual transformers being 1: 1, the bank ratio
is made 1:1, then the factor \j3 is removed from every entry in the last
column of the table, and the Fourier series becomes
/s
(18)
*'a = h - ic
iA = 1.103/d(cOS 9 + L COS 50 y COS 70 yy cos 110 +

cos 130 + -fa cos 170 - fa cos 190
im
.O
00
co
IT)
en
(N
+ + +
en
<N
On
+ + +
m
p.
r-H
en
1i
i
.
en
F5
11
/"S
r-
On
N
fi

(19)
&
O
%
It differs from Eq. (14) only in the signs of the terms representing the fifth, seventh, seventeenth, nineteenth, etc., harmonics. Although all the harmonics
in Eqs. (14) and (19) are equal in amplitude, the two series represent different
wave shapes because of the difference in signs (or in phase) or certain orders
of harmonics.
Alternating Line Currents on Network Side ofl2-pulse Converter. A 12-pulse

group in a hv dc converter is composed of two 6-pu3se groups fed from sets
of valve-side transformer windings having a phase difference of 30 (or 90)
between the fundamental voltages. Since a is normally the same for both
6-pulse groups, the fundamental valve-side currents have the same phase
difference as the voltages, and the fundamental network-side currents are in
phase with one another. The no-load voltage ratios between the network-side
windings and each of the two sets of valve-side windings are equal ; hence the
fundamental network-side currents are also equal.
The resultant network-side current of the two groups is then given by the
sum of Eqs. (14) and (19). To keep the power rating of the 12-pulse converter
equal to that of a 6-pulse converter, however, both the direct voltage and the
alternating current of each of the two bridges of the 12-pulse converter should
o
o
ON
CI
CO
CO
O
ID
r-
i/1
<*3
CO
~"
co
>-< c
83
I-.
a*
-H
CO
ON
(S
-i
-<
OO
<-
co
co
1
j-V1
</->
rv-i
OO
00
4)
*o
p-
304
--
be half of the corresponding quantities of the comparable one-bridge 6-pulse

converter. Half of the sum of Eqs. (14) and (19) is
i12 = 1.103/rf(cos 0 tt cos 110 + yj cos 130

~Y3
cos 230 + 2V cos 250
) (20)
This contains only harmonics of orders 12q + 1. Currents of orders 5, 7, 17,

19, etc., circulate between the two banks of transformers, but do not enter
the ac line. If there are two valve windings and one network winding on each
transformer, these harmonics appear only in the valve windings. In practice,
such harmonics in the two 6-pulse groups are not always exactly equal in
magnitude nor exactly in phase opposition; hence their cancellation is in
complete, and they appear to some degree in the network-side line currents.
They are uncharacteristic harmonics in a 12-pulse converter.
Figure 3 shows the wave shapes of current in each component 6-pulse
group and in the 12-pulse group. The fundamental waves are shown in
broken lines.
305
AC Harmonics at Overlap
In Figure 1 the wave shapes for positive overlap appear as better approxi
mations to sine waves than do the wave shapes for no overlap. Hence we make
the qualitative deduction that the effect of overlap is to decrease the ampli
tudes of the harmonics.
Quantitative results are computed from the following formulas. They are
valid only for characteristic orders h. For overlap not exceeding 60, the
complex rms value, with phase referred to the respective commutation
voltage E is
Ih KtFx(a, 5, h)
(21)
amperes
where
=
(E\ = V64
<22>
and
j (h l)a / (A 1)5
/-(h + l)a-l-(h + l)5
- --Fx = h 1
h 1
(23)
Sometimes it is convenient to express the harmonic as a fraction of one of the

following currents :
hi = yJzflEjX = crest value of ac component of current in line-to-line short
circuit on valve side.

base
= rms fundamental alternating current corresponding to
~
h Is2 with no overlap.
ho = rms fundamental alternating current with no overlap.
ho = holh = rms harmonic current with no overlap.
Id = direct current.
The results are as follows:
-f=K2Ft
ls2
(24)
(25)
where
Fig. 3. Alternating line currents of a two-bridge 12-pulse converter with no overlap:

(a) current of six-pulse bridge with YA-connected transformer; (b) current of six-pulse
bridge with YY-connected transformer; (c) total current.
base
= K3Fx
per unit
(26)
306
8-2
307
where
where
**=k
(27)
-f =
(28)
/fsin[(A - l)w/2])2
Mr
MO
where
-2/hr
<29>
i
= K5F
(30)
ln
OD
ho
where
K> =
i
t = K6Fi
(32)
*'=23
(33)
h~
[sin[(A + l)uj2]}2
h+ 1
) +(
h- 1
-2(sinp-l)M/2]}{sin[(lM2]|cos(2g + K)y/2
(41)
The last equation has the same form as the law of cosines for the length of
one side of a triangle in terms of the lengths of the other two sides and the
included angle.
Computed results for 4/4 0 versus u are plotted in Figures 4 to 11. Results
for l'h = 4/4ase versus l'd at the usual value of a or y, 15, are plotted in
Figure 12.
Overlap Greater than 60. In the region bounded by 60 30, and 8 < 150, Eqs. (21) to (41) apply if a is replaced by a' and d by 8',
where, as before (Eq. (20) of Chapter 4),
a' = a 30
8' =8 + 30
m' = m + 60
(42)
where
Direct-voltage Harmonics
where
a
D = cos a cos 5 = 2 sin -
+ <5 sin -u = ,
A formula for complex values of the harmonics of the direct voltage is the
following:
= iF3(,M)
"
(34)
(43)
dO
Usually only the magnitude of a harmonic is wanted, the phase being of

no interest. Convenient formulas for computation are the following :
4 = 2K1F2(u, u, ft)
V6is
2K2F2=
ir2KA=-h
4
A. = 2K,FZ=
2KS2=%"
-'base
2K=
i=2
amperes
F-
\-1
Vdh = FJa, u, h)
(35)
(36)
(37)
(40)
(44)
V.do
where
l(h+l)a+l(h+l)8 /(h-l)a+l(k-l)8
TM
h+ 1
/(ft+i)(i+/(ft + i)) /(ft -i)g(i+ /(ft -i))
ft - i
ft + l
_ r(cos [(ft - 1)/2]Y (cos [(ft + 1)m/2]Y
Fs =
(38)
(39)
4 = 2K F =
-?
2K6F2
and a formula for the rms values is
LI
cos
A-l
[fl -
+l
l)/2]yos P + l>/2]j
cos(2g +
K)j1/2
(46)
16
h= 5
12
\\\
\v
\\
\ \
\\\
V\15\
\\\
30/\\
h=7
10
wv\\\1
\\
-
\\
W
Vv\
w\ \
24
309
\5#
\
60 to
.a = 0
V
90'\\
W
K
\
V
\\ \\
\\\\
\\
'
Fig. 4. Fifth harmonic current o-f a six-pulse converter as a function of converter angles.
\\
\\
Vv \
\\
\.
90/"
60
10
20
30
40
Fig. 5. Seventh harmonic current of a six-pulse converter.
50
60
10
8-2
h= 11
7
6
k\ \
w\
\\\\ \\
w
\\\ \
v\
\
v\ \N
V\
\\\
32
V\\\
V\
60 to
\a = 0
0"
15
60
30
20
10
60>/
90
A = 17
=-> 5 -
311
40
Fig. 8. Seventeenth harmonic current of a six-pulse converter.
60
50
Fig, 6. Eleventh harmonic current of a 6- or 12-pulse converter.

1
II
\n
XO
V\l5
-0
\60
20
XN
-
II
cn
'
V\5
156-v_
a=0
Nsj_6Q>/
30
40
50
Fig. 7. Thirteenth harmonic current of a 6- or 12-pulse converter.
60
, 60*
\\ NV5L
10
Fig. 9. Nineteenth harmonic current of a six-pulse converter.
8-2
313
h= 1
1
h= 5
a = 15' or 7 = 15"
h = 23
6R
\a = 0
1
Q
6C)Ki5
60
50
40
30
20
10
60
Fig. 10. Twenty-third harmonic current of a 6- or 12-pulse converter.
40
20
60
70
Fig. 12. Characteristic ac harmonics as functions of the direct current fora = 15: I't is
in percent of Is2 ; IL is in percent of /base-
Tile equation for F4 is like that for F2 if the sines in the latter are replaced
by cosines. Again, only the characteristic values of h are applicable.
Graphs of VdhjVd0 for h = 6, 12, 18, and 24 are given in Figures 13 to 16,
v
respectively.
It is notable that, unlike the ac harmonics, the direct-voltage harmonics
depend on a even if u = 0.
h- 25
Phase Sequence of DC Harmonics
. \a = 0
i
The pole-to-ground voltages va and vb of a bipolar dc line can be analyzed

into symmetrical components as follows :
Zero sequence:
va+vb
v0 = -
J*
Fig. 11. Twenty-fifth harmonic current of a 6- or 12-pulse converter.
Positive sequence:
vx
va
vb
= -
Similar expressions hold for the currents.

Consider a converter having an even number, 2b, of bridges in series on the
dc side with the midpoint grounded. Let the transformers of these bridges
314
8-2
25
315
12
70
-6020
_
k=6
10
50-N
-/90
40
30
15
90
50/
30
10
8 40
jC// /
20'
A
/// /s'/
151.
(AO
20
1/130
/
//
[////
20
vf
1
40
10
50
60, 9C
1
20
30
40
50
60
Fig. 13. Sixth harmonic of direct voltage of a six-pulse converter as a function of con
verter angles.
be connected alternately in YY or YA or in any other connections that give

a phase difference of 30 between the valve-side alternating voltages of the
two bridges of each pair. Then, if the number of bridges per pole, b, is odd,
the sequences of the characteristic harmonics are as follows :
Sequence
Zero
Positive
Harmonic Orders
Effective Number of
Bridges per Pole
, 12# + 6
6, 18, 30,
0, 12, 24, ..., 12q
1
b
The zero-sequence voltages are caused by only one bridge per pole, because
the remaining bridges per pole (an even number) comprise pairs in which the
zero-sequence voltages cancel. If the number of bridges per pole is even, there
are no zero-sequence characteristic harmonic voltages. This is the preferred
10
20
30
40
50
60
Fig. 14. Twelfth harmonic of direct voltage of a 6- or 12-pulse converter.
arrangement from the standpoint of minimizing noise in telephone lines

exposed to the dc line, because zero-sequence voltages produce zero-sequence
(ground-return) currents.
Effect of Finite Smoothing Reactor
The foregoing formulas for the characteristic harmonics in the alternating

currents and direct voltage of 6- and 12-pulse converters were derived on the
basis of several assumptions, the most questionable of which may appear to
be that of perfectly smoothed direct current. This assumption implies a smooth
ing reactor of infinite inductance. Since any practical reactor has finite in
ductance, usually not exceeding 1 H, the question of the accuracy of the
formulas in a practical case may be raised. Also, if the reactor is not infinite,
the current .wave shape varies with the nature of the load.
\
316
8-2
317
80
60
60
60'
h- 18
50
60
40
50
40
50.
'30
'20
30
20
10
30
10
-V
\
(b)
0
\\
Fig. 15. Eighteenth harmonic of direct voltage of a six-pulse converter.
Fig. 17. Alternating-current wave shapes, six-pulse rectifier with negligible commutating
and dc reactances and with resistive load: (a) a = 0, (6) a = 30.
6
60
60
30
60
40
40
40
60
30
20
30
20
30
10
10
20
Table 3. AC Harmonics of a Six-pulse Rectifier with Zero and Infinite Smoothing

Reactors, No Commutating Reactance, and Resistance Load
10
20
10
Magnitude of Harmonic Current

Difference
Order of (%)
Harmonic
Ld = 0
Li = co
10
50
0
Fig. 16. Twenty-fourth harmonic of direct voltage of a 6- or 12-pulse converter.
The other extreme is the complete absence of a smoothing reactor. Formulas

have been derived for ac harmonics under this condition, combined with zero
commutating reactance, zero ignition delay, and resistance load on a rectifier.3
The ac wave shape of such a 6-pulse rectifier is shown in Figure lla, and in
Table 3 the magnitudes of its harmonics are compared with the corresponding
7 ,
11
13
17
19
23
25
+1.000
-0.226
+1.000
-0.200
-0.113
-0.143
+0.091
+0.065
+0.091
+0.077
-0.0567
-0.0588
-0.0454
-0.0526
+0.0412
+0.0349
+0.0435
+0.0400
+13
-21
0
-16
-4
-14
-5
-i3
318
values obtained under the assumption of an infinite smoothing reactor. The

differences are fairly small. Results for a practical size of smoothing reactor
are intermediate.
For a rectifier of lower pulse number the harmonics would be more affected
by size of reactor and nature of load. In particular, a two-pulse rectifier with
resistance load has no harmonics of alternating current.
8-3 UNCHARACTERISTIC HARMONICS
The conditions postulated in the foregoing analysis of characteristic

harmonics of a converter are never exactly fulfilled in practice. Consequently,
not only are the harmonics of characteristic orders slightly changed from their
theoretical magnitudes and phases, resulting in small components of opposite
phase sequence from their characteristic sequences, but also and this is more
importantharmonics of uncharacteristic orders are produced. Indeed, a
converter is likely to produce harmonics of all orders and some dc component
on the valve winding of the transformers.
The harmonics of low uncharacteristic orders are normally much smaller
than those of adjacent characteristic harmonics in the converter itself. Filters
are usually provided for the low characteristic orders, however, and on the
network (or line) side of the filters, the uncharacteristic harmonics may be of
about the same magnitudes as those of the characteristic harmonics. For high
orders, the magnitudes of both characteristic and uncharacteristic harmonics
are small and approximately the same, even before filtering. For the highorder characteristic harmonics, the equations presented in Section 8-2 cannot
be depended on for accurate results. The magnitudes of these harmonics and
of all the noncharacteristic harmonics can be found only by measurement.
Causes
The ignition delay angle of a rectifier is usually measured from a zero of the
commutating voltage. If the three-phase alternating voltages are unbalanced,
their zeros are not equally spaced, and, consequently, the valves are not fired
at equal time intervals. Probably, even with balanced voltages, there is some
"jitter" in the electronic circuitry of the current regulator that produces un
characteristic harmonics. The variation of firing angles from their normal
values is usually cited as 1 or 2. Reeve and Krishnayya58 state, however, that
on the Cross Channel link the variation was +3 for rectifier operation and
1.5 for inversion.
It was shown in Section 5-11 that the combination of high gain and short
time constant in the current regulator would cause alternate early and late
8-3
UNCHARACTERISTIC HARMONICS
319
ignitions. As a result, harmonics of orders 3q are produced in the direct

voltage, and harmonics of orders 3q + 1 in the alternating currents. These
orders are uncharacteristic if q is an odd integer. For example, a third har
monic and its odd multiples appear in the direct voltage, and even harmonics
appear in the alternating currents.
Inverters normally operate on C.E.A. control, and unbalanced three-phase
voltages can again lead to unequally timed firing. The C.E.A. control has
no feedback. As a rule, inverters on C.E.A. control produce smaller un
characteristic harmonics than does a rectifier on C.C. control.
Another suggested cause of uncharacteristic ac harmonics is interaction of
characteristic-harmonic and fundamental currents in nonlinear elements of
the power system.14'29 The theory of modulation shows that such interaction
produces sum and difference frequencies, which, in the case in question, are
uncharacteristic. This cause appears to be unimportant, because the principal
nonlinear elements of a power system are transformers, in which only the
small exciting current is affected by the nonlinear relation between current and
flux. Of course, transformers do generate harmonics, but there is no evidence
that these interact significantly with converters. The same could be said for
corona, which is also a shunt nonlinear element.
Amplification of Uncharacteristic Harmonics, Several hv dc terminals on
going into service experienced trouble from a low-order uncharacteristic
harmonic of large amplitude causing improper operation, and even instability,
of the C.C. control. At Lydd42 it was the third harmonic; at Benmore, the
ninth.52 Analyses of these troubles have led to the following explana
tion:55,56'58 The addition of harmonics to the fundamental three-phase
voltage waves shifts the times of voltage zeros from the zeros of the funda
mental waves alone. These shifts of zeros cause unequally spaced firings of
valves, which, in turn, generate uncharacteristic ac harmonics. If any of these
current harmonics meets a high impedance, significant voltage harmonics of
like orders are produced. It may happen that one of these uncharacteristic
alternating-voltage harmonics has the same harmonic order and phase
sequence and nearly the same phase as one of the voltage harmonics assumed
at the beginning of this explanation. That particular harmonic is amplified
by positive feedback.
If the loop gain is high enough, a harmonic oscillation of increasing ampli
tude is produced: this is instability.
Consequences
Uncharacteristic harmonics (1) increase telephone interference, because it

is not feasible to provide adequate filtering of each order of them, and (2) in
some instances cause instability of C.C. control, as explained above.
320
8-3
Suppression or Diminution
In the instances of control instability cited above, a three-phase hv ac

filter bank was provided for the offending harmonic (third or ninth). Such a
filter provides a low shunt impedance at the frequency of the offending
harmonic, with the result that a given current of this frequency produces less
voltage of the same frequency, and thus the loop gain is reduced to a degree
that gives but little amplification of this harmonic; consequently the control
becomes stable.
This method is expensive. A modification of the control system, which
operates at a low level of power, would be less expensive. For example,
filters to block the offending harmonic could be placed in the three-phase
low-voltage circuit that provides a replica of the commutating voltages to the
control system. No reason is apparent why a shunt filter in this location could
introduce any different transfer function into the loop from what a similar
filter placed in the hv circuit would.
A better control system is isochronous control such as the phase-locked
oscillator described by Ainsworth.57 This generates a series of equally timed
firing pulses that is locked to the correct average delay angle by the current
regulator, actuated by the difference between the set and measured values of
direct current. This, or an equivalent scheme, for getting equally timed firing
of valves should be used if it is desired to diminish uncharacteristic harmonics
as much as possible.
Relation of Uncharacteristic Harmonics to Errors in Ignition Angles
Even AC Harmonics Suppose that, as previously discussed, the ignition

times of valves in a six-pulse bridge are alternately late and early. To be more
specific, assume that the odd-numbered valves, constituting one half bridge,
are ignited early by an angle e, while the even-numbered valves, constituting
the other half bridge, are ignited late by the same angle. The wave shapes of
the alternating currents consist of alternate positive and negative pulses of
unaltered duration 120, but the intervals between a positive pulse and the
following negative pulse are increased by 2e from the normal value.
On pages 297 to 300, the Fourier series for trains of positive only, and
of negative only, rectangular pulses were derived, and it was shown that, with
the correct relationship between the two trains, the odd harmonics are
doubled but the even harmonics vanish. The same relationship holds for the
nonrectangular pulses observed where there is overlap. In the event of relative
displacement 2s between the two trains, the resultant harmonics can be
found by vector addition, with the odd harmonics separated by 2hs instead
of zero and the even harmonics separated by % 2he instead of by n. The
UNCHARACTERISTIC HARMONICS
321
sums are respectively 2 cos he and 2 sin he times the respective harmonics of
one train. For small he the decrease of magnitude of the odd harmonics is
negligible; the even harmonics except those of order 6q, which do not appear
in the individual trains, increase from zero to a nonzero value, which we
proceed to estimate.
The ratio of an even harmonic of order h to the fundamental wave at small
overlap is
--
+ TTo(e)5
h{\
%{he)2
T1~2hcose~
+ {hef
If, 2 sin he
he %(hs)3
= [1 "+ i(Ae)2
+ -(hs)4 -\
]=e
* *
radians
(47)
For e = 1, corresponding to 2 relative shift between the positive and

negative pulses, the second and fourth harmonics are each approximately
1/57.3 = 0.0174 per unit = 1.74% of the fundamental current. This value is
further decreased by overlap.
Triple AC Harmonics. It was shown in Section 8-2 by Eq. (11) that a train
of rectangular pulses of normal width (120) has no triple harmonics. If,
however, the pulse width is longer or shorter than normal, triple harmonics
are generated. Again let us estimate the magnitude of such harmonics as a
function of the angular ignition error. Suppose that the ignitions of two valves
connected to the same phase are late by e and that the other four valves of
the bridge ignite on time. Then the alternating current of that phase consists
of positive and negative pulses both of which are shorter than normal by e.
The current of the phase leading that one has pulses longer than normal by e,
and the current of the remaining phase has pulses of normal length. Assume
zero overlap, so that the series of Eqs. (10) and (12) are applicable. In each
of these series and in their sum, the ratio of odd harmonic to fundamental is
h _ sin {hw}2)
h h sin (w/2)
Now put w = 2n/3 e and h = 3q. Then
Ih sin (#7i + l.Sqe)

IL 3# sin(7i/3 e/2)
sin qn cos 1.5qe cos qn sin 1.5#s
3#[sin (ti/3) cos (s/2) cos (ti/3) sin (e/2)]
SinL5ge
3tf[( V3/2)cos(e/2)
sin (e/2)]
(49)
322
8-4
For small e, cos (e/2) 1, sin (e/2) = 0, sin 1.5gs 1.5<?e, and
323
8-4 TROUBLES CAUSED BY HARMONICS38
= *0.577*
h 3q3}2
TROUBLES CAUSED BY HARMONICS
i
(50)
V3
List of Troubles
For e = 1 = 0.0174 rad, IJI1= 0.01 = 1%.

Magnitudes of Uncharacteristic Harmonics Found in Field Tests
Measurements of harmonics from the converter at Lydd (at the English
end of the Cross Channel link) were made with the filters disconnected, giving
the results shown in Table 4. They were obtained by Fourier analyses of
Troubles in the Converter and on the AC Power System

1. Extra losses and heating in machines and capacitors
2. Overvoltages due to resonance
3. Interference with ripple control systems
4. Inaccuracy or instability of the constant-current control of converters
Troubles on the Telecommunication Systems
5. Noise on voice-frequency telephone lines
oscillograms.
Table 4. Harmonic Currents Measured on AC Side
of Converter at Lydd39,49
To an arbitrary scale
Order of
Converter
12-pulse
400-A DC
6-pulse
Harmonic
Blocked
Operation
Operation
29.7
9.3
25.9
10.2
10.9
21.6
92.5*
6.2
66.5*
44.3
I00-A DC
2
3
4
5
6
7
8
9
10
11
12
13
2.0
2.1
0.3
6.0
1.3
4.0
0.8
0.1
0.5
2.0
0.6
1.0
26.4
9.2
16.2
31.7
57.8
22.3
119.6*
67.9
3.8
19.2*
0.3
15
0.2
16
0.04
18.4
17
18
0.2
13.4
4.4
11.1
7.5*
0.3
19
20
0.08
10.4
8.6
11.7
3.4
5.4*
4.9
* Characteristic harmonics.
43.6
75.3*
14
0.1
23.8
21.5*
28.4
17.9
15.0
Noise on telephone lines is the most difficult trouble to eliminate and forms
the subject of most of the rest of this chapter. There are reasons for concern,
however, over the effects of harmonics on the power system itself. Item 4 was
discussed in Section 8-3, page 318. Items 1, 2, and 3 are briefly discussed
below.
j
i
I
:
Extra Losses and Heating in Machines7'-8'50

Harmonic currents in induction and synchronous machines cause additional
losses in, and heating of, these machines. These effects are chiefly attributable
to the harmonics of low orders, which can have large magnitudes. Lowresistance damper windings serve to shield the rotor iron from harmonic
fluxes that would overheat it, and the heating of the damper windings is small.
Damper windings are used on salient-pole synchronous machines but not on
round-rotor synchronous machines.
Large harmonics in induction motors reduce the torque available from
them at rated speed and cause parasitic torques at lower speeds that can
prevent a motor that is being started from attaining its rated speed.
There have been proposals for feeding rectifiers from generators isolated
from the rest of the ac system, and this is done at Volgograd. Among the
advantages claimed are that the provision of ample damper windings on the
generators would cost less than the provision of ac harmonic filters and that
the generators could be allowed to vary in speed more than generators usually
do in an interconnected ac system with no concern about detuning of these
filters. Studies of this possibility on the Nelson. River project59 showed, how
ever, that filtering was more economical than building the generators to carry
the harmonic currents continuously. An additional complication was that,
because the project would be built by stages, additional generating plants
being connected to the rectifiers at each stage, there was no assurance that
324
8-5
the harmonics would divide among all these plants in any definite ratio with
out putting undue restrictions on the design of the future generators, ac
lines, etc.
Although filters are provided at the rectifier station, the frequency generated
by the plants connected to this station is subject to large variations during
system disturbances. The filters are then out of tune, and a greater part of the
harmonic current output of the rectifiers passes through the generators,, which,
however, were designed, at small additional cost, to withstand the temporary
additional heating caused by these increased harmonic currents.
Measurements of losses caused by harmonics in a cylindrical-rotor syn
chronous generator of normal design, loaded by nonlinear reactors were
reported by Easton.50 Measurements of harmonic impedances of a 27.5-MVA
generator were reported by Gardiner.51"
Extra Losses and Heating in Capacitors
The increase of losses in capacitors due to harmonics is9'24
fc(tanKFf
(51)
h=2
where
DEFINITIONS OF WAVE DISTORTION OR RIPPLE
325
Trouble from such resonance is most likely at a frequency close to a harmonic

frequency for which no filter is provided. For example, third-harmonic
resonance could occur if Qc s 0AQs. Such resonance could have several
undesirable effects: (a) overheating of the capacitors, (b) overvoltage at the
capacitor bank, and (c) instability of the constant-current regulator of a
converter.
Interference with Ripple Control Systems47

Some electric-power utilities sell electric energy at especially low rates for
off-peak loads, such as water heaters, and control the hours during which
such loads can be connected by transmitting audiofrequency tones, in the
range of 290 to 1650 Hz, from substations over power-distribution circuits
to customers' premises to control contactors in series with such loads. Simi
lar control is used for street lighting by some utilities. The receiving devices
for the control signals are broadly tuned and can accept harmonics from
high-power converters, which may cause undesired operation of the con
tactors or prevent desired ones.
The remedies are adequate filtering of ac harmonics or decreasing the
susceptibility of the ripple control system to harmonics.
C = capacitance
tan 5 = loss factor
8-5 DEFINITIONS OF WAVE DISTORTION OR RIPPLE
coh = 2n times frequency of kth. harmonic

Vh = rms voltage of hth harmonic
The dielectric stress is proportional to the crest voltage, which may be either
raised or lowered by the harmonic voltages.
The total reactive power, including fundamental and harmonics,
CO
Q=1Q
A=1
(52)
should not exceed the rated reactive power of the capacitor,

Overvoltages from Resonance
A complete description of a periodic current or voltage wave that is neither

sinusoidal would require either an oscillogram or a list of the
harmonics present, together with their phases and magnitudes, or, at least,
constant nor
their magnitudes.
For some purposes, such as filter design, it is convenient to have a more
concise expression a single number serving as an index of the degree of
departure of the wave from its ideal shape. Several such indices have been
defined.
Total RMS Harmonics
The following method may be used to estimate the possibility of resonance

between a large shunt capacitor bank on the power system and the rest of the
48
system at a harmonic frequency
Let Qs equal short-circuit power of power system at point where capacitor
bank is connected, Qc equal rating of capacitor bank and h equal order
of harmonic at which resonance may occur. Then
/&Y
For alternating current, this quantity is
VSa2
Hi = -z-= -
Ii
per unit
(54)
per unit
(55)
and for direct current it is
-JYlYJi Yi2~i}
H2 = -- = -.-
326
8-5
327
values of all the harmonics divided by the crest value of the fundamental
wave. Of course, rms values can be used instead of crest values in both
numerator and denominator, giving
where / = effective (rms) current

Id average direct current
It = rms fundamental current
Ih rms harmonic current of order h
Similar expressions hold for voltages.

Because the harmonics are squared, the largest ones, that is, the lower
orders, dominate the result.
Wasserab32 calls these indices Welligkeit (waviness).
ff6 = Lh=2
h
j
h
(59)
This index weighs all harmonics equally. It is not used for dc quantities. In
practice, we can neglect harmonic orders above the twenty-fifth with very
little error.
Deviation from a Sine Wave (or a Constant)
For alternating current, it is
jl
Zj | max
riz = --hm
._
per unit
(56)
(57)
and for direct current

tt
dl max
Ha = --where
per unit
i =instantaneous current
z'j = instantaneous value of fundamental current wave
Ilm = crest value of fundamental wave
Id average direct current
In words, the index is the maximum difference of ordinates of the wave in

question and of the corresponding fundamental wave or average value. The
result depends on the phase position of the harmonics. All harmonics have
equal weights.
A closely related index is the peak-to-peak value of ripple of direct current
or voltage. The maximum positive and negative deviations are taken sepa
rately, and their difference is found. Again, it is usually divided by the average
value :
tt
Hs =
'min
-'max
Id
per unit
(58)
Maximum Theoretical Deviation from a Sine Wave

In many cases the magnitudes of harmonics are known or calculated, but
their phase angles are not known. For this reason, it is usually more practical
to assume that the crests of all the harmonics occur simultaneously at least
once per cycle. The index thus modified is the arithmetical sum of the crest
Psophometric and C-message Weightings
We now come to a group of wave-form indices in which the various fre

quencies or harmonic orders are weighted according to their effectiveness in
interfering with telephone conversations. The sensitivity of the human ear,
the response of the telephone receiver, and the coupling between power and
telephone circuits all vary with frequency, and these variations can be taken
into account by appropriate weighting factors.
Two systems of weighting are in wide use : (a) that of the Bell Telephone
System (B.T.S.) and the Edison Electric Institute (E.E.I.) is used in the United
States and Canada; (b) another system promulgated by the International
Consultative Commission on Telephone and Telegraph Systems (C.C.I.T.T.)
is used in Europe. (The C.C.I.T.T. resulted from a merger of C.C.I.F. and
C.C.I.T. in 1956.)
One set of weighting factors of each system, based on the sensitivity of the
ear and the response of telephone equipment, applies only to currents and
voltages on the telephone circuit. This is called C-message weighting by B.T.S.
and E.E.I., andpsophometric weighting by C.C.I.T.T. These weights are shown
in Tables 5 and 6 and in Figure 18. In both systems maximum weight occurs
at 1000 Hz. In the C.C.I.T.T. system, weight 1000 (0 dB) occurs at 800 Hz;
in the B.T.S.-E.E.I, system, unit weight occurs at 1000 Hz.
In both systems, the weights have been revised from time to time to reflect
the increasing bandwidth and higher standards of quality of telephone trans
mission. The C-message weighting was adopted in 1960, when it superseded
the earlier F1A weighting, which had already superseded the 144-line weight
ing in 1941. The psophometric weighting presented here is from the 1963
edition of C.C.I.T.T. Directives.35
Other sets of weighting factors pertaining to the coupling between power
and telephone lines and to the interfering effect of currents and voltages on
power systems are described next.
328
Reference frequencies
C-MSG
CCITT
800 Hz
1000 Hz
/
n/ / /
'V
20
30
//
'/c- Vl< G
g?
"n"
//
c6
t-4
no
no
in
Ti in

$ <5
NO N CO
NO CS oo It NO cs OO
it f in NO NO l> r- oo On ON
ro CO CO CO co CO CO if
if
on on
NO CS
.1
Coupling Factors
(ZJ
In both the B.T.S.-E.E.I. and C.C.I.T.T. systems, the coupling between a

power circuit and a telephone circuit is assumed to be directly proportional
to frequency. The coupling is taken by convention as 5000 units at 1000 Hz
in the B.T.S.-E.E.I. system and as 1 unit at 800 Hz in the C.C.I.T.T. system.
Change in coupling as a function of separation or length of circuits is not
included in these factors.
NO NO f- C-
r-
CS in n
in in in in in
1-4 44
rH
1-4 44 44
44
44
OO
44
'
""
1-4
n in n NO
r-;
44
C9
1
.?
i
g
NO t 00 On
1-4 1-4
1-4
If
cs cs cs'
Nro'tvir-coONO-iNf23
u-l
SwoNOinvNNOiNONONomiflOiogwNOMh/ico
iTfOOONMNOlD
N Tf NOON COf-NhCOCOWOir-rtNNcOM't'flfNlfll'llfllON01OlONO
m' oi i no
i
If)
NNmcO(OMCNNOONlqfs, t":
oooiO'tmNrtddddddoddbo-H
.a
44
ON cS n OO
r-,'Qin"n0
ao
t/i
rtNoPt8njv>o\t-->nr-rj(Nr-r-eor--3;cnci-
Snoo'nN-:<ONOOi?NOONO\o-HiogoooNNNON
000-,fiClfl|nrCOO'ONONOOONONONONONONOO
dddoddddeddddo'dd-HodcJoooo
NO N CO
s:
(60)
in in

NO N h
oo in
If SO NO if CO cs
ON ON ON
IONONONMCOONONO--<NNfOtlOlONOOOOOOlO
Eo(/Ft/2

co ON it On if
co <n oo co
On on On
NO (S OO it O
'ONoo'fONONoo'tONON'tgNgNco.;;.-;.;;;,_
1"N
t/N CS <S
Tir
44
1
!
l
I
1
o1
is iq iq N
defined by
1-4 1-4 T-4
>n
72
'* "
io
These factors give an approximation to the effect of wave shape of current

or voltage of a power line on telephone noise, excluding the magnitude of
power-system current or voltage and the geometrical aspects of coupling.
Each of these factors is a root-sum-square combination of the products of
coupling factors and weighting factors for each frequency. In the B.T.S.-E.E.I.
system the result is called telephone influencefactor (abbreviated TIF), and is
rtOQrt-HHrt-HrtHHrtHrtrt-MN'tNOO'fNCOI'N
MOOOOOOCOOOCOOOCOOOOOOOMCOCOOOCOCOOOIhrNqNONq
Telephone Influence Factor (TIF) and Telephone Harmonic Form Factor
cs r~
m oo o
t- 00 OO
><
*o
Fig. 18. Comparison of B.T.S. C-message and C.C.I.T.T. weights.
NO N CO if O O
N ci to in
N
'
in
t Nf "t 4if
62 63 3 65 99 67 89 69 70
in>ninininninnnNONONONO
S.
On i'
On
NN N N
N N
m on N O On h m
r- On in cS no co in
ON t*
cs oo no m if cs -"-i co m
ii on m
OO in NO

CS CS
m if if it if on fS
' ' ' ' *' ' '
ON
HNClTfl/lNONCOONO-vn'l
10
23
1
1
Frequency, kHz
40
50
0.1
CC1
10

on oo t- OO NO <s oo
oo
NO
N h- Tf O NO if rNO n <n
f- NO
ON CO 00 OO 00 r-
OOOOOO
fOm'fNflnNONOf-'hCOQNONOO
N N co m
rsif,nNotccONcsifinNor--ooON44
cs
M N M N
u- m
(S
oo
330
8-5
Table 6. CC.I.T.T.: Psophometric Weights (1000/)/) and Telephone Interference

Weights fpf
h
/(Hz)
1
2
3
50
100
150
200
4
6
8
10
12
16
dB
fpf
0.71
63.0
0.000044
20
8.91
35.5
-41.0
0.00111
24
-29.0
-21.0
0.00665
30
40
50
1000/)/
300
400
295
-10.6
0.0223
0.111
484
-6.3
0.242
500
600
800
661
794
1000
-3.6
-2.0
0.0
0.413
0.595
1.000
89.1
60
70
80
100
/(Hz)
1000/7/
1000
1200
1500
1122
1000
2500
861
708
617
3000
525
3500
4000
5000
2000
dB
fpf
+ 1.0
1.40
1.50
1.62
0.0
-1.3
-3.0
1.77
-4.2
1.93
-5.6
1.97
376
-8.5
178 -15.0
15.9 -36.0
1.65
0.89
0.10
kf = 5000(//1000) = 5/
(61)
pf = C-message weighting
(62)
Vf = rms voltage of frequency / on power line
-Jfyf = rms voltage, unweighted
(63)
In the C.C.I.T.T. system the result is called telephone harmonic form factor
(T.H.F.F.) and is defined by a similar expression with
kf
800
Pf
psophometric weighting
1000
(64)
331
In the B.T.S.-E.E.I, system, the quantity is called C-message weighted voltage.

In the C.C.I.T.T. system, the longitudinal induced voltage is called psopho
metrically weighted voltage, and the resulting transverse voltage is called
psophometric voltage:
y
where
V=
-JY]Lp,v,?
(66)
where Vf is the rms longitudinal or transverse voltage of frequency / on the

telephone line.
* T Product, kV T Product, Equivalent Disturbing Voltage, and Equivalent
I
Disturbing Current
These are weighted currents or voltages in the power systems. In the

B.T.S.-E.E.I. system, the I-T product is a root-sum-square combination of
products of currents (in amperes) of various frequencies, each multiplied by
the corresponding TIF :
I-T=
VE(r? = /-(TIF)
(67)
where If = rms current of frequency /

Tf corresponding single-frequency TIF
The kV-T product is a similar combination of power-line voltages (in kV).
In the C.C.I.T.T. system, the disturbing current or voltage is similarly defined.
In the use of either system, the power-line current or voltage must be specified
as either balanced, that is, positive and negative sequence, or unbalanced,
that is, zero sequence.
Analogous quantities on the two systems are tabulated in Table 7.
(65)
Generally the sums are for a finite number of discrete frequencies, which
include the power frequency (50 or 60 Hz) and its multiples.
C-message-weighted or Psophometrically Weighted Voltage on the Telephone

Circuit
These quantities are either longitudinal or transverse voltages on the tele

phone circuit weighted psophometrically or by C-message weights (see
page 327). They also are root-sum-square combinations of the noise effect of
discrete frequencies produced in the telephone by other sources than the
speaker's voice. Our attention is limited to voltages induced by a power line.
Table 7. Corresponding Quantities in B.T.S.-E.E.I. and C.C.I.T.T. Systems

B.T.S.-E.E.I.
C-message weighting
Telephone influence factor (TIF)

C-message-weighted voltage:
Longitudinal
Transverse
I*T product
kV T product
C.C.I.T.T.
Psophometric weighting
Telephone harmonic form factor
(THFF)
Psophometrically weighted voltage
Psophometric voltage
Equivalent disturbing current
Equivalent disturbing voltage
332
8-6 MEANS OF REDUCING HARMONICS
Increased Pulse Number
In low voltage high-current rectifiers, high pulse numbers have sometimes

been used, ranging from 24 to 108. This means of reducing harmonics is very
effective as long as all valves are in service, but it requires complicated trans
former connections. In hv high-current converters for dc transmission,
problems of insulation of the converter transformers to withstand high
alternating voltages in combination with high direct voltages dictate simple
transformer connections. A pulse number of 12 is easily obtained with simple
connections of two six-pulse valve groups, as we have seen, and 24 pulses
can be obtained with four six-pulse groups by use of a phase-shifting trans
former bank in conjunction with two 12-pulse converters. The required phase
shift is 15. Two ways of obtaining it are shown in Figure 19.
Fig. 19. Voltage vector diagrams of autotransformers for shifting three-phase voltages by
angle (drawn for = 15): (a) ring; (b) zigzag.
<f>
8-7
TELEPHONE INTERFERENCE
333
harmonic output with three groups in service would consist of both the
12-pulse and six-pulse harmonics of three groups. Obviously, the 12-pulse
converter has some advantage over the six-pulse converter even when one
bridge is out of service though less than when all are in service.
Filters
Any necessary reduction in harmonic output of the converter beyond that

accomplished by increase of pulse number must be done by harmonic filters.
Most experts on hv dc transmission feel that it is more economical to use a
12-pulse converter with filters than to use a converter of higher pulse number
with the permissible reduction in filters. Therefore, the emphasis in the rest
of this chapter is on design of adequate harmonic filters.
Filters are almost always needed on the ac side of the converter and, some
times, on the dc side also. The ac filters serve two purposes simultaneously:
supplying reactive power of fundamental frequency in addition to reducing
harmonics. Hence the part of the cost of filters chargeable to the need for
reducing harmonics is usually near to the cost of the filter inductors, the
filter capacitors being required for supply of reactive power. Thus we are led
to the concept of the minimumfilter, which is required for harmonic reduction
only in installations where the reactive power required by the converter can
be supplied by the ac system without reinforcing the latter. A filter costing
more than the minimum filter not only supplies additional reactive power but
also generally gives better filtering. Care should be taken, however, that not
too much reactive power is supplied during operation of the dc link at light
load.
In addition to the ac harmonic filter at the converter stations, such filters
could also be placed in any sections of transmission line giving rise to especi
ally bad telephone noise. This is seldom, if ever, done, because it is usually
cheaper to modify or relocate the telephone line.
<f>
The effectiveness of 12- or 24-pulse converters in. reducing harmonics is

somewhat decreased when one valve group is out of service. Consider, for
example, a converter with four valve groups. If two groups have one trans
former connection and the other two groups have a different connection,
giving a 30 phase difference, then, with all groups in service, the ac har
monics consist of the 12-pulse harmonics of four groups; but, with only three
groups in service, there are the 12-pulse harmonics of three groups and the
six-pulse harmonics* of one group. If all groups were connected alike, the
* Sixth harmonic and its odd multiples.
8-7 TELEPHONE INTERFERENCE
General
The frequencies used in commercial voice transmission range from 200 to

3500 Hz. In this range lie many harmonics of the power-system frequency
which are usually of small magnitude but which, because of the high TIF
weightings and the great difference between the power levels at which power
circuits and telephone circuits operate, may, nevertheless, result in perceptible
or even unacceptable telephone noise. The power on a voice-frequency
334

3
8-7
5
335
telephone circuit is from 10 to 10 W. By contrast, that on a power dis

tribution circuit is 103 to 105 W, and on a major transmission line is 107 to
109 W.
If the ground has finite and uniform conductivity, the foregoing statements
are still substantially true with respect to the electric field. The magnetic
field, however, can now penetrate the earth, and its effect on self- and mutual
inductances is as if the image conductors were lowered to a greater depth
History
below the surface of the ground. The equivalent depth of ground return is
proportional to the skin depth-, both of these depths vary inversely as the
square root of the frequency and of the conductivity.
If the ground is nonuniform, the foregoing is still true qualitatively.
Because the distance between the overhead conductors of a ground-return
circuit and their image conductors are much greater than the distances
between conductors of a metallic circuit and because the two conductors of
a ground-return circuit are in a vertical plane, the coupling between two
ground-return circuits is very much greater than the coupling between two
metallic circuits separated by the same distance as the ground-return circuits.
Although ground-return circuits were used for dc telegraphy and voice
telephony when these arts were new, they are seldom if every used now
because of the severe problems of noise and cross talk. Power circuits are
also all metallic except for some hv dc lines, because the power loss and
telephone interference from ground-return ac circuits are both high. It would,
therefore, appear that the coupling to be calculated is that between a metallic
power circuit and a metallic telephone circuit. But, on the contrary, the
practice is to calculate the coupling between a ground-return power circuit
and an open-ended ground-return telephone circuit. In other words, one
calculates the longitudinal voltage induced in the telephone circuit by residual
current in the power circuit. The reasons for this practice are now given.
The subject of inductive coordination of power and telephone circuits has

been thoroughly studied by both power-system and telephone engineers for
the last 50 yr or more. In the United States those studies have been made
principally by the B.T.S. and E.E.I.; in Europe, by the C.C.I.F. (now the
C.C.I.T.T.) and by governmental power and telephone authorities. Improve
ments in the quality of telephone service during this period have resulted
in a decrease of the tolerable psophometric voltage. Rectifiers in industrial
and railway service are not new, but the converters for hv dc transmission
have now attained much higher power ratings than those.
Coupling, Electric and Magnetic
Coupling between power circuits and telephone circuits is through both
electric and magnetic fields. Unless the spacing between the two circuits is
close, howeverfor example, if both circuits are on the same poles the
magnetic coupling predominates, and the electric coupling is negligible. The
magnetic coupling may be expressed as a mutual impedance, that is, as the
voltage induced in the telephone circuit per ampere of current in the power
circuit.
Spacing. The coupling between two circuits having parallel conductors
increases with increased spacing between conductors of the same circuit and
decreases with increased distance between circuits.
Length. The coupling between parallel circuits is directly proportional to

the common length, known as the length of exposure.
Metallic and Ground-return Circuits. If the ground were a perfect conduc
tor, there would be no electric field or varying magnetic field in the ground,
and the fields above the surface of the ground caused by overhead conductors
carrying currents and charges would be the same as that caused by the actual
conductors and their image conductors with the ground removed. On the
assumption that the surface of the ground is a horizontal plane, each image
conductor is a fictitious conductor like the corresponding real conductor,
located directly below the latter as far below the surface as the real conductor
is above it and carrying current and charge equal in magnitude but opposite
in direction to those in the real conductor.
Unbalances. The currents in a metallic three-phase power circuit are

nominally balanced. In telephone parlance balanced currents have a sum that
is zero. In power-system parlance, the phase currents have positive- and
negative-sequence components but no zero-sequence components. Inpractice,
the line-to-ground fundamental-frequency voltages of the power system are
almost entirely of positive sequence. The impedances of the three phases,
however, are not perfectly balanced because of the inequality of interphase
spacing and the lack or infrequency of transpositions. Consequently, some
zero-sequence current exists. The power circuit is usually grounded at the
source end, and some are multigrounded. At least, there are shunt capaci
tances to complete the zero-sequence path. The balance of a power circuit
seems to be poorer for the higher harmonics than for fundamental currents.
The residual current, which, in a three-phase circuit, is 310, generally
induces both longitudinal and transverse voltages in a parallel telephone
circuit. If the power and telephone circuits are well separated, however (say,
0.5 km or more) and if the two wires of the telephone circuit are close
336
8-7
together and frequently transposed, the induced transverse voltage is negligible,

but the induced longitudinal voltage is much greater. This induced longitudinal
voltage, however, would still produce no noise in the telephone circuit if the
latter were perfectly balanced ; but small unbalances in the telephone circuit
give rise to unbalanced currents, accompanied by a transverse voltage.
Unbalances on the telephone circuit may consist of any unequal series
impedances or shunt impedances to ground ; for example, ringers connected
from one wire to ground and slight differences in resistance of wires or of their
capacitances or leakage to ground.
To sum up, the following series of events causes balanced voltages in
metallic power systems to produce transverse noise voltage in metallic tele
phone circuits :
Balanced voltage on power circuit, through

Unbalance of power circuit, causes
Residual current on power circuit, which, through
Coupling between two ground-return circuits, induces
Longitudinal voltage in telephone circuit, which, through
Unbalance of telephone circuit, causes
Transverse voltage in telephone circuit
Shielding (Screening), Passive ground-return circuits between the power
and telephone circuits or near to either of them can affect the coupling and
the balance. Among such circuits are ground wires on open power lines and
metallic sheaths on power cables and on telephone cables, all of which are
multigrounded. Their usual effect is to decrease the coupling. Any parallel
circuits, however, that are unsymmetrically located with respect to the power
circuit and near it also increase the unbalance of the latter. Either the screen
ing effect or the unbalancing effect may predominate. Other parallel, ener
gized power lines can have both of these effects and also be an additional
source of telephone influence. For underground cable circuits, the earth has
a screening effect, and for submarine cables, the water has an even greater
effect.
The screening factor is the ratio of the induced telephone noise with the
screening to that without the screening.
Frequency. The coupling increases with frequency. For circuits having

fixed current paths, the inductive mutual reactance is directly proportional
to frequency. Because the equivalent depth of ground return decreases with
increasing frequency, however, the coupling between two ground-return
circuits increases more slowly than as the first power of the frequency.
The coupling is usually calculated at 1000 Hz (B.T.S.-E.E.I.) or at 800 Hz
(C.C.I.T.T.). Then the variation of coupling with frequency is calculated as a
337
separate factor. As previously noted, in the calculation of TIF, the coupling

is assumed to vary in direct proportion to the frequency, which is an
approximation.
The mutual impedance between two groundreturn circuits is usually computed
by Carson's formula,1 which assumes a homogeneous earth and parallel
conductors. Krakowski54 has extended Carson's work to crossing conductors.
Riordan and Sunde4,11 have extended this work to a two-layer earth, that is,
to surface layer of one resistivity separated by a horizontal plane from an
infinite volume having a different resistivity.
If the separation between power and telephone circuits is much greater
than that between conductors of the same circuit, the conductors of each
circuit may be replaced by one equivalent conductor at the center of gravity
of the several conductors. The coupling between widely separated groundreturn circuits increases with the resistivity of the ground.
Unbalance of Telephone Circuit
The balancefactor of a telephone pair is the ratio of the longitudinalvoltage

induced in the telephone pair by. current in a power circuit to the transverse
voltage resulting from unbalance of the telephone circuit. For a subscriber
loop in modern cable, a typical value of this factor is said to be 316, or
50 dB.40 An open-wire circuit would be more poorly balanced and thus
have a lower balance factor.
Unbalance of Power Circuit
A common configuration of the three conductors of a three-phase trans
mission line is for them to be in a horizontal or vertical plane with equal
spacing between adjacent conductors. Transpositions are rarely used. If such
a line carries balanced (positive-sequence) currents, the zero-sequence voltage
drop is approximately one-thirtieth of the positive-sequence voltage drop,
corresponding to 30 dB.
If the conductors have a triangular configuration, the balance is better,
being theoretically perfect for an equilateral triangle. Transpositions, when
used, help to balance the line for the fundamental. For the high harmonics,
the number of transposition cycles per wavelength is usually too small to be
effective. Ground wires on the power line usually improve the balance slightly,
but the ground wires most commonly used have a much higher resistance
than do the main conductors; hence, their effect is small.
Direct-current overhead lines of the usual configuration are inherently well
balanced.
338
8-7
339
Acceptable Levels of Telephone Noise
Calculation of Telephone Noise due to Two-terminal DC Line40'53,60
The C.C.I.T.T. Directives35 state that the psophometric emf should not
exceed 1 mV. The psophometric emf is twice the psophometric voltage
(psophometrically weighted noise voltage) measured across a noninductive
resistance of 600 Q, terminating the line at the place of measurement, when
the other end of the line is terminated in its characteristic impedance. The
corresponding noise power in a 600-Q resistive termination is
This discussion is limited to overhead dc lines because underground or

submarine dc cables are so well shielded that no noise problem exists unless
power and telephone conductors are in the same cable, which is unlikely :
V2/R = (0.5 x
1. Calculate the harmonics of the direct voltage on the valve side of the dc
assumed operating conditions. Among these conditions
are both rectification and inversion with all bridges in service and with one,
or more bridges out of service. In a bipolar line, monopolar operation with
either metallic or ground return should be studied if such operation is con
templated. The control angles a of the rectifier and y of the inverter should be
assumed to have their normal values ; for example, a = 15 and y = 16, and
values of overlap u corresponding to several values of direct current between
minimum and maximum values may be tried to see which produces the
greatest TIF. In the usual monopolar scheme, all harmonics have earth
return. In a bipolar scheme, the harmonics should be divided into zeroand positive-sequence components. Generally, only the zero-sequence
components are important. Both characteristic and uncharacteristic har
monics should be included. Uncharacteristic zero-sequence harmonics
corresponding to characteristic positive-sequence harmonics of the same
order are not so much the result of conversion from positive to zero sequence
by unbalance of the line, for most bipolar dc lines are inherently well balanced,
but are mainly due to generation of uncharacteristic harmonics by the con
reactor for various
10 ~3)21600 = 4.2 x 10 "10 W.
The B.T.S. standard of acceptable noise37,40 is more strict. The noise

contribution on the customer loop must not exceed 20 dBrnc, that is, 20 dB
over the reference level of 10 ~12 W (1 pW), which is 10 ~10 W. Not more than
half of this (0.5 x 10" 10 W) should be from any one source. Since such lines
are usually terminated in 600 Q, the transverse noise voltage (C-message
weighted) corresponding to this noise power would be
jPR =
V0.5 x
10"lox 600 =
%/3 x
10~8 = 1.73 x 10~4 V = 0.173 mV.
On the assumption of 50 dB balance on the telephone line, the maximum

permissible longitudinal induced noise voltage from one source is 102*5x
0.173 mV = 55 mV.
verters.
2. Draw the terminating circuits, including dc reactor and dc harmonic
filters and surge capacitor if used, and calculate their impedances and
Calculation of Noise Voltage (Psophometric Voltage)
The transverse noise voltage in a telephone circuit caused by

rent in a parallel power circuit is calculated as follows:38
Khk-hPh)2
residual cur
weighted volts
(68)
where I= length of exposure, m

Ih = residual current, A, of harmonic order h in power line = 3/0 for
three-phase line
impedance between power line and telephone line, Q/m,
mutual
=
Zmh
at frequency of Mh harmonic, calculated by Carson's formula
ksh = screening factor

kh balance factor = ratio of transverse noise voltage to longitudinal
voltage, presumed to be independent of frequency
or psophometric weighting of hth harmonic with
respect to reference frequency, 1000 or 800 Hz, respectively (see
ph = C-message
For calculation of longitudinal voltage, the factor kh is omitted.
open-circuit voltage ratios for each harmonic frequency in order to establish

a Thevenin's equivalent circuit for each terminal.
3. Calculate the dc transmission-line parameters attenuation and phase
constants and characteristic impedancefor each harmonic frequency.
4. Consider one harmonic frequency at a time, assume the known harmonic
voltage to be applied to the rectifier end of the line through the terminating
circuit for this end, the line being terminated at the inverter end in its ter
minating circuit, with the inverter harmonic voltage assumed to be zero.
Calculate harmonic residual current in the line as a function of distance from
the energized terminal.
5. Repeat with energization from the inverter end with the rectifier har
monic voltage set at zerg.
6. Find the resultant harmonic current on the line as a function of distance
from the sending end. Here, three different possible conditions must be
distinguished:
a.
Different ac frequencies at
the two ends; for example, 50 and 60 Hz.
340
8-7
Most of the harmonic frequencies produced at one end will be different from
those produced at the other end. Since each harmonic frequency is to be
considered separately, the resultant values of suchharmonics are not calculated
at this stage. Certain frequencies, however, such as 300 Hz and its multiples,
are generated at both ends, being characteristic of the 50-Hz end and un
characteristic of the 60-Hz end. If the uncharacteristic harmonic is negligible
in comparison with the characteristic one of the same frequency, it can be
neglected. If not, it can be calculated as in case b.
b. Same ac frequencies at both ends but no parallel ac tie; hence, asyn
chronous operation. The same nominal harmonic frequencies are generated
at both ends, but the phase difference between them varies continually and
randomly. In a sense, it may be said that the corresponding harmonics
generated at the two ends have slightly different frequencies but still have the
same noise weighting. Hence, at any given point of the line, the components
of the same nominal frequency should be combined by taking the square root
of the sum of the squares.
c. Parallel ac tie, synchronous operation. There is still a variation of phase
difference between fundamental voltages at the two ends, according to
loading of the parallel ac tie, and this difference is greatly multiplied in the
higher harmonics. Even for the lowest characteristic dc harmonicthe sixth
a 180 change of phase difference accompanies a 30 change in the funda
mental phase difference; and the 180 change is enough to make the vector
sum of the two components vary from an arithmetic difference to an arith
metic sum. Hence the components may reasonably be combined, as in
case b, by the root-sum-square method.
7. For each harmonic frequency, compute the coupling between the dc

line and a particular telephone line as the mutual impedance between two
ground-return lines. See "Coupling," page 334. If this computation is made
with screening neglected, the screening factor can be computed separately,
and the mutual impedance can be corrected by multiplying it by this factor.
Multiplication of each resultant harmonic zero-sequence current by the
corrected mutual impedance for this harmonic frequency gives the induced
longitudinal voltage in the telephone line.
8. Multiply each frequency component of longitudinal voltage by its
noise weighting (C-message or psophometric weighting), and compute the
sum-mean-square. This is the longitudinal noise voltage.
9. Compute the transverse noise voltage by dividing the longitudinal
noise voltage by the balance factor. This step may be omitted if the acceptable
telephone noise is given as longitudinal noise voltage.
If the distance between the power line and the telephone line is less than
several times the spacing between power conductors, the longitudinal voltages
341
in the telephone line due to harmonic currents in each conductor of the power
line should be computed separately and then added vectorially. A computer
program for doing this has been described.60
Calculation of Telephone Noise Due to Three-phase Line
If the harmonic currents in the conductors of an ac line were accurately

known, the noise voltage in an exposed telephone line could be calculated in
the same way as described above on page 339, for a dc line. But the problem
of calculating the harmonic currents on an ac line is very difficult. Some of the
reasons for this difficulty are the following:
1. The converter of the dc link is connected not to just one ac transmission

line but to an extensive transmission network or grid. The solution of such a
network is obviously more difficult than the solution of a single line.
2. Moreover, the transmission network is connected through transformers
to various subtransmission networks and medium- and low-voltage power
distribution lines. Much of the telephone noise problem resides in these
lower-voltage circuits, most of which are carried on the same poles with
telephone circuits.
3. Conditions in the power system are continually changing with time of
day, season of year, and long-term load growth. These changes consist mainly
in magnitude and character of loads but also in outages of lines, generators,
and transformer banks for repair or maintenance and connection of new
ones as required by load growth.
4. Although the connections of converter transformers do not pass zerosequence harmonics, the unbalanced impedance of ac lines converts a part of
the balanced harmonics to zero-sequence. Unbalances of lines are more
pronounced in the distribution circuits than in the high-voltage transmission
system.
5. Although zero-sequence harmonic waves are attenuated more than

balanced harmonics, they are continually renewed by conversion through
unbalanced impedances.
6. The ac power system has many sources of harmonics besides the dc
transmission link under consideration. Among these additional sources are
industrial rectifiers, generators, transformers, and corona.
Measures for Reducing Telephone Interference
Outline
I. Reduce harmonics on the power system by filters.
II. Reduce coupling between the power system and the telephone system.
342
8-8
A. Relocate lines for greater separation.

B. Improve the balance of overhead lines by transposition.
C. Change overhead power lines to underground cables.
D. Change ground-return telephone lines (if used) to metallic circuits.
E. Change open-wire telephone lines to cable.
III. Reduce the susceptibility of the telephone circuits.
A. Improve the balance.
1. If ground-return ringing circuits are used :
a. Change low-impedance ringers to high impedance;
b. Change ground-return ringing circuit to metallic circuit.
B. Change voice-frequency circuits to carrier or microwave.
Discussion of Items IIA and IIIB. Since most long-distance communication
circuits nowadays are microwave radio or carrier cable, the principal prob
lems of inductive coordination are with the local subscriber loops, which
could not operate economically at other than voice frequencies. Also it is on
local party lines that selective ringing is often obtained by connecting some
ringers from one side of the metallic voice circuit to ground and some from
the other side to ground. If some telephone subscribers live near the power
line, adequate separation cannot always be achieved by moving the lines that
serve them. The parallel exposures, however, can be made short.
Policy on Reduction of Telephone Interference
By and large, the same people and firms are both telephone subscribers and
consumers of electric power, They are interested in obtaining good service of
both kinds at the lowest total cost. Measures to reduce telephone interfer
ence to an acceptable level should be taken on the power system, on the
telephone system, or on both systems, as required, to achieve this objective
at lowest total cost. In other words, the solution adopted should be the best
one both technically and economically.
In most cases of interference caused by converters for hv dc transmission,
the consequences of this policy would be that the power company would take
the measures required to reduce the harmonics to a level satisfactory from
the standpoint of the power system. Then, if the telephone noise were still
above the acceptable level, the required additional changes would be made
on the telephone system.
Who should pay for these changes on the telephone system is mainly a
legal question that is outside the scope of this book. In the end, the same
public pays through the charges made for both power and telephone service.
HARMONIC FILTERS
343
8-8 HARMONIC FILTERS
Purposes
The ac harmonic filters serve two purposes: (1) to reduce the harmonic
voltages and currents in the ac power network to acceptable levels and (2) to
provide all or part of the reactive power consumed by the converter, the
remainder being supplied by shunt capacitor banks, by synchronous con
densers, or by the ac power system. The dc harmonic filters serve only to
reduce harmonics on the dc line.
Types
The filters at a converter station may be classified by their location, their

manner of connection to the main circuit, their sharpness of tuning, and the
number and frequencies of their resonances.
Location. Filters are located on both ac and dc sides of converters. Filters
on the ac side may be connected either on the primary (network) side of the :
converter transformers or on the tertiary winding if one is provided for this
purpose. Filters are never connected to the secondary (valve side) windings.
Since the tertiary windings, if provided, have a lower voltage than the
primary windings, the filters are insulated for lower power-frequency and
surge voltage and, therefore, cost less. The tertiary windings, however, add
to the cost of the transformers. These windings usually have a high leakage
reactance, which inherently forms a common branch in series with all the
shunt filters and complicates the computation of possible resonances between

the filters and the ac network.
Series or Shunt. Harmonics may be (a) impeded in passing from the

converter to the power network or line by a high series impedance, (b) diverted
by a low shunt impedance, or (c) both. Figure 20 illustrates the first two kinds.
Each is a dual of the other.
The series filter must carry the full current of the main circuit and must be
insulated throughout for full voltage to ground. The shunt filter can be
grounded at one end and carries only the harmonic current for which it is
tuned plus a fundamental current much smaller than that of the main circuit.
Hence, a shunt filter is much cheaper than a series filter of equal effectiveness.
Ac shunt filters have another advantage over series filters in that at funda
mental frequency the former supplies needed reactive power but the latter
consumes it.
344
HARMONIC FILTERS
r-nm~
(<0
(b)
Fig. 20. (a) Series filter and (b) shunt filter.
For these two reasons shunt filters are used exclusively on the ac side. On
the dc side, the dc reactor, which is obviously a series element, constitutes all
or part of the dc filters. It must withstand high direct voltage to ground
and high direct current. It serves several additional functions (Section 7-2),
however, that require series connection. The remainder of the dc filters (if
used) consists of shunt branches.
Ac filters could be A-connected, but this connection offers no advantage;
therefore, the Y connection with grounded neutral is used.
Sharpness of Tuning. Two kinds are used : (a) the tunedfilter (high Q filter),
which is sharply tuned to one or two of the lower harmonic frequencies, such
as the fifth and seventh, and (b) the damped filter (low Q filter), which, if
shunt-connected, offers a low impedance over a broad band of frequencies
embracing, for example, the seventeenth and higher harmonics. The second
kind is also called a high-pass filter. Figures 21 and 22 show typical circuit
diagrams and characteristics of the two types. They are analyzed under
"Design of Tuned Filter," page 355, and "Design of High-pass Damped
Filters," page 375, respectively.
Cost of Filters
The capital cost of ac filters is in the range of 5 to 15% of the cost of the
terminal equipment.* This is high enough to justify careful design from the
standpoint of economy as well as adequacy. The cost of losses should also be
taken into consideration. The cost of filters may be partly charged to reactive-
* For example, the cost of the filters of the New Zealand scheme was said to be 12%.52
Fig. 21. Single-tuned shunt filter: (a) circuit; (b) impedance versus frequency.
2Rf~
0
(a)
Frequency f
(b)
Fig. 22. Second-order damped shunt filter: (a) circuit; (b) impedance versus frequency.
power supply and partly to filtering though there is no logical basis of the
division.
A minimumfilter is one that adequately suppresses harmonics at the least
cost and supplies some reactive power but perhaps not all that is required.
A minimum-cost-filter is defined under " Minimum-cost Tuned Filter," page
368. It may or may not give adequate filtering.
About 60% of the capital cost of the filters is that of the capacitors. Hence,
substantial savings are possible by judicious choice of kind of capacitor.
346
8-8
Criteria for Adequacy of AC Filters

Ideally, the criterion should be the absence of all detrimental effects from
harmonics, including telephone interference, which is the most difficult effect
to eliminate entirely. This criterion is impractical from both technical and
economic standpoints. From the technical standpoint of filter design, the
distribution of harmonics throughout the ac network is too difficult to deter
mine in advance. From the economic standpoint, the reduction of telephone
interference can generally be accomplished more economically by taking
some of the measures in the telephone system and others in the power
system.
The practical criterion would be an acceptable level of harmonics at the

converter terminals, expressed in terms of harmonic current, of harmonic
voltage, or of both. The filter designer would prefer a criterion based on
harmonic voltage at the converter terminals because he can more readily
guarantee staying within a reasonable voltage limit than a reasonable cur
rent limit despite changes in the network impedance seen from the converter
terminals.
Unfortunately, there is no general agreement on the acceptable limit of
either harmonic current or harmonic voltage. Presently, we can look only at
the limits that have been proposed or attained.
Stumpf40 stated that, from the experience of the Bell Telephone System
with industrial rectifiers, an I T product greater than 25 kA would be
likely to cause severe interference problems; one less than 5 kA would be
unlikely to cause any interference problems.
Several others have proposed limits for harmonic voltage.
HARMONIC FILTERS
347
b. Limit the arithmetic sum of the characteristic harmonic voltages of

orders 5 to 25 to 2.5% with any one harmonic as in (a) and the rest
from the most unfavorable impedance locus.
4. Filters for the New Zealand scheme47 were designed so that each charac
teristic harmonic would be less than 0.7%.
Effect of Network Impedance on Filtering
The converter approximates a constant-voltage harmonic source on the dc
side and a constant-current harmonic source on the ac side. More accurately,
the converter is a low-impedance harmonic source on the dc side and a highimpedance harmonic source on the ac side. We now consider, on the ac side,
the effect of filter impedance and network impedance on the harmonic
voltage Vh at the converter terminals and on the harmonic current Ihn entering
the network.
Figure 23 shows an equivalent circuit for the purpose. The harmonic
Ihn
Filter
Converter
Ac network
Fig. 23. Circuit for computation of harmonic currents and voltages on the ac side of a
converter.
1. Ainsworth41 has suggested the following limits:

a. Maximum theoretical deviation from a sine wave (H6, Section 8-5,
page 327) is not to exceed 3 to 5%.
b. Telephone harmonic form factor (THFF, Section 8-5, page 328) is not
to exceed 1 or 2%.
2. Iliceto38 reported that, for the Sardinian project,
a. H6 had been specified as 4% (a value said to be satisfactory for turbo
generators, induction motors, and fluorescent lamps) and
b. Maximum THFF as 1%.
3. The values proposed by C.E.G.B. for the Kingsnorth scheme are the
the following44 :
a. Limit every single characteristic-harmonic voltage to 1%. This should
be calculated with the most unfavorable network impedance within the
chosen impedence bounds (Figure 28).
he generated in the converter is assumed to be known. It splits

between two paths, the shunt filter and the network. The harmonic voltage
across this parallel circuit depends on the impedance of these two branches in
parallel. Let
current
= impedance of filter to harmonic of order h

Z,m = impedance of network to harmonic of order h
Y*/=Z'hf
and y
'hn
-A
7
&
hn
Then the harmonic voltage is
V =
ZhfZhIhc
Yftf+Y,
lhn
hf
and the harmonic currents in the network and filter, respectively, are
In =
V*
Zh,
V,
=
2W
Zhf+Zhn
Zh/hc
hf +
Z*hnhc
Zhf
Yft/ + Yhn
Yh/Iftc
Yfl/ + Yfin
(69)
(70)
(71)
348
Since the impedance of the network to harmonics is subject to change and is

seldom accurately known, the effect of some extreme assumptions is inves
tigated :
1. If the network impedance were nil to all harmonics, there would be
Vh = 0 and Ihn =Ihc. Shunt filters would have no effect. All the harmonic
current generated by the converter would enter the network. Filtering would
appear perfect ifjudged by voltage but bad if judged by current. This assump
tion of Zta = 0 is unrealistic. If it were approximately true, filters with series
elements would be required.
2. If the network impedance were infinite, all the harmonic current gen
erated by the converter would pass through the filter. There would be
Jh = 0, \hf = lhci and Vh = Zhflhc . Filtering would be perfect if judged by
current and could be good if judged by voltage, for the design of suitable
filters would present no great problem. This assumption of Zk = co, although
obviously untrue, might give reasonable results as regards harmonic voltages.
3. There is, however, a more pessimistic assumption: that the network and
filter are in parallel resonance. The resulting impedance would be a high
resistance; and \h, IftB, and lhj would all be high. Indeed, the harmonic
network current and voltage could be increased by the presence of the
filter. The filtering could be bad,' whether judged by current or voltage or
both. Moreover, the filter could be overloaded ; that is, its elements would be
subjected to both high harmonic current and high harmonic voltage.
Since tuned filters are customarily provided for the low characteristic
harmonics and since the impedance of such a filter at the frequency to which
it is tuned is a low resistance, severe parallel resonance of filter and network
to such a harmonic is unlikely unless the filter passband is too narrow and
unless either the system frequency is abnormal or the filter is detuned. Such
resonance is likewise unlikely at the higher frequencies for which the highpass damped filter provides a low impedance and high power factor. It is more
likely to occur at a low uncharacteristic harmonic. It is unlikely to occur at
more than one harmonic frequency at the same time although, because of
changes in the network, it could occur at another harmonic frequency at
another time.
The severity of resonance depends on the amount of damping due to
losses both in the filters and in the network. Therefore, some knowledge of
the response of the network to harmonics is desirable.
Impedance of the AC Network

The impedance of the ac network seen from the converter terminals, as a
function of frequency, may be either measured or calculated. Both methods
offer certain difficulties.
8-8
HARMONIC FILTERS
349
Measurement. Measurements must be made with the power system alive

the ac network contains other sources of harmonics,
measurements of network impedance at harmonic frequencies require a highpower source of harmonics. An adjustable-speed motor-generator set may be
used for generating a single frequency, adjustable, perhaps, from 180 to 1200
Hz. Alternatively, a rectifier short-circuited on the dc side can be used to gen
erate all its characteristic harmonics simultaneously. Either source may be
single-phase and may be connected to the ac network through a step-up
transformer. For greatest usefulness, the complex impedance must be meas
ured, not merely the scalar value.
Measurements apply only to conditions at the time of measurement and
not to future conditions under which a converter may operate. Nevertheless,
measurements give some useful information that cannot be obtained by
computation. Results of measurements are discussed later.
at high voltage. Since
Calculation. Calculation can be done by either a model (network analyzer)

or a digital computer. Calculations can be made for light and heavy loads,
outages of lines or equipment, and planned future conditions. The principal

uncertainties are due to inadequate knowledge of the circuit parameters at
harmonic frequencies and to the effect of unbalances of the circuits.
The following are some suggestions for details of the power-system repre
sentation :
Represent only the positive-sequence network.

Overhead lines may be represented by one equivalent it section, whose
branches are corrected for each frequency, or, alternatively, by several
nominal it circuits in tandem.
Transformers may be represented by a fixed leakage inductance in series with
a resistance that is a function of frequency. Their capacitances are neglected.
Generators are represented by an inductance equal to between 0.8 and 0.9 of
their subtransient inductances.
Loads may be represented by Tesistances in series with the transformer
inductances.
In the use of a network analyzer, it is more feasible to use a constantfrequency source and to readjust the circuit impedances for each harmonic
than to use an adjustable-frequency source with a fixed network.
Examples and Conclusions. Three loci of network impedance in the com

plex impedance plane for the 220-kV Italian mainland network at the San
Dalmazio terminal of the Sardinia dc link are given in Figures 24, 25, and 26.
These loci were measured on a model. They illustrate several points :
1. Alternation of resonance (low resistance) and antiresonance (high

resistance) as the frequency increases.
350
8-8
HARMONIC FILTERS
351
500
500
5613
550
540
250
530
610
*250
250
1085
109Q
250,
r300
1278
i5?3
1630
1 100
.1101
.633
650.
850'
'
iG50
950
>625
1120
710
760'
670
720
595
-250
-500
0
250
500
750
1000
R(fi)
Fig. 25. Calculated impedance of 220-kV network at minimum load with one line dis
connected. (From Ref. 38 by permission.)
500
250
500
750
Fig. 24. Calculated impedance of 220-kV network at maximum load. (From Ref. 38 by
permission.)
2. Lower maximum impedance at heavy load than at light load.

3. Great change in locus due to a line outage.
A graph of the measured scalar impedance of the 132-kV ac system at

Lydd49 is given in Figure 27. The Lydd converter of the Cross Channel link
was used as a harmonic source.
Various calculations and measurements have led to the following obser
vations:
4. Resonances in the ac network are the rule rather than the exception.
5. The harmonic impedances bear no relationship to the fundamentalfrequency short-circuit level.
6. Loads provide some damping; that is, they make resonances less sharp.
This effect is more apparent on circuits of low voltage than on those of high
voltage.
7. This damping increases with frequency.
8. On cable systems, the impedances to higher harmonics (fifteenth to
twenty-fifth) are lower than on overhead-line systems.
It can be seen that the loci in Figures 24, 25, and 26 would fit closely
within circles on the right side of the vertical axis and tangent to it at the
origin. In addition, a very great part of all these loci would be to the right of
straight lines making angles of +80 with the horizontal axis. At lower
voltages the entire loci would lie to the right of such lines. Figure 28a shows
a boundary adopted by C.E.G.B. for its studies of 132-kV filters for the
Kingsnorth project,44 the radius being 55 Q. A similar boundary, but with
500- (1 radius, was adopted for the 400-kV system. Such diagrams facilitate
filter design by limiting the range of values of network impedance to be
assumed.
Figure 28b shows the boundary in the complex admittance plane corresponding to that in Figure 28a in the impedance plane. .
352
540
500
550
520
250
270,
970
'950,
50"V85,1000
(IxT -415 V
300 1010"
32?>330 .
i
'920
1020
310
11280
650
620
560
.1028
-250
1040
1050
Fig. 27. Measured harmonic impedance of the 132-kV 50-Hz system at Lydd. (From
Ref. 49 by permission.)
568
'660
570
-500
250
500
750
1000
1250
*(B)
Fig. 26.' Calculated impedance of 220-kV network at minimum load with two lines dis
connected. (From Ref. 38 by permission.)
<
AC Filter Design General Remarks

Rn
Aims are (a) to achieve adequate harmonic reduction and (b) to supply
the required reactive power at fundamental frequency, (c) achieving both at
minimum cost.
O'f
Composition. The ac filters in each phase usually comprise:

1. Tuned filters for several (2 to 8) lower harmonics
2. A damped filter for higher harmonics
3. Switchable shunt capacitors
The lower characteristic harmonics have the largest current magnitudes

and, therefore, require filters that have low impedances at and near the
frequencies of these harmonics. It is more economical to use a separate tuned
branch for each of these harmonics than to provide a wide-band filter of
sufficiently low impedance.
(a) Z plane
(b) Y plane
Fig. 28. Bounds of loci of (a) impedance and (b) admittance to harmonics in an ac
network.
......
=-
354
8-8
The higher harmonics have smaller magnitudes, and the frequency ratio of
successive harmonics approaches unity. A great many tuned filters would be
required, and their passbands would overlap anyhow. A damped high-pass
filter is more economical for these higher harmonics.
The number of tuned filter arms varies from one dc link to another, the
commonest number being four. Logically, the combination of tuned arms
and high-pass arms should be the cheapest one that satisfies the filtering
requirements. Provision of tuned filters for the seventeenth and nineteenth
harmonics may depend on the number of bridges in the converter.
The relative magnitudes of characteristic harmonics of 2- and 4-bridge
12-pulse converters are shown in Table 8 for all bridges in service and for one
Table 8
Number of
Bridges
Single-tuned Filters. A single-tuned filter is a series RLC circuit (Figure 21)

tuned to the frequency of one harmonic (generally a low characteristic
harmonic). Its impedance is given by
1
Zf = R+j[coL- )
a>C
2/2 or 4/4
3/4
1/2
=1
1.000
0.750
0.500
11
13
17
19
23
25
0.091 0.077 0
0
0.044 0.040
0.050 0.036 0.068 0.058 0.015 0.013 0.033 0.030
0.100 0.071 0.046 0.038 0.029 0.026 0.022 0.020
(72)
At its resonant frequency, its impedance is a low resistance R. Its passband

is commonly regarded as bounded by frequencies at which \Zf\ = yJlR. At
these frequencies the net reactance equals the resistance, and the impedance
angle is 45.
Let the quantities co, R, L, C in Eq. (72) be replaced by the following :
co =
Current
I_ = tuned angular frequency (rad/sec)
(73)
Vlc
= - = deviation (per unit) of frequency from tuned frequency (74)
X0 = conL = =
I = reactance of inductor or capacitor (ohms)

>
when
bridge out of service. Filters for the fifth, seventh, seventeenth, and nine
teenth harmonics are needed only when a bridge is out of service.
Shunt capacitors are used mainly for varying the reactive power when the
load on the converter changes. They also improve the filtering of high
harmonics.
Size. The size of a filter is defined as the reactive power that the filter
supplies at fundamental frequency. It is substantially equal to the funda
mental reactive power supplied by the capacitors. The total size of all the
branches of a filter, including shunt capacitors, is determined by the reactivepower requirement of the converter station and by how much of this require
ment can be supplied by the ac network and by synchronous condensers,
if any.
The size of individual arms depends on filtering requirements, but seldom
is it less than the size for minimum cost (see "Minimum-cost Tuned Filter,"
page 368).
The design of tuned filters involves selection of their size and sharpness of
tuning (0, and is discussed immediately below.
The design of high-pass damped filters involves selection of their size,
sharpness of tuning, and resonant frequency; it is discussed beginning on
page 375.
355
Design of Tuned Filters
Harmonic Current of Order h in Per-unit of Full-Load
Fundamental
_
HARMONIC FILTERS
(o
co (75)
quality factor of inductor or sharpness of tuning of filter

Q=
R
(dimensionless) (76)
From these,
<0=co(l+5)
(77)
c=5o=vRe
(78)
COn
(79)
(On
Substitution of Eqs. (77), (78), (79) into Eq. (72) gives
Z/Rl+jQd
(80)
For the small frequency deviations (<M 1) in which we are now interested,
the impedance is given very nearly and more simply by
Zf s R(1 +j2d 0 =
|Z/| Wl + 4SZQ2 = Z0 V Q-2 + 4S2
(81)
(82)
!|
356
8-8
HARMONIC FILTERS
357
The admittance, conductance, and susceptance under like conditions are
~ 1 +J25 Q)
fR(
|Yf I
l-j2SQ
Q-J28Q2
~
R(1 + 482 Q2) X0(l+ 48ZQ2)
(83)
~ _ 1_ _
Wl + 452Q2 X0/l +482Q2
f~R{\+ 482Q2)
f =
28Q
R(1 + 452 Q2)
Q
Z0(l
X0(l
+ 432Q2)
28 Q2
+ 482Q2)
(85)
(86)
Inductive susceptance is positive ; capacitive, negative.

Frequency Deviation (Detuning). In practice a filter is not always tuned
exactly to the frequency of the harmonic that it is intended to suppress.
1. The power-system frequency may change, thus causing the harmonic
frequency to change proportionally.
2. The inductance of the inductor and the capacitance of the capacitor may
change. Of these two, the capacitance changes more because of aging and
change of temperature due to ambient temperature and self-heating (see
" Capacitors," page 365).
3. The initial tuning may be off because of finite size of tuning steps.
A change of L or C of 2 % causes the same detuning as a change of system

frequency of 1%. The total detuning or equivalent frequency deviation is,
consistent with Eq. (74),
E
A,
1 (XL
A C\
In subsequent analysis, <5 is assumed to be wholly attributable to Af
Graphs
of Impedance.
Figure 29 shows three curves of filter impedance
|Zy| versus frequency deviation 8. Curves A and B are for the same R; they
have the same minimum impedance. Curves B and C are for the same X0 ;
they have the same asymptotes D (corresponding to R = 0). The equation of
the asymptotes is \Xf\= 2X0\8\. Curves A and Careforthe same
th.y
have the same passband PB. From Eq. (81) the edges of the passband are at
8 1/2Q, and the width of the passband is IjQ.
From these curves it is apparent that the impedance of the filter at its
resonant frequency can be decreased by decreasing R. In order to keep the
impedance low over a frequency band bounded by the points of maximum
0
S, percent
Fig. 29. Impedance of tuned filter as function of frequency deviation. Curve D consists
of the asymptotes of curves B and C.
Curve
A
B
C
R
{&)
10
10
5
X0
(a>
500
250
50
25
50
250
Passband
(PB)
2%
4%
2%
expected frequency deviation, however, it may be necessary to decrease X0

also, thereby decreasing Q.
Figure 30 has a generalized dimensionless impedance curve with coordi
nates y = \Zf\ QIX0 versus x = Q8. In these coordinates, the mihimum
impedance is 1, the width of the passband is 1, and the asymptotes are
y = 2|x|. The curve is a hyperbola, given by y2 = 4x2 + 1.
Minimization of Harmonic Voltage Vh (see " Effect of Network Impedance
on Filtering," page 347) requires minimization not of the filter impedance Zhf
alone but of the impedance Zh resulting from the parallel combination of
filter impedance Zhf and the impedance Zhn of the ac network(Eq. 69) and
Figure 23) :
Iy
(88)
L,= IU
= ivi = iz"
Jh\lhc
|Y|
|YV + Yhn\
-J
358
8-8
HARMONIC FILTERS
359
The harmonic voltage is then
cfiy
vh = i\x,a,
In practice there is a maximum Q for which a coil of given inductance can be

built to operate at a given frequency, and economy dictates a somewhat
lower Q. If the harmonic voltage is unacceptably high at this Q, it becomes
necessary to decrease X0 by increasing the size of the filter.
The assumption of infinite network impedance is optimistic and unrealistic,
because it rules out the possibility of resonance between the network and the
filter, which increases the harmonic voltage.
Case 2. Purely Reactive Network. We now pass to consider the most
pessimistic assumption about the network. Equation (88), with admittances
expressed in terms of their components, becomes
L_ Asymptote,
slope = 2
V, =
0
Fig. 30. Generalized impedance curve of tuned filter.
The variables that are not under the control of the filter designer are chosen
pessimistically; that is, so as to give the highest Vh . Then the variables that
are under his control are chosen optimally to give acceptable Vh. The
variables for which pessimistic values are assumed are the frequency devia
tion S and the network impedance
Harmonic voltage is shown to in
crease with <5; hence, the pessimistic value of 5 is the greatest value that is
expected to persist: <5m. The network impedance is taken as the worst value
within reasonable limits (see "Impedance of the AC Network," page 348).
The variables that the designer can vary, within reasonable limits, are the Q
and the " size " of the filter. There is an optimum value of Q that gives mini
mum harmonic voltage for the assumed network conditions, and this value,
denoted by Q0 should be used. It is independent of filter size. Then size
is chosen for acceptable harmonic voltage and for the desired amount of
reactive power. Since Q0 depends on the assumptions about the network
impedance, several cases must be examined.41
Case 1. Infinite Network Impedance. In this case the resultant impedance
is merely that of the filter: Zh =Zhf. By substitution of Eq. (82) for \Zf\ into
Eq. (88), the harmonic voltage is given as
Q = Qo = CO
V (Ghf
Ihc
(92)
+ Gim)2 + (Bhf + B,m)2
In the present case we may put Ghn = 0 and also, on the assumption of reso
nance, Bhf + Bhn = 0. Then, simply,
x = QS
Vh = \Zkf\Ihc = IhcX0(Q-2 + 4S2m)112

For given X0 and Sm , Vh is minimized by making
,_
(89)
(90)
v = -
'
G>,f
and substitution of Eq. (85) for
(93)
Ghf , with <5 = 8m, gives
V = X0Ihc(Q-'
+ 4dlQ)
(94)
This is minimized if
Q=Q.=X
20m
(95)
Vh=.45mX0Ihc
(96)
giving the harmonic voltage as
which is twice the valueEq. (91) obtained in case 1.

The present case is unduly pessimistic, because every power network has
some conductance that decreases the voltage at parallel resonance.
Case 3. Network with Limited Impedance Angle. Let the network imped
ance angle < be limited to values between <j)m, where 0 <(j>m < 90. It is
shown that the highest harmonic voltage occurs if 4> = (f>m and has opposite
sign from that of <5. Since no limit was placed on |YJin], we must find and use
the value that minimizes |Yft| and, hence, maximizes Vh. Here a graphical
analysis is informative.
360
8-8
As before, the greatest value of <5, <5m, must be assumed, and optimum
must be found, this being the value that maximizes |YA|.
Figure 31a shows the locus of filter impedance Eq. (81)
ZV=o(e_1+725J,
with fixed Xf 28mX0 and variable Rf = X0jQ as a horizontal line in the Z
plane. In the Y plane (part b of the figure) this line, inverted, becomes a
semicircle of diameter l((25mX0) tangent to the G axis at the origin. Points on
the semicircular locus in the Y plane, corresponding to points on the recti
linear locus in the Z plane for the same values of Q, are found by drawing
radial lines from the origin of each plot with equal but opposite slopes; for
example, the points for Q = 11(25J lie on lines of slope 1 (angle 45).
Vectors from the origins to points on the loci represent filter impedance and
admittance, Zv and Yhf, respectively.
In Figure 316 vector Yv is tentatively taken as that for Q l/(2<5m), and a
tentative vector Yh is added to it to give Y* . The terminal points of Yhn and
HARMONIC FILTERS
361
Yh must lie within or on the boundary of the shaded area, drawn there for
4>m 60. It is readily seen that the tentative choice of vectors does not give
minimum Yh . For the assumed Yhf the shortest vector Yh is perpendicular to
sb
the boundary and terminates on the boundary above the vertex. Furthermore,
it may be seen that the tentative choice of Yhf is not that which maximizes Yft
with respect to Q. The proper Yhf is that which ends on the semicircle at a
point where the boundary at angle +(j>m is tangent to the semicircle.
In Figure 32 the vectors are redrawn so that Yhf maximizes Yh, and Yh
X,
2 snlx,
Zhf/
26mX0
R,
(a)
Fig. 32. Construction for finding optimum Q and worst network admittance Yh. Drawn
for <f>m = 60".
2<5_X
Fig. 31. Loci of (a) filter impedance Zhf and (6) filter admittance
Yhf with constant X
and varying R and Q; (b) also shows tentative choices of Yhf and Yhn
minimizes it. The truth of the statement made earlier, that \(j>\ =</> and that
the sign of < is opposite to the sign of 5, is proved by the vector diagram,
drawn for positive 5 = 8m and for negative $ = <pm . (Remember that <t> was
defined as an impedance angle and that ~4> is the corresponding admittance
angle.) In addition, the effect of varying 5 may be shown. Decrease of 5
increases the diameter of the semicircular locus of Yhf, thereby increasing
Yh and decreasing Vh. Hence \5\ =8m is more pessimistic than |<5| <dm.
Negative 5 turns the figures upside down.
Formulas for optimum Q and for the corresponding Yhf , Yhn, and Yh
can be derived by trigonometry, starting with the known diameter of the
semicircle, the known angle <f>m , and other angles related to (f>m Triangle
362
8-8
OCD is isosceles, with the two smaller angles COD /_ CDO 4>J2. In
right triangle ODF, OD = OF cos (0m/2); hence
|Y-! =
cos ((bml2)
i&r
(97)
In right triangle OED, OE = OD cos (0m/2) and DE = OD sin (<f>J2).
K- |YV|
IY ! IY I coco.
|Y,|
0S2 (0/2)
and
IYJ
- YW|
|
sin
os
(0/2) sin (0/2)
ZOmA0
363
(f>m on the optimum Q of the filter and on the maximum guaranteed har
monic voltage Vh. In particular, limitation to +75 reduces the harmonic
voltage for a given size of filter, or the size of filter for a given harmonic
voltage by about 21% from Case 2 (purely reactive network, <j)m = 90).
Typical values of Q in practice range from 30 to 60 with series resistors.
Double-tuned Filters. One double-tuned filter (the circuit of which is shown

in Figure 336) is substantially equivalent, near the resonant frequencies,
+1
C0S
HARMONIC FILTERS
m
W
_ srnm
40mX o
Yv = |Yw|/-90MJ2
Yt = 17-90 + ,
(101)
Yte = [Yl /+<
(102)
(100)
La
\Lb
Ra
Rb
c3
The value of Q corresponding to the chosen Xft/ is found from Figure 31a:
o3>
and from Figure 316,
tan <j)f = cot ((f>J2)
(a)
Equating the last terms of the two equations for tan <pf , we find the optimum
value of Q to be
cot(&,/2)
Qo = -
= IT*-r-r-
25m
cos (j)m+l
(104)
'
25m sin
The corresponding harmonic voltage is
\Yh\
4Wc_
cos (f>m + 1
8mQo
Vh!8mx0lhe
co
2.00
30 45
1.87 1.21
2.14 2.35
those tuned
seventh harmonics. The equations for the parameters of the
double-tuned filter are given by Ainsworth.41
The advantages of a double-tuned filter over two single-tuned filters are the
following:
to the fifth and
Such filters are used at both terminals of the Cross Channel link.42 The
impedance of that at the French terminal is given in Figure 34a and 6. Its
parameters are
Table 9
15
3.80
2.03
to two single-tuned filters in parallel (Figure 33a); for example,
1. Its power loss at fundamental frequency is less.

2. One inductor, instead of two, is subjected to full impulse voltage.
Table 9 shows the effect of limiting the impedance angle of the network to
(b)
Fig. 33. Transformation from (a) two single-tuned filters to ( b) double-tuned filter.
60
0.87
75
0.65
2.67
3.17
80 85
0.60 -0.55
3.41 3.68
90
0.50
4.00
R2
*3
= 4.2 Q
= 1.656 Q
=2.11 Q
Cj = 1.51F
C3 = 12.08 /iF
Lj =208 mH
L2 -- 24 mH
600
8-8
HARMONIC FILTERS
365
'i
500
C3
400
8.
300
Filter Components and Their Ratings
<8 200
u
100
150
200
250
300
350
400
450
Frequency f, Hz
Fig. 34a. Double-tuned filter at Echinghen for fifth and seventh harmonics: calculated
scalar impedance versus frequency.
'
400 J L500
300 _ .450
200
285
N90
>400
/270
100
350C3
x-
0
240'
-100
i375/
\ /26O
1295
250
J 340
/ \330
T220 \320
-200
"200
-300
"180
/300
310
305
-400 .160 l!|t

0
100
300
200
400
500
600
R, ft
Fig. 34b. Double-tuned filter at Echinghen for fifth and. seventh harmonics: calculated
locus in Z plane. Points are marked with the frequencies (in Hz;.
Capacitors. Capacitors account for the major part of the cost of filters.
They are composed of standard unitstypically rated at 100 or 150 kvar,
8 to 14.4 kV, at 50 or 60 Hz connected in series and parallel for obtaining
the desired overall voltage and kvar ratings.
Each unit consists of several rolls made of alternate layers of aluminum
foil and sheets of insulation, tightly enclosed in a sheet-steel box filled with
an insulating liquid. The solid insulation consists of either (a) several sheets
of special paper impregnated with the liquid or (b) a sandwich of one sheet of
such paper placed between two sheets of thermoplastic material. There is also
an unavoidable thin film of liquid between the solid insulation and the metal
foil; the thickness of the film depends on the pressure with which the rolls
are formed. Three kinds of liquid impregnants are in use: (a) mineral oil,
(b) trichlordiphenyl, and (c) pentachlordiphenyl. The last two are generically
called askarels. Thus the dielectric properties depend on those of the paper,
the impregnant, and the plastic (if used) and on the amounts used of each.
The density of the paper can be varied, ranging from 0.8 to 1.2 g/cm3, and
the paper may be impregnated with any one of the three liquids. Of these,
trichlordiphenyl is the most used at present, having superseded pentachlor
diphenyl, which has a higher freezing point, a lower dielectric strength, and a
lower dielectric constant.
Two of the most important properties of the capacitors are (a) temperature
coefficient of capacitance and (b) reactive power per unit of volume. The
latter is usually proportional to the dielectric constant and the square of the
maximum safe voltage gradient. Other important properties are (c) power
loss, (d) reliability (or life), and (e) cost. Approximate values of some of these
properties are listed in Table 10 for various dielectric materials. These values
should be interpreted as indicative rather than exact, because they vary with
the density of the paper, the thickness of the liquid film, the temperature at
which they are measured, quality control of the materials, etc.
A very low temperature coefficient of capacitance is desirable for tuned
filters in order to avoid detuning caused by change of capacitance with
ambient temperature or with self-heating of the capacitors; but this property
is unimportant for damped filters or for power-factor capacitors. Capacitors
filled with mineral oil can have either positive or negative temperature
coefficients, depending on paper density or film thickness, and, by proper
design, can be made to have essentially zero coefficient. In the past such
capacitors have been used for tuned filters almost to the exclusion of other
kinds. Low temperature coefficient is obtainable also by use of high-density
paper impregnated with pentachlordiphenyl. Both of these kinds, however,

are substantially more bulky and expensive than those of equal rating having
a dielectric of plastic and paper impregnated with trichlordiphenyl. The latter
366
HARMONIC FILTERS
367
surges. The required insulation level may be greatly reduced, with attendant
savings in cost, by protecting the inductor by connecting a lightning arrester
Table 10. Typical Properties of Capacitors for Power Systems partly based
on Ref. 46
of suitable rating in parallel with the inductor.

Temperature
Solid
Dielectric
Paper
Density
(g/cm3)
Coefficient
of Capaci
tance (10-6
Impregnant
per deg C)
Dissipa Relative
Dielectric tion
kvar
Constant Factor per Unit
at 20C
(%) Volume
Arrangement. The most economical sequence of the components of a tuned

filter, from ground to line, is R, L, C.
Overall
Paper
1.0
Mineral
oil
+250
3.6
0.17
41
Paper
1.2
+400
4.2
0.19
39
Paper
0.8
Mineral
oil
Pentachlor-
0.20
87
0.28
:V9
0.20
100
i\
diphenyl
Paper
1.0
Pentachlor-
-460
5.2
-50
5.4
5.5
diphenyl
Paper
1.2
Pentachlordiphenyl
Paper
0.8
Trichlordiphenyl
-750
1.0
Trichlordiphenyl
-500
5.5
Paper
1.2
Trichlordiphenyl
-100
5.5
Trichlordiphenyl
-710
3.1
paper
Paper
Plastic and
>!
0.28
1. Highest power-frequency alternating voltage, say,

or normal voltage.
10% above nominal
2. Higher effective frequency deviation than that assumed in determining

adequacy of filtering, say, 5%.
3. Highest harmonic current, caused by resonance of the respective filter
branch with the network and the other filter branches. Logically, harmonic
currents from other sources than the converter in question should be assumed;
however, experience has shown that, with high-power converters, the har
monics from other sources can be neglected.
66
Tuning
0.10
Conditions Under Which Required Ratings of Filter Components Are Deter

mined. For preventing damage to the filter components, their ratings must
be based on the most severe conditions to which they may be exposed ; for
example, one should assume the following :
150
200
is acceptable for use not only as power-factor capacitors and in high-pass

filters but also in automatically tuned filters (page 372).
Capacitors obtain their high reactive power per unit volume by having low
losses and operating at very high voltage stress. For this reason, prolonged
operation at moderate overvoltage must be avoided to prevent thermal de
struction of the dielectric; and even very brief operation at high overvoltage
must be avoided to prevent destructive ionization of the dielectric.
The required reactive-power rating of a capacitor is calculated as the sum
of the reactive powers at each of the frequencies to which it is subjected.
Inductors. These are built with nonmagnetic cores. The inductance usually
has a fixed value. The Q at the predominant harmonic frequency may be
selected for lowest cost and is usually between 50 and 150. If lower Q is
desired, a series resistor is used. The cost of the inductor depends mainly on
the maximum rms current and the insulation level for withstanding switching
'
;
;
The manufacturing tolerances of inductors and capacitors are 2 to 3%.

Final tuning of the tuned filters must be done after installation. The capaci
tance can be varied in small enough steps by changing the number of units in
parallel in the tier next to ground.
The most convenient indicator is a harmonic phase-angle meter, which
measures the phase difference between the harmonic voltage across the entire
filter and the harmonic current through it. Such instruments have been
installed in the converter stations at Sakuma, New Zealand, and Konti-
Skan.48
The converter current should be at least 0.8 of rated current. The fifth,
sixth, and seventh-harmonic filters (also seventeenth, eighteenth, and nine
teenth, if used) should be tuned while the converter is in six-pulse operation.
The eleventh, twelfth, and thirteenth-harmonic filters should be tuned during
12-pulse operation.
Operation with One Tuned Branch of Filter Out of Service
Such operation may become necessary or, at least, desirable when it is

necessary to repair a filter component or substitute a spare component. The
possibility of resonance between the remainder of the filter and the power
=5
368
8-8
system at the frequency for which the disconnected branch is tuned should be
assessed. If this possibility appears unlikely, the converter may be operated
with this branch disconnected while the harmonic voltage of this frequency is
measured to determine its acceptability. Perhaps it will be necessary to run
the converter at reduced load in order to make this harmonic voltage
acceptable.
Minimum-cost Tuned Filter

The cost of a filter tuned for a particular harmonic varies with the size of
the filter in the manner shown in Figure 35, and is least at a particular size.
HARMONIC FILTERS
369
A filter capacitor is subjected to currents and voltages of two frequencies

the fundamental power-system frequency and the harmonic frequency (of
order h) for which the filter is tuned. The rating of the capacitor, in Mvar,
must be the largest sum of the fundamental-frequency reactive power and the
harmonic reactive power for which the filter is designed. Other harmonics
in a tuned filter are negligible. The cost of the capacitor is assumed to be
directly proportional to its rating. The size of the filter, by definition, is the
reactive power of the capacitor at fundamental frequency only.
The fundamental-frequency source is essentially a constant-voltage source;
the harmonic source is essentially a constant-current source. Therefore, the
fundamental-frequency reactive power of the capacitor is directly propor
tional to its size ; the harmonic reactive power is inversely proportional. The
capacitor rating is
100
/2
PrC = Vlcot C +
h(D
Total
80
Minimum
Harmonic
Fundamental
component, AS
B/S
megavars
(107)
The rating of the inductor may be assumed to depend similarly on the sum
of the fundamental and harmonic reactive powers :
20
PrL~h2+
Size of filter, Mvar/phase
Fig. 35. Cost of filter versus its size, illustrated for fifth-harmonic filter for 600-MW
4-bridge 12-pulse converter.
The shape of the curve is attributable to the presence of two components of

cost, one of which is directly proportional to size and the other, inversely
proportional to size :
K=AS + BS~1
where
V2I2
+ A6
where C = capacitance (F)

co1 =2n x fundamental frequency
Vi = fundamental voltage (kV)
= harmonic current of order h (kA)
S = size of capacitor (Mvar)
8 60
component
=S
K = cost (kj)
S = size (Mvar)
A, B = constants (k$/Mvar and k$-Mvar, respectively)
(106)
(108)
hS
The uncertainty of this assumption is considerably offset by the facts that the
fundamental reactive power of the inductor is much less than its harmonic
reactive power and that the total reactive power and cost of the inductor are
smaller than those of the capacitor. The cost of the inductor is greatly
affected by its insulation level, but this component of cost should be inde
pendent of filter size.
Neglecting the cost of the resistor, the total cost of the filter is
K = PrCUc + PrLUL
(109)
where XJC and UL are the unit costs of capacitor and inductor, respectively.
Substitution of the values of PrC and PrL from Eqs. (107) and (108) gives
K=s(uc
l(Uc+ UL)
= AS
+ BS~\
(110)
370
8-8
The size for minimum cost is found by equating the derivative

zero:
dK
-
= A-BS-2 = 0
dK/dS to
(ill)
(B\1/2
Smin = b)
(112)
and substitution of this value of S into the equation for cost gives the minimum
-I
Smin = (AJ
/3.32 x 105\1/2
..
J = 9-32 Mvar Per Phase
~(
=
\3.82 x 103
iin = 2fAB = 2
= $71,200 per
V3.82 x
103 x 3.32 x 105
phase
For any size S the cost is
Find the minimum-cost fifth-harmonic filter for a bipolar four-bridge

12-pulse converter rated LOO kA, 300 kV on the dc side. The filters are to
be connected to the 235-kV 60-Hz three-phase line. The fifth-harmonic
filter is to be designed for the operation of the converter with one bridge out
of service. Assume the unit cost of capacitors to be $3.50/kvar and that of
inductors $8.00/kvar. At full load with all bridges in service, a = 15, u = 25,
and cos $ = 0.866. The limiting network impedance angle may be taken as
75.
$332,000
= $38205+ -2
O
The result is plotted in Figure 35.

Capacitance of the minimum-cost filter is
C=
9-32
-
a>iV\
377 x (235/ V3)2
= 1.34 x
10-6F= 1.34/F
Its inductance is
SOLUTION
The rated power of the converter is
Pn = VdJdn = 600 x
L=
1.00 = 600 MW
The full-load fundamental alternating line current is
IL1 = 3
(235/ V3)2(88.5 x 10"3)2 11,500

5
is the size for minimum cost.

The corresponding cost is
(113)
Example 1
371
105 $- Mvar
= 3.32 x
as
Kmla = 2AB
---= -
'BV/2
whence
. cost
"
+ UL)
HARMONIC FILTERS
Pd
Vt
600
,= p-= 1.70 kA = 1700 A

cos <f> 3(235/ V 3)0.866
The fifth-harmonic current from one bridge of the converter is approximately

1700/(5 x 4) = 85.0 A, but a more accurate value, allowing for ignition delay
and overlap, is 0.165 x 1700/4 = 70.2 A. From Eq. (98), we find that the
harmonic current in the filter is larger than that from the converter by a
factor |YAy |/|Yft[ = sec (<m/2) = sec 37.5 = 1.26; it is, therefore 1.26 x
70.2 = 88.5 A.
Uj
A = Uq +
= 3820
tl
8.00
3.50 + 2 = 3.82 $/kvar
$/Mvar
'
(D)
CihwJ2
1.34(5 x 377)2
= '21 H
Optimum Q for (j>m = 75 is 0.65/<5m . If m is taken as 0.02, Q0 = 0.65/0.02

= 32.5, and the resistance of the filter is
3,
Q0
(L2
Qo\Cj
0-210 V/2
1 /
32.5 \1.34 x 10~6/
395
32.5
The fifth harmonic voltage is (by Table 9)
V5 =3.17SmX0/5 = 3.17 x 0.0? x
395 x 88.5 = 2220 V
= 2.22 kV
This is 1.64% of the fundamental line-to-ground voltage V1 and is greater

than the usually acceptable value of 1%. For decreasing it to 1%, the size of
the filter would have to be 1.64 times as great as that calculated, or 15.3
Mvar per phase. The increase of cost (Figure 35) would be moderate,
$10,000/$71,200 = 14%.
From the curve in Figure 35 it is seen, on the one hand, that the cost of the
372
373
The capacitance can be varied by switching a variable number of capacitor

units in parallel in the tier nearest to ground potential. The inductance can be
varied by the use of a tapped coil and tap-changing mechanism or by a
variometer (a fixed coil in series with a movable coil so that the coupling
between the two coils is variable). A range of 5% is usually adequate.
One proposed method of control measures the harmonic-frequency
reactive power of the entire branch and decreases L (or C) whenever 'this
exceeds a preset value or increases L (or C) whenever it is less than another
preset value. In other words, there is an on-off servomechanism with a dead
1
]
zone.43
|
|
8-8
filter increases sharply with a decrease of size below that for minimum cost
and, on the other hand, that the cost increases more slowly with an increase
of size above that for minimum cost. Moreover, the quality of filtering in
creases with size. Thus, it is reasonable to assume that a filter smaller than
that for minimum cost is almost never used and that ones greater than that for
minimum cost are often used.
The foregoing example points to one reason why six-pulse operation of a

large converter is undesirable. The fifth-harmonic filter for full six-pulse oper
ation would be four times as great as that designed for dne bridge of four
out of service. Its three-phase size would be 3 x 4 x 15.3 = 184 Mvar. Since
the full-load reactive power consumed by the converter is P tan (f> =
600 x 0.577 = 346 Mvar, the fifth harmonic filter would supply 185/346 = 53%
of the reactive power required. The seventh-harmonic filter, designed on
the same basis, would supply 27%; and the two filters together would sup
ply 80% of the full-load reactive requirement. The whole bank of filters would
supply too much reactive power, especially atlight load.
Automatically Tuned Filters43
It was shown earlier (" Design of Tuned Filters," page 355) that for each
tuned filter branch there is an optimum value of Q, depending on the assumed
values of maximum frequency deviation 5m and maximum network imped
ance angle <f>m. Of these two variables, Sm is the one that has the greater
effect on Q, for Q varies inversely as <5m . Thus high Sm requires low Q, which
increases the continual power-frequency losses and which either impairs the
filtering by increasing the harmonic voltage or requires a greater capacitance
and, consequently, greater cost for maintaining the same quality of filtering.
Only partially offsetting this increase of cost is the fact that low Q decreases
the harmonic current at resonance and thus decreases the reactive-power
rating of a of a given capacitance.
One can get cheaper or better filtering by limiting the equivalent maximum
frequency deviation. A large part of this equivalent deviation is caused by
variation of capacitance with temperature. This part of the deviation can be
limited by using capacitors with low temperature coefficient of capacitance,
but this feature increases the cost of the capacitors.
Two methods have been proposed for limiting the equivalent frequency
deviation.
One of these maintains the average temperature of the capacitors nearly
constant by cooling them with air currents from a fan controlled thermo
statically or by capacitance measurement. The other method varies either the
inductance or capacitance by small steps so as to maintain the frequency
deviation at small values ideally at zero.
HARMONIC FILTERS
The advantages of the automatically tuned filter over a fixed filter are that,
for equally good filtering :
1. A capacitor of lower rating may be used.

2. The capacitor may be of a kind that has a high temperature coefficient
of capacitance but also has a high reactive-power rating per unit of volume
and per unit of cost.
3. Since it has a higher Q, the power loss is smaller.
1
|
]
|
|
|
|
Advantages 1 and 2 reduce the cost of the capacitor, which is the most
expensive component of the filter. Advantage 3 reduces the cost of the resistor
and the cost of the system losses. These cost savings are offset partially by the
cost of the tuning control.
In some cases, advantage 1 cannot be realized because the filter must
supply a large amount of reactive power at fundamental frequency. In such
cases, however, advantages 2 and 3 are still realizable, and, in addition, the
quality of filtering is improved.
Filter design follows the procedure already outlined for fixed tuned filters
except that a smaller <5m is specified, which depends on the accuracy of the
automatic tuning.
Figure 36 compares the reactive-power rating of the capacitor of an auto
matically tuned filter for <5m = 0.01 to that of a fixed-tuned filter having the
same filtering performance, that is, the same harmonic voltage at maximum
frequency deviation.
Example
Calculate the rating of the capacitor required for the fifth-harmonic filter
bank of Example 1 (a) if the filter has fixed tuning and if the maximum esti
mated detuning varies from 1 to 5%; and (b) if the filter is automatically
tuned so as to limit the detuning to 1%. In both cases the fifth-harmonic
voltage is to be limited to 1% of the fundamental voltage, and the maximum
angle of the network impedance at 300 Hz is assumed to be 85.
i
I
374
8-8
HARMONIC FILTERS
375
and the capacitor rating should be
PrC PrCl + PrC5
950<5m +
0.0352
Mvar
The result is plotted in Figure 36.

SOLUTION, Part b
This is the same as the result of part a for <5m = 0.01.

Design of High-pass Damped Filters
Automatically tuned
Figure 37 shows three kinds. The first-order filter, a series RC circuit,
3
2
Maximum detuning 5m,
Ox
Fig. 36. Comparison of capacitor ratings required for fixed-tuned and automatically
tuned filters.
SOLUTION, Part a
From Example 1, the fundamental line-to-ground voltage on the filters is

Vt = 235/ 3 = 135.5 kV. The fifth harmonic voltage should not exceed 1%
of this, or V5 = 1355 V. The fifth-harmonic current put out by the converter
is 70.2 A, and the fifth-harmonic current in the filter is /5 = 70.2 sec (</>m/2)
= 70.2 sec 42.5 = 70.2/0.737 = 95.5 A. By Eq. (105),
X,n =
F5(cos <f>m
~4dJs
+ 1)
"
1355 x 1.087 3.86

~
45m x 95.5
ohms
The capacitance of the capacitor of the filter is
c=
1
5 Xq
1
5g)j Xq
The fundamental-frequency reactive power (" size ") of the filter is

rC1 =
V\(QtC =
vfa
_ V\ _
y\K
5!,
51
5X0
5 x 3.86
=(i35.5)2aw
19.3
950<5m Mvar
The harmonic-frequency reactive power is
Prcs = 40 = (95.5)2 x
I
(a)
(b)
(c)
Fig. 37. High-pass damped filters: (c) first order, (b) second order, (c) third order.
requires a large capacitor and has excessive power loss at fundamental

frequency. The second- or third-order filter/is used. Both are built with low
Q, 0.7 to 1.4, and have capacitive reactance at fundamental frequency and
low, predominantly resistive, impedance over a wide band of higher fre
quencies. See Figures 38 to 43. The resonant frequency1can be chosen near
the first pair of characteristic harmonics for which tuned filters are not
provided.
Impedance. The impedance of the second-order filter is
_
_ 1 (I J- w
2nd~jcoC+\R+j(oL
(114)
and that of the third-order filter with two equal capacitors is

vars
Z3rd y'coC,.+
\R +
+
l/jcoC jcoL
_i
(115)
376
8-8
377
HARMONIC FILTERS
12
- Q=
2.000
0.707
- 1.414
1.000
J2
1.000
1.414
0.707
3
2
Frequency //per unit
0.707
Asymptotes
2.000
Fig. 38. Scalar impedance of second-order high-pass filters. Per-unit /' and Z' are defined
by Eqs. (116) to (120).
For the sake of generality, let the following dimensionless variables be
introduced:
45
Frequency //per unit
Fig. 39. Scalar impedance of third-order high-pass filters with equal capacitors.
1
"
-Jlc
(116)
(117)
-r
Q=
Z' =
Xo
Z
(118)
(119)
(120)
These definitions are consistent with those used in connection with the tuned
filterEqs. (73) and (75) except that Q is defined as the reciprocal of the Q
previously definedEq. (76). The consistency overriding this seeming

inconsistency is that, in the series RLC circuit, the lower the resistance, the
sharper the tuning ; but, with the resistor in parallel with the inductor in the
. second-order high-pass filter, the higher the resistance, the sharper the tuning.
Thus in both cases Q is an index of sharpness of tuning. In the third-order
filter, unfortunately, although the definition adopted here for Q is like that
for the second-order filter, the effect of the increasing Q is to sharpen the
resonance but to broaden the antiresonance.
The dimensionless impedances become as follows in terms of f and Q:
Z2nd=;r + (e+jr)
M
iX-1
1
1 , (
jf'+\Q + m'+jf)
(121)
(122)
8-8
HARMONIC FILTERS
379
2.0
2
= 2.00
1.5
= 1.41
Q = 0.71
1.0
1.00
1
1.41
2.00
0.5
>
0.3
/' =
= 1.41
0
Frequency f, per unit
Fig. 40. Scalar admittance of second-order high-pass filters.
= 2.00
-1.0
2.00
0.5
1.5
2.0
G'
Fig. 42. Second-order high-pass filtersloci in Y' plane.
1.41
Asymptotes
= 0.71
1
1.00
1.41
.00
0.5
1.0
2.5
These are plotted in Figures 38 and 39, respectively. The corresponding

admittances are plotted in Figures 40 and 41 as functions of frequency, and
in Figures 42 and 43 in the complex admittance plane.
All these families of curves show how the characteristics of a high-pass
filter of given size varies with Q (or R) and frequency. In every case the
impedance approaches the value R at high frequencies.
Design. There is no question of optimum Q of high-pass filters, as there
is with tuned filters, because the Q is chosen for giving the best characteristic
over the whole band of frequencies higher than those covered by the tuned
filters. The performance of high-pass filters is not sensitive to frequency
deviations even for harmonics near their resonant frequencies.
Supply of Reactive Power by AC Filters and Shunt Capacitors

0
Frequency
per unit
Fig. 41. Scalar admittance of third-order high-pass filters with equal capacitors.
The maximum amounts of reactive power that can be accepted by and

drawn from the ac network should be determined from such considerations as
voltage regulation at the converter terminals and reactive-power capability
380
8-8
HARMONIC FILTERS
381
during the worst operating condition with respect to generation of harmonics.

The filter, as already stated, presents capacitive reactance at power frequency,
and thus supplies at least part of the reactive power required by the converter.
The rest can be drawn from the ac network or supplied by making the filter
larger or by adding shunt capacitors.
If switchable shunt reactances are required, the alternatives are the
following :
+ 1.0 --6#0.6
1. A subdivided filter; that is, two or more duplicate filters.

2. Switchable high-pass filters in addition to those required for minimum
satisfactory filtering.
3. Switchable shunt capacitors.
1.41
Intheory, situations could arise in which shunt inductors would be required,

but have not yet been used on existing dc links.
A larger filter than minimum is less costly than a minimum filter plus
shunt capacitors and gives better filtering. If switchable elements are required,
a filter plus switchable shunt capacitors is cheaper than switchable filters.
The switchable shunt reactances can be switched so as to maintain either
net reactive power or alternating voltage within specified limits.
An alternative to switchable shunt reactance is a synchronous condenser.
A condenser is more expensive than shunt capacitors but has several
advantages :
Q = 0.71
+ 0.5
= 1.41
1. It can absorb reactive power as well as supply it.

2. It has smooth control of reactive power or voltage instead of a few large
-0.5
steps.
3. It provides more stability of the alternating voltage.
Q = 2.00
-1.0
0.50
1.00
1.50
2.00
2.50
G'
Fig. 43. Third-order high-pass filtersloci in Y' plane.
of nearby generating stations. The range of this reactive power should be

compared with the range of reactive power drawn by the converter between
minimum and maximum loads. If the latter range does not exceed the former,
it can be fitted into the former by addition of fixed shunt reactances. If the
latter range exceeds the former, some of the shunt reactances must be
switchable and must be switched as the load on the converter varies.
The shunt reactances must at least comprise the minimum size of filter
that satisfies the requirement for satisfactory suppression of ac harmonics
Synchronous condensers have been installed at several of the existing dc

links at the terminal that has no considerable nearby generating plant. They
are usually connected to low-voltage tertiary windings of the converter
transformers.
Final Design
This can take the form of (a) minimizing the cost (and size) of the filters
for a specified quality of filtering, according to one or more specified criteria,
or (b) optimizing the quality of filtering for a given total size of filter. In the
second form, the best distribution of size among several filter arms must be
determined. In either case the Q of each tuned filter must be optimized, and
the corresponding guaranteed-maximum harmonic voltages must be found.
In finding optimum Q, each tuned branch may well be considered by itself.
In finding the harmonic voltages,, however, it seems better to consider the
382
8-8
entire filter as a unit after those <2's have been, determined. For those har
monics for which tuned branches are provided, the design value of maximum
detuning (frequency deviation) should be assumed.
An additional refinement is now introduced in finding the harmonic
voltages; namely, by assuming that the possible load admittance is not only
bounded by extreme angles +4>m but is also limited by the minimum conduc
tance shown in Figure 28b, corresponding to the circular locus in Figure 28a.
At each one of the harmonic frequencies the vector admittance of the whole
filter lies in one of the five regions demarcated on the admittance plane in
Figure 44. At any one of the frequencies for which a tuned branch is provided
Region 1
Region 2
HARMONIC FILTERS
383
pessimistic, that is, which give the lowest resultant admittance. From the
geometry of the figure, it is apparent that the resultant admittance vector is
as described below:
In region 1 or 5 it is perpendicular to the nearer angular limit.
In region 2 or 4 it terminates on the nearer corner of the boundary.
In region 3 it terminates on the vertical (minimum conductance) line of
the boundary.
In the determination of the proper size for each filter arm for a 12-pulse
seventh-harmonic filters
for acceptable deviation from a sine wave and divide the remaining size
among the remaining branches so as to obtain the lowest telephone inter
ference factor. The fifth and seventh harmonics have very little effect on
TIF. The total number of tuned arms may vary from two to six.
The optimization of filter design requires many successive trials and is
therefore accomplished most efficiently by a digital computer. A suitable
program for this purpose is outlined by Parker.45
A comparison between calculated and measured harmonics on the KontiSkan link, as reported by E. Uhlmann,516 is given in table 11.
converter, it is generally best to make the fifth- and
Table 11
Region
of Yn
Region 3
Region 4
Region 5
Fig. 44. Determination of network admittance Y for minimum resultant admittance Y

corresponding to filter admittances Yr lying in five different regions.
the filter admittance is almost certain to lie in region 1 or 5, as previously

assumed. For harmonics of other orders, however, it may well lie in regions
2, 3, or 4. For filter admittance in each region, we must determine the network
admittance and the corresponding resultant admittance which are most
Harmonic order
Calculated value (%)
Measured value (%)
11
0.4
0.33
13
0.3
0.24
23
0.28
0.10
25
0.28
0.10
Design of DC Filters
As stated before, the dc reactors, although designed primarily for other

functions (Section 7-2), constitute all or part of the filters on the dc side of the
converter. If the dc line is in cable, generally no additional filtering is required
on the dc side, because the cable sheath and the ground or seawater adequately
shield telephone lines from noise induced by the dc harmonics. With over
head dc lines, additional filtering is usually required.
The additional filtering, if used, has the same general form as that of ac
filters, including one or more tuned branches (e.g., for the sixth and twelfth
harmonics), and, perhaps, also a high-pass filter, from each dc pole to the
ground or to the electrode line.
Design procedure is similar to that for ac filters except that the rating of
capacitors for combined direct and harmonic powers cannot be calculated
from the total reactive power. The same kinds of capacitors, however, can be
384
8-8
used for dc as for ac. The ratings of the capacitors are determined primarily
by the maximum direct voltage and by the required capacitances. The
harmonic voltages and currents in dc filters are relatively smaller than in ac
filters because of the effect of the dc reactor.
Under extreme conditions of telephone interference, the inductance of the
dc reactor may be increased, or two reactors may be connected in tandem.
In the latter case, the shunt branches should be connected to the point between
the two reactors.
Summary of Data or Asssumptions Required for Design of Filters
I. Ac network
A. Preexisting harmonic levels

B. Variation of alternating voltage
C. Unbalance of alternating voltage (one of the factors affecting magni
tudes of the uncharacteristic harmonics)
D. Power frequency and its variation
E. Requirements on power-frequency reactive power maximum that
can be supplied by or absorbed by the network
F. Impedances at harmonic frequencies
1. For various load conditions
2. With various outages of lines and equipment
3. Limiting phase angle
II. Converter
A. Number of poles
B. Number of bridges
C. Rated direct current and voltage
D. Use of ground return
E. Pulse number or numbers
F. Normal converter angles (a, y, u)
G. Modes of abnormal operation contemplated (for example, one or
more bridges out of service, monopolar operation of bipolar line,
current over rated value)
H. Overvoltage protection (which limits surge voltages on filters)
III. Criteria adopted for adequate filtering (page 346)

A. Maximum theoretical voltage deviation
B. Maximum voltage of an individual harmonic
C. Voltage TIF
HARMONIC FILTERS
385
Protection of Filters
Inductors. The voltage division across a tuned filter is as follows :
1. The power-frequency voltage appears almost wholly (and direct voltage,
wholly) across the capacitor.
2. The harmonic voltage appears equally across the capacitor and inductor.
It is the predominant voltage across the inductor but a minor one across the
capacitor.
3. Surge voltages appear almost wholly across the inductor.
Thus the only component subjected to transient overvoltages much greater
than the steady voltage is the inductor. Direct-current links constructed
before 1970 use filter inductors insulated for the same level (B.I.L.) as the
circuit to which the filters are connected. The cost of an inductor, however,
depends very largely on its insulation level. Considerable saving of cost is
possible by protecting the inductor by a lightning arrester connected in
parallel with it. The arrester must be able to "seal off" (stop conducting)
against the steady-state voltage, which may be assumed as the arithmetic
sum of the crest values of the power-frequency voltage and the harmonic
voltage. This voltage is very much lower than the crest surge voltage deter
mined by lightning arresters on the main power circuit. The arrester across
the inductor may be an ac type even if used on a dc filter, because the harmonic
voltage alternates. When the early dc links went into service, commercial
low-voltage lightning arresters capable of withstanding the required crest
currents and of dissipating the required energy were not available. For
satisfying these requirements, arresters of higher voltage rating were neces
sary, and, consequently, the economy in the inductor could only be partially
realized.
Similar considerations hold regarding protection of the inductors of highpass filters except that the steady-state voltage across these inductors is so low
that a simpler form of arrester or protective gap is feasible.
Resistors. The resistor in a high-pass filter of any order must withstand

the same surge voltage as the reactor and can be protected by the same
arrester or gap. In tuned filters, the arrester may be connected across the
series combination of resistor and inductor.
Capacitors. Capacitors need no overvoltage protection beyond that pro
vided for the main circuit to which the filters and shunt capacitors are connected. Protection against short circuits in individual units is provided by
fuses. The fuse of a faulty capacitor unit should blow before the case of the
unit bursts. Faulty units are replaced periodically after several of them occur
;
i
j
j
386
BIBLIOGRAPHY
in the same capacitor bank. The bank must be taken out of service for
replacing units.
General. If the dc ground is distant from the converter station, with con
nection to it by an electrode line, the dc filters are connected to the electrode
line instead of to ground. A dc lightning arrester or gap must be connected
from the electrode line to ground at the station. This must seal off against the
direct voltage caused by IR drop in the electrode and electrode line.
The ac filters as a whole are usually connected to the three-phase supply
circuit through a circuit breaker, and individual filter branches have dis
connecting and grounding switches. Use of resistance-step circuit breakers
reduces transient overvoltages during energization of the filters.
In some converter stations, faults in the filter capacitors are detected by
relays in which the fundamental currents are compared in two filter branches
connected to the same phase.48
PROBLEMS
1. Prove that the wave form of the network-side line current of one six-pulse
bridge is the same whether the transformer bank is connected YA or AY.
2. Make a Fourier analysis of the ac wave shape in Figure 3c.
3. Make a Fourier analysis of the ac wave shape in Figure 176.
4. Bipolar dc links are sometimes tested in round power. This means that
direct current is circulated in the two conductors, as usual, but both poles
have the same polarity of voltage. Therefore power flows in one direction
on one pole and in the opposite direction on the other pole. At each ter
minal one pole of the converter works as a rectifier, the other, as an inver
ter. Determine the phase sequence or sequences of the six-pulse and
twelve-pulse harmonics of the direct pole-to-ground voltages if the con
verters have one bridge per pole and the two bridges are supplied with
three-phase voltages (a) with no phase difference between them and (b)
with 30 phase difference. Which of the these arrangements is preferable
with respect to telephone interference (c) from the dc line and (d) from
the ac network ?
BIBLIOGRAPHY
1. "Wave Propagation in Overhead Wires with Ground Return," by John R. Carson,
Bell System Tech. Jour, Vol. 5, pp. 539-554, October 1926.
2. Engineering Reports of the Joint Subcommittee on Development and Research of the
Edison Electric Institute (formerly the National Electric Light Assn.) and the Bell
Telephone System (American Telephone and Telegraph Co.), New York, 5 volumes,
July 1926 to January 1943.
387
3. "Current and Voltage Wave Shape of Mercury Arc Rectifiers," by H. D. Brown

and J. J. Smith, A.I.E.E. Trans., Vol. 52, pp. 973-984, 1933. Disc., pp. 984-986.
4. "Mutual Impedance of Grounded Wires for Horizontally Stratified Two-Layer
Earth," by John Riordan and Erling D. Sunde, Bell System Tech. Jour., Vol. 12,
pp. 162-177, April 1933.
5. " Influence of H.V.D.C. Lines on Communications Plant," by H. Klewe, C.I.G.R.E.,
Report 333, 1935.
6. "Interference with Neighbouring Circuits Caused by High-Voltage Direct-Current
Power-Transmission Lines," by Ch. Degoumois, Brown Boveri Rev., Vol. 28, pp.
330-332, October 1941.
7. " Operation of Nonsalient-Pole-Type Generators Supplying a Rectifier Load," by
M. D. Ross and J. W. Batchelor, A.I.E.E. Trans., Vol. 62, pp. 667-670, November
1943.
8. "Inductive-Co-ordination Aspects of Rectifier Installations," A.I.E.E. Committee
Report, Elec. Eng. (A.I.E.E. Trans.), Vol. 65, pp. 417-436, July 1946.
9. "Harmonic Currents Produced by Rectifiers in A.C. Supply Systems and Methods

of Preventing Resonance Phenomena Due to These Currents in Systems Including
Capacitors," by Harry Forssell, C.I.G.R.E., 1946, Report 102, 15 pp.
10. "Notes on Harmonics Returned by Mutators into the Supply System," by P. G.
Laurent, C.I.G.R.E., 1946, Report 125, 16 pp. Tuned shunt filters.
Earth Conduction Effects in Transmission Systems, by Erling D. Sunde, Van Nostrand,

New York, 1949; esp. Chapter IV, "Mutual Impedance of Insulated Earth-return
Conductors," pp. 98-139.
and Flow of Harmonics in Transmission Systems," by S. Whitehead
"Generation
12.
and W. G. Radley, Proc. I.E.E., Vol. 96, Part II, pp. 29-48, 1949.
13. "Effects of Supply Line Unbalance on the Filtered Output Ripple of Polyphase
Rectifiers," by C. K. Hooper and C. H. McAdie, A.I.E.E. Trans., Vol. 69, Part II,
pp. 766-770, 1950.
14. " Interactions between Harmonics, Transformer Saturation, and Operation of Recti
fiers and Inverters," by E. Friedlander, C.I.G.R.E., Report 302, 1950, 19 pp.
11
15. " Inductive Coordination Aspects of D-C Systems Supplied by Rectifiers," A.I.E.E.
Committee Report, A.I.E.E. Trans., Vol. 70, Part 1, pp. 1034-1055, 1951.
16. "The Influence of the Number of Phases Used in a Converter on Harmonics," by
F. Busemann, Direct Current, Vol. 2, pp. 65-72, December 1954.
17. " Contributions to the Study of Rectifier and Network Harmonics and Their Evalua
tion for Tracing the Source of Wave Distortions," by E. Friedlander and W. G.
Thompson, C.I.G.R.E., Report 312, 1954.
18. "Higher Harmonics in a Power System Supplying Mercury Rectifiers," by B. G.
Orkina, translated by B. F. Kraus, Direct Current, Vol. 2, pp. 115-121, June 1955.
From Elektrichestvo, No. 2, 1955.
19. Interference between Power Systems and Telecommunication Lines, by Hans Robert
Julius Klewe, London, Edw. Arnold, 1958. An E.R.A. report.
20. "Telephone Interference and Other Effects Caused by the Gotland H.V.D.C. Trans
mission," by G. A. Pettersson, L. Ahlgren, and H. Forsell, C.I.G.R.E., Report 324,
1958.
388
BIBLIOGRAPHY
21. "The Calculation of the Harmonics of Audio-Frequency Currents in D.C. Power

Lines," by A. M. Pintsov, Direct Current , Vol. 4, pp. 8-13, June 1958. Formulas for
harmonic emf's on dc side of 12-pulse, two-bridge converterwith common system reac
tance and with four and five valves conducting alternately, five conducting simul
taneously, and five and six conducting alternately. Graphs. Formulas for equivalent
ac inductances.
22. "The Proceedings of the Direct Current Study Committee of C.I.G.R.E., Leningrad,
1957, Part III," Direct Current, Vol. 3, pp. 267-272, March 1958. Harmonics, pp.
267-269.
23. "Harmonics in Electrical Systems and Their Reduction through Filter Circuits," by
E. Bornitz, M. Hoffman, and G. Leiner, C.I.G.R.E., Report 304, 1958.
24. "Capacitors in Power Systems with Rectifier Loads," by A. Schmidt, Jr., C. E.
Rettig, and H. A. Gauper, Direct Current, Vol. 4, pp. 116-119, March 1959. Pre
sented to C.I.G.R.E. Study Committee 10, 1958. Based on A.I.E.E. Tech. Paper
53-24, revised. Parallel resonance between shunt capacitors and power-system reac
tance excited by harmonics from rectifiers may cause overloading of capacitors and
telephone interference.
25. "Evaluation of Message Circuit Noise," by A. J. Aikens and D. A. Lewinski, Bell
System Tech. Jour., Vol. 39, pp. 879-909, July 1960.
26. The Telephone Influence Factor of Supply System Voltages and Currents, Supplement
to Engineering Report No. 33, Joint Committee on Development and Research,
Edison Electric Institute and Bell Telephone System, E.E.I. Publication 60-68,
Sept. 12, 1960.
Hingorani, Garraway, London, 1960, Chapter 10, "Harmonics."
28. "Problems Arising from Higher Harmonics in H.T. D.C. Power Transmission," by
E. Schulze, Appendix II to C.I.G.R.E., Report 417, 1960. Also in Direct Current,
Vol. 4, pp. 129-130, June 1959.
29. "Reactive Power Compensation and Harmonic Filters," by A. Stalewski, Appendix
III to C.I.G.R.E. Report 417, 1960. Also in Direct Current, Vol. 4, pp. 130-133,
June 1959.
30. " Computation of the Harmonics of Rectifier Current in the Single-Conductor Mode
ofD.C. Transmission" (in Russian), by Yu. S. Kraichik, N.I.I.P.T., Vol. 5, pp. 180192, 1960.
31. " Telephone Influence Factor (TIF) and Its Measurement," by W. C. Ball and C. K.
Poarch. A.I.E.E. Trans., Vol. 79, Part 1, pp. 659-664, January 1961.
32. Schaltungslehre der Stromrichtertechnik (Converter Circuit Theory), by Theodor
Wasserab, Berlin, Springer, 1962, Chapter VIII, "Oberwellen der Gleichspannung
und des Wechselstromes " (Harmonics of the Direct Voltage and of the Alternating
Current), pp. 361-393.
33. "Some Problems in Connection with the Commissioning of the Lydd Converter
Station," by L. Csuros and G. S. H. Jarrett, Direct Current, Vol. 7, pp. 114-121,
May
1962.
34. "D.C. Interconnection between France and Great Britain by Submarine Cables:
Behaviour of the Boulogne-Echinghen Converting Station from the Point of View of
Telephone or Radio Disturbances," by P.-G. Laurent, C. Gary, and J. Clad6,
C.I.G.R.E., 1962, Report 331, 27 pp.
389
35. Directives Concerning the Protection of Telecommunication Lines Against Harmful

Effects from Electricity Lines, International Telegraph and Telephone Consultative
Committee (C.C.I.T.T.), published by the International Communication Union,
Geneva, 1963, 414 pp. loose leaf.
36. "Static Converters and Distortion of the Power System," Direct Current, Vol. 8,
pp. 145-148, May 1963. Based on I.E.E. Symposium on Transient, Fluctuation, and
Distorting Loads.
37. "A New Objective for Message Circuit Noise," by D. A. Lewinski, Bell System Tech.
Jour., Vol. 43, pp. 719-740, March 1964.
38. "Some Design Aspects of Harmonic Filters for H.V.D.C. Transmission Systems,"
by F. Biceto, J. D. Ainsworth, and F. G. Goodrich, C.I.G.R.E., 1964, Vol. 3, Report
405, 25 pp. Sardinian link.
39. " Propagation of Harmonic Currents and Voltages through a Transmission System :
CEGB tests," Direct Current, Vol. 9, pp. 156, 161, November 1964.
40. "Induction. Effects of HVDC Transmission on Telephone Communications," by
F. M. Stumpf, I.E.E.E. 1965 International Convention Record, Part 9 ("Power"),
pp. 77-88. Abstract in Direct Current, Vol. 10, p. 71, May 1965.
41.
42.
"Filters, Damping Circuits, and Reactive Volt-Amps in H.V. D.C. Convertors," by

J. D. Ainsworth, Chapter 7, pp. 137-174, of High Voltage Direct Current Convertors
and Systems, edited by B. J. Cory, Macdonald, London, 1965.
"Some Design Aspects of the Cross-Channel Power Link," by L. A. Harris, ibid.,
Chapter 8. Filters at Lydd and Echinghen are described on pp. 196-198 and 217.
43. "The Application of Self-Tuned Harmonic Filters to HVDC Converters," by C. D.

Clarke and M. J. Johanson-Brown, I.E.E. Conference Publication No. 22, Conference
on High Voltage D.C. Transmission, held at the University of Manchester, Sept. 19-23,
1966, Part 1, pp. 275-276, Paper No. 55.
44. "Design Considerations for A.C. Harmonic Filters," by G. L. Brewer, C. D. Clarke,
and A. Gavrilovic, ibid., Part 1, pp. 277-279, Paper No. 56.
45. "An Analytical Study of Harmonic Filter Design," by A. M. Parker, ibid., Part 1,
pp. 280-283, Paper No. 57.
46.
47.
48.
49.
50.
51.
"The Design of Capacitor Components of Large High Voltage A.C. Filter Net
works," by R. O. M. Powell, ibid., Part 1, pp. 284-288, Paper No. 58.

" Experience with HarmonicsNew Zealand HVDC Transmission Scheme," by G. H.
Robinson, ibid., Part 1, pp. 442-444, Paper No. 89.
"Experience of A.C. Harmonics from HVDC Installations," by H. Holmborn and
H. Mirtensson, ibid., Part 1, pp. 445-449, Paper No. 90.
"Factors Influencing the Harmonic Impedance of a Power System," by K. W.
Huddart and G. L. Brewer, ibid., Part 1, pp. 450-452, Paper No. 91.
"Turbo-generator Operational Parameters on Rectifier Loads," by V. Easton, ibid.,
Part 1, pp. 453-454, Paper No. 92.
Ibid., Part 2 (Discussions), pp. 83-90, 109-119. a. Discussion by G. E. Gardiner,
p. 115 and Fig. UU, p. 119. b. Discussion by E. Uhlmann, p. 85.
52. "Harmonic Phenomena Associated with the Benmore-Haywards H.V.D.C. Trans

mission Scheme," by G. H. Robinson, New Zealand Engineering, Vol. 21, pp. 16-29,
Jan. 15, 1966.
390
53. "Harmonics on an E.H.V. AC/DC Transmission Line," by D. H. Welle, A. G.

Phadke, and D. K. Reitan, I.E.E.E. Conference Record, Region 3 Conference,
April 1967, pp. 331-336. Also " Evaluation of Harmonic Levels on an HVDC Trans
mission Line," by same authors, American Power Conference, Chicago, April 1967.
54. "Mutual Impedance of Crossing Earth-Return Circuits," by M. Krakowski, Proc.
I.E.E., Vol. 114, pp. 253-257, February 1967. Disc., Vol. 115, p. 724, May 1968.
55. " Harmonic Instability between Controlled Static Convertors and AC Networks," by
J. D. Ainsworth, I.E.E. Proc., Vol. 114, pp. 949-957, July 1967.
56. "Generation of Abnormal Harmonics in High Voltage AC DC Power Systems," by
Arun G. Phadke and James H. Harlow, I.E.E.E. Trans, on P.A.&S., Vol. 87, pp.
873-882, March 1968. Disc., pp. 882-883.
57. "The Phase-Locked OscillatorA New Control System for Controlled Static Con
vertors," by J. D. Ainsworth, I.E.E.E. Trans, on P.A.&S., Vol. 87, pp. 859-865,
March 1968. Disc., p. 865.
58. "Unusual Current Harmonics Arising from High-voltage DC Transmission," by
John Reeve and P. C. S. Krishnayya, I.E.E.E. Trans, on P.A.&S., Vol. 87, pp. 883-892,
March 1968. Disc., pp. 892-893.
59. " The 450 kV Direct Current Transmission System for the Nelson River Project,"
by L. A. Bateman, L. S. Butler, and R. W. Haywood, C.I.G.R.E., Report 43-02,
1968, 9 pp. See Part 4, p. 5. Similar information in " Nelson River D.C. Transmission
Project," by L. A. Bateman, R. W. Haywood, and R. F. Brooks, I.E.E.E. E.H.V.
Conference, Montreal, October 1968.
60. "Telephone-Interference Calculation for Multiconductor Power Lines," by W. Scott
Meyer and Hermann W. Dommel, I.E.E.E. Trans, on P.A.&S., Vol. 88, pp. 35-41,
January 1969. Disc., p. 41.
61. "An Interconnected AC Filter for High Voltage DC Converters," by T. Gilsig,
I.E.E.E. Trans, on P.A.&S., Vol. 89, pp. 463-467, March 1970. Disc., pp. 467-469.
62. "A General Approach to Harmonic Current Generation by HVDC Converters," by
John Reeve, John A. Baron, and P. C. S. Krishnayya, I.E.E.E. Trans, on P.A.&S.,
Vol. 88, pp. 989-994, July 1969. Disc., pp. 994-995.
9
Ground Return
9-1 ADVANTAGES AND PROBLEMS
Mostdc transmission lines use a return path through the ground or seawater
or both, either continuously or for short times of emergency. For brevity,
such return paths are called ground returns even if the sea constitutes all or
part of the path. The ground path has a very low resistance and corre
spondingly low power loss in comparison with a metallic line conductor of
economical size and equal length if the ground electrodes are properly
designed. The resistance is low because direct current in the earth in a steady
state, unlike transient or alternating current, does not follow closely the
route of the metallic conductor but spreads over a very large cross-sectional
area in both depth and width. The resistance of this path is essentially
independent of the length of the line and may be regarded merely as the sum
of the resistances associated with each electrode unless the electrodes are near
one anotherwhich certainly would not be true in long-distance transmission.
These resistances can be made low.
It has been stated frequently in the literature (e.g., Ref. 35) that a groundreturn dc line is more economical than a line with metallic return, because
ground return saves most of the cost of one metallic conductor and the
losses in it. (It is, of course, understood that these savings are slightly reduced
by the cost of the ground electrodes and the power loss in the ground re
sistance.) This contention is true only if a ground-return circuit is compared
with a monopolar metallic circuit that has the metallic return conductor
grounded at one point; but the latter kind of circuit is almost never used.
A fairer comparison would be one between a monopolar ground-return
circuit and a bipolar metallic circuit on the basis of cost of line per unit of
power transmitted. Let us assume that both lines have the same current per
conductor and the same voltage from conductor to ground. The bipolar line
has twice as many conductors and can carry twice as much power at a
slightly higher efficiency than the monopolar ground-return line. The cost of
the bipolar line is about twice that of the monopolar line if submarine cables
391
392
GROUND RETURN
are used; but if an overhead line is used, its cost is surely less than twice that
of the monopolar line.
The advantage of the ground-return line is not its capability to transmit
power at a lower unit cost. There are, however, two definite advantages of
ground-return fines.
The first is that a dc fine can be built in two stages if the power capability
required in the initial stage is half or less of that required ultimately. It can
operate in the first stage as a monopolar fine with one metallic conductor
(overhead or in a cable) with ground return and in the second stage as a bi
polar fine with two metallic conductors and normally without ground
current. Thus a considerable part of the ultimate capital cost is postponed
until the second stage. The initial expenditure is somewhat more than half of
the total expenditure. The additional costs for the first stage are for the
electrodes and the electrode fines and, in the case of an overhead fine, for
the towers and right of way.
After the completion of a bipolar line, a second advantage is enjoyed;
namely, that, in the event of an outage of one pole of the line or of a converter,
the fine can be operated temporarily at almost half of its rated power by the
use of the other pole with ground return. For this reason the reliability of a
bipolar dc fine is substantially equal to that of a double-circuit three-phase
line, although it has only two conductors instead of six.
A homopolar fine with two or more independent conductors of the same
polarity has the same advantages as a bipolar fine. The difference between
these two kinds of line is that the homopolar fine with two poles in operation
has twice the ground current that it has with only one pole in operation; the
bipolar fine with both poles in operation has much less ground current
(theoretically none) than it has with one pole in operation.
The Gotland, Konti-Skan, and New Zealand fines were built, or intended
to be built, in two stages. The Nelson River fine, which is to have several
bipolar circuits, is being built in several stages.
Ground return has its problems as well as its advantages. The problems are
(a) the design of the ground electrodes for low resistance and low cost of
installation and maintenance, (b) the location and screening of the electrodes
so that ground currents cause negligible electrolytic corrosion of buried or
immersed metallic structures and harmless step and touch voltages, and (c)
prevention of the ground current's interference with the operation of other
services, such as ac power transmission, railway signals, or ship's compasses.
Before considering these problems in detail, we consider the distribution of
ground currents, especially in the vicinity of an electrode. An understanding
of the current field in the earth is necessary for understanding the problems
caused by earth currents.
9-2
THE
CURRENT FIELD NEAR AN ELECTRODE
393
9-2 THE CURRENT FIELD IN THE EARTH NEAR AN ELECTRODE
The current field at any point in the earth is described in terms of two vector
quantitiesthe current density J and the electric field strength $, and one
scalar quantity: the electric potential V with respect to remote earth. The
field as a whole is characterized by how these quantities vary in a system of
space coordinates having as its origin the ground electrode that is conducting
into the ground. From a practical standpoint we are interested
a current I
mainly in these field quantities at or near the surface of the earth.
Uniform Earth
The only property of the earth that determines the field of direct current in
it is its resistivity. Hence, in this connection, "uniform earth" means an earth
of uniform anisotropic resistivity.
Actually, the earth is very nonuniform, having a core of radius 3500 km
(0.55 of earth's radius) consisting of hot molten magma of low resistivity,
surrounded by a mantle 2860 km thick of hot but solid material, also of low
resistivity. The mantle is overlain under the continents by a relatively thin
crust about 40 km thick made of rocks of higher resistivity, but this crust is
thin or absent under the oceans. The crust is also nonuniform, consisting of
strata of different rocks, overlain in places by various soils and by bodies of
water. The surface layers of the land have many local irregularities of differing
resistivity, such as rivers,marshes, deserts, and buried pipe fines. Approximate
resistivities of some materials are listed in Table 1.
In spite of this general nonuniformity, there may be relatively uniform
volumes of earth in the vicinity of an electrode. Since the assumption of
uniform earth gives the simplest case for analysis, it is a good starting point.
It also affords a basis with which the various cases of nonuniformity can be
compared.
The following assumptions are made in addition to that of uniform
resistivity :
1. The surface of the earth is a horizontal plane.

2. The electrode is hemispherical with center at the surface of the earth
(Figure 1) and radius a meters.
As a consequence of the symmetrical geometry, the current flow fines are

radial, and the current density is uniform everywhere on the surface of any
394
GROUND RETURN
9-2
Table 1. Typical Resistivities of Materials Encountered by Ground Currents
Material
Copper
Typical Resistivity
(Q m)
1.72 x 10-8
Remarks
1 x 10-7
1 to 7 x 10-7
2.2 x 10-7
Lead
Carbon
The electric field intensity, also known as the voltage gradient, is
At 20C; increases with

$ =pj=
0.3
0.2
Seawater
Varies with composition
Rocks
(2)
volt per metre
where p = resistivity (Q m) of earth

The electric potential at any point in the earth with respect to a point at
infinite distance ("remote earth") is
Varies with salinity and
V=
temperature
Fresh water
Soils
Average land surface
-2
7tr
3 x lO"5
Granulated coke
395
where I
= current (A) entering ground from electrode
r = radius (m) of hemisphere
27tr2 = area (m2) of hemisphere
temperature
Iron
Steel
THE CURRENT FIELD NEAR AN ELECTRODE
5 to 300
10 to 1000
150
103 to 108
[ idr=~-[
)r
2n)r
%r
= ~~
2nr
volts
(3)
These values hold at the surface of the earth as well as along vertical oi
inclined rays. Figure 2 shows equipotential lines on the surface of the earth
I
77Wm
Fig. 1. Current flow from hemispherical electrode in surface of uniform earth is shown
by solid lines. Broken lines are intersections of equipotential surfaces with vertical plane.
hemisphere in the earth having its center at the center of the electrode. This
through the hemisphere divided by the area
of the hemisphere :
current density is the total current
/ = -z2
amperes per square metre
0)
10 m
Fig. 2. Plan view of equipotential circles on surface of the ground near a hemispherical
electrode of 1-m radius on which 100 V is impressed.
around an electrode. Figure 3 shows current density, voltage gradient, and

potential plotted in ordinary coordinates, and Figure 4 shows them in log-log
coordinates. The log-log plot is advantageous in its ability to cover a wide
range of the variables legibly and in the simplicity of the curves.
396
GROUND RETURN
9-2 THE
CURRENT FIELD NEAR AN ELECTRODE
397
The resistance of the electrode with respect to remote earth is
100
P
I 2na
ohms
(4)
Half of this resistance is that of the earth between the electrode and a hemi
sphere having twice the radius of the electrode. Low grounding resistance is
required for hv dc transmission and can be obtained by using a large electrode.
For practical reasons, very few, if any, electrodes are hemispherical.
However, the formulas given, except Eq. (4), are still valid for distances
exceeding several times the greatest dimension of the electrode but much
smaller than the distance to the other electrode.
>>
6
Electrode in Local Volume of Low Resistivity
Equation (4) indicates that low earth resistivity aids in obtaining low
resistance to ground, but, under our former assumption of uniform earth,
this would imply lowering the resistivity of the whole planet. A more realistic
assumption would be to locate the electrode, still assumed hemispherical, at
the center of a hemispherical volume of material of resistivity px much lower
than the resistivity p2 of the rest of the earth (see Figure 5). This assumption
5
10
15
Distance r from center of electrode, m
Fig. 3. Current density J, electric field intensity 3, and potential V in uniform earth of
resistivity 100 O- m near hemispherical electrode of 1-m radius conducting current of
6.28 A into the ground.
Fig. 5. Hemispherical electrode in hemispherical volume of soil of radius b and of

resistivity pi less than resistivity p2 of the rest of the earth.
O)
i-2
,"3
,-4
10
102
103
104
105
106
Fig. 4. Log-log plot of the quantities in Figure 3.
preserves the symmetry that makes analysis very simple. The current density
is still given by Eq. (1), and the gradient by Eq. (2) with the appropriate value
of p. The potential for r > b is still given by Eq. (3) with p =p2 and for
a < r < b by
'bdr
~J2
a'
+P2
.
dr
Pi } Pi ~Pi
2n\ a
(5)
398
GROUND RETURN
9-2
The current density, electric field intensity, and potential for this case are
shown in Figure 6 for a = 1 m, b = 10 m, pt = 10 Q m, and p2 = 100 O m.
10'
10
'
10"
10"
however, in decreasing not only the grounding resistance and potential of the
electrode but also the heating of the earth, the power loss, and the step voltage
near the electrode.
Consider a spherical electrode of radius a, buried in uniform earth (or

immersed in a uniform liquid) of resistivity p, the center of the electrode being
at depth h below the horizontal surface (Figure 7). Current is fed to the
10"
399
Buried Electrode17
\\ \ \
\ \ \
10"
FIELD' NEAR AN ELECTRODE
\a
\
_
2?
>& S
CO
.- c
THE CURRENT
Image
10
102 10 104 105 106
Fig. 6. Log-log plot of current-field quantities produced by 6.28 A into ground from
hemispherical electrode embedded in hemisphere of 10-Q, m soil of 10-m radius, which
in turn is embedded in 100-H m earth. Broken lines are for uniform 100-H m earth.
The curves are like those in Figure 4 for uniform earth at distances greater
than b but differ at smaller distances, that is, within the volume of low re
sistivity. The curve of electric field intensity S is discontinuous at r = b, but
those of J and V are continuous.
The resistance of the electrode to remote ground is
1 (p
K = 2n\ ai i", P2 -Pi
ohms
r\~-
(6)
which is less than that for uniform earth of resistivity p2 by the ratio
Pi
P2
p iV*
1-
P2>b
(7)
With the values used in Figure 6 (b - 10a and p2 = lOp, R'jRe = 0.19, and
for the ground current 1= 6.28 A the electrode potential is reduced from
100 to 19 V.
Again, a hemispherical electrode embedded in a hemispherical volume of
material of low resistivity is impractical. Lowering of the electrode resistance
and potential, however, can be obtained by locating electrodes in marshy
ground or in a trench back-filled with coke.
It must be emphasized that this effect is purely local and has no effect at
points beyond the area of low-resistivity material. It has an important use,
Buried electrode
Fig. 7. Geometry for calculating electric field of spherical electrode buried in uniform
earth.
electrode through a small insulated wire from a source above the surface. In
order to calculate the electric field around this electrode, we resort to an
imaginary device.
First, we imagine the medium (earth or water) to be extended above the
original surface until it fills infinite space. The present case then resembles the
first case considered (page 393), that of a hemispherical electrode at the
surface of a half-infinite conducting medium, except that now the current is
distributed through twice the solid angle of the earlier case; hence, for a
given current, the current density is now only half of its previous value at a
given distance r from the center of the electrode. Likewise the electric field
and potential are halved.
Second, we assume a second spherical electrode like the first one to be
placed directly above the latter at distance 2h, center to center, that is, at a
distance h above the original surface of the medium. The second electrode
injects into the medium a current equal to that injected by the first electrode.
400
9-2
GROUND RETURN
The resultant field of the two electrodes can be calculated by superposition;

that is, by the vector addition of current densities or electric field intensities
or, more easily, by the algebraic addition of potentials. In particular, at any
point on the original surface, which, clearly, is equidistant from the two
electrodes, the resultant current density and field intensity are horizontal. In
other words, no current crosses this surface, and, thus, the field satisfies the
boundary condition of the actual problem. All the current below the surface
comes from the actual buried electrode. Also, the potential at the surface is
simply twice that attributable to each electrode separately. It is
r.=
tr
volts
10'
Electrode
surface
10
Surface electrode
where r = \jx2 + h2 (see Figure 7). The electric field intensity at the surface
is horizontal and of magnitude
dV plx
=
dx 2 3
/t
(9)
Per metre
>
hemispherical electrode at the surfaceEq. (2). At great distances, & becomes

very small.
$ versus x from Eq. (9) is plotted in Figure 8 on log-log coordinates for
<3=1 m, h 10 m, I
6.28 A, and p = 100 Q m. The field intensity caused
by a hemispherical electrode with center at the surface (Figure 4) is plotted
for comparison. From a point directly over the electrode out to a radius of
about h, the field of the buried electrode is considerably less than that of the
surface electrode, but at greater distances the difference is negligible.
The surface field intensity has a maximum value
10"
at
pJ
0.0614pi
*max=
Buried
e lectroc e
Directly above the electrode, x = 0, and hence S = 0. If x

h, xr, and
& s plllnr2, having almost the same value as the field intensity from a
max
-2
401
volts Per metre
>/2A = 0.71A
(10)
(11)
but here the field is only 4/3 V 3 = 0.77 times the field of the surface electrode
at the same point. The maximum field intensity of the surface electrode,
however, occurs at its surface, where x = a; hence the maximum surface field
is reduced in the ratio
(12)
by burying. In our example ajh = 0.1, and the maximum field intensity is
10"
0.1
10
1
x, horizontal distance from electrode, m
\\
100
Fig. 8. Electric field intensity at surface of earth caused by spherical electrode of 1-m
radius buried with center 10 m below surface in uniform earth of 100-Q m resistivity and
conducting current of 6.28 A into the earth, compared with the field intensity caused by
hemispherical electrode at surface (Figure 4).
reduced by the ratio 3.85 x 10 ~3. Although the reduction in electric field
intensity due to burial of the electrode is only a local effect, it has an import
ant practical use. Burial to adequate depth is a good way of reducing hazardous
step voltages to safe values. The step voltage is the voltage between two feet of
a person or an animal.
In Figure 8, the curve for the buried electrode is asymptotic to two straight
lines, one with slope of +1, the other with slope of -2, intersecting at
x = h = 10 m, 6 1 V/m. The use of asymptotes for sketching curves of field
intensity versus distance from the electrode is another advantage of log-log
coordinates. The limiting,values of $ are found by substituting r3h3 =
402
9-2
GROUND RETURN
constant for small x (x<h) and r 3 s x3 for large x (x>h). Then & varies
as x1 for small x and as x-2 for large x.
Curves qualitatively like those in Figure 8 were obtained by measurements
made near Tracy, California, using as ground electrodes (a) the ground mat of
Tracy Substation of U.S.B.R. and (b) an electrode in a well 480 m deep.49
Resistance. The potential of the buried electrode is found by the addition
of the potential due to its own current field in an infinite medium and the
potential due to the image electrode. This sum, divided by the current of one
electrode, gives the resistance of the buried electrode to remote earth as
hmS
(13)
403
The cross-sectional areas of water and earth on the spherical shell are,
respectively,
Ax
= 2otr2
A2 2nr2
and
(14)
and the total current passing from the electrode into both media is
I
= J1A1+J2-2-2r2(/la + J2K)
t15)
From the equality of the tangential gradients, we find that

$ = (f x = $2 = P
Jz =i
whence
Pz2
Pz
Substitution of Eq. (17) into Eq. (15) gives

Electrode on Shore Line
Suppose that a spherical electrode is centered on the edge of a long,
straight shore line (Figure 9) and that both the bottom of the sea or lake and
Air
Wh6nce
Water,
pi
(18)
=2r\*LpM
9)
J2 = -17
; ctp2ipi)
2r (n +
Likewise
Earth, p2
I=2r2J1(+7yj
-
The electric field intensity in water and ground is
Si = 1 = p,J, = P2J1 = 2r2(alPi + n/p2)

and the potential is
Fig. 9. Shore-line electrode.
the land surface slope at an angle a to the horizontal. Assume, further, that
both the land mass and the sea bottom have equal, uniform resistivity p2 and
that the water has a much lower resistivity p x .Consider the part of land, water,
and air bounded by a sphere of radius r centered in the electrode. The water
inside the sphere is a spherical wedge, shaped like one segment of an orange;
the land and sea bottom constitute a hemisphere; and the air fills the re
mainder of the sphere. The current flow lines are again radial, and no current
crosses the plane boundary between any two of the three media. The current
densities in water and earth are inversely proportional to their respective
resistivities. The gradients and potentials are the same functions of r in both
media, the equipotential surfaces again being spherical shells. We now pro
ceed to find the gradient and potential as functions of r.
(20)
_y
2r(alp1+nlp2)
(22)
The resistance from electrode to remote ground is
5R'e = 2a(ajpx
+ %lp2)
(23)
In comparison with the corresponding resistance in an earth of uniform

resistivity p2 Eq. (4)it is
1
R'
(*M(P2IPi)
+1
Re
(24)
The potential and potential gradient at any point on the land are decreased
by the same ratio. Some values of the reciprocal of this ratio are given in
404
9-2
GROUND RETURN
Table 2. For example, with land resistivity p2 = 100 Q m, sea resistivity

Pi = 0.2 Q m, and slope 0.1 rad (5.7), a shore electrode has 1/17 = 0.06 of
the resistance and voltage gradient of the corresponding land electrode.
Table 2. Efficiency (1 IP) of Shore Electrode in Comparison with Land Electrode

Slope Angle a (rad)
pl
10
20
50
100
200
500
1000
0.01
0.02
0.05
1.03
1.06
1.16
1.32
1.06
1.13
1.32
1.64
2.25
4.18
1.16
1.32
1.64
2.59
4.18
7.37
2000
5000
16.9
7.37
13.5
32.8
0.1
0.2
0.5
1.32
1.64
1.64
2.25
4.18
2.59
4.18
8.96
16.9
32.8
80.6
1.80
2.59
2.59
4.18
8.96
16.9
32.8
80.6
4.18
7.37
16.9
32.8
64.7
160.0
7.37
13.5
32.8
64.7
160.0
128.0
319.0
319.0
797.0
=
w2
$ = measured electric field intensity at some point

u = calculated electric field intensity at same point for uniform
earth of resistivity ps
ps = surface resistivity at point of measurement
pa - apparent resistivity at point of measurement
Suppose that
Pa
Pz
(26)
where rj is defined by Eq. (24). With p2 = 100 Q m and r) = 0.06, as in the

example below Eq. (24), page 404, $l$u = 0.06, and pa2 = r/p2 = 0.06 x 100 =
6 1 m. The same result would be obtained by a measurement made at any
point on the sea, where px =0.20 - m. The proof is left to the reader.
The apparent resistivity of the earth, measured at any point on the surface
of the earth in the field caused by one or more electrodes, is,the resistivity of
a fictitious uniform earth which, for the same electrodes and currents, would
give the same field intensity as that measured at the point in question. Let
<?
2~P*1
2
%r
Division of Eq. (21) by this gives
Apparent Resistivity
405
were recalculated for resistivity pa, the result would be S, the observed field
intensity. Thus Eq. (25) is consistent with the foregoing definition of apparent
resistivity and may be taken as an alternative definition of that quantity.
Equation (21) gave the electric field intensity at distance r from a shore
electrode. Suppose that this intensity were measured on the landward side of
this electrode, where the surface resistivity is p2 . The field intensity in a
uniform earth of resistivity p2 would be
u2
Thus, a shore electrode, and especially one at salt water, is seen to have a
considerable advantage over a land electrode in soil of average or even low
resistivity.
7u Ps
7
(25)
Since the field intensity in a uniform earth with given current injections is
directly proportional to the resistivity of this earth, if SH , calculated forps,
Anisotropic
Earth67
Some parts of the crust of the earth consist of numerous thin horizontal
strata having different resistivities. Suppose that several kinds of such strata
occur in a cyclic order. The electrical properties of such strata in bulk can be
described approximately in terms of two resistivities :
the beds
the
beds
resistivity,
across
pt, transverse
pj, longitudinal resistivity, parallel to
We also define """"
=[ \m
) = coefficient of anistropy
(p
(27)
Since pt> ph ka> 1.

Keller and Frischknecht67 have shown that the equipotential surfaces
around an electrode in an anisotropic earth, instead of being spherical shells
as in an isotropic uniform earth, are ellipsoids of revolution about a vertical
axis through the electrode; these ellipsoids have their vertical semiaxes
shorter than their horizontal ones. Consequently, the electric field intensity
on the surface of the earth is greater by a factor ka than it would be with an
isotropic earth having resistivity />,. Accordingly, the apparent resistivity of
the anisotropic earth with vertical resistivity exceeding horizontal resistivity is
Pa = Kpi =
jplPt
(28)
406
9-2
GROUND RETURN
If, on the other hand, the strata were vertical instead of horizontal, making
the horizontal resistivity greater than the vertical, the apparent resistivity
would be
407
mm-
.V/AW.V
(29)
Pa=Pt
Here/?,, the longitudinal resistivity, is the vertical resistivity, although it may

appear to be the result of a horizontal measurement, that is, a measurement
made between points on the surface of the earth, one of which is the ground
electrode.
P2
< Pl
Two-layer Earth
The simplest assumption about the earth's structure over moderate dis
tances, after that of a uniform earth, is that of a two-layer earth or single
overburden. A horizontal layer of constant thickness h and uniform resistivity
Pi is assumed to lie on a " basement" rock of infinite depth having a different
uniform resistivity p2 , which may be either greater or less than px . Figure 10
Fig. 11. Current flow from electrode at surface of two-layer earth with lower layer
more conductive than upper layer; also the corresponding equipotentials.
the upper layer into the lower, less

resistive layer within a given distance from the electrode than it would in a
uniform earth of resistivity pv Consequently, the surface voltage gradient is
attenuated more than it would be in the uniform earth.
The limiting case of the first kind is that in which the lower layer is a perfect
insulator (p2 oo). This case is readily solved if the electrode is assumed to be
a vertical cylinder of radius a and depth equal to the thickness h of the upper
layer. By symmetry, the flux density is uniform over any cylindrical surface
of radius r and height h coaxial with the electrode. The electric field
intensity at this same cylindrical surface, including its circular intersection
with the horizontal surface of the earth is
p2<px. More of the current plunges from
pi
P2
> Pl
= PiJ =
Fig. 10. Current flow from electrode at surface of two-layer earth with lower layer more
resistive than upper layer; also the corresponding equipotentials.
shows qualitatively the lines of current flow where p2 >pv In comparison

with Figure 1 for a uniform earth, more of the current is crowded into the
upper, less resistive layer. Consequently, the voltage gradient at the surface
of the earth is attenuated less with increasing distance than it would be if the
earth had uniform resistivity pv Figure 11 is for the contrary case, in which
pj
2nrh
(30)
volts per metre
Figure 12 compares the electric field intensity at the surface of such a twolayer earth with that at the surface of a uniform earth.
The potential difference between the electrode and a point on the surface
of the earth at distance r is
F= -
$dr
dr
27ih
-y2nh a
volts
(31)
As r increases without limit, V decreases without limit; hence the concept of
potential of the electrode or of any point in its field with respect to remote
408
GROUND RETURN
9-2 THE CURRENT FIELD NEAR AN ELECTRODE
before, by an arrangement of electrodes that is symmetrical with respect to the

surface where it is satisfied. The only way to satisfy such a condition at two
parallel surfaces, however, is to make the arrangement of electrodes sym
metrical with respect to both of those surfaces. A little thought shows that an
infinite number of electrodes are required, as suggested by Figure 13, and
10'
Y"
10
tf\\
2?
409
10"1
10
(b)
10*
1
10
Distance r, m
10'
Distance r, m
10'
(a)
Fig. 12. Comparison of electric field intensity at surface of the earth: (1) in two-layer
earth and (2) in uniform earth. The upper layer of the two-layer earth has resistivity
Pi = 100 n m thickness h= 10 m and cylindrical electrode of height 10 m and radius
1 m; the lower layer has p2 ~ co and infinite thickness. The uniform earth has p =100 Q mand hemispherical electrode of radius 1 m; I
= 6.28 A.
Upper layer
earth is useless. If there are two or more electrodes conducting no total

current to earth, however, every point in the field has a finite potential with
respect to remote earth. The simplest and most practical such case is that of
two electrodes of equal size conducting equal but opposite currents. The
potential of a point distant v from an electrode conducting current I
and
distant r2 from another electrode conducting current lis
An electrode long enough to reach all the way through the upper layer, as
assumed above, is not generally practical. A more realistic electrode is one
at the surface of the earth, such as the hemispherical one assumed earlier in
this chapter; but such a one presents more mathematical difficulty.
One approach to the problem would be, as in the case of a buried electrode,
page 399, to seek a configuration of fictitious electrodes in a homogeneous
earth that would satisfy the boundary conditions. For the two-layer earth
with nonconducting lower layer, the boundary conditions are that no current
should cross either of two parallel, horizontal surfacesthe surface of the
earth and the surface separating the two layers. This condition is satisfied, as
Fig. 13. Fictitious array of electrodes in an infinite uniform medium that produces the
same field of current in the upper layer as the actual electrode (shown in black) when that
layer is bounded by insulating media.
these are equally spaced by 2h on a vertical line through the center of the
actual electrode, which is expanded from a hemisphere to a sphere to preserve
the symmetry. Each electrode conducts current 21 into the fictitious earth, /
being the current of the actual hemispherical electrode.
The potential at any point in the upper layer can be found as the algebraic
sum of components of potential due to the separate electrodes. Thus, at
410
9-2
GROUND RETURN
411
point P on the surface of the real earth at distance x from the center of the
real electrode, the potential is
... + >-.
r.emi+>+l
2n \x rt r2 +
r
=*v{l + 2xl})
where
V0lts
(34)
The electric field intensity at P can be found either as -dVjdx or by taking

the vector sum of the components of intensity due to each of the electrodes.
The sum may be formed by considering pairs of electrodes equidistant from
the surface of the earth. The resultant field of a pair is horizontal and is two
times the horizontal projection of the inclined component due to one member
of the pair. The nth pair contributes
cos
i
nrt
/E
(33)
+ (2nh)2y12
r = x2
Upper
layer
nx2
2IK
2IK
The vector sum is

$
p ,I(
00
+ 2x3
nx2I\ *
n= 1
volts per metre
z I
rnJ
(35)
The series in Eq. (35) converges but that in Eq. (33) does not.
The electric field intensity relative to that which would be produced in a
uniform earth of resistivity pt is
$
co
=1
+ 2x3
co
1+2
pT(wf
<36>
The equation is now in terms of dimensionless variables

as a function
of xjh. As shown in Eq. (25), <$j<u =pjpu where pa is the apparent resistivity.
An alternative concept to that of an infinite array of electrodes is that of
reflections at the bounding planes. A ray of current is assumed to act like a
ray of light, being reflected at a surface, with angle of reflection equal to angle
of incidence. Equivalence of the two concepts is shown in Figure 14. The total
distance of travel rn as well as the angle of arrival 0 is the same by both
concepts. For example, the path OAP with one reflection is equal in length
and in direction of arrival 0X to BAP; likewise, path OCDEP with three
reflections, two of which are from the lower surface, is equal in length and in
direction of arrival to FEP.
We are now ready to generalize with regard to the ratio of resistivities
PilPv We have considered the case of p2{px = co. The other extreme special
Fig. 14. Equivalence of current rays arriving at P from fictitious electrodes B and F
or from reflections of rays from real electrode O.
case is that of p2fpt = 0, or p2 = 0, that is, the lower layer a perfect conductor.
The pattern of current flow is similar to that sketched in Figure 15. The lines
are normal to the lower boundary. This pattern can be produced multiply by
reversing the current in alternate electrodes in Figure 13 or 14, that is,
by making K= - 1 instead of + 1 in Figure 14. It can be produced also by
negative reflections at the lower boundary of the upper layer. The arrowheads
on rays AP, CD, and DE would be reversed, so that the vector sums OA + PA,
OC + DC, and ED + EP become vertical instead of horizontal.
2/0
Fig. 15. Part of a fictitious array of electrodes similar to that of Figure 13 except that
the injected currents alternate in sign. The lines of current flow are normal to the boundary
plane midway between electrodes.
412
GROUND RETURN
Another special case is p2/Pi = 1 or Pi =Pi- this is the case of uniform

earth. The fictitious electrodes B and fin Figure 14 become unnecessary, so
that we can put K = 0. Or the reflected rays disappear, and the only ray
reaching P is the direct ray OP.
The three special cases are all covered by putting a factor K" in the
numerator of the sums in Eqs. (33), (35), (36). Thus Eq. (36) becomes
co
&
=1
+ 2x3
Zt
co
jsn
=1
+2
100
jfn
[;1 + (2nhjx)1]3l2
(37)
10
where AT has the value +1, 1, or 0 according as p2 =co, 0, or px. Other

values ofp2 are also covered if we define the reflection coefficient K by
K=~
(38)
P2 + P1
It obviously fits the three special cases already considered. Proof of the
correctness of K in the general case rests on the assumption of a transmitted
ray AG as an extension of incident ray OA in the same direction, and on
equating the sum of the normal (vertical) components of incident and
reflected rays of current density in medium 1 to the normal component of
transmitted current density in medium 2, and similarly equating the sum of
the tangential (horizontal) components of incident and reflected electric field
intensity in medium 1 to the tangential component of electric field intensity in
medium 2, and observing the vector relationships $x pxJx and
= p2J2
Note that K in Eq. (38) has the same form as the reflection coefficient of
electromagnetic waves at the end of a transmission line.
Results of calculation of $u as a function of hjx in log-log coordinates are
shown in Figure 16. There is observed a lack of symmetry between the upper
and lower halves of the figure. The upper half has an envelope of unit slope
for p2 = 00, as in Figure 12b; the lower half has a curved envelope of everincreasing slope.
Asymptotes. Intuitively, the apparent resistivity at short distances from the
electrode must be pls the resistivity of the upper layer; at long distances, it
must be p2 , the resistivity of the lower layer. Let these intuitive deductions be
proved mathematically. In Eq. (37), as x 0, SjSu = pjpx -* 1, or pa~*PiFor x ;> h, r -> x, and
~\+2
But
(39)
Yjk" = k+k2 + k3 + --- = t
1~ K
K2<1
(40)
"
2
.e
c;
*1
> c
'(O
to
0.5
\v
E
. g
0.2
"D
Q>
0.1
\v_
0.1
0.05
Q>
II
0.02
0.01
0.01
vv
0.005
0.002
0.001
0.1
10
x _ distance from electrode
ft "thickness of upper layer
0.001
100
Fig. 16. Electric field intensity at the surface of a two-layer earth as a function of the
,
,
distance from the electrode.
414
9-2
GROUND RETURN
From the definition of KEq. (38),

K
1- K
1(p2
2\pl
(41)
415
terized with good accuracy by two parameters representing transverse resistance per unit area : Tx = pxhx and T2 =p2h2 ohm-metres2. The dimensionless ratio T2jTx is used as a parameter in Figure 17, which shows the ratio
and hence
1 + 2 y Kn = 1 + - - 1 = Pi
Pi
<$u
Hence
Pi
Pa-Pi
Pi
(42)
(43)
(44)
Figure 16 shows that, with increasing distance from the electrode, the
apparent resistivity changes smoothly from the resistivity of the upper layer
to that of the lower layer. This transition occurs in a shorter distance (x = h
to 10/z) if the lower layer is more conductive than the upper layer than it does
if the contrary is true.
10
Three-layer Earth 66,67

In many places geologists have measured relative electric field intensity (or
apparent resistivity) as a function of distance from a current electrode and
have found that the results are reasonably in accord with those calculated
from a theoretical three-layer earth in which the middle layer has a much
higher resistivity than either of the other layers.

The reason for the agreement is that the crust of the earth generally has
three layers of rock, described as follows :
1. The uppermost or surface layer, several kilometres thick, consists of

soil and cool rock of many kinds inwhich the cracks or pores contain moisture.
The conductivity of soil and rock near the surface of the earth is attributable
almost entirely to moisture in which are dissolved various chemicals, so that
the moisture acts as a weak electrolyte. The resistivity of this layer is low or
moderate (10 to 104 l m).
2. The middle layer, 10 to 30 km thick, consists of rock under such high
pressure that the pores and cracks are closed. The resistivity is very high
(104 to 108 H m).
3. The lowermost layer has no moisture, but the rocks are highly con
ductive because of their high temperature. This layer, extending to the center
of the earth, may be assumed infinitely deep, for the purpose of calculation.
Its resistivity is low (1 to 100 m).
Although the assumed three-layer structure is described exactly by three

resistivities (pu p2, p2) and two finite thicknesses (hL, h2), it may be charac-
Fig. 17. Electric field in three-layer earth with highly resistant second layer:
and T2 = p22- (Adapted from Ref. 34 by permission.)
Tx = pjii
of electric field intensity on the surface of a three-layer earth to that in a

homogeneous earth of resistivity px as a function of distance from the elec
trode expressed in units equal to the thickness of the uppermost layer.
These curves show at small and moderate distances a characteristic similar
to that of a two-layer earth in which the lower layer is more resistive than the
upper layer. At greater distances the curves plunge abruptly, as do those for
a two-layer earth, with the lower layer more conductive than the upper layer.
The distance within which half of the current has leaked from the middle
layer to the lower layer is66
It lies near the maximum ordinate.

Figure 18 shows measured electric field intensities in several parts of the
United States expressed as a ratio to the value that would exist at the surface
416
9-3 THE CURRENT FIELD BETWEEN THE ELECTRODES
GROUND RETURN
electric field intensity at the surface caused by a given current into an electrode
at a given location. Geophysicists, however, use surveys of apparent re
sistivity made with varying spacing between current electrodes as part of
the evidence for deducing the geological structure.
The principal conclusion that we can draw from their work and apply to
ground-return dc circuits is that at distances up to several hundred kilometres
from the electrode the electric field gradient at the surface is greater than those
in a uniform earth of resistivity equal to that of the upper layer. At greater
distances the contrary is true, but at such distances the field intensities
attributable to the ground return current are already negligible.
New England
Adirondacks
X)
10
03
TD
C
CO
2
Q)
>
Columbia River Plateau
irt
9-3 THE CURRENT FIELD BETWEEN THE ELECTRODES
0)
1
.2
417
Southern Nevada
0.1
Central California
t3
0)
LU
0.01
0.01
0.1
1
10
Distance from electrode, km
100
1000
Fig. 18. Comparison of electric field measurements made in various parts of the United
States. Those in the Columbia River plateau, southern Nevada, and central California
were made in investigations for the Pacific Interties, 1963 to 1965. The rest were made by
the U.S. Geological Survey, 1961 to 1963. (Adapted from Ref. 66 by permission.)
of a standard earth having a uniform resistivity of 150 Q m. (This value is

the median of 10,000 measurements of the resistivity of superficial rocks in
the United States.67) Most of the curves in this figure exhibit a hump at a
distance of 10 to 100 km from the electrode and drop rapidly at distances
between 100 and 1000 km. The curve for the Columbia River plateau in
eastern Washington and Oregon is exceptional, being almost flat out to
100 km. The surface rocks there are volcanic and have an average resistivity
close to that of the standard earth.
Measurements made in Sweden19 with current electrodes separated by
distances up to 1100 km were consistent with a three-layer earth described as
follows :
Upper layer, 1 km thick, pt 4000 m

Middle layer, 30 km thick, p2 = 14,000 fi m
Lower layer, infinitely thick, p3 = 800 Q m
The middle layer consists mainly of granite.

This section has been written from the standpoint that, knowing the geo
logical structure of the earth in a certain region, one wishes to compute the
A distinction must be made between the path in the ground followed by

alternating current and that followed by direct current between widely
separated electrodes.
Alternating current returning in the ground follows the route of the
metallic conductors whether they are overhead or underground. Alternating
ground current does not take short cuts. It spreads out both downward and
sideward through distances depending on the frequency of the current and on
the resistivity and magnetic permeability of the ground. At audio and power
frequencies, these distances are of the order of several hundred metres. More
exactly, the skin depth is
S=
metres
(46)
WAV
where f frequency of current, Hz
p = ground resistivity, Q m
p. ground permeability, H/m
For nonmagnetic earth it becomes
5 = 503
metres
(47)
At these frequencies, the resistance of the ground-return path is independent

of resistivity and is directly proportional to frequency; it is
= 2.96 x
106/
ohms per metre
(48)
For example, at 300 Hz (the fifth harmonic of 60 Hz) and with p = 150
a- m, 5 355 m and RJl = 0.888 Q/km; at 3000 Hz, 5 = 112 m and
Rjl= 8.88 Q/km.
418
9-4
GROUND RETURN
Current flow lines are parallel to one another and to the metallic conductors
except near the ends of the transmission circuit, where these lines diverge
from, or converge to, the electrodes. Here is an end effect, in a zone having
dimensions of 23 to 33, which results in an additional component of resistance
associated with each electrode.
As frequency is decreased, the ground currents spread farther below and to
the sides of the metallic conductors, and the end effects occupy a greater
length of the transmission line as well as zones beyond the ends of the line.
At zero frequency (direct current), nothing is left but end effects. The re
sistance of the ground-return is merely the sum of the resistances associated
with each electrode. The current flow lines become independent of the route
of the metallic conductors. The current leaves the electrode in all directions
without respect to the direction of the metallic conductors.
Both alternating and direct currents follow ground paths offering the least
impedance. For direct current, the impedance consists only of resistance, and
the paths are determined by the relative resistivities of different parts of the
earth. Ac impedance has both resistive and reactive components, and the
THE NATURAL CURRENT FIELD IN THE
EARff
419 ,/
higher the frequency, the more dominant the inductive reactance becomes.
Alternating current tends to follow paths of low reactance, which are paths
near the metallic conductors. The crowding of the ground current into a small
cross-sectional area, however, increases the resistance of the circuit.
In dc transmission we are concerned with both direct ground current and
harmonic current, the former widely dispersed and the latter following
closely the route of the transmission line, which, fortunately, decreases the
harmonic voltages induced in other circuits.
If the two electrodes of a ground-return dc line are far apart compared
with the size of the electrodes themselves, the electric field intensity anywhere
except at points nearly equidistant from the two electrodes is very nearly
equal to the intensity due to the nearer electrode. At equidistant points, the
intensity is twice that due to one electrode. Figure 19 shows, in unconven
tional log-log coordinates, the field intensity at the surface of the earth along
a straight line from one electrode to the other, 2 km away. One curve is for
a homogeneous earth; the other, for a thin conducting layer.
9-4 THE NATURAL CURRENT FIELD IN THE EARTH66 67
One method of assessing possible adverse effects of ground-return currents

from dc transmission is to compare the voltage gradients at or near the sur
face of the earth due to the transmission scheme, at various distances from
a ground electrode, with the voltage gradients due to natural ground currents
at the same locations. If the gradients due to dc transmission do not exceed
the gradients due to natural currents, it can be assumed that the adverse
effects of the dc transmission are no greater than the effects of natural
currents, which, presumably, the affected structures or services are already
withstanding or must withstand in the absence of dc transmission.
>
c-
A'
CJ
Cause
O)
LlJ
Currents called telluric currents are induced in the earth by variations of the
,-3
>"4
1
10
100
100
1000
Distance from nearer electrode, m
10
Fig. 19. Electric field intensity along straight line on surface of earth between the two
electrodes 2 km apart. Curve A is for uniform earth, p = 100 Q m. Curve B is for con
ductive layer 10 m thick of pL = 100 H m, below which p2 = <x>. Hemispherical electrodes
of 1-m diameter: 1= 6.28 A,
magnetic field of the earth. The variations of the magnetic field are believed to
be caused by fluctuations inthe stream of charged particles emitted by the sun.
Some particles have positive charge ;others, negative charge. As these particles.
approach the earth, the earth's magnetic field deflects them eastward or west
ward, depending on the sign of the charge. They form a ring of current around
the earth in the ecliptic plane. If this ring current were constant in magnitude,
it would contribute only a static component to the earth's magnetic field, the
major part of which is believed to be caused by currents inside the earth.
However, the density of the stream of charged particles intercepted by the
420
9-4
GROUND RETURN
earth, as it travels in its orbit varies with time. Therefore, the ring current and
the component of magnetic field due to it vary with time. Moreover, the meet
ing of the two streams of particles on the back side of the earth sets up turbu
lence which leads to additional oscillations in the mmf.
THE NATURAL CURRENT FIELD IN THE EARTH
p(il m)
1000 100 10
1
1000
100
Frequencies and Amplitudes
~NS\\k
\\
Frequency
(mHz)
10 to 40
60 to 100
120 to 420
25 to 100
10 to 17
2.4 to 8.3
100 -
Some of the periods are long enough to justify the assumption that the effects
of the natural currents are not unlike the effects of ground-return currents
from dc transmission.
These oscillations undergo amplitude modulation at several different
\\
10 -
100
10 -
10
\"
\ \
Wbvv
100 ...
Period
(sec)
\\ \\
The oscillations of the earth's magnetic field, called micropulsations, have

periods and frequencies lying principally in the following three ranges:
421
10
-I -
periods :
10
0.1
100
Percent of time exceeded
Fractions of an hour.
Daily. The amplitude at midday is 10 to 20 times that at night.
28 days (period of rotation of the sun), amplitude ratio from 2 : 1 to 20 : 1.
11 yr (sunspot cycle).
The peak-to-peak amplitudes range from 0.5 to 3000 gammas* (0.4 to

2400 mA/m) or more. The larger amplitudes are associated with the so-called
magnetic storms, which are associated with heightened sunspot activity and
with strong auroral displays. The average amplitude also increases with
latitude somewhat as shown in Figure 20, which is necessarily based on
rather incomplete information.
The fluctuating electric field intensity induced at the surface of the earth by
the fluctuating magnetic field intensity His66
$
H\J5pf
microvolts per metre
where H = amplitude of fluctuation of magnetic field, gammas

p = earth resistivity, Q m
/ = frequency of fluctuation, Hz,
* One gamma = 10~s oersted 0.8 mA/m is the unit used by geopbysicists.
(49)
Fig. 20. Graph for estimating the probability that geomagnetic micropulsations of mag
netic field intensity Hwill exceed specified levels, together with scales for converting these
to the amplitudes of electric field intensity 3 induced in earth of resistivity 10, 100, or
1000 H m. (Adapted from Ref. 66 by permission.)
and, of course, the corresponding current density is

&
J= ~
P
/5/Y/2
W
H\ )
microamperes per square metre
(50)
Thus, increased resistivity increases $ in proportion to \jp and decreases J

by the same factor for a given H. The micropulsations of magnetic field H,
although varying widely with time and place, are relatively unaffected by
local variations in resistivity. Such variations affect and J substantially in
the same manner for ground-return currents as for natural currents. Figure
20 has additionalscales for electric field intensity in materials of resistivities 10,
100, and 1000 2 m for micropulsations of period 38 sec, which may be taken
as the typical period.
, H, and J in Eqs. (49) and (50) may be taken as peak-to-peak or as
effective values, provided that the same kind of value is used for all.
422
GROUND RETURN
9-5 COMPATIBILITY WITH OTHER SERVICES
Comparison with Currents from Ground-return Transmission

Keller66,67 has estimated, for three electrode locations in the western
United States, the distances from the electrode beyond which the electric
field intensity developed by micropulsations of the earth's magnetic field
exceeds the electric field intensity due to 1.8 kA of ground current at the
electrode for various percentages of time. These results are shown in Table 3.
Table 3. Distance from Electrode at Which Peak-to-peak Amplitude of Electric
Field Pulsation due to Natural Ground Current Equals That Due to 1.8 kA From
Electrode
Percent of
Time during
which Natural
Field Exceeds
Field from
Ground Return
50
10
1
0.1
Central Valley
of California
Southern
Nevada
Oregon)
(Tracy)
(Hoover Dam)
(km)
(mi)
(km)
(mi)
(km)
(mi)
90
29
9.5
56
18
6
66
17
4.5
1.2
41
120
52
13
75
32
2.0
than 0. 1% of the time, the same separation of 2 mi would be adequate even if

the ground return were to be used 100% of the time.
Since the comparison of dc ground-return transmission with natural
ground currents is only an approximate criterion, and usually a pessimistic
one, of the required separation between electrode and other facilities, we
must examine the question of compatibility between ground-return trans
mission and other facilities in more detail. The only thing that has been
shown with certainty by this comparison is that it is unnecessary to consider
any effects at distances over 100 km from a ground electrode carrying
1.8 kA.
Columbia
River plateau
(The Dalles,
3.2
423
11
2.8
0.7
2.9
All utilities and structures on which ground-return current from dc trans

mission might possibly have an adverse effect are now operating, or must be
designed to operate, in the presence of natural ground currents. Hence, if the
levels of electric field intensity in the ground caused by dc transmission do not
exceed those already existing, the effect of ground-return dc transmission on
these utilities and structures may be safely assumed to be negligible.
Those effects, such as electrolytic corrosion, which are cumulative, should
be compared according to the average period of use of ground return; for
example, if ground return is to be used only during emergencies, and if such
emergency conditions do not exist more than 9 h/yr (on the average), the
effect of ground-return current is comparable with that of natural ground
currents which exist 1.0% of the time. From Table 3, the corrosive effect of
1.8 kA 9 h/yr from an electrode near The Dalles, Oregon, for example, would
exceed that of natural ground currents only within a radius of 2 mi (3.2 km).
Other effects, which have a definite threshold, should be judged by the
maximum ground-return current regardless of the period of use. For example,
dc ground-return telegraph circuits are known to be rendered unusable during
infrequent, unusually severe magnetic storms. Since such storms occur less
General
Ground-return dc lines have the potentiality of adversely affecting the

services of plants of other public utilities, such as ac power systems, telephone
systems, local water and gas distribution systems, long-distance oil and gas
pipelines, railroads, and maritime shipping. Considerable apprehension has
been expressed by spokesmen for the utilities that might be affected. Although
it is only natural that they should be concerned, tests and investigations have
shown that all the detrimental effects of ground current can be either elimina
ted or made negligible at moderate additional cost by proper engineering.
The most important measure to take is to locate the ground electrodes at
sufficient distances from thickly settled areas and from pipelines, buried
cables, seaports, etc. Second in importance is the use of cathodic protection,
as needed, on buried metal objects closer to the-electrode.
The effect of zero-sequence dc harmonics in producing noise in voicefrequency telephone circuits was discussed in Chapter 8. The other problems
are treated in what follows.
The general policy for preventing or correcting such troubles should be that
already advocated in Chapter 8, to wit: (1) close cooperation between the
engineers of the involved utilities in making the necessary technical studies
and tests and (2) changing the plant of one utility or the other or both as may
be required for furnishing satisfactory service of both types at the lowest total
cost.
Electrolytic Corrosion
The most serious problem that may be aggravated by direct ground current
is that of electrolytic corrosion of buried or immersed metallic structures,
such as underground and submarine power and communication cables; gas,
424
GROUND RETURN
water, and oil pipes, ships' hulls, docks, tower footings, and other foundations. The phrase " may be aggravated " was carefully chosen. There would
still be a serious problem of electrolytic corrosion of buried orimmersed metal
lic structures even if no dc lines were built. First, there are natural ground
currents, already discussed in Section 9-4. Second, even if there were none,
there would still be local ground currents caused by emfs of contact between
different metals, between local spots of different amounts of impurities in the
metal, and between spots in contact with ground moisture (electrolyte) containing different ions or the same ions in different concentrations. One
expects a piece of unprotected iron buried in the ground to rust, at first
superficially and eventually, completely.
A superposed uniform current field may aggravate corrosion, but under
other circumstances it may prevent corrosion. The different lies in the direc
tion of the current.
Metallic ions are positive and move through an electrolyte in the direction
of the electric current. Where current leaves a metal object, such as an iron
pipe, metal ions go into the electrolyte, and some of them combine with neg
ative ions; for example, a typical reaction between an iron anode and soil is:
Fe+ +
+ 20H
-> Fe(OH)2
(51)
The formation of the Fe++ ion releases two electrons that may traverse the
pipe to some other point where current enters it. At this point the electrons
combine with positive ions; for example,
2e- + 2H+
->
H2
(52)
This reaction coats the pipe with a layer of hydrogen that protects it from
oxidation and other corrosion. Metal from which current leaves to enter an
electrolyte is an anode. Metal that current enters on leaving an electrolyte is a
cathode. Both anodes and cathodes are electrodes. Metallic anodes are
corroded; most cathodes are protected from corrosion. Corrosion does not
occur where current passes directly from one metal to another metal or solid
conductor, such as carbon, with which it is in good contact, for here conduc
tion takes place by electrons, not by ions.
Different parts of the same metal object, such as the iron pipe considered
above, may be anode and cathode; and the same part may be sometimes
anode and at other times cathode because of reversal of the current. Metal
removed by leaving current is not redeposited by entering current unless the
reversals of current occur rapidly. Alternating currents of commercial power
frequency cause only about 1% as much corrosion as does direct current of
equal rms value.
The basic law of electrolysis is Faraday's : the mass of material that leaves
the anode and goes into solution in the electrolyte is proportional to the
'
quantity of electricity or charge. The constant of proportionality is called the

electrochemical equivalent of the material:
m = Kecq
425
Kec\i dt
(53)
where m = mass, kg, of metal removed in time t sec

q = charge, C, issuing from metal in same time
i- electric current, A
Kec = electrochemical equivalent of anode material, kg/C
Since the charge carried by each metallic ion is proportional to the valence of
the element in the particular electrochemical reaction and since the. mass of
the ion is proportional to its atomic weight, the electrochemical equivalent is
Kec =
where
YYl
(54)
kilograms per coulomb
vKF
ma = atomic mass or "weight" (oxygen = 16)

v = valence
KF = Faraday's constant = 9.65 x 107 C/kg-molecules

For example, iron has an atomic weight of 55.85 and, in the usual reaction
that forms rust, has a valence of 2; hence its electrochemical equivalent is
55.85/(2 x 9.65 x 107) = 2.89 x 10 ~7 kg/C. The electrochemical equivalents
of iron and of several other common metals used as electrodes or in buried
objects are given in Table 4.
The values of Kec shown in Table 4 are theoretical. The loss of weight of
a buried metal object due to electrolytic corrosion is often less and sometimes
Table 4.
Electrochemical Equivalents and Electromotive Forces of Common Metals
Element and
Its Symbol
Usual
Atomic
Valence Weight
24.32
26.97
65.38
55.85
207.21
EMF*
(V)
0.126
0.454
0.0932
0.335
-0.762
0.339
-0.440
-0.126
0.289
1.22
1.04
3.87
0.0374
1.18
2.37
Lead (Pb)
Hydrogen (H)
Copper (Cu)
Copper (Cu)
63.57
+0.345
63.57
+0.522
1.008
1.67
(g/A h) (kg/A-yr)
(mg/C)
-2.34
Magnesium (Mg)
Aluminum (Al)
Zinc (Zn)
Iron (Fe)
3
2
Electrochemical Equivalent Kec
1.074
0.0104
0.329
0.659
* The American Chemical Society uses the opposite sign convention.
3.97
2.94
10.66
9.13
33.87
0.328
10.39
20.77
426
9-5
GROUND RETURN
more than the theoretical value. The ratio of actual loss to theoretical loss is
known as corrosion efficiency,15 anode efficiency, or coefficient of deteriora
tion.31 Tests have shown that, for ferrous materials buried in soils, the corro
sion efficiency at high current densities (0.5 to 50 A/m2) varies from 0.2 to
1.4; at moderate current densities and average soil moisture, from 0.5 to 1.1;
at low current densities (say, 1 to 10 mA/m2), the loss of mass closely approxi
mates the normal corrosion due to local galvanic action added to the theoreti
cal value due to stray current. The recommended practice in the calculation of
loss of anodic metal is to assume corrosion efficiency of 1.0.
Thus the rate of loss of anodic metal can be calculated, at least approxi
mately, from the current, time duration of its flow, and the electrochemical
equivalent. If the current density were uniform over a given part of the
anodic surface, the rate of decrease of thickness at that surface could also be
calculated. If, however, the current density were nonuniform, the corrosion
would likewise be nonuniform, and only the average decrease of thickness
could be readily calculated.
An iron cathode does not waste away, as does the anode, but a lead cathode
may deteriorate in certain soils if it has too great a negative potential with
respect to the soil.
If two metals are in contact with each other, only the one higher in the
electrochemical series corrodes and protects the other; for example, leadsheathed cable with steel armor or run in steel ducts is protected from corro
sion, but the steel is corroded.
Ceramic and plastic pipes do not corrode electrolytically.
The permissible density of the current leaving a buried metal object depends
on the desired useful life of the object, the thickness of metal that can be lost
without impairing the serviceability of the object, and the electrochemical
equivalent and density of the metal. For buried pipes and cable sheaths the
permissible maximum current densities due to stray currents from groundreturn transmission electrodes is usually taken as 1 to 10 mA/m2. For thicker
objects the permissible current density is greater.
Example
A certain buried telephone cable has a lead sheath 137 mils (3.48 x 10~3m)
thick with an outer diameter of 2.54 in (6.45 x 10 ~"2 m). What is the per
missible leakage current density for loss of 10% of the thickness of the sheath
in 40 years?50
The following characteristics of lead are known:
Specific gravity = 11.34 g/cm3 (density = 1.134 x 104 kg/m3)

Electrochemical equivalent = 33.8 kg/A yr
COMPATIBILITY WITH OTHER SERVICES
427
SOLUTION
Permissible penetration in 40 yr is 0.1 x 3.48 x 10 ~3 = 3.48 x 10-4 m, or

3.48 x 10"4/40 8.70 x 10-6 m/yr.
Volume lost per yr per m2 of surface = 8.70 x 10 ~6 m3.
Mass lost per yr per m2 of surface = 8.70 x 10~6 x 1.134 x 104 = 9.90
_2
x 10 kg.
~3
Current density that accomplishes this = 9.90 x 10~2/33.8 = 2.9 x 10
A/m2, say, 3 mA/m2. (Answer )
Rusck40 states that an iron object subjected to direct current for 15 yr

loses 0.174 mm of iron by corrosion with a current density of 10 mA/m2.
From this he concludes that current densities in the range of 1 to 10 mA/m2
could be permitted. He states that in the U.S.S.R. a density of 15 mA/m2 is
permitted for armored cables, and 75 mA/m2 for mild-steel pipelines.
Materials. Most structures in the earth or sea are made of mild steel.
Pipelines are made of steel or cast iron. Sheaths of power cables are extruded
from pure lead, and those of telephone cables are of lead alloyed with a small
amount of antimony or calcium. Corrosion of these lead alloys may be
calculated as if they were pure lead, and ferrous metals as if they were iron.
Power and telephone cables are usually installed in ducts of fiber, tile, or
iron. The nonmetallic ducts give some insulation between lead sheath and
ground, and iron ducts help protect the lead from corrosion. Submarine
cables and cables buried directly in the ground are usually protected by
layers of asphalt-impregnated paper, then by jute, and by steel-tape armor.
The steel, by virtue of its position above lead in the electromotive series
(Table 4), protects the lead from corrosion as well as giving it mechanical
protection.
Protection. The principal means of protection of buried or immersed

metal objects from corrosion due to ground currents from dc transmission
lines are (a) adequate distance of separation between such objects and the
ground electrodes, (b) insulating coatings, and (c) cathodic protection.
Separation decreases the current densities and voltage gradients in the
ground. Insulating coatings make it impossible, or at least difficult, for
current to leave the protected object. Cathodic protection forces the current
to leave the protected object by a metallic conductor.
The shortcoming of insulated coatings is that, through damage in installa
tion or through subsequent deterioration, small breaks may occur. Leakage
currents are localized at these breaks, usually giving a higher local current
density and higher rate of corrosion than those which would occur on an
uncoated object. It is difficult to inspect or repair coatings on a buried object.
428
GROUND RETURN
9-5
429
Cathodic protection should therefore be used on coated objects to ensure

that leakage currents are into the object, not out of it.
On long pipes both insulating coatings and cathodic protection are used.
The coating greatly reduces the current that the cathodic protection must
supply, and by reducing IR drops in the pipe it increases the feasible spacing
between current drains. It is feasible, however, to protect more compact
objects by cathodic protection alone.
Consideration is now given to estimation of the current density leaving or
entering certain bare metal objects buried or immersed in a conducting
medium (soil or water) in which there is a known field of current in the
absence of the metal bodies.
Compact Buried Object. The surface current density of a metal object of
small dimensions in comparison with its distance from the main electrode
can be estimated by the use of a formula derived for a sphere having a much
lower resistivity than that of the surrounding medium. At first, the resistivity
of the sphere is assumed to be zero, making its surface equipotential. It may
be shown that the sphere, when placed in an otherwise uniform field, has a
maximum surface current density three times as great as that of the undis
turbed field. Now, on the assumption that the sphere has a finite, low resist
ivity, its internal current density is found to be uniform and parallel to the
density of the undisturbed field but three times as great. The current that
enters one hemisphere and leaves the other is, therefore, three times the current
of the undisturbed field that would pass through the area ina2) of the maxi
mum cross section of the sphere. The internal electric field intensity S2, of
of the undisturbed field, being
course, is much smaller than that
2= $0
(55)
Pi
where p2 is the resistivity of the sphere and px is that of the medium.
In estimating the maximum surface current density of nonspherical objects,
one must expect saliencies in the direction of the external field to have density
several times greater than the maximum surface current density of a sphere
at the same location.
Bare Pipe Crossing the Field of the Main Electrode. A bare metal pipe, or a
cable with bare metal sheath or armor, is assumed to run straight across the
field (Figure 21) and, therefore, to have a point P closest to the electrode Eand
at a perpendicular distance y from it. Any other point Q on the pipe distant x
from P is distant r = (x2 + y2)1/2 from E. It is assumed that the earth has
uniform resistivity p and that the field is not distorted by the presence of the
PipeThe latter assumption is strictly true only if the resistance per unit length
of the tubular pipe or sheath is equal to that of the cylindrical part of the dis-
-1.00
Fig. 21. Leakage-current density J on buried pipe relative to density Jo at a point on the
pipe nearest to current electrode and distant y from it as a function of distance x from
that point.
placed medium. It is approximately true, however, if the medium has a low

resistivity, as does seawater. For a medium of higher resistivity, such as soil,
this assumption is poorer, but since it leads to pessimistic results and gives a
simple calculation, it may be used as a first approximation.
The longitudinal electric current in the pipe at any point of its length is
proportional to the voltage gradient along the pipe at the same point; thus
dV
A-"*'
volt per metre
(56)
where ip = longitudinal current in pipe, left to right, A

0t = resistance per unit length of pipe, Cljm
V = potential of pipe at distance x from P
The leakage current causes a change in longitudinal current, thus :
*L
dir*p
dx
ampere per metre
(57)
where iL = leakage current leaving pipe per unit of its length

If this equation is substituted into the first derivative of Eq. (56) with respect
to x, the result is
d2V
dip
dx 2
dx
(53)
Tl
430
T5
9-5
GROUND RETURN
The potential V of the pipe is the same as that of the soil with which it is in
contact, and, for electrode current / amperes into the homogeneous earth,
this is
V-f
2nr
(59)
=r
where
k=Pi
d2V
, ,
dx
k
r
,,1
Radial Bare Pipe (Figure 22). This is a special case of the previous case,
with y 0. Termination of the pipe at r = r0 causes no theoretical difficulty,

and the practical implications are discussed later. The voltage gradient along
the pipe is now the radial gradient of the field, and its first derivative (second
derivative of the potential) is
volt per metre2
j.
d2V
dr 2
* is
2k
TT
(66)
whence the leakage current density is

(61)
pi
(67)
Jl = 7i2D0tr2
The leakage current density is
JL =
ampere per square metre
nD
431
Reversal of current in the main electrode, which then becomes a cathode,

gives worse corrosion of the pipe, especially at the point nearest the main
electrode.
(60>
The second derivative of V with respect to
(62)
with greatest value (not a mathematical maximum) at the end of the pipe near
the electrode, where r = r0 . This density is twice that in the previous case for
equal distances from the electrode to the nearest point of the pipe.
where D is the diameter and nD the circumference of the pipe. Substitution

of Eq. (58) for iL and of Eq. (61) for d2Vjdx2 gives
JL = 2 1
3)
3( - )
ampere per square metre
(63)
At point P where x = 0 and r = y, the leakage current density is

pi
Jm = Hemp
Pipe
ro
(64)
which is the maximum absolute value. The negative sign denotes current
entering the pipe.
Adoption of dimensionless ratios J' = JLIJL0 and x' = xjy leads to the
equation
2(*')2 ~ 1
[(x')2+l}5/2
(
}
and to the graph in Figure 21.

The pipe is cathodic out to points x = +0.107y and is anodic at greater
distances. Maximum current density leaving the pipe is J' = 0.202 at
xjy = 1.222.
Fig. 22. Buried pipe running radially from the electrode E.
According to the last equation, the current is leaving the pipe everywhere ;
and this appears impossible. If, in accordance with the assumptions, the pipe
were merely cut off or capped, the current would have to enter the butt end or
cap, not the cylindrical surface ; but in practice, although not in this rough
theory, current would enter the cylindrical surface in a short section of pipe
near the end. Perhaps a more practical case is one in which the pipe terminates
in a partly buried tank or makes a right-angled bend to go to a tank or
other equipment on the surface of the earth. Then most of the current would
enter the surface of the tank or of the vertical run of pipe.
432
9-5
GROUND RETURN
Example 2
Effect of Distortion of Field Caused by Pipe. Rusck40 has shown that the
expression for leakage-current density with field distortion neglected
Eq. (64) should be multiplied by the dimensionless correction factor,
plotted in Figure 23 and calculated as
my) =
(/W2{i - (fyylSotfy) iV0(ftO]}
433
COMPATIBILITY WITH OTHER SERVICBS
What is the minimum permissible distance between the telephone cable

described in Example 1 and a ground cathode conducting 2000 A if the cable
is buried in a straight line in soil where the apparent resistivity is 180 Q m?
The longitudinal resistance of the cable sheath is 0.36 ft/km.
(68)
SOLUTION .
where
From Eq. (64), the minimum permissible distance is
JIn (0.7y/
\p
v Dh)J
metre-1
(69)
ymin =
S0 = Struve's function7
N0 = Bessel function of second kind for real arguments7
The graph shows that the approximate formula gives good results for large
separations between the electrode and the pipe and for low resistivity of the
medium but gives much too high a value of leakage current density for small
separations and high resistivity.
1
|
|
{ii?Dmz)
(a)
where p = 180 fi m
/ = 2.0 x 103 A
D = 6.45 x 10 ~2 m
m = 3.6 x 10-4 n/m
JL0= 2.9xl0-3A/m2
Substitution of these numerical values gives

ymin = 6470 m = 6.5 km = 4.0
mi
This result is based on the approximate theory, which neglects the distortion
of the electric field caused by the presence of the cable. It rests also on the
assumption that the dc line operates at full load with ground return for the
whole 40 yr.
Example
Examine the accuracy of the approximate solution in Example 2. Assume

the depth of burial to be 1 m.
SOLUTION
ByEq. (69),
0.01
0.1
(7i
180 In
10
ftFig. 23. Correction factor for density of leakage current of bare buried pipe. See Eqs.
(68) and (69). (Adapted from Ref. 40 by permission.)
x 3.6 x 10~4
\1/2
0.7 x 6.47 x 103 | x 6.47 x
= (8.02 x
a/6.45
lO"2 x
103
1/
10"4)(6.47 x 103) = 5.2
(b)
From the graph in Figure 23, at fiy = 5.2, read F(fiy) = 0.85. The corrected
434
9-5
GROUND RETURN
value of the surface current density is 0.85 x 2.9 = 2.5 mA/m2. For 2.9
raA/m2, the closest distance could be reduced to (0.85)1/3 6.47 = 0.95 x 6.47
= 6.15 km. The error is only 5%, and the accuracy must be considered very
good in view of the uncertainty of other factors.
Example
If, in Examples 1 and 2, the ground return is used only one-thousandth of

the time, on the average, in the 40-yr period, how much closer may be cable
to the electrode ?
SOLUTION
The surface current density may be a thousand times as great as before, or

2.9 A/m2. On the assumption that the anode efficiency is unchanged, applica
tion of the approximate method shows that the closest distance may be
decreased to (1000)-1/3 = 0.1 of the previous value, becoming now 0.65 km.
For this new distance, fiy = 0.59, and F(fiy) = 0.13; that is, the actual surface
current density is 0.13 x 2.9 = 0.38 A/m2. The distance that would give
2.9 A/m2 can be found by successive trials. The next approximation to the
distance would be (0.13),/3 0.65 = 0.33 km. We let the matter rest here. It is
obvious that the approximate method becomes inaccurate at such short
distances.
Cathodic
Protection12'15'23'44
In this method of protecting a buried metal object from electrolytic

corrosion, the protected object is kept at a negative potential with respect
to the surrounding soil (or water) so that current enters the object from the
soil. In order to accomplish this, current must be drained from the object
by a metallic conductor, raised in potential by a source of emf, and delivered
to an anode from which it returns to the earth. In this method, the protected
object is a cathode. The anode or anodes are made of cheap material which
can be easily replaced and whose integrity is not so necessary as that of the
cathode for containing pressure, excluding moisture, etc.
Three alternative methods are used for raising the potential of the drained
charge :
1. In forced drainage, a dc source, such as a battery, dc generator, or
rectifier is used.
2. In galvanic-anode drainage, the anodes are made of zinc, aluminum, or
magnesium, which are negative with respect to iron or lead. The ground
moisture is the electrolyte of a battery having the protected object as cathode,
and no separate battery is needed. Inside any battery on discharge, the current
435
is raised in potential. Actually the battery emf is the difference of changes of

potential occurring at the contact surface between the electrodes and the
electrolyte.
3. In bus drainage, used principally to protect pipes near electric railways,
the pipe is bonded to the most negative point available, which may be the
ground bus at the rectifier substation or the negative terminal of a negative
feeder booster.
Forced drainage is the only method used for protecting long cables or
pipelines. The necessary emf is usually provided by rectifiers if ac power is
available. Wind-driven generators with floating storage batteries are used in
some areas where the wind blows almost constantly. Generators driven by
gas engines are used along some natural-gas pipelines. Rectifiers are mounted
on poles; the engine-driven generators, in cabins.
Three arrangements of anodes are shown in Figure 24. The single anode is
7A
(a) Single anode
<<
(b) Continuous anode
jj
(c) Multiple anode
|j
Fig. 24. Three arrangements of anodes for cathodic protection of pipe. Multiple anodes, if
far apart, would have separate dc supplies.
applicable only to compact objects. Multiple anodes or continuous anodes

are suitable to extended objects, such as cables or pipelines. Continuous
anodes are made of old rails or abandoned pipelines buried parallel to the
pipeline to be protected. Multiple anodes are generally more practical than
continuous ones. The spacing between elements of the anode can be increased
if their distance from the protected pipeline is increased. The greater the
distance, however, the higher the resistance between the anode system and the
protected pipe and the greater the danger of damage to other structures,
especially in congested areas. An attempt is made to place the anodes so as to
get a reasonably uniform leakage current entering the protected pipe per unit
of length. Sometimes the anodes are buried deeply, well below the pipeline,
principally for reducing the right-of-way requirements.
The anodes are usually made of scrap iron. Corrosion of this iron can be
greatly decreased by surrounding it by fine coke, carbon, or graphite. This is
discussed further in Section 9-7. Anodes of special corrosion-resisting alloy
are being used increasingly.
Coatings. Pipelines with anodic protection are usually coated with insulat
ing materials. The best coatings are made of polyvinyl chloride tapes. Other
|j
11
j
j\
j]
|:
jj
ij
|j
;
j
436
GROUND RETURN
coatings used are concrete, cement mortar, bituminous materials such as

asphalt, with or without asbestos felt, coal-tar enamel, and mastic. Coatings
greatly decrease the current leaking into the pipe and thus decrease the
current to be drained by the dc power sources. At the same time, the longi
tudinal IR drop in the pipe is decreased: hence the spacing between drainage
points (rectifiers) can be greatly increased while still keeping all intermediate
parts of the pipe at sufficiently negative potential. The effectiveness of coatings
depends on high resistivity, low permeability to water, good bond between
coating and metal, and freedom from holes and cracks. Most coatings
deteriorate with age but still have considerable value. The spacing between
rectifiers may be as low as 150 m on bare or poorly coated pipes and as great
as BO km (50 mi) on pipes with good thermoplastic coatings.
Typical values of leakage conductance of a 12-in (0.3-m) pipe are given in
Table 5. Assuming the average potential of the pipe with respect to the soil
437
the leakage conductance. Unfortunately, many pipelines with deteriorated

coatings are still in use; and their potentials to the soil must be kept in a
narrower band. Some corrosion engineers say 0.85 to 1.20 V. There is some
opinion, however, that the maximum acceptable voltage may vary with the
resistivity of the adjacent soil. For example, one authority46 has suggested
that a straight line on log-log coordinates from 0.85 V in a soil of 0.1 Q*m
to 8.5 V in soil of 104 O-m (see Figure 25) roughly separates the points
-10
Table 5. Typical Values of Leakage Conductance of Buried 12-inch Pipe15
Gil
Resistivity, fl-m
(mho/1000 ft) (mho/km)
Bare pipe in water (p = 1.4 Q m)

Bare pipe in soil of low resistivity (14 O m)
Bare pipe in soil of medium resistivity (140 O m)
Bare pipe in soil of high resistivity (1400 O m)
or pipe with normal coating
Pipe with very good coating
100
10
1
300
30
3
0.1
0.3
0.01
0.03
the leakage current in amperes per unit length is numerically

equal to the conductance in mhos per unit length.
With the passage of time, polarization effects reduce the leakage conduc
tance, especially on coated pipes.
to be 1.0 V,
Permissible Range of Potential Difference Between Coated Pipe and Soil.

The potential of an iron pipe with respect to the adjacent soil, for the avoid
ance of corrosion, should be negative by at least 0.85 V as measured with a
copper-copper sulfate half-cell. Good insulating coatings can stand greater
voltages, say, up to 10 V or more, with no adverse effect on pipe or coating.
Inferior coatings, however, that are perforated, are permeable to water, or
have high leakage conductance should not be subjected to such a high voltage.
Highvoltages tend to form hydrogen on the surface of the iron at holes in
the coating, thereby loosening the bond between the pipe and the coating.
Electroosmosis drives water into permeable coatings, thereby increasing
Fig. 25. Greatest negative potential of pipe to soil suggested by Parker46 :

log F= 0.2 log/5 + 0.13.
corresponding to serious corrosion from the points corresponding to no

corrosion. Further investigation of this matter is desirable.
The range of voltages permissible for a pipe with poor coating, 0.85 to
- 1.2 V or even to - 3 V, is narrow. Drainage points may have to be closely
spaced in order for cathodic protection to maintain voltages within this
range. If some point on the pipeline is within a few kilometres of a dc trans
mission electrode conducting current of 1000 A or more, the variation in.
soil voltage in the vicinity of this point may take the pipe-to-soil voltage out
of this range if the cathodic protection rectifiers supply substantially constant
voltage. For example, with 1= 1000 A, p = 150 Q m, r = 10 km,
V=
+ 103 x
150
= +2.4 V
In x 104
Under such circumstances one or more of the rectifiers may well be of the
kind having their voltages automatically controlled so as to maintain the
desired pipe-to-soil voltages.
Calculations. The variation, with distance from current drain point, of
difference in pipe-to-soil voltage due .to cathodic protection, and of longi
tudinal current and leakage current can be calculated as for a leaky dc line.
On a very long pipe with one drainage point, these quantities all decrease
438
9-5
GROUND RETURN
exponentially with distance. In a section between two drainage points, the

midpoint is similar to an open circuit inasmuch as the longitudinal current is
zero. The potential varies as cosh ax, and current as sinh ax, if x is measured
from the midpoint.
439
Insulated joints
Track rails
Rheostat
AC Transmission
Many three-phase transmission lines terminate at both ends at substations
where there are grounding transformers usually transformer banks with
Y-connected windings and grounded neutral points. Zero-sequence direct
currents may flow in such lines. The magnitude of such currents depends on
the difference of potential of the ground mats at which these transformers are
located. These currents tend to saturate the magnetic cores unsymmetrically.
As such currents increase, the first indication is increase of the audible hum
from the affected transformers. This increase of hum is noticeable when the
neutral current per phase exceeds the normal alternating exciting current by a
ratio of 1.2 to 1.5 and is pronounced at a ratio of 2. At higher currents the
transformers may overheat because of increase of iron loss both in the core
and in the paths of stray flux (core clamps, tank, lid, etc.). Such currents, if
they exceed the relay setting, can also cause false tripping by overcurrent
ground relays of the plunger type.
Such troubles may be prevented by locating the dc ground electrode at
sufficient distance both from the converter station itself and from other sub
stations. Adequate separation of the electrode from substations is advisable
for another reason also for reducing electrolytic corrosion of the substation
structures and ground mat.
Excessive direct currents, however, occur infrequently in lines and trans
formers as a result of unusually great natural ground currents, especially
during the so-called magnetic storms. (See also Section 9-4.) The lines most
susceptible to such troubles run easterly and westerly in high northern
latitudes.
Series capacitors in a transmission line would block stray direct currents
whether from natural sources or from dc ground-return transmission. There
is no record to date of series capacitors having been installed for this purpose.
In some instances the transformer neutrals can be grounded through
resistors of a few ohms for reducing the direct neutral current.59
Railway Signals19'20
70
A typical dc track circuit is shown in Figure 26. It is isolated from adjacent

track sections by insulatingjoints in the rails. A battery is connected from one
rail to the other at one end of the section, and a relay coil is connected rail to
Battery
Relay coil
1 to 6 V
Fig. 26. Direct-current track circuit.
rail at the other end. Contacts of the relay may control block signals, inter
locking devices at railway crossings, junctions, and terminals, or indicators in
the dispatcher's office for centralized traffic control. If no train is in the sec
tion, the relay is energized and, if used for a block signal, puts the signal in the
"clear" or " proceed " aspect. If, however, the section is occupied by a train
or even by a single car, the wheels and axles form a short circuit, deenergizing
the relay and making the block signal indicate " stop." Note that this system
fails safe: if the battery is dead or if a rail is broken, the signal indicates
" stop" even if the section of track is not occupied.
During some ground-current tests made in Sweden about 1952, false
operation or conditions that could give false operation of signals were ob
served at distances as great as 20 km (12 mi) from an electrode conducting
170 A, and calculations indicated that an electrode current of 1000 A might
cause false operation of signals at distances up to 60 km (37 mi).19 Naturally,
this caused alarm. The track circuits used in Sweden at that time, however,
had only one insulated rail, the other rail being continuous. The continuous
rail readily picks up stray currents, which by their IR drops introduce
unwanted voltages into the track circuits. A stray current of 2.5 A or more
might cause misoperation.
In Great Britain, the United States, and probably elsewhere both rails are
sectionalized by insulating joints, making the track circuit relatively immune
from stray currents. Not only does a sectionalized rail pick up much less
ground current than a continuous one, but also, under normal conditions,
the currents in the two rails, the resistances of the rails, and the IR drops in
the two rails are approximately equal; therefore, the net voltage drop around
the track circuit caused by stray current is very small. The balance can be
upset by inequality of the leakage resistances of the two rails to the ties and
ballast or, even more, by a broken rail. Then, if the stray current were great
enough and in the right direction, it could prevent a relay already energized
from dropping out and thus cause a signal to give an unsafe false indication
of "clear."
440
GROUND RETURN
9-5 COMPATIBILITY WITH OTHER
Tests made by B.P.A. in 1967 on an electrical model of an 8000-ft (2,44-km)

block of track with a 1.5-V battery and a sensitive (4-2) track relay, which
picked up at 75 mA and dropped out at 35 mA, showed that this could happen
only if all the following conditions were satisfied simultaneously:
1. Ground surface gradient of 82 mV/m in the direction of the track, or a
potential difference of 300 V across the section
2. Break in rail between 2600 and 4700 ft (0.79 and 1.44 km) from the
relay end of the section
3. Rail-to-ballast resistance of unbroken rail about forty times as high as
that of the broken rail
4. Positive (negative) terminal of track battery connected to the broken
rail if stray current enters (leaves) the block at the relay end
The implication of condition 1 is apparent from the fact that an electrode
conducting 1.8 kA to ground and having a resistance of 0.1 2 to remote
ground has a potential of only 180 V to remote ground. An unbalance of
rail-to-ballast resistance as great as that of condition 3 is very improbable.
Kendall and Auer70 reached a different conclusion, which prompted the
construction and testing of the model mentioned above.
Hayes53 states that a separation of several miles between electrode and
track provides suitable protection. The B.P.A. studies described above show
that a separation of 1 mi should be ample for currents up to 2 kA or more.
The effect on railway signals is surely one of the least important problems of
hv dc transmission with ground return.
About 75% of the track circuits in the United States use low-voltage direct
current. If a problem ever arose of the possibility of false operation of such
circuits in a few blocks close to electrodes conducting large direct currents, it
could be solved by using a higher battery voltage and a less sensitive track
relay. If more drastic measures were ever required, ac track circuits or coded
track circuits20 could be used.
SERVICES
441
where Hdc = horizontal component of magnetic field strength attributable to

direct current in cable
component of earth's field*
horizontal
=
He
A = angle that cable makes with direction of He
dc
h
I
a
2izr r
Idh
2nr2
Lh
2n(h2
ampere per metre
+ x 2)
(71)
where Id = direct current, A

h = height of compass above cable (Figure 27), m
x = horizontal distance of compass from cable, m
r = slant height of compass from cable, m
Magnetic north
Compass
Hdc cos X
(b)
S: (a) plan; (b) elevation through
Compass Error21,35,62
Fig. 27. Trigonometry for calculation of compass error

line PC.
The magnetic field of direct current in a submarine cable can deflect ships'
magnetic compasses from their normal direction. This effect varies with the
current in the cable, the depth of the cable below the surface, the azimuth of
the cable, the horizontal distance of the ship from a position directly over the
cable, and the strength of the horizontal component of the earth's natural
magnetic field. It is greatest when the cable runs parallel to the undisturbed
direction of the compass that is, approximately north and south and when
the ship is directly over the cable. The compass error is
The field due to currents in several cables can be found by superposition.

If the dc line is a metallic circuit in a two-conductor cable or in two singleconductor cables laid close to one another, the components of field due to the
positive and negative current very nearly cancel one another at the surface
of the sea. If the two single-conductor cables are laid far enough apart,
however, to make damage to both of them improbable from the same event
(trawling or dragging anchor), or even if they are only far enough apart
(150 m) to make it feasible to grapple one cable for repair without disturbing
&=
tan-1 cosA
(70)
approximately
* H varies with position, especially with latitude. Inthe temperate zones it is
0.2 Oe= 16 A/m.
442
GROUND RETURN
the other, the cancellation of their fields is almost negligible. Instead, the
compass of a ship sailing over the two cables in succession shows almost the
full deflection due to one cable, first in one direction and then in the other.
Ground currents are so widely dispersed that the effect of a ground-return
circuit on a compass is essentially that of the cable or cables alone.
The question of compass error was raised by the British Admiralty in con
nection with the Cross Channel scheme because of the dense ship traffic in
the Strait of Dover. Preliminary tests were made in the Solent.62 Two ships
were used in the tests, one having a nonmetallic hull and a compass near the
water level and the other having a steel hull and a compass well above the
water level. Each ship made eight runs across a submarine cable conducting
100 or 200 A in direction nearly magnetic north-south, with sea return.
Compass errors ranging from 3 to 11 were both calculated and measured
with good agreement (median difference 0.5 = 0.009 rad). Compass errors
measured on the steel-hulled ship were only 0.5 to 0.6 of those measured on
the ship with nonmetallic hull, the difference being attributable to the shield
ing effect of the steel hull and to the greater height of the compass above
level.
The error of a ship's compass directly over a cable of the Cross Channel
scheme at 46-m (25-fathom) depth and conducting 800 A was calculated
as 7, which was deemed intolerably great.35 Consequently ground-return
operation was not permitted, and the two cables were laid only 3 m apart so
that their fields would effectively cancel each other.
In the Gotland scheme, in which one submarine cable carries 200 A with
sea return, the maximum compass error was calculated as 0.3, and this value
was confirmed by a measurement made on the ice in the winter of 1955. The
small deflection results from the direction of the cable, which does not deviate
by more than 11 from east-west, and from the smaller current. The fact that
the cable sheath and armor carry about 15% of the return current is probably
of little importance as regards compass error.
No problem of compass error has been reported on three other schemes
using submarine cable and sea return: Konti-Skan, crossing the Kattegat,
with its dense ship traffic; the New Zealand scheme, crossing Cook Strait;
and the Sardinian scheme, crossing the Tyrhennian Sea.
Although the compass error may be considerable directly over a dc cable,
it decreases rapidly with horizontal distance and, thus, is significant only in a
narrow strip of water along the route of the cable. Ships habitually plying a
nearby course would be aware of the error, while most other ships in the
strait would be sailing almost perpendicularly to the cable and thus would be
subject to significant cable error for only a very short time. Besides, nowadays
most big ships use gyro compasses normally with magnetic compasses only
water
as backup.
9-6
DESIGN OF ELECTRODES GENERAL
443
For the reasons stated, the magnetic-compass error caused by a dc sub

marine cable, although deemed significant in one scheme, is not generally
regarded as a problem.
9-6 DESIGN OF ELECTRODESGENERAL
Because the design of land electrodes differs considerably from the design
of sea or shore electrodes, the two categories are discussed separately. First,
however, we discuss those considerations which are common to both
categories.
Requirements
The requirements for both kinds are the following :
1 Low resistance
2. Adequate current-carrying capacity
3. Low maintenance cost
4. Accessibility
5. Limitation of damage to other facilities
6. Safety for persons, livestock, and fish
Each of these requirements is discussed briefly.

Low Resistance. Low resistance is necessary for low power loss. It is
obtained by the use of electrodes of large surface area located in soil or water
of low resistivity. The resistance must remain low during the whole life of theelectrode.
Adequate Current-Carrying Capacity. An adequate current-carrying capac
ity also requires a large surface area and low current density for avoiding
overheating and drying of adjacent soil or boiling of adjacent water. Thus
low resistance can be maintained during long periods of operation at full
rated current.
Low Maintenance Cost. An important item of maintenance cost could be

the replacement of electrodes decomposed by electrolysis. Since metal ions
are carried in the direction of the electric current, metal anodes are consumed
while cathodes are protected. Anodes are preferably made of cheap nonmetallic material, such as coke. In schemes in which the current is never
reversed, cathodes may be simpler and cheaper than anodes. In monopolar
schemes, however, in which power can be reversed by the reversal of current,
444
9-7
GROUND RETURN
and in bipolar schemes in which either pole may be temporarily out of service,
the electrode at each end can function sometimes as anode and at other times
as cathode; therefore, both electrodes must be designed as anodes.
Accessibility. Electrodes must be accessible for inspection and main
tenance. Most electrodes are made of several separate units, arranged so that
the current to each unit can be measured and so that any unit can be discon
nected and, when necessary, replaced. Some shore electrodes have their
elements suspended in tubular wells. The electrode elements can be pulled up
for inspection and replacement.
Location. Electrodes should be located far enough away from other buried
or immersed metal structures so that electrolytic corrosion attributable to the
dc line is negligible in comparison with other causes of corrosion. In particu
lar, the electrodes should not be close to the terminal substation of the dc
line itself because of the hazards of corrosion and of transformer saturation.
Usually the electrodes are placed from 8 to 50 km (5 to 30 mi) from the
terminals and are connected to the converters through electrode lines lightly
insulated from ground. Depending on local circumstances, the electrode line
may be carried on the same towers as the main dc line, perhaps serving the
additional function of overhead shield wire, or on separate poles or towers,
or in cable, on a separate route. For avoiding corrosion of tower footings near
the electrode, these may be insulated from ground or wooden poles may be
used.
Safety. The maximum surface gradient near a land electrode should be
limited to about 5 V/m so that a walking or standing man cannot receive an
annoying electric shock. If the land is to be used for grazing, the same limit
ing gradient (5 V/m) should be acceptable. The maximum potential difference
between two feet is called the step voltage. This matter is discussed in more
detail in "Safety," page 455.
Attention must be given also to touch voltage between a person's feet and
such metal objects as fences. Fences enclosing an electrode site can be
designed to follow equipotential lines, at least approximately. Also, these
fences and others in the vicinity can be sectionalized by insulators, grounded
at many points, or made of nonconducting material.
In the vicinity of a sea or shore electrode, areas of high voltage gradient
must be enclosed or screened to exclude fish and unauthorized persons.
Limitation of voltage gradients near a dc electrode is easier than near an
ac electrode, because dc short-circuit currents are effectively limited to a
smaller multiple of rated current. On the other hand, those gradients caused
by load current may persist for a long time.
DESIGN OF LAND ELECTRODES
445
Duties
Three different kinds of duty for dc ground electrodes are listed below in
order of decreasing severity :
1. Continuous duty
2. Intermittent duty ground return used only during emergencies for a
limited time
3. No load current.
The requirements of the electrodes of different categories differ chiefly
with respect to the following factors.

Electrolytic Corrosion. Electrolytic corrosion of the electrode itself and of
foreign metallic objects, being cumulative, depends on the ampere-hours of
use. Hence, the construction and location of the electrode is most critical in
category 1 and least critical in category 3.
Heating and Drying of the Soil. These effects depend on the duration of
each period of use, provided that other conditions, such as temperature,
rainfall, and porosity of soil, are similar. Again, the order of criticality is
1, 2, 3.
The only electrodes of category 3 to date are those of the Cross Channel
scheme. The purpose of such grounding is twofold :
1. To limit voltage stresses on insulation
2. To facilitate detection and clearing of ground faults
In this category, ground current is negligible except during ground faults.

Ground electrodes in categories 1 and 2 must be designed much differently
from grounds for ac systems.
9-7 DESIGN OF LAND ELECTRODES
General
The factors to be considered in the design of a land electrode are the choice
of site, the choice of electrode materials, the form or layout of the electrode,
the computation of its dimensions for the avoidance of overheating of the
soil, the computation of its resistance and time constant, and the assurance
of low step and touch voltages.
446
GROUND RETURN
9-7
447
quate area of land having soil of low resistivity and (b) far enough from the
converter station and other substations and from densely populated areas,
power plants, pipelines, railways, etc., so that problems of electrolytic cor
from a plain iron anode, and, besides, the cost of coke per unit mass is much
less than that of iron.
If carbon or graphite rods are used instead of iron rods for distributing
current to the coke, the loss of material from the anode by electrolysis is
almost entirely eliminated.
The method of construction is to dig a trench (Figure 28), to fill it with coke
rosion or interference are avoided as much as possible. The distance generally

should be at least 8 km (5 mi).
Surface of ground
Choice of Site
The location of the electrode should be chosen (a) where there is an ade
Choice of Anode Material
It is assumed that the electrode is an anode, at least part of the time. The
principal problem is to find a cheap material that does not suffer excessive
electrolytic corrosion.
Although iron, especially scrap iron, is cheap, it does not last long enough
when carrying large direct currents (1000 A or more). The theoretical loss of
mass from an anode carrying 1000 A (time-average) is 9000 kg (9 tonnes) per
yr, which is clearly an excessive amount. The actual loss from an iron anode
buried in the soil may range from 0.2 to 1.4 times the theoretical loss, depend
ing on the nature of the soil, but even the lower figure is still excessive.
Loss of graphite is said to be about 0.6 that of iron and hence is also
excessive.
Experiments carried out in connection with the Kashira-Moscow dc line31
showed that if iron anodes, instead of being buried in direct contact with the
soil, were buried in a pit filled with crushed coke, the loss of iron was reduced
by a factor ranging from 68 to 21,000, depending on the time of year and,
presumably, on the moisture in the soil. Using the more optimistic factor, the
annual loss of iron at 1000 A dc is reduced to 0.4 kg, or, with the more
pessimistic factor, to 140 kg, which is an acceptable amount.
The explanation of the great reduction in corrosion of the iron is that
conduction from iron to coke is mostly electronic rather than ionic. If
moisture reaches the iron, however, a small part of the conduction from the
iron is ionic, resulting in some electrolytic corrosion. This supposition is
confirmed by observations made during the same experiments, that the lowest
rate of wastage of iron occurred when the ground was frozen, and the highest
rate when the soil was wettest.
Fromthe standpoint of loss of iron, ideal operation is with a current density
at the surface of the iron that keeps the surface dry. In the cited experiments,
this density was found to be 15 to 25 A/m2. The best coke to use has a low
sulphur content.
In an electrode so constructed, the coke becomes the true electrode, and
the role of the iron is merely to distribute current to the coke. There is some
wastage of coke, but less mass of coke is lost per kiloampere-year than of iron
Backfill
Distribution
conductor
Coke
Fig. 28. Cross section of land electrode made with fill of granulated coke in the bottom
of a trench: depth of burial h; sectional area ax b.
to half the final depth, to lay the distributing conductors on this layer of
coke, to put the remaining half of the coke on top of the lower half, and
finally to backfill the remaining space in the trench to the original grade. At
each stage, the coke or backfill should be compacted by tamping.

The arrangement described has been used for practically all the land
anodes yet constructed for high-power dc lines. Such anodes are located at
both ends of the Volgograd-Donbass lines, near Benmore (South Island of
New Zealand), and at Rice Flats near the Celilo substation of the CeliloSylmar line.
Form of Electrode
The layout of an electrode in plan view may take various forms, several of
which are illustrated in Figure 29.
The ring form, if located in uniform soil, gives the most uniform current
into the ground per unit length of trench and thus uses the materials to best
448
9-7
GROUND RETURN
Star
(a)
449
of n Equiangularly Spaced Branches of Equal Length l/n

(74)
12
N(n)
0.53
1.45
3.42
5.5
10.0
Notation
Re = electrode-ground resistance, Q.
(b)
Fig. 29. Plan views of several forms of land electrode: (a) straight; (b) ring; (c) sixpointed star; (d) branched star.
advantage. It requires, however, a large tract of land with uniform soil and is
rather inflexible.
p ground resistivity, Q m
I= total length of conductors, m
D diameter of ring, m
b = yjdh, m
d = diameter of conductor, m
h = depth of burial of center of conductor, m
In x = natural logarithms of x = 2.303 log10 x
The straight line and the star have current densities from coke to soil that
increase at the ends of the arms. The length and shape of the arms can be
varied as required by local conditions.
In the star-shaped electrodes the decrease in electrode-to-ground conduc
tance for each additional arm decreases with an increasing number of arms.
Rusck40 suggested the six-armed star as a good compromise between con
ductance per unit area of site and conductance per unit volume of electrode
material. A six-pointed star of distorted form was used at Benmore.54'57
A ring was used at Rice Flats.78
A branched star appears to use the area of site efficiently.
The foregoing formulas rest on the assumption thattheconductor surface is

equipotential. For greater accuracy, the drop in potential from the point of
feed to the open end or ends may be taken into account. By using the formula
for input resistance of an open-ended leaky dc line, we get
Electrode-ground Resistance3,6'16111,18
where R = corrected electrode-ground resistance

Re uncorrected electrode-ground resistance
Rt = longitudinal resistance of one arm of conductors
Formulas are given below for the resistance, with respect to remote earth,
161,1
All these electrodes are assumed to be
of buried electrodes of several forms.
made of conductors having a constant circular cross section and buried to
uniform depth in soil of uniform resistivity.
RiY/a
coth ( ~
(75)
This relation should be applied separately to each arm of a star electrode

having center feed or to each half of a straight electrode having center feed.
For a ring, it should be applied to an arc between each feed point and halfway
to the next feed point; that is, the ring should be considered open halfway
between each pair of adjacent feed points.
The foregoing formulas rest also on the assumption of uniform earth.
Equations have been published for the resistance to ground of a ring elec
trode buried in the upper layer of a two-layer earth.24,41 For installations
where the ground has distinct layers of thickness smaller than the horizontal
extent of the electrode, it may be advisable to take the stratification into
..
Straight Rod(Cylinder)
R.=
Rin = R,11
sAe)
when
nl
(72)
Ring (Torus)
4D
ll Tib
It
~b
(73)
account.
450
9-7
GROUND RETURN
Resistivity of Soil
Since the electrode-to-ground resistance and the voltage gradients near any
given electrode are both proportional to soil resistivity, low resistivity (high
conductivity) is important.
The conductivity of the soil is attributable almost entirely to the moisture
in it and to the minerals dissolved in this moisture. Therefore, an adequate
supply of moisture to the soil in the vicinity of the electrode must be assured.
Accordingly, land electrodes can be classified into two kinds irrigated
and unirrigated. The soil around an irrigated electrode is kept moist by
pumping water to it; the unirrigated electrode relies on the natural supply
of water. The land electrodes constructed to date have been unirrigated. Of
course, irrigation could be added later if it were found necessary.
The retention of moisture is affected by gravity, capillarity, hygroscopy,
evaporation, and electroosmosis. Gravity affects the movement of free water,
which is that in excess of 15 to 20% of the dry weight of the soil. Capillarity
can raise moisture to a level higher than gravity does by an amount that
increases as the sizes of interstices decreases. It is more effective in clay than
in sand. Hygroscopic water is that held by chemical attraction. Evaporation
depends on soil temperature, air humidity, and surface covering. Plants
increase the evaporation; paving retards it.
An electrode tends to dry the soil in two ways : first, by heating it, thereby
increasing evaporation, and, second, around an anode, by electroosmosis.
Electroosmosis moves water in the direction of the current. It has been used
11,42,56
but it is
to advantage in stabilizing the soil in cuts or excavations,
detrimental around an anode. The tendency of an electrode to dry the soil must
be offset by the effects of rainfall, gravity, and capillarity in restoring soil
moisture.
The current density at which electroosmosis begins to have a marked
detrimental effect depends on the character of the soil and on its supply of
water. An idea of the order of magnitude of this current density is provided
by two sources. Russian investigators31 found the critical current density to
be 9.8 A/m2 and recommended densities of 1.5 to 2.0 A/m2. The New
Zealand Electricity Department found that at the electrode site near Benmore54 the critical current density was 1.0 A/m2. An accurate value of critical
electroosmotic current density need not be known, because usually a lower
current density is imposed by the requirement of thermal stability, discussed
below.
Compaction of the soil by methods well known in the construction industry
aids in retention of moisture by decreasing evaporation and increasing
capillarity. The pressures usually applied are in the range of 200 to 300
lbf/in2 (105 to 1.5 x 105 N/m2).
451
The resistivity of soil decreases with increasing temperature up to the

boiling point of water. The hottest points in the soil are at the surface of the
electrode. When the temperature there reaches the boiling point, arcing
between soil and electrode occurs, the resistance from electrode to ground
increases abruptly, the I2R loss increases, and more moisture boils. This
condition, called thermal instability, must be avoided.
When the soil freezes, its resistivity increases, but it still retains some con
ductivity. Fortunately, in most climates the soil does not freeze deeply.
The amount of moisture increases and the variation of temperature of the
soil decreases with depth below the surface. These facts, plus the reduction
of electrode-to-ground resistance and of step voltage, indicate the advisability
of burying the electrode to as great a depth as can be economically achieved.
Heating of SoU2,6, 15111,17,40
The analogous equations of the current field and the heat-flow field in the
earth can be used in the derivation of a relation between the potential of the
electrode with respect to remote earth and the temperature rise of the elec
trode and of the soil in contact with it. Analogous quantities are electric
potential and temperature rise, electric current and heat flux, electrical
resistivity and thermal resistivity.
When the two fields coexist in the same soil, there is one important differ
ence between the otherwise analogous relations. In the current field, the only
source of current is the electrode, there being no sources of current in the soil.
In the heat-flow field, no heat is supplied by the electrode, but heat is gen
erated in the soil by the loss of electrical energy in the soil's resistance. In
other words, the current field has no divergence except in the electrode ; the
heat-flow field has divergence everywhere else but has none in the electrode.
Another aspect of the analogy is that certain boundary conditions must be
similar. The boundary conditions for the electric field are (1) that the elec
trode surface be equipotential, (2) that the potential of all very distant points
be equal, having some arbitrary value, taken as zero, and (3) that no current
pass through the surface of the ground. The boundary conditions for the
thermal field are very similar except that some heat does pass through the
surface of the ground into the atmosphere. For facility of analysis we assume
the atmosphere to be a perfect heat insulator as well as a perfect electrical
insulator. This assumption gives a pessimistic value of the temperature rise
of the electrode, because actually, when the soil near the electrode is heated,
some heat is lost to the atmosphere.
The electrode is at the highest potential and temperature. Both electric
charge and heat flow away from the electrode. The lines of flow of both
quantities coincide. Electrical equipotential surfaces are also surfaces of
equal temperature.
452
GROUND RETURN
9-7
The relation between potential and temperature is found from the prin
ciple of conservation of energy. As one goes farther from the electrode, the
flow of electric energy through the equipotential surfaces decreases and the
flow of heat energy increases correspondingly, the sum of the two being
constant and equal to the electric power leaving the electrode, which is
VeI watts.
On a given equipotential surface at potential V with respect to remote
earth, the electric power (rate of flow of electric energy) is VI watts. On the
same surface, the heat-energy flow density is the product of thermal conduct
ivity X and temperature gradient or X dBjdr watts per square metre, where 6
is the temperature and r is the distance along the field vector. The heat-energy
flow over the entire surface is
, , de
X \ dA
J 4
dr
(76)
watts
where A is the area (square metres) of the equipotential surface. The total
energy flow through the surface is equal to the electrode power, or
dr
dA
+ VI = VJ
(77)
watts
453
At the surface of the electrode, 9 = 9e and V = Ve, whence

(83>
B-=wP
=
V2JpBe
(84)
where Ve = potential of electrode with respect to remote earth, V

9e = temperature rise of electrode and contiguous soil above ambient
temperature, C
p = resistivity of soil, ft m
X = heat conductivity of soil, W/m C
This relation is independent of the shapes of the electrode and of the surface
of the ground. It does rest, however, on the assumption of uniformity of both
electrical and thermal resistivities.
Let typical numerical values be substituted in Eq. (84) (Rusck40):
Maximum soil temperature in summer in a temperate climate, 25C
Boiling point, 100C
Maximum permissible temperature rise, 9e = 100 25 = 75C
Thermal conductivity of soil, X = 1 W/m C.
Equation (84) then becomes
By rearrangement
X I
dA = (Ve - V)I
(78)
watts
Now
/=| JdA = -\
PJa
P J a dr
dA
(79)
so that
"\A7rdA (V--V)l\Jr-dA
=
(80)
Ve = ReId = V2 x
1 x 75 x
Vp S 12Vp
volts
(85)
for example
if p = 10 Q m,
ifp = 100 a -m,
if p = 1000 ft m,
Ve = 38 V
Ve = 120 V
Ve = 380 V
Thus, in order to avoid overheating of the soil near the electrode under
continuous ground-return operation, the voltage of the electrode is restricted
to moderate values.
Let us suppose, further, that the ground current is' Id= 1.2 kA. Then the
electrode resistance and the power loss in the soil are as follows:
Differentiation of both sides with respect to A gives

p (Cl ' m)
xJ=~{Ve~V)T
dr
dr p
Re (Q).
(81)
and integration of both sides with respect to r gives
10
0.03
100
1000
0.10
0.31
I&Re (kW)
46
150
460
(82)
Ordinarily, an electrode site is chosen where the ground resistivity is not much
454
9-7
GROUND RETURN
/ R
(2
\1/2
volts
(85a)
in which Re is the electrical resistance between electrode and remote ground,

and Re is the corresponding thermal resistance. The necessity of calculating
these two resistances for each particular electrode geometry increases both
the accuracy and the labor.
Thermal Time Constant16"118'54
The temperature does not rise exponentially; hence, the soil and electrode
have no true time constant. We may define their time constant T, however,
as the time required to reach the final steady-state temperature Bf if the tem
perature increased linearly at its initial rate.
Initially the soil and the electrode are at a uniform temperature: there is no
temperature gradient and hence no flow of heat. The electric energy converted
into heat in any element of volume serves to raise the temperature of the
element. Therefore, the initial rate of the rise of temperature of the element
is the rate of heat production divided by the heat capacitivity y. Heat capacity
is heat energy per unit increase of temperature, and heat capacitivity is
heat capacity per unit volume. It differs from specific heat, which is heat
capacity per unit mass, by the density of the substance.
The rate of heat production is the rate of dissipation of electric energy,
which, in unit volume, is pJ2 watt per cubic metre. The initial rate of rise of
temperature is
dd
whence the time constant, as
pJ2
T
*
defined, is
T=
(86)
yQf
jj[
(87)
At the electrode, the final temperature is given by Eq. (84). On substitution

of this, the time constant becomes
T=US) =2$j)
I'
l.
seconds
(88)
For the last equation to be true, J must be interpreted as the average current
density at the electrode surface. Actually, the initial rate of the rise of temperature varies from one part of the surface to another in proportion to the
square of the current density, but the flow of heat inside the electrode tends
to equalize the temperature.
The expression in parentheses has the dimension of length. For a spherical
electrode, it is the radius.6 For the electrodes of Figure 29, page 448, it is
about six to nine times the diameter d.
Typical values of soil parameters are
2 = 1 to 3 W/m C
y = 1.3 x 106 J/m3 C for sand and clay, corresponding to specific
heat 0.2 cal/g and specific gravity 1.6 g/cm3
p = 10 to 100 Q m
Land electrodes of the size used for high-current dc transmission have time
constants of the order of 100 days.
Safety4'10'22,25,27'29,34'38
I
I
f,
J
0r
455
where y heat capacitivity of soil, J(m3 C

X = thermal conductivity of soil, W/m C
p = electric resistivity of soil, Q m
Re = resistance from electrode to ground, 1
Ae = surface area of electrode, m2
greater than 100 Q m. With low soil resistivities, the surface area of the
electrode is determined by temperature rise, but with high resistivities it may
be determined by acceptable power loss.
A more recent analysis83 replaces Eq. (84) by the following:
Ve
f
I
I
|
I
|
f
1
1
|
Voltage gradients high enough to be uncomfortable, or even dangerous,

the ground near
electrodes for high-current dc transmission, and this aspect of electrode
design must be investigated. If the gradient turns out to be too high, there are
at least two alternatives for making the installation safe.
One method is to enclose the area of high gradients by a fence so as to
impede access to unauthorized persons. There are several objections to this
method :
to persons and to livestock may occur on the surface of
1. Obviously it is not foolproof.

2. The land area required by a high-current electrode is considerable, and
so is the cost of even an ordinary fence of the required length.
3. In most cases, to allow the land to be used for grazing or for agriculture
would be more economical than to reserve it solely for the electrode.
The better method is to bury the electrode deeply enough to reduce

maximum surface gradients to a safe value. A depth of 1 or 2 m is already
advisable for other reasons, including (a) mechanical protection of the electrode, (b) reaching soil that is suitable as to moisture content, temperature,
456
GROUND RETURN
9-7
and absence of freezing, and (c) decreasing the resistance from electrode to
remote earth. Usually, the depth of burial required for safe surface gradient
is no greater than, or only a little greater than, the depth required for other
purposes.
Tolerable Body Currents. The current through the body is more significant
than the voltage across the body. Body current has the following effects,
according to the magnitude of the current. Listed in order of increasing
current, they are (a) perception, (b) annoyance, (c) discomfort and strong
muscular contractions, (d) cessation of breathing, (e) ventricular fibrillation
(irregular heart action), and (f) burns. Danger begins with the third item;
the victim, grasping a live conductor, is unable to let go of it. He can be
rescued by deneergizing the circuit or by pulling him from it. Cessation of
breathing can usually be remedied by artificial respiration. Ventricular
fibrillation is very dangerous. The only known remedy is a controlled countershock of short duration within a few minutes. The equipment, and skill for
administering such a countershock are almost never available.
Considerable experimentation has been done on let-go currents with
human subjects25 and on the threshold of ventricular fibrillation with
animals.4,34 Most of this work has been done with 60-Hz current, but enough
has been done with direct current to show that the dc let-go and fibrillation
currents are five times as great as the corresponding 60-Hz currents. Let us
assume that this same ratio applies to current values causing perception and
annoyance.
The perceptible body current is 0.1 to 0.2 mA at 60 Hz; assume 0.5 to

1.0 mA dc. Annoying body current is 1 mA 60 Hz; assume 5 mA dc.
Experiments with a large group of human volunteers showed that 99.5%
of the men could let go at 9 mA, 60 hz, and 99.5% of the women, at 6 mA,
60 Hz. The corresponding direct currents are 45 and 30 mA. The concept of
let-go current is not applicable to livestock or to a prone human except that a
current of this magnitude gives a painful shock.
Experiments with animals showed that the threshold fibrillation current
depends on the following:
1. Species and size roughly proportional to body weight and heart
weight. Effect on men is calculated by interpolation to body weight of 70 kg.
2. Current pathway, lowest current for fibrillation being between a foreleg
and opposite hind leg; that between two forelegs was slightly higher and that
between two hind legs much higher.
3. Frequency. Threshold current for 25 Hz is 1.25 times that for 60 Hz,
and that for direct current, 5 times that for 60 Hz.
4. Duration of shock, varying as t~i/2 for t from 1 to 5 sec.
457
The theoretical 60-Hz rms threshood fibrillating current for normal men is
I
= 165/ 1 mA, where t is duration of shock in seconds. The corresponding
%/
dc value is
825/r mA.
Design Criteria. The body current of men and livestock should be below
the let-go level. Suitable criteria might be (a) the annoyance level for standing
or walking men and (b) half the danger (let-go) level for prone men or
standing animals, both to be with good contact to the soil (wet hands and
feet) at the worst place in the field with rated direct current in the electrode.
With a. fault on the dc line, the ground current might be several times its
rated value but only for a very short time (fraction of a second). Men or
animals could readily move a short distance from an uncomfortable place to a
comfortable one.
The body currents corresponding to these criteria are (a) 5 mA for a
standing man, (b) 22.5 mA for a prone man, and (c) for a standing animal,
22.5 mA multiplied by the ratio of its body weight to 70 kg.
Body and Contact Resistances. It is clear that the current shunted through
a body has no perceptible effect on the electrode current nor on any part of the
electric fieldof the anode except very close to the points of contact of the body
to ground. Therefore the current through the body is equal to the opencircuit voltage between the points of contact divided by the sum of the body
resistance and the two contact resistances in series.
Various authorities have suggested values of human-body resistance
ranging from 500 to 3000 Q. Obviously, it depends very much on the condi
tion of the skin. Dalziel22 measured body resistance to direct current from
one hand, wet with salt water and grasping a metal electrode, to both feet
in 3/4 in (2 cm) of salt water; the minimum value of nine subjects was 1135 Q.
With one arm immersed in salt water to the elbow and both legs in 10 in (25
cm) of salt water, resistances were 400 to 600 Q. For our purpose, a value of
1000 D is conservative, perhaps somewhat pessimistic.
Contact resistances between a bare foot and moist soil were found to be
equivalent to that between a circular metal plate of 16 cm diameter and the
soil.6,29 This value is 3ps, where ps is the resistivity, in ohm-metres, of the
surface soil.18 Thus the resistance of a standing or walking man (foot to
foot) is 1000 + 6ps Q. The resistance of a standing man grasping a metal
object with one hand is 1000 + 1.5pf, on the assumption that the resistance
of the contact between hand and metal object is negligible. This situation is
not likely to arise near a land electrode unless there are metal fence posts or
wires in the area of high voltage gradient. The resistance between the hands
and feet of a prone man is 1000 + 3ps, on the assumption that the contact
458
9-7
GROUND RETURN
459
resistances from hand to soil are equal to that from foot to soil. The resis
tance between front and hind feet of a standing or walking animal can be
calculated in like manner if the body and foot-to-soil resistances are known.
If we assume that the average density and resistivity of an animal body
is equal to those of the human body, the body resistance is inversely pro
portional to the weight. Then the body resistance of a horse or cow weigh
ing 1000 to 1400 lb (400 to 600 kg) is about (70/500)1000 = 140 H. The
hoof-to-soil contact resistance is inversely proportional to the contact area
and is about the same as that for a man's foot. Thus the resistance between
fore and hind feet of a cow or horse is about 140 + 3ps Q.
Table 6 shows data from which the tolerable step voltage and gradient are
computed according to the criteria adopted above. It is clear that, under these
criteria, the maximum permissible gradient is 5 -t- 0.03ps V/m, determined by
annoyance to a man walking with wet feet and 1-m steps near the electrode.
sections of infinitely long horizontal cylinders instead of spheres. Consider

first only one of these cylinders embedded in an infinite medium. By sym
metry the current density and voltage gradient vectors lie in planes perpen
dicular to the axis of the cylinder, are radial in direction, and have magnitudes
that are a function of only one coordinate the radius r. The gradient due
to each cylinder is pljlnrl The gradient due to both cylinders is the vector
sum of the two components. At points on the surface of the ground, the two
components are equal in magnitude, make angles 6 with the horizontal,
and have a horizontal resultant that is
Table 6. Tolerable Body Currents, Step Voltages, and Voltage
It is proportional to the ground current per unit length, ///, and to the resis
tivity and is inversely proportional to the depth of burial h to the center of
the cylinder. This method of calculation is accurate enough for electrodes of
finite length much greater than the diameter and for square or rectangular
cross section as well as round. It is adequate also for star or ring layouts.
The surface gradients near the ends of a cylinder of finite length or near
the ends of the arms of a star, however, are somewhat greater than that
calculated by Eq. (90).
Figure 30 shows how the potential and its gradient vary over a small range
of horizontal distance from the electrode. It is apparent that the gradient
varies considerably over the length of a step if the step is not much smaller
than the depth of burial. It would be more correct to compute the maximum
difference of potential between two points separated by the length of step than
to compute the maximum gradient and multiply it by the length of step.
The latter procedure, however, is simpler, errs on the pessimistic side, and is
just as accurate as are the tolerable values of body current.
Gradients near a DC Ground Electrode

Creature
Position
Weight (kg)
Resistance (H) :
Body
Contact
Criterion
Man
Walking
Prone
Cow or Horse
Standing
70
70
500
1000
1000
3ps
140
6ps
Annoyance
Pain
Pain
+ 0.03/>,
22
22 + 0.07ps
160
22 + 0.48ps
+ 0.03/>s
11 +0.035ps
11 +0.24p,
11.0
11
35
59
3Ps
Tolerable
current (mA)
Voltage (V)
Step (m)
5
5
Gradient (V/m)
General
Ps= 0
p., = 100 n m
Ps = 200 n m
5.0
8.0
11.0
14.5
18.0
Computation of Maximum Gradient. It will be shown that the maximum

voltage gradient on the ground occurs at a horizontal distance from the
electrode equal to the depth of burial of the electrode and this gradient is
readily computed.
Because the maximum gradient occurs so near the electrode, it suffices to
assume the latter to be straight and infinitely long. With reference to Figure
7, let the circles marked "buried electrode" and "image" represent cross
pi
pi x plx
=
nrl r nrl
cos 6 =
(pl\
nrl
\ zr lix +h
It is easy to show that the maximum gradient occurs at x h and is
pl_
max = 77
2nlh
(89)
(90)
Design of Land Electrode by Rusck40
Rusck published a proposed design for a land electrode for 1200 A into a
soil of resistivity 100 fl m. He chose to use a coke fill with square cross
section 0.5 by 0.5 m in ditches 1.5 m deep and to arrange the ditches in the
form of a six-pointed star. The average depth of burial of the center of the
coke fill is then 1.25 m. He chose to distribute the current to the coke by a
round iron bar.
The electrode-ground resistance already determined by the heating limit
(table on page 453) is 0.1 O.
460
9-7
GROUND RETURN
1.5
461
their mass has corroded away, their useful life is 11,200/(160 x 2) = 35 yr at

100% load factor or correspondingly longer at lower load factor. This seems
a reasonable life.
r*i
1.0
0.5
Design of Benmore Electrode54,57

T>\__Cy]indricai
electrode
-0.5
-1.0
1.5
-4
1
-2
-1
x/h
Fig. 30. Surface potential V and gradient in the vicinity of a long, cylindrical, hori
zontal buried electrode : unit of V is its value at xjh= 3 ; unit of is its maximum value.
The total length / of all the electrode arms is found by trial substitution for
/ into Eq. (74) for n = 6 to be 3500 m. Hence the length of each of the six
arms is 3500/6 = 585 m. The area of the interface between coke and soil is
3500 x (4 x 0.5) = 7000 m2, and the average current density at this surface is
1200/7000 = 0.17 A/m2, which is well below the permissible limit of 1 or
2 A/m2 determined by electroosmosis.
A current density of 0.5 A/mm2 was chosen for the iron distribution con
ductor. Since the current per arm is 1200/6 = 200 A, the required crosssectional area is 200/0.5 400 mm2 = 4 cm2, corresponding to that of a
square bar 2 by 2 cm. The resistance of such a conductor 585 m long is about
0. 19 2. Assuming the current to diminish linearly from 200 A at the input end
to zero at the extremity, the power loss in the iron conductors is 6 x 0.19
x (200)2/3 = 15 x 103 W = 15 kW. The power loss in the ground is
120 V x 1200 A = 1.44 x 10s W = 144 kW, and the sum of these losses is
15 + 144 = 169 kW. The power loss in the coke is probably much smaller
than that in the iron.
The volume of iron is (2 x 10-2)2 x 3500 = 1.4 m3, and, with a density of
8 tonnes/m3, its mass is 11.2 tonnes. If we assume the loss of iron by elec
trolysis to be 0.015 of the theoretical value, it would amount to 160 kg/yr.
On the assumption that the iron conductors remain serviceable until half of
This was the first land electrode built except those for the VolgogradDonbass line, about which littlehad been published. Its rated current is 1.2 kA.
The site was chosen in a valley situated 4.75 mi (7.6 km) south of Benmore
power plant and having a tertiary deposit of clay, silt, and sand. Tests made
on the soil at the site showed that its effective resistivity was 61.5 2 m, that
its thermal conductivity varied from 1 to 3W/m C, with an average value of
2 W/m C, that its heat capacitivity was of the order of 2 x 106 J/m3 C,
and that its maximum ambient temperature was 25G. The effective resistivity
was taken as that of a uniform earth, which would give the same resistance
from a large buried ring electrode to ground as it would if buried in the upper
layer, of resistivity 33 2 m and thickness 36.5 ft (11.1 m) resting on a lower
layer of 100 2 m and at least 500 ft (150 m) thick. Several test holes were
bored at depths down to 70 ft (21 m), and soil samples were analyzed for
grain sizes. Moisture content of the soil varied from 4 to 21% of dry weight,
with a mean value over the site of 8%.
A test electrode buried in a straight trench 50 ft (15 m) long and having the
same cross section as that proposed for the main electrode was located in
soil similar to that on the site of the main electrode. Its resistance, calculated
from Eq. (72) with p = 68 2 m, was 3.5 2. When the test electrode was loaded
with the same current per unit length as that proposed for the main electrode,
its resistance was initially 3.2 2 and dropped slowly to 2.8 2. The temperature
rise of the soil near the electrode was found to be about 60C. Time constant
of the test electrode was about four days. When the current density was raised
to 1 A/m2 into the soil, electroosmosis was indicated by the behavior of
measured resistance and temperature.
Because it was difficult to fit a ring electrode onto the site, a six-pointed
star was adopted. As shown in Figure 31, the arms are not straight or of
equal lengths. When rock was encountered on one arm, this arm was ter
minated and another arm was extended. The junction point of the star was
placed in an area of lower than average soil resistivity.
The maximum resistance for thermal stability was computed by Eq. (84) as
Re
2pX9
T
V 2 x 61.5 x 2(100 - 25)

1200
136
= = 0.1132
1200
462
9-7
GROUND RETURN
463
Life Inspection after four years' operation, during which the electrodes
conducted 7.4 x 106 Ah as cathode and 1.6 x 106 Ah as anode, showed
evenly distributed pitting with very little reduction in diameter.
Design of Rice Flats Electrode78
Electrode line
I
0
0.5
'
1.0 km
Fig. 31. Plan of Benmore ground electrode as built. Ground surface equipotential lines
in percent of electrode voltage to remote earth. (Adapted from Refs. 57 and 81 by per
mission of the I.E.E.E.)
Use of Eq. (74) with n = 6 showed that each arm must have a length of at
least 400 yd (365 m) if the cross section of the coke were 20 by 20 in (51 by
51 cm) and if the depth of burial were 5 ft (1.5 m).
The area of the coke-soil interface was computed as 4500 m2, and the
corresponding current density as 0.27 A/m2. This is well below the critical
density (1 A/m2) found for electroosmotic drying.
The time constant was computed from Eq. (88) as 100 days.
Because the electrode would occupy an area of 200 acres (80 hectares),
fencing it would be expensive. It was decided to bury the electrode deeply
enough so that surface voltage gradients would be safe and to permit the land
to be used for grazing. A step voltage of 20 V in 5 ft (13 V/ml was taken as
safe for animals. The depth of burial of 5 ft (1.5 m) was foundao be adequate.
The maximum surface gradient, computed from Eq. (90) yvith ps = 33. 12 m
and with end effects neglected, is (33 x 1200)/(2je x 365 -x 6 x 1.5) = 2 V/m.
The calculated resistance of the electrode was 0. 11 II; the measured resis
tance varied between 0.29 and 0.32 12. The discrepancy between these two
values may be attributable to departure from the intended star configuration
and to a greater effect of the lower, high-resistivity layer on the effective
resistivity. The highest measured voltage gradient was 6 V in 4 ft (5 V/m).54
The cost of the electrode, including the cost of investigation, was about
15,600 ($44,000).
The chosen location is on the route of the dc line 6.6 mi (10.6 km) south of
Celilo, the southern terminal of the dc Pacific intertie. It is reasonably
distant from other plants that might be affected by corrosion, being 18 mi
(29 km) from one major gas pipeline and 35 mi (56 km) from another, and
7.5 mi (12 km) from the city of The Dalles, Oregon, with its underground gas
and water distribution systems.
The soil is a fine wind-deposited silt (loess) up to 50 ft (15 m) thick, resting
on basalt. Annual precipitation is only 16 in (42 cm), but the fine texture of the
soil permits it to retain moisture. Measurements of soil moisture made at
various points during the course of a year show that it varies from 5 to 14%
of the dry weight. Tests made with small electrodes over 18 months showed
that satisfactory operation could be achieved with this rather dry soil if the
current density at the electrode-soil interface were kept somewhat below
1 A/m2. Therefore, the expense of providing irrigation or of adopting a
moister site 20 mi (32 km) farther away was judged to be unjustified. The
chosen site is used for dry wheat farming.
Measurements of soil resistivity on the site showed 70 12 m at 5-ft (1.5-m)
depth, down to 36 Q m at 40 to 50 ft (12 to 15 m), up again to 50 2 m at
300 ft (92 m) and on up to 100 12 m at 700 to 1000 ft (210 to 300 m). The
design value was taken as 70 12 m. The thermal conductivity of the soil was
measured as 0.87 to 1.13 W/m C. Soil temperature at depths of 3.5 to 4.5 ft
(1.1 to 1.4 m) was measured at 15C or less in May, 1964, and as 16C in
August, 1965. The design value was taken as 25C, giving an allowable rise
of 75C. The specific heat of the soil was assumed to be 0.19 cal/g.
The following requirements were set as a basis for design :
1. Current capability of 1.8 kA continuously, although only occasional use

of ground return was expected
2. Current density at the electrode surface not to exceed 0.5 A/m2, so as
not to dry the soil by electroosmosis
3. Maximum surface potential gradient, 8.5 V/m, in the interest of safety
to people and animals
The maximum permissible electrode voltage was calculated by Eq. (84) as
Ve = V2 x
75 x 1.0 x 70 = 102 V
464
9-8
GROUND RETURN
DESIGN OF SEA AND SHORE ELECTRODES
465
whence the maximum electrode-ground resistance is
Re = VJI = 102/1800 = 0.057 Q

A ring electrode was chosen as suitable for the site. The cross section of the
coke was first taken as 0.5 by 0.5 m, and the depth of burial as 1.5 m. From
Eq. (73), with b = VO.5 x 0.75, the minimum diameter of the ring was cal
culated as D = 1050 m (3200 ft). Later the dimensions were increased some
what. The coke section became 2 ft by 2 ft (0.6 by 0.6 m), and D = 3400 ft
(1110 m), from which the computed resistance is Re = 0.056 Q. The measured
resistance is 0.105 fi.
The surface area of the coke is 7tl 110 x 4 x 0.6 = 8400 m2, and the cor
responding current density is 1800/8400 = 0.21 A/m2, which is well within
the specified value of 0.5 A/m2.
The permissible step voltage was calculated to limit body current to 6 mA
with body resistance of 1000 Q and two contact resistances of 3ps = 3 x 70
= 210 Q each, or a total of 1420 O. Step voltage = 0.006 x 1420 =8.5 V.
For a step of 1 m, the corresponding gradient is 8.5 V/m. The maximum
gradient calculated by Eq. (90) is
pi
max
2nTh
70 x 1800 .
= 4 Vm
2it x 1IOOtt x 1.5
= 3.04 x
General
Most of the dc transmission schemes built before 1970 have sea or shore
electrodes at one or both ends. These schemes include Gotland, KontiSkan, New Zealand, Sardinia, Vancouver Island, and the Pacific NorthwestSouthwest Intertie.
For the purpose of this discussion a shore electrode is defined as either an
electrode on the seashore near the sea or one in seawater near the shore, as in
a cove or lagoon. Only one true sea electrode has been built to date: this is in
the Pacific Ocean 6000 ft (1.8 km) off the shore near Santa Monica, Cali
fornia.
Advantages. The principal advantages of a sea or shore electrode over a
land electrode are shown by the analysis in Section 9-2, page 402.
1. It is possible to obtain a given low resistance to earth by means of a

smaller electrode installation.
2. For a given ground current, the electric field at a given distance from the
electrode is considerably smaller.
The same advantages hold for an electrode in or at the shore of a large
body of fresh water, but to a lesser degree because of the higher resistivity
of fresh water than of seawater.
which is less than half the design limit.

By Eq. (88), the thermal time constant is
1.3 x 10-V0.057 x 8400V
9-8 DESIGN OF SEA AND SHORE ELECTRODES
=6-5*10-W
107 sec = 352 days
The current is distributed to the coke in the trench through a series of 1067
high-silicon cast-iron cylinders each 5 ft (1.53 m) long by 1.5 in (3.8 cm)
diameter. Groups of five cylinders were fabricated into strings with a 5-ft
insulated cable between adjacent cylinders. A 5-ft flexible lead at each end of
each string was connected to No. 4/0 AWG (108 mm2) copper distribution
cable laid in the same trench in a large closed circle fed at three equally
spaced points through buried insulated cables radiating from a point inside
the circle where the electrode line terminates at a steel tower, which also
carries the main dc line.
Cost. The total cost of the Rice Flats electrode, including surveys, tests,
design, land easements, materials, labor, and overhead, was about $300,000.
Problems. Three problems found with sea or shore electrodes but not with
land electrodes are the following:
1. Additional corrosiveness of seawater, especially at the anode.
2. The necessity for the protection of fish.
3. The protection of the electrode installation from mechanical damage
caused by waves, ice, and driftwood during storms, especially at high tides.
Chlorine is released at an anode in seawater. It attacks most metals, wood,

rubber, and some plastics. Suitable anode materials are graphite and plati
nized titanium. Suitable anode enclosures may be made of polyvinyl chloride
(PVC) or concrete. Conductors feeding the electrodes may be insulated with
PVC or other suitable plastics. These corrosion problems do not appear at
the cathode, although hydroxides may form on it.
Fish are attracted to the anode but not to the cathode. A threshold gradient
is required to attract fish, and a considerably higher gradient stuns or paralyzes
fish. Dell55 states that "fish over 12 in long are stimulated by a voltage
466
9-8
GROUND RETURN
gradient as low as 0.17 V/in" (6.7 V/m). Elder and Whitney79 state: "Some
observers note that it is the head-to-tail voltage rather than gradient that
affects fish. Voltages in the order of 1 to 5 V have been observed to make a
fish unconscious." Quinn and Howard58 state that a gradient of 2.5 V/m
causes discomfort to a man in the water, but that 1.25 V/m can be regarded as
safe for man and for large fish. Areas of high voltage gradient can be enclosed
by screens or perforated tubes or boxes, or they can be separated from the
main sea by a rock breakwater.
Protection against wave action may be obtained by covering the electrodes
with large rocks, placing them in a cove separated from the main sea by a
breakwater, or putting them on shore at a safe distance from the sea. Another
method is to place the electrodes on the sea bottom where the water is so deep
that wave forces are weak.
After the pioneering Gotland scheme, which began operation in 1954
with 200-A sea return, no more dc sea or shore electrodes were built until
1965, and these were for much greater currents (1.0 and 1.2 kA).
Danish Shore Anode of Konti-Skan Scheme65
This scheme had a rated current of 1000 A and was initially monopolar
with sea return. It was decided that the current in the sea cables should always
flow westward, from Sweden to Denmark. Consequently, the electrode on
the Danish side is always the anode and that on the Swedish side always the
cathode. The cathode is very simple, consisting of about 300 m of bare
copper cable of 600 mm2 cross section laid in a ring on the sea bed in 10 to
15 m of water 3 km offshore. The anode, however, posed some interesting
design problems.
The site was chosen on the shore of the Kattegat about 8 km south of the
cable terminal at Stensntes. There a short, lightly insulated branch was built
from the main dc line to the coast at a fairly isolated spot 1 km northeast of
the village of Sora, remote from urban areas, harbors, cables, etc. The site
comprises 1.5 acres (0.61 hectare) of marshy flat land with coast line of 137 m
and width of 46 m. The sea is very shallow far from the land. At low tide the
water recedes as much as 400 m. It was decided to put the electrode on the
shore rather than in the sea because of the serious risk of damage from ice on
the sea in hard winters.
The subsoil consists of 1.4 m of sand, with blue clay underneath to a depth
of at least 44 m.
The first plan considered was to excavate a basin and suspend the anode
elements in the water of this basin from wooden frames. Because of the wide
areas of shallow water, the natural supply of new seawater to such a pool
would be inadequate. Indeed, the flow of underground fresh water into the
467
pool would make the salinity of the water in the pool considerably smaller
than that of seawater. Pumping of seawater into the pool would be necessary.
Pumping being necessary in any case, a cheaper alternative plan was
chosen. This plan was to suspend each anode element in a separate small well.
Each well would be lined with a pipe of material that would conduct current
but would keep the soil from filling the well.
Tests were made to find suitable materials for the anode element and the
well casing, both of whichwould have to be practically immune from deteriora
tion caused by the products of electrolysis of salt water (principally chlorine
and oxygen gases).
The anode elements are impregnated graphite rods 2.44 m long and
0.100 m in diameter and weigh 38 kg, including 5 m of cable. Tests showed
that when such elements were operated at 50 A for 2 months in pumped
seawater, the loss of material was only 5 g per 1000 Ah. Without pumping,
however, in hot water with increased concentration of chlorine and oxygen,
the rate of loss was 10 times as great, or 50 g per 1000 Ah.
For well casings hard polyvinyl chloride (PVC) pipes were chosen; PVC
pipes had been used by well drillers for some years for the sand filters placed
at the bottom of a well and from which the water is pumped out. These pipes
were available with an inside diameter of 0.230 m, wall thickness of 0.010 m,
and length of 6 m, with the number and size of slits required for passage of
current and of water. The slits cover 16% of the wall in a length of 2.8 m,
the total area of slits per pipe being 0.35 m2.
The suitability of anode-element and casing materials was ascertained by
two test wells using a somewhat smaller anode (1.50 by 0.075 m) and a
shorter casing (3 m). The test anodes could carry 50 to 60 A with water flow
as little as 0.2 liter per sec per well with very little temperature rise and with
no problems. The measuredresistance was 0.8 Q for one anode. Without water
flow, the temperature in the test wells rose 40C, and the corrosion of the
graphite anodes was increased tenfold, as already noted; but the PVC pipe
showed only a slight discoloration but no loss of mechanical strength in 8 to
12 months.
The resistance of the full-scale anode was calculated by proportion to be
0.5 Q per well. It was decided to use 25 wells, spaced 5 m apart in a straight
line. If there were no interaction between elements, the resultant resistance
would be 0.5/25 = 0.02 O. The interaction, however, was expected to increase
the resistance by a factor 1.6, making it 0.032 O.
The number of anode elements was chosen to give an average current of
40 A per well. It was expected, however, that the current of the elements at or
near the ends of the row would be greater than that of the other elements.
This tendency was partially offset by using closer spacings between wells
near the ends of the row. The spacing between elements 1 and 2, likewise
468
9-8
GROUND RETURN
between 24 and 25, was made 3 m ; and that between 2 and 3 and 23 and 24,
4 m. Another factor tending to equalize the currents was the greater length
of the cables from a central point to the end elements.
The 25 wells were dug in a row along the shore about 20 m from normal
high-water line. A steel tube of 16-in (41 cm) diameter was drilled 5 m into
the ground. Then the 10-in (25.4-cm) PVC pipe was inserted, the steel tube
was withdrawn, and the space between the PVC pipe and the 16-in hole was
filled with filter sand.
A section of one well is shown in Figure 32, and a plan of the site, in
Figure 33.
Junction box
Hff
- +1.00 once a year

- +0.75 seven times a year
+ 0.50 40 times a year
PVC-insulated cable'
25
mm2
(X
'
Electrode line, 910
mm2 ACSR
25
(normal)
==<>*rMain pipe
Sand
salt water
'.'Outflow!
supply
ditch >
cm diam li t
PVC tube, 23/25
-Graphite electrode
Current-permeable zone
mm2 PVC-insulated copper
Road
! 13
,-iU
mnuniim Hirt tuiuji
Electrode wells
Pump
Symbols
Dam
43*%- Outflow ditch with

direction of flow
PVC suction pipes
for salt water
* pvc pressure pipes
fQr salt water
Salt water wells in

sea bottom
..........
...................
.,.'i
10
....
10
20
30
11
1ni
40
ir
1 1
50
Metres
Fig. 33. Site plan of the Danish shore electrode of the Konti-Skan link. (From Ref. 65
by permission.)
insulated suspension cable

Clay
469
. cable. One for each well.

Number 16 shown only
Seawater level to be expected
- +1.52 (flood observed 1893)

- +1.25 once in five years
Road
.
Meters
Fig. 32. Longitudinal section of an electrode well for the Konti-Skan link. (From Ref. 65
by permission.)
The PVC tube projects above the surface of the ground and carries a
junction box where the neoprene-insulated suspension cable is connected to a
PVC-insulated 25-mm2 copper cable that runs underground to the guyed
terminal tower of the electrode line. On that tower is a copper busbar 1 m
long to which are connected the 910-mm2 ACSR electrode-line conductor
and 25 insulated cables that are connected to the individual anode elements.
The lengths of these cables vary from 18 to 78 m.
Water Circulation Seawater from 9 wells in the seabed about 25 m off

shore is brought by a self-priming centrifugal pump placed about 1.5 m
above normal sea level. Seawater enters each well through a small inlet pipe
connected to the wall of the PVC tube near its lower end. The bottom end
of the PVC tube is closed by an oak plug. Part of the water (10 to 20%)
passes through the slits of the PVC tube and percolates through the sand.
The rest of the water overflows through a short pipe into an outflow ditch
that takes it back to the sea. All the pipe is of plastic. The water supply is
about 0. 1 liter per well per sec. The I2R loss in each well is about 400 W, and
if the heat produced were carried away by the circulating water, its temperature
rise would be about 1C. Much of the heat, however, is conducted to the
earth and to the air. The temperature rise of the well water at 50 A without
water circulation is 15C.
Measurements on Completed Station The ground resistance is

40 V/1000 A = 0.040 Q.The resistance of a singlewell is 0.63 ft. Both of these
values are about 1.25 times the corresponding predicted values. The current
per well varied from 53 A at well 1to 35 A atwell 23 when the total currentwas
1000 A. If the soil had been uniform, the expected current distribution would
have been: wells 1 and 25, 48.5 A each; wells 2 and 24, 43 A each; wells 3 to
23, 39 A each,
'{
|
|
|i
470
GROUND RETURN
The highest surface voltage gradient is 2 V/m near the outermost wells. In
the sea at normal shore line, the greatest gradient is 0.3 V/m, which is too
small to attract fish. The area south of the electrode station is marshy land
used for grazing. The maximum gradient in this area is 0.2 V/m, which has
no visible effect on the cattle. Gradients were measured out to distances of
about 1 km for estimating the densities of currents leaving hypothetical
buried pipes and cables. At 600 m, currents entering such facilities would be
30 to 150 mA/m2, depending on wall thickness, diameter, and kind of metal,
and in the more extended anodic zone the calculated current densities are
only 5 to 20 mA/m2. There are, in fact, no such uninsulated pipes or cables
near the electrode station, except some short water pipes for domestic supply.
The electrode station is surrounded by a fence that approximately follows
equipotential lines. The fence is built of plastic-covered wire mesh. It is
divided into 10-m sections insulated one from another but not grounded.
The electrode station is inspected at intervals of 1 or 2 weeks. Occasionally
it has been necessary to remove sand or mud from the bottoms of some of
the wells. The electrode, when drawn out, has a maximum voltage of 15 V
to adjacent ground. The use of rubber gloves, though not necessary because
of shock, is advisable because the electrodes are slimy and dirty. It has
never been necessary to interrupt transmission for performing maintenance
on the electrode station, although the pump failed and had to be replaced.
The savings in investment, including capitalized losses, of using ground
return on Konti-Skan instead of metallic returns, has been estimated at
16.5 x 106 Danish Kronor (about U.S. $2.2 x 106).
Shore Electrode on North Island, New Zealand55,60
Site. A site was chosen on the west coast of the North Island 15.8 mi
(25.4 km) from the dc terminal substation at Haywards, 8 mi (13 km) from
the dc cable terminal at Oteranga Bay, and far enough from the city of
Wellington and its harbor. The site was named Te Hikowhenua, which means
in Maori "electricity of or in the earth." A preliminary calculation showed
that the leakage current density in. the armoring of the dc cables would be not
more than 4.7 mA/m2 when the electrode was carrying 1200 A (the rated
current of the dc link). This is less than the value (10 mA/m2) deemed
acceptable.
The site is a narrow strip of beach covered with a mixture of sand, shingle,
and large boulders some up to 6 ft (2 m) across to a depth of about 10 ft
(3 m) over bedrock, which also forms the sea bottom. The designers decided
to bury the electrodes in the beach and to protect them with the large boulders.
The seabed sloped about 1.5 (0.026 rad). The resistivity of the seawater
was measured as 0.23 fl-mat 15C, and the design value was taken as 0.25.
9-8
471
The resistivity of beach material saturated with seawater ranged from 0.3
ra for 4-in (10-cm) stones.
m for fine sand to 0.9
As in the design of the Danish electrode, one of the first questions was that
of suitable materials for the electrode units, for insulation of cables attached
to the electrode units, and for porous enclosures around the electrode units.
Tests were made on various materials.
Impregnated graphite, used by a pulp and paper manufacturer for elec
trolytic production of chlorine, was chosen for the electrode elements. A test
cell with a graphite electrode in a steel drum filled with seawater was used to
test insulating materials. The free chlorine content was kept at 1000 parts per
million higher than the concentration expected in the electrode cells
to serve for accelerated life tests. The following insulating materials were
found to be suitable : Hypalon, high-density polythene, and polytetrafluorethylene. Suitable sealing and insulating compounds were found to be phenolic
and epoxy resins. Epoxy resin mixed with a small amount of powdered
graphite provided good adhesion to graphite and to most plastics.
For cell enclosures, hard, dense, cast-concrete pipe, made with pozzolan
sulphate-resisting cement, and drilled with 3/16-in (5-mm) holes to produce
an effective porosity of 6%, was selected. Calcium chloride was used as a
strength accelerator.
Design. The following dimensions were chosen for the electrode units and
their enclosures :
Graphite electrode unit: diameter, 6 in (15 cm); length, 7 ft (2.1 m)

Concrete pipe: internal diameter, 24 in (0.61 m); wall thickness, 3 in (7.6 cm);
length, 8 ft (2.44 m)
A section through one well is shown in Figure 34a.
A reasonable value of electrode-earth resistance was considered to be
0.1 to 0.2 Q.
A formula was derived for calculating the resistance from a single electrode
element to ground. It considers as concentric cylinders the graphite element,
the seawater between this element and the perforated concrete container, and
the shingle saturated with seawater between the outer surface of the concrete
container and the surface of the excavation in the beach rock. These cylinders
are resistances in series. The last component of resistance is between the
surface of the excavation and remote ground; this is treated as in Section
9-2, page 402, with the sea as a spherical wedge and the excavation as hemi
spherical. The result is
Re
2id
p! In
+ p2 In + p3 In + 271T3
(91)
472
___
GROUND RETURN
9-8
Electrode line
terminal poles
Switch house
,Final beach level
.it cell
,
+1.75' highest high-water levels.
VSShT"'' ii2/S0Uidef layet
Junction pit
4' G.E.W. pipes set
in concrete
58'- 7"
/ foi M fig|
jH-
f 9'-0"
12 -6
9'- 8"
=srl4'-0"
datum-VX-
0'
"""
-
-5.7' lowest low-water level
Low-water _
,wr~
1""
120'-0" approximately
mar
(a)
665'- 0'
Switch house
)" concrete pipes
1' G.E.W. pipes
Approximate L-. ;
_
high-water s
Jine_-_L__.
Junction pits
T
V T TT
-ir-6-(y-6-i4-<5-6-4-i-
"""""IJmits of trench
in rock
Fence
Approximate
low-water
line
(b)
Fig. 34. (o) Section and (b) site plan of shore electrode of the New Zealand scheme.
(From Ref. 55, N.Z. Engineering, by permission.)
where / = length of electrode unit = 2.13 m

r = radius of electrode unit = 0.076 m
rx = inner radius of concrete pipe = 0.305 m
r2 = outer radius of concrete pipe = 0.382 m
r3 = radius of excavation = 3.66 m
Pi = resistivity of seawater = 0.25 2 m
p2 = resistivity of perforated wall of concrete pipe = 4.0 2 m
p3 = resistivity of shingle saturated with seawater = 10 2 m
p4 = resistivity of rock = 1000 2 m
as defined by Eq. (24), using px and p4 and a 0.026 rad
7]
The computed result was 1.9 2. If there were no interaction between elements,
13 elements would be needed for attaining a resistance of 0.15 2. The inter-
473
action may be estimated from Figure 3.2, page 79, in Ref. 16III, which indi
cates that 24 elements are required. The number chosen was 25, and the
spacing between adjacent elements was 25 ft (7.6 m), except for the last two
spaces at each end, which were reduced to 20 and 15 ft (6.1 and 4.6 m).
The distance between centers of the end elements is thus 570 ft (173 m).
A plan of the site is given in Figure 34b.
Construction. Beach material was pushed toward the sea to form a coffer
dam 10 ft (3 m) high. A trench was excavated in the bedrock, and two fullsized test cells were constructed. A small concrete pad for each cell was cast
on the floor of the excavation, and the perforated concrete pipes were set on
the pads. The trench was filled to the top of these pipes with large boulders.
Concrete slabs 18 in (46 cm) thick and 18 ft (5.5 m) in outer diameter were
cast around the top of the pipes. Their purposes were to prevent sideward
movement of the pipes and support inspection pipes of 4-ft (1.2-m) diameter.
After the latter pipes were placed, the rest of the trench was filled with beach
material. Graphite electrode elements were hung in the cells, and their leads
were brought out to two temporary junction pits. Tests were run with 100 A
for 12 h between the beach electrodes in parallel and secondary electrodes
placed in the sea. The resistance was measured continuously and was found
to vary 10%with the tide. The highest voltage gradient measured was 6.2 V
in 3 ft (6.8 V/m). The current division between the two cells was never worse
than 45 A : 55 A.
On the basis of these tests, it was decided to proceed with construction of
the remaining 23 cells. Construction proceeded from July to November,
1964. A severe storm with 9-ft waves breached the coffer dam and flooded
250 ft (76 m) of open trench, but this did not seriously delay the construction.
Leads from each cell were brought to a concrete-block switch house,
where the electrode line terminated at copper busbars. An isolating switch
was provided for each two cells.
At the end of the job, the beach was restored to its normal level.
Final Tests. The measured resistance varied from 0.23 to 0.30 2, according
to the tide. This value is 1.5 to 2.0 times the design value; The maximum
voltage gradient was 7 V in 3 ft (7.7 V/m) and occurred on the seaward side
of the trench line. This indicates that most of the current flowed in the upper
layers of shingle. The two outer cells carry about 30% more current than do
the adjacent cells.
Cost. The cost of the electrode, including investigation, was approximately
57,000 ($160,000).
Life. After 22 months' service with 6 x 10~6 Ah usage as anode, the elec
trode elements were found reduced in volume by about 25 %, partly due to
corrosion by sea water. Additional useful life of 10 to 15 years was estimated.
1
474
9-8
GROUND RETURN
The Ocean Electrode at Los Angeles79
-This electrode serves the southern terminus (Sylmar) of the Pacific North
west-Southwest dc intertie. It is the first major dc electrode to be placed in
the open ocean rather than on a beach or in a sheltered cove. Its rated current
is 1.8 kA. It can serve intermittently as either anode or cathode.
Because the entire Los Angeles area is underlain with many pipelines and
other substructures, it was difficult to find a suitable site for a land electrode
within a reasonable distance of the converter station at Sylmar without
causing serious problems of electrolytic corrosion. It was, therefore, decided
to place the electrode in the Pacific Ocean, in spite of the fairly long distance
(30 mi or 47 km) from Sylmar, so as to reduce greatly current densities on the
land in that area.
The site chosen is 6000 ft (1.8 km) offshore in Santa Monica Bay. This is
not only the closest part of the ocean to Sylmar but it is also at a considerable
distance from Los Angeles harbor, from the steam power plants on the
ocean shore, and from oil pipelines in the ocean. Another factor in the choice
of the site is that the electrode line could be run much of the way from
Sylmar (22 mi or 35 km) to the shore on towers of existing ac transmission
lines. This part of the electrode line consists of two 1272-mcm (646 mm2)
ACSR conductors, taking the place of the usual overhead ground wires and
insulated by two suspension insulators. The remaining 8 mi (13 km) to the
shore is by two 1250-mcm copper-conductor lead-sheathed 15-kV cables run
in duct banks. These terminate at a bus in a vault near the beach, where six
underwater conductors to the electrode also terminate. The resistance of the
electrode line is about 1.1
The electrode consists of a linear array (Figure 35) of 24 elements resting
on the bottom of the ocean, which is at least 40 ft (12 m) deep. At this depth
the effect of wave motion is expected to be very weak. The spacing between
elements is 25 ft (7.6 m) except near the ends, where the spacing is reduced to
23, 20, 16 ft (7.0, 6.1, 4.9 m). The overall length of the array between centers
of the end elements is 543 ft (166 m). Each element has two horizontal metal
rods, one above the other, in a concrete box. The lower rod is about 1.5 ft
(46 cm), and the upper rodis 3.5 ft (107 cm) above the ocean floor.
The rods are made of Durichlor 51, a high-silicon iron alloy (14.5% silicon,
4.5% chromium, 0.9% carbon, and 0.65% manganese) manufactured by the
Duriron Company for use as anodes in salt water. Each rod is 5 ft (1.52m) in
length and 3 in (76 mm) in diameter. The estimated life of these rods is 85
yr under an operating cycle of 54 A, each polarity half of the time, constantly
except for 12 h/yr at 1800 A.
Each pair of rods is enclosed in a rectangular box (Figure 36) 11 by 7 by
5.17 ft high (3.4 by 2.1 by 1.6m)made of unreinforced concrete and weighing
l6'-0"
25'- 0" \
-20'- 0"
1
V /
\,
17 at 25'-0" = 425'- 0 \
20'- 0"N
23'-0'\
543'- 0"
,23'-(r
2A/
shields s.
'*|
(Latitude 34" 01' 20"

j Longitude 118 33' 18"
?7 ?8 1V?10 ?11 ?12 ?13 Tl4 ?i5 jlsTuligTiaO
475
,16'- 0"
25' 0"
a-n-THfi-
( Unlighted
buoy
Untighted J
buoy
Three-conductor undersea cable

Fig. 35. Plan of Los Angeles ocean electrode. (From Ref. 79 by permission of authors.)
Top section
Center
sections
Lead wire
Bottom section
Fig. 36. Detail of electrode rods and electrode shield, Los Angeles ocean electrode.
(From Ref. 79 by permission of authors.)
ssppr
476
PROBLEMS
GROUND RETURN
9 tons (8.2 tonnes). Each box is made in four sections, separated by horizontal
tongue-and-groove joints. The box has in its bottom, top, and sides a total of
45 screened circular holes of 13.6-in (34.6-cm) diameter, having a total area
of 0.377 m2. The rods are suspended from polypropylene rope slings attached
to the top section of the box. The boxes serve (a) to prevent access of fish to
regions of high voltage gradient, (b) to protect the rods from underwater
turbulence and from the movement of silt, and (c) to decrease electrode-toground resistance by providing separation of the rods from the ocean bottom.
Maximum calculated gradient outside the box is 2.5 V/m in the first few
centimeters from the box.
The continuous rating of the electrode is 1410 A, set by thermal limitation
of the underwater cable. The electrode can carry 1800 A for 14 h without
exceeding this thermal limitation.
Two three-conductor 300-mcm (152 mm2) 5-kV copper cables connect
the electrode to the vault on the shore. Each conductor is spliced to
four electrode elements. The leads from each element are 2/0 AWG
(67.5 mm2) copper with polyethylene insulation.
The calculated resistance of the electrode to earth is of the order of 0.01
to 0.02 2. The electrode can operate satisfactorily with less than 0.02 2 resis
tance even if one of the three-conductor cables should be cut; for example, by
a dragging anchor of a small boat. Large ships are not expected to anchor
in this area.
Shore Electrode in Mud Flats
A design for an electrode for 1.2 kA situated in shallow seawater (4 m

deep at low tide) was published by Quinn and Howard.58 Graphite rods
were to be suspended in a circle of 10-m diameter from wooden or concrete
piles. The piles would be driven in a circle of 15.2-m diameter concentric with
the circle of electrodes and would support a polypropylene net for excluding
small fish. The voltage gradient at the net was calculated at 1.25 V/m. The
electrode-to-ground resistance was estimated as 0.042 2.
PROBLEMS
1. Compare the full-load losses of a bipolar line having a resistance of

18 2 per conductor and a current of 1.8 kA per conductor with that of a
homopolar line using the same conductors and same current and having
ground-electrode resistances of 0.05 2 each.
2. Compare the efficiency of the line in Problem 1during bipolar operation
477
with that during monopolar operation (a) with ground return and (b)
with metallic return using the other conductor. Voltage to ground is
400 kV at the sending end.
3. Derive an equation for the conductance per unit length from a buried
pipe to a parallel pipe of different size. Use the method of images.
4. Prove that, in the current field of a shore electrode (Section 9-2, page
402), the apparent resistivity measured at the surface of the sea is equal
to that measured on land.
5. Find an expression for the apparent resistivity of the sea when a shore
electrode is used and the land and sea bottom have infinite resistivity.
The assumptions made in the text should be used.
6. Prove that the expression for reflection coefficient in Eq. (38) is correct.
Find the relationship between the angle of arrival and the angle of
refraction of a ray of current passing from a medium of resistivity pt
through a plane boundary to a medium of resistivity pz . The arriving
ray is defined here as the resultant of the incident and reflected rays
in medium 1.
7. Check Rusck's calculation on rate of loss of iron at JL = 10 mA/m2
(page 427).
8. Compute the minimum acceptable spacing between rectifiers and the
current and voltage to be furnished by each rectifier in order to keep the
potential of a long, insulated pipeline between the limits of 0.85 and
1.20 V. Assume that the soil is chemically uniform and has a uniform
resistivity of 100 2 m. Neglect foreign currents. The following data
pertain to the pipe:
Outer diameter, 3.0 ft (0.915 m)
Wall thickness, 0.438 in (1.11 cm)
Resistivity of pipe, 2.1 x 10 ~7 2 m
Coating, felt and asphalt, leakage resistance of unit area, 6 x 104
2 ft2 (5.6 x 103 2 m2)
Depth of burial (top of pipe), 3 ft (0.9 m)
9. Plot curves of pipe potential, longitudinal current, and leakage current
per unit of length foT the insulated pipe in Problem 8.
10. Repeat Problem 8 for an uncoated pipe otherwise the same.
11. Find the decrease of conductance from electrode to ground for each arm
added to a star-shaped electrode.Assume each arm to be a cylinder 500 m
long and 0.5 m in diameter. Assume soil resistivity to be 100 2 m.
12. Calculate the power loss in the coke of the electrode designed by
Rusck (Section 9-7, page 459).
478
GROUND RETURN
BIBLIOGRAPHY
479
16. Earth Conduction Effects in Transmission Systems, by Erling D. Sunde, Van Nostrand,
New York, 1949.
Chapter II, "Earth Resistivity Testing and Analysis," pp. 38-63.
Chapter III, "Resistance of Grounding Arrangements," pp. 66-97.
Chapter VI, "D-C Earth Conduction and Corrosion Protection," pp. 177-214.
17. . Transient Performance of Electric Power Systems, by Reinhold Riidenberg,
McGraw-Hill, 1950, especially Chapters 24 and 25, pp. 310-339. Step voltage,
multiple electrodes, buried electrode, method of images, resistances to ground, and
heating of ground.
18. "Les Bases Generale de la Technique des Mises a la Terre dans les Installations
filectriques (General Technical Principles of Grounding Electric . Plant)," by P.-G.
.
Laurent, Revue Generale d"Electricity, 7th series, Vol. 1, No. 7, pp. 368-402, July,
1951. In French. Excellent general treatise on principles and methods of grounding,
including purposes and characteristics of operating and safety grounds, electrical
properties of soils, ground electrodes, hazards to life, and heating of the soil. English
translation of all except part VI, "Heating of Soil," published as Appendix IV of
Ref. 39.
BIBLIOGRAPHY
1. A Method of Measuring Earth Resistivity, by F. Wenner, Report 258, Bulletin of the
U.S. Bureau of Standards, Vol. 12, No. 3, pp. 469-482, Feb. 3, 1916.
2. Erdstrome (Earth Currents), by Franz Ollendorff, Julius Springer, Berlin, 1928. In
German. Reprinted by University Microfilms, Ann Arbor, Mich., 1962. 260 pp.
3. " Calculation of Resistances to Ground," by H. B. Dwight, Elec. Eng., Vol. 55, pp.
1319-1328, December 1936.
4. " Effect of Electric Shock on the Heart," by L. P. Ferris, B. G. King, P. W. Spence, and
H. B. Williams," A.I.E.E. Trans., Vol. 55, pp. 498-515, 1263, May 1936. Tests at
Columbia University on the effect of shock on heart action of anesthetized animals
(guinea pig, rabbit, cat, dog, pig, sheep, and calf). Current causing fibrillation increases
with body weight and heart weight. For calf (70 kg) 3-sec 60-Hz minimum fibrillation
current is 0.21 to 0.47 A; for sheep (56 kg), 0.16 to 0.39 A; sheep, 3-sec dc, 0.75 to
2.1 A.
5. " Practical Aspects of Earthing," by E. Fawssett, H. W. Grimmitt, G. F. Shotter, and
-H. G. Taylor, I.E.E. Jour., Vol. 87, pp. 357-390, October 1940. Disc., pp. 390-400.
Comprehensive article.
6. "Grounding Principles and Practice: I, Fundamental Considerations on Ground
Current," by Reinhold Riidenberg, Elec. Eng., Vol. 64, pp. 1-13, January 1945.
General principles of ground currents. Formulas for resistance of various forms of
ground electrode. Republished as Appendix III of Ref. 39.
7. Tables of Functions with Formulae and Curves, by Eugene Jahnke and Fritz Emde,
4th ed., 1945, Dover Publications, New York. Struve function So, p. 218; N0, p. 190.
8. "The Experimental Sending of D.C. through the Earth in Sweden," by R. Lundholm,
C.I.G.R.E., 1946, Report 134. 12 pp. Four tests in which direct current was circulated
between pairs of earth or sea electrodes and currents were measured in railroad rails
and in sheaths of telephone cables.
9. "Soil and Water Resistivity in Finland," by E. K. Saraoja, C.I.G.R.E., 1946, Report
201, 4 pp. Average measured resistivities of soil and water, including values for frozen
19. "Return Current through the Earth for High-voltage D.C. Transmission," by
R. Lundholm, Direct Current, Vol. 1, pp. 79-86, March 1953. Experiments in Sweden
with direct current circulated through ac transmission lines. Effects on railway signals
and currents in rails and cable sheaths.
20. "Direct Current Track Circuits for Railways," by D. G. Shipp, Direct Current,
Vol. 1, pp. 118-121, June 1953.
21. "The Magnetic Compass Error Caused by D.C. Single-core Sea Cables," by F. Busemann,E.R.A. report B/T 116, May 1953. Also in Direct Current, Vol. 1, pp. 208-213.
"A Study of the Hazards of Impulse Currents," by Charles F. Dalziel, A.I.E.E. Trans.,
Vol. 72, Part 3, pp. 1032-1041, October 1953. Disc., pp. 1041-1043. Direct-current
resistance of body from hand to both feet in 3/4 in of saltwater is 1135 Ci, minimum
of nine subjects.
22.
23. "Cathodic Protection: Some Basic Considerations," by A. C. Vivian and J. S.

Gerrard, Direct Current, Vol. 2, pp. 138-142, September 1955.
soil and water.
24. "Efficiency of Grounding Grids with Nonuniform Soil," by J. Zaborsky, A.I.E.E.

Trans., Vol. 74, Part 3, pp. 1230-1233, December 1955. In Appendix 2 is derived a
formula for the resistance of a ring-shaped electrode buried in the upper layer of a
two-layer earth.
10. "Dangerous Electric Currents," by Charles F. Dalziel, A.I.E.E. Trans., Vol. 65, pp.
579-585, August-September 1946. Disc., pp. 1123-1124.
11. "Applications of Electro-osmosis to Soil Drainage," by A. H. D. Marwick and A. F.
Dobson, Engineering, Vol. 163, No. 4230, pp. 121-123, Feb. 7, 1947.
12. " Cathodic Protection of Buried Metallic Structures," by R. de Brouwer, C.I.G.R.E.,
1948, Report 205.
13. "D.C. Transmission with Return Current through Earth," by C. E. Soderbaum,
I. Beckius, M. Bockman, and R. Lundholm, C.I.G.R.E., 1948, Report 401, 24 pp.
14. "Comparative Properties of Grounding Electrodes," by Reinhold Riidenberg, Elec.
World, Vol. 129, No. 5, p. 72, Jan. 31, 1948. Brief comparison of surface, volume,
current capacity, and gradient of electrodes of various shapes having equal ground
25. "Let-go Currents and Voltages," by Charles F. Dalziel and F. P. Massaglia, A.I.E.E.
Trans., Vol. 75, Part 2, pp. 49-55, May 1956. Disc., pp. 55-56. 60-Hz let-go currents
for 99.5% of 134 men, 9 mA; of 28 women, 6 mA. For dc exactly five times as great.
Mean dc release current for 28 men was 74 mA; value for 99.5% of these men, 0.82
x 74 mA. Safe voltages for wet hands and feet, 10.2 V 60 Hz, 51 V dc.
26.
Underground Corrosion, by Melvin Romanoff, U.S. National Bureau of Standards,

Circular 579, April 1957, U.S. Govt. Printing Office, Washington, D.C., iv + 227 pp.
27. "Electric Shock Hazard Analysis," by K. S. Geiges, A.I.E.E. Trans., Vol. 75, pp.
1329-1331, February 1957. Disc., p. 1331. Perception current, 0.1 to 0.2 mA; muscular
contraction, 6 to 20 mA; fibrillation of heart, 100 mA; these are for ac. Tolerable dc
somewhat higher.
resistance.
15. The Corrosion Handbook, edited by Herbert H. Uhlig and sponsored by the Electro
chemical Society, Wiley, New York, 1948, 1188 pp.
T|
i
480
GROUND RETURN
28. "Telephone Interference and Other Effects Caused by the Gotland H.V. D.C. Transmission," by G. A. Pettersson, L. Ahlgren, and H. Forsell, C.I.G.R.E., 1958, Report
324. Corrosion and compass variations are discussed.
29. "Hazards of Electric Shock at Stations during Fault and Method of Reduction,"
by A. Elek, Ontario Hydro Research News, Vol. 10, No. 1, pp. 1-6, January-March
1958. Includes use of crushed stone on ground surface and grounding of fences
around station.
30. "Corrosion Problems Arising from High Voltage D.C. Transmission with Earth
Return," by J. H. Gosden, Appendix V to C.I.G.R.E. Report 417, 1960, pp. 24-28.
Also in Direct Current, Vol. 4, pp. 159-161, September 1959. Disc., p. 161.
31. "Investigation of Some Problems Concerning the Earth-Return H.V.D.C. System,
Kashira-Moscow Line" (in Russian), by S. A. Bazhenov, V. P. Pimenov, and M. R.
Sonin, Elek. Stantsii, Vol. 30, No. 9, pp. 54-59, September 1959.
32. "EarthReturn Effects," by A. Stalewski, Direct Current, Vol. 4, pp. 155-158, Sep
tember 1959. Disc., p. 161. Report to C.I.G.R.E. Study Committee 10, June 1959.
33. "Examples of Geoelectric Surveys," by L. S. Palmer, l.E.E. Proc., Vol. 106A, pp.
231-241, June 1959. Disc., pp. 241-244.
34. "Alternating-Current Shocks of Varying Parameters Affecting the Heart," by W. B.
Kouwenhoven, G. G. Knickerbocker, R. W. Chestnut, W. R. Milnor, and D. J. Sass,
A.I.E.E. Trans., Vol. 78, Part 1, pp. 163-167, May 1959. Disc., pp. 167-169. Experiments on anesthetized dogs with 60-Hz shocks of various durations from 1/120 to
5 sec. Defibrillation by countershock.
35. "Use of Earth and/or Sea Return," Chapter 9 of High Voltage Direct Current Power
Transmission, by Colin Adamson and N. G. Hingorani, Garraway, London, 1960,
pp. 133-142.
36. Recommendations for the Protection of Underground Cables against Corrosion,
C.C.I.T.T., New Delhi, 1960.
37. " Soil Thermal Characteristics in Relation to Underground Power Cables," A.I.E.E.
Committee Report, A.I.E.E. Trans., Vol. 79, Part 3, pp. 792-844, December 1960.
Disc., pp. 844-856. Following is a list of the subdivisions of the report and the respec
tive authors:
I. "Problem and Objective," by R. J. Wiseman and R. W. Burrell, pp. 792-795.

II. "Soil Types: Identification and Physical Properties," by William A. Del Mar,
R. W. Burrell, and C. A. Bauer, pp. 795-803.
III. "Soil Moisture Characteristics," by L. H. Fink, pp. 803-819.
IV. "Soil Thermal Resistivity: Typical Field Values and Calculating Formulas," by
W. A. Sinclair, F. H. Buller, and C. B. Benham, pp. 820-832.
V. "Practical Application: Trench Design and Construction," by A. S. Brookes
and H. H. McGrath, pp. 832-836.
VI. "Measurement Techniques," by A. W. W. Cameron and A. S. Brookes, pp.
836-844.
38. "Threshold 60-Cycle Fibrillating Currents," by Charles F. Dalziel, A.I.E.E. Trans.,
Vol. 79, Part 3, pp. 667-673, October 1960. Disc., p. 673. Fibrillation currents follow
normal distribution curve (straight line on probability graph paper). Fibrillation
currents of various animals are plotted against their body weights : 70-kg values are
assumed to apply to man.
39. Guide for Safety in Alternating-Current Substation Grounding, A.I.E.E., Publication 80,
March 1961, 136 pp. A comprehensive treatise on grounding for safety. Includes
reprints of four articles on this subject and an annotated bibliography of 78 items.
'
BIBLIOGRAPHY
481
40. "H.V.D.C. Power Transmission: Problems Relating to Earth Return," by S. Rusck,

Direct Current, Vol. 7, pp. 290-298 and 300, November 1962. Risk of corrosion in
field of electrode on sea shore and design of land electrode.
41. "Preliminary Evaluation of Station Grounding Conditions from Soil Resistivity
Tests," by J. Endrenyi, Ontario Hydro Research News, Vol. 14, No. 2, pp. 15-22, 1962.
Effect of layers of different resistivities on resistance from electrode to ground.
42. "Soil Stabilisation by Electro-osmosis," by N. W. Lister, Elec. Rev., Vol. 171, pp.
463-467, Sept. 21, 1962. Application to excavations.
43. "Electrochemical Corrosion by Direct Current," by W. G. von Baeckmann, Direct
Current, Vol. 8, pp. 81-86, March 1963. Elementary discussion of protection of
buried metal from corrosion.
44. Corrosion and Corrosion Control, by Herbert H. Uhlig, Wiley, New York, 1963.
Chapter 12, "Cathodic Protection," pp. 182-197.
45. Earth Resistances, by G. F. Tagg, George Newnes, London, 1964.
46. "More on Coating Damage," by Marshall E. Parker, Gas, Vol. 40, pp. 17, 129,
June 1964.
47. "Ground Current in HVDC Transmission I, Preliminary Report [on] Columbia
Basin Tests," by T. Cantwell, P. Nelson, J. E. Webb, Arnold Orange, A. L. Kinyon,
R. F. Stevens, and C. L. Waugh, I.E.E.E. International Convention Record, New
York, Mar. 22-26, 1965, Part 9 (Power), pp. 1-9. Also WESCON, August 1964 and
I.E.E.E. Trans, on P.A.&S., Vol. 85, pp. 240-247, March 1966.
48. "Direct Current Ground Electrode Behavior," by Allen L. Kinyon, ibid., pp. 10-16.
Measurements of ground-circuit resistance as functions of time and conclusions on
permissible current densities at surface of electrodes for 6-h operation.
J
j
j
j
(
|
|
\
I
j
j
4
;
49. " Investigations of Ground Current Phenomena Associated with HighVoltage Direct
Current Transmission," by Fred E. Cornwell, ibid., pp. 17-22. Also in I.E.E.E. Trans.
on P.A.&S., Vol. 85, pp. 254-258, March 1966. Disc., pp. 258-260. Also presented at
I.E.E.E. WESCON, August 1964. Tests in Sacramento Valley of California on poten
tial gradients, transformer neutral currents, and pipe-to-soil potentials. One current
electrode was in a deep well.
50. " Corrosion Aspects of HVDC on Buried Telephone Plant," by L. E. Fiorito, ibid.,
pp. 23-30. Also in I.E.E.E. Trans, on P.A.&S., Vol. 87, pp. 35-40, January 1968.
Formulas are derived and illustrated for leakage current on a cable sheath near
current electrode and for minimum separation of cable from electrode for acceptable
rate of corrosion.
51. "Electrodes for EHV D-C Transmission in a Metropolitan Area," by S. G. Pann,
A. A. Galindo, and G. R. Elder, ibid., pp. 31-36. Calculations and tests in the Los
Angeles area indicated that a sea electrode should be used.
52. "H.V.D.C. Transmission: Earth Electrode Arrangements," by Anders Kohler,
Direct Current, Vol. 10, pp. 18-24, February 1965.
53. The Influence of H.V.D.C. Transmission on General Utility Operations, a report prepared for the Oregon Public Utility Commissioner by Cornell, Howland, Hayes, and
Merryfield, Corvallis, Oregon, 1965.
54. "The Benmore Land Electrode," by D. G. Dell, New Zealand Engineering, Vol. 20,
pp. 165-175, May 15, 1965.
55. " The North Island Sea Electrode," by D. G. Dell, ibid., pp. 213-222, June 1965.
56. "Practical Applications of Electro-osmosis," by J. K. Gladwell, New Zealand
Engineering, Vol. 20, No. 2, pp. 66-72, 1965.
:
j
J
j
ij
I
j
I
482
57.
" The Benmore Land Electrode for the Benmore-Haywards H.V.D.C. Transmission
Scheme," by D. G. Dell, I.E.E. Conference Publication 22, HVDC Transmission,
58. "Earth Electrodes," by P. Quinn and P. E. Howard, ibid., Paper No. 83, pp. 419-421.
Proposed designs of electrode in mud flat and of one in an arid region near a reservoir.
59. "Some Basic Concepts of HVDC Ground Return Currents," by E. V. Lindseth and
F. E. Cornwell, ibid., Paper No. 84, pp. 422-426. Potential gradients around Tracy
and Mead; transformer neutral currents; transients in tests with pulses.
60. "The North Island Sea Electrode for the Benmore-Haywards HVDC Transmission
Scheme," by D. G. Dell, ibid., Paper No. 85, pp. 427-430.
61. "Design of the Sea Electrode System, Sardinia-Italian Mainland 200 kV Scheme,"
by D. S. Thorp and D. MacGregor, ibid., Paper No, 86, pp. 431-434.
62. "Experimental Study of the Effects of Using Sea-Return," by L. Csuros and J. M.

63.
64.
65.
66.
Cranmer, ibid., Paper No. 87, pp. 435-438. Experiments in the Solent between England
and the Isle of Wight.
"A Study of the Electric Field in the Mediterranean Sea," by F. M. Cahen and
M. H. Macaire, ibid., Paper No. 88, pp. 439-441. Model and digital-computer studies
of field associated with Sardinia-Italy link.
Discussions of foregoing papers by R. F. Stevens, F. E. Cornwell, H. R. Gunn, and
D. MacGregor, ibid., Part 2, pp. 111-114.
"The Anodic Earth Electrode for the Konti-Skan H.V.D.C. Link," by E. Anderson
and N. R. Neilson, Direct Current, Vol. 11, pp. 54-56 and 58-63, May 1966. Shore
electrode in Denmark.
Statistical Study of Electric Fields from Earth-Return Tests in the Western States and
Comparison with Natural Electric Fields, by George V. Keller, report prepared for
the Bonneville Power Administration by the Colorado School of Mines Foundation,

Golden, Colo., June 1, 1966. See also Ref. 77.
67. ElectricalMethods in GeophysicalProspecting, by G. V. Keller and F. C. Frischknecht,
Pergamon, Oxford, 1966, 527 pp.
68. " Railway Signal Systems," by J. R. DePriest, Paper No. I,A Symposium on Effects
of HVDC Transmission on Railway Operation, Communication and Signal Section,
Ass'n. of American Railroads, Denver, Colo., Nov. 15-16, 1966.
69. "Track Circuit Fundamentals," by C. E. Staples, Paper No. II, ibid.
70. " Theoretical Considerations of the Effect of a D-C Earth Voltage Gradient on Neutral
D-C Track Circuits," by H. C. Kendall and J. H. Auer, Jr., Paper No. Ill, ibid.
71.
BIBLIOGRAPHY
GROUND RETURN
"The Challenge of Instrumentation and Measurement of Current Flow in the Earth

and Its Impact on the Energy Transmission Industries," by C. John Essel, Paper
No. XII, ibid.
72. "The Effects of High Voltage Direct Current Transmission on Buried Metallic Pipelines," by John Van Bladeren, Paper No. X, ibid.
73. " Corrosive Effects of HVDC Transmission," by Fred E. Cornwell and T. E. Backstrom, Materials Protection, Vol. 5, pp. 8-13, February 1966. Disc., pp. 13-14.
74. "Electrodes for EHV DC Transmission in a Metropolitan Area," by S. G. Pann,

A. A. Galindo, and G. R. Elder, l.E.E.E. Trans, on P.A.&S., Vol. 85, pp. 260-264,
March 1966. Preliminary ground-current tests leading to the decision to build an
ocean electrode.
f
]
j
:
\
\
{
\
|
I
?
j
j
|
=
j
|
j
f
1
I
\
j
{
|
|
f
ji
|
|
I
I
I
|
483
75. Ground Current Behavior as Related to Buried Pipelines, by Ira C. Bechtold, Richard
H. Jahns, Robert V. Langmuir, and William R. Smythe, a report prepared for Pacific
Lighting Service and Supply Co., Southern California Gas Co., and Southern Counties
Gas Co., Apr. 15, 1967.
76. " Electrophysical Study of the Soil for the North End DC Grounding Electrode of the
Northwest-Southwest Intertie," by A. L. Kinyon, l.E.E.E. Publication 31C46, pp.
52-64, January 1967.
77. "Statistical Study of Electric Fields from Earth-return Tests in the Western States
and Comparison with Natural Electric Fields," by George V. Keller, l.E.E.E. Publi
cation 31C46, pp. 35-51, January 1967. Also in I.E.E.E. Trans, on P.A.&S., Vol. 87,
pp. 1050-1057, April 1968. Similar to Ref. 66.
78. " The Rice Flats DC Ground Electrode for the Pacific Northwest-Southwest HVDC
Intertie," by Carleton L. Waugh, Proc. Western Water and Power Symposium, Los
Angeles, Apr. 8-9, 1968, pp. D-53-60.
79. "The Los Angeles HVDC Ocean Electrode," by G. R. Elder and D. B. Whitneyibid., pp. D-61-68.
80. "Major Features of the Vancouver Island 260-kV HVDC Submarine Link," by
H. M. Ellis and W. Chin, Proc. American Power Conf, Vol. 30, pp. 1017-1034,
April 1968; copyright, 1969.
81. "A Survey of Corrosion Aspects Related to the Operation of Electrodes for HVDC
Ground Return," by Ingvar Lid6n and Heine M&rtensson, a report to C.I.G.R.E.
Study Committee 36 (Interference), Volgograd, 1969.
82. Appendix of " Electric Power Transmission by H.V.D.C. Submarine Cable across
the Adriatic Sea from Yugoslavia to Southern Italy," by M. Visentini, A. Asta, and
F. Trimani, C.I.G.R.E., 1958, Report 210, pp. 13-22. Compass error.
83. " Thermal Analysis of Earthing Rods at the Ends ofD. C. Lines," by K.R.Rakhimov,
Electric Technology U.S.S.R., Vol. 4, pp. 36-43, July 1970. Translated from
Elektrichestro, No. 10, pp. 23-26, October 1969.
485
EFFECTIVE VALUE OF ALTERNATING CURRENT OF A SIX-PULSE CONVERTER
III ! IV
III
Appendix A
Effective Value of Alternating Current of

a Six-pulse Converter
Figure A1 shows the wave shapes of one-half cycle of a line-to-neutral

voltage and of the corresponding line current on the valve side of the converter
transformer for single overlap, that is, for overlap angle not exceeding 60.
The origin of the time angle 6 = cot is at the positive crest of the voltage wave.
The next half cycle is the same except that the instantaneous values of voltage
and current are negative instead of positive.
The instantaneous voltage is given by
e = Em cos 0
(Al)
and the instantaneous current of each segment of the wave by
IS2 COS 0
>~L0t
Fig. Al. Positive half-waves of (a) alternating line current and (6) alternating line-toneutral voltage of a six-pulse converter.
with omission of iIV , which obviously makes no contribution to the integral.

Each integral is now evaluated in turn, after which the other indicated
operations are performed.
Segment I
Seg
ment
+<- u
AS 60
Equation
Limits
I a - 60 < 9 < 5 - 60
II 5 60 < 9 < a
+ 60
III a + 60 < 9 < 5
+ 60
a - cos (i9 + 60)]
(A2)
ia = Id = /s2(cos a cos 5)
im = 7s2[cos (0 - 60) - cos 5]
(A3)
i, = /s2[cos
IV 5 + 60 <9 <a + 120
nI
(A4)
The effective (or root-mean-square) value of any periodic current (period

2n) is
0m]2d9j
Hi
LVJa-eO9
AS +60
Ax +60
J (5-60
484
2dd+
te/j2 = Is22
J a+60
(A6)
\-U/2
hn2 dQ J
/J
(A7)
[cos a cos (9 + 60)]2 d9
(A8)
By the substitution 9' -6 + 60, dB' dO, which displaces the origin hori
zontally, this becomes
W2
For any wave such as this in which the two successive half cycles are alike
except for sign, each 2n in Eq. (A6) may be replaced by x. For the wave in
question, the integral may be divided into segments, thus :
n / AS-60 d0
a 60
(A5)
=0
~I
s
i2
f (cos
cos 9')2 dB'
j
Isi2[*' cos2
422 cos2
(cos2 a
/s22
(<5 a)
<5 a
+-
2 cos a cos 9' -f cos2 &') dB'
9' sin 26
2 cos a sin B + +
a 2 cos a(sin <5
sin 25 sin 2a
4
sin a)
(A9)
By use of the auxiliary angles,
<5
a =u
and
<5 4- a = X
(A10)
486
and the trigonometric identity Dwight,1 page 73, Eq. (401.09)

sin A sin B 2 sin %(A
and by Dwight, page 76, Eq. (404.22),
B) cos %{A + B)
cos2 A
(A11)
we obtain
>r
= hi u cos
idi
a+~
2 cos a
2 sin ~ cos
A*
-I- (sin u cos 2)

A*
cos2 a + cos2 b = cos X cos u + 1
f*<5+60
2n
2n
d9 = Id2(a-b+)=Id2{-u)
~3
3
5-60
(A13)
[cos (0 60) cos b]2 d9
jw(2 + cos X cos u) + cos X sin u 8 cos
= IS22
Sn
* sin u
2 cos <5(sin <5
Id2
sin a) + u cos2 <5
/i
(A14)
- 4(cos a + cos 5) sin - cos
c.
(A15)
=
By the identity Dwight, page 73, Eq. (401.10)
cos a
T2n
-J
u(2 + cos X cos u) sin u(2 + cos 2)

rr,;
(cos a cos b)
u{2
|_ 3
/I
Ad
+ cos <5 = 2 cos - cos -
(A16)
(A17)
/ A
(A22)
21
d
2n
+ cos 2 cos u) sin u(2 + cos 2)~|

(cos a cos b)2
cos 2 cos u) + sin u(2 + cos 2) ,
1T
(cos a cos <5)2
7=7d{f[l-(2,H)]}1/2
where
(cos2 a 2 cos a cos <5 + cos2 b 2 "|
The effective current is
+ cos B = 2 cos \{A +) cos \(A -B)
(A21)
The whole sum is obtained by adding Eq. (A13) for segment II. It is
The partial sum, n(I2 + /m2) is
"(cos2 a + cos2 b) + sin u cos A
hi2tw(2 + cos 2 cos u) sin u(2 + cos 2)]
This expression can be put in terms of 7d instead of Is2 by use of Eq. (A3),
becoming
2 cos b sin 9" + 9" cos2 b
+ \ sin u cos 2 2 cos 5 2 sin 2- cos -2 + u cos2 b

2
cos A
s22["(2 + cos 2 cos u) + cos 2 sin u 2(1 + cos 2) sin u]
= 7s22( -
2s22jw +
cos - sin - = 2
2 cos b cos 9" + cos2 5) dB"
sin 2b sin 2a
and
Cd
hi2i 2 +
sin -
/S22j (cos2 9"
- cos -
cos2 - = (cos 2 + 1)
a+60
(cos 9" cos b)2 dB"
(A20)
The partial sum becomes
But
Cd
/A1A.
(A19)
whence, by Eq. (A16),
Segment III
hi"
(A18)
Ad
7s2
+ 1)
(A12)
I
'a +60
%hll
= (cos 2A
,
cos 2a + cos 2b + 2
-cos2 a + cos2 b = -
Segment II
nla2=ld2
487
F(X,u) = x [last fraction in Eq. (A23)j

2it
'
f k
(A24)
(A25)
'
\
488
The coefficient of u in this fraction is
cos2 a + cos2 <5 2 cos a cos 5 2 cos 2 cos u
= (cos2 a 1) (cos2 5 1) cos (a (5) cos (a 5)

sin2 a sin2 5 cos A cos u cos A cos u
=
cos 25) cos A cos u cos A cos u
= 1 (cos 2a
1 cos A cos u cos A cos u cos A cos u
cos u)
= (1 cos A
+
+
+
cos 2 cos u
2k
3
sin u(2 + cos A) u(1 + cos A

(cos a cos 5)2
+ cos u)
sin u{2 + cos A) u{1 + cos A

sin2 (2/2) sin2 (w/2)
+ cos u)
1.00
0.99
0.98
0.97
0.96
0.95
a>
0.94
0.93
a or 7
0.92
0.91
20
30
15
30s to 90
40
Fig. A2. Effective value of alternating current of a six-pulse converter.
489
The approximate expression for effective alternating current, based on the

assumption of a rectangular pulse of height Idand duration 2itj2 Eq. (28) in
Chapter 3is
Hence
/0=WI
= Vl F(A,u)
l0
(A27)
(A28)
in which F is given by Eq. (A26). A plot of 7//0 versus u forvarious values of a

is presented in Figure A2.
Hence
F{A,u) =
Reference
1. Tables of Integrals and other Mathematical Data, by H. B. Dwight, Macmillan, 1947.
(A26)
491
FUNDAMENTAL CURRENT, POWER, AND REACTIVE POWER

-
ji
Appendix B
V2
/*<5-60
-*s2
J a 60
j- =
[2 cos a - 2 cos (9 + 6O)]/-0 d9
'a +60
Fundamental Current, Power, and

Reactive Power of a Six-pulse Converter
(2 cos a 2 cos 5)/ 6d9
Ia+60[2 cos (9 - 60) - 2 cos 6]/~ 9 d9
6-60
'<5+60
pa+60
2 cos a/ 6d9
1
J a 60
Refer to Figure A1 and Eqs. (Al) to (A5) for the wave shapes of voltage
and current during a half cycle with overlap angle less than 60.
Any periodic quantity may be represented by a Fourier series. The complex
form of such a series is
(*<5+60
<5-60
2 cos b/9d9
'560
if Dh/hB
h= <x>
a 60
(Bl)
(*5+60
+
where
and
Dfi =
f*
F(9)/-h9d0
fx = expO'x) = cos * 4-y'sin a:
(B2)
(B3)
Substitution of 2 cos x
The crest value and phase of the hth harmonic wave are given by 2Dh . The
limits of the integral need not be as shown but may span any range of 2n rad.
If F(9) = F{9 + 7i), as is true for the wave in question, and if h is odd, the
limits may span a range of n rad, in which case the coefficient of the integral is
doubled. Thus the crest value and phase of the fundamental current (h = 1)
are given by
fa+ 120
V2I1==-71 J a -60 i(9)/-9d9
60)/ 9d9
=/x / x gives
(*5+60
(/S - 9 -/-6 - 9) d9
(/a-9-/-a~9)d9~
60
5-60
5-60
'5+60
(/60 / 20 60) d9 +
(J 6Q / 20 + 60)
a+60
a- 60
Integration, by use of the formula
/a bO d9 = /a b9
(B4)
where It is the complex rms value of fundamental current and i{9) is the
instantaneous current. The phase of the current is expressed as advance with
respect to the line-to-neutral voltage. The integral is subdivided into segments
I, II, and IIIin Eqs. (A2) to (A5). Multiplication of each side of Eq. (B4) by
7t//s2 and substitution of these equations give
490
2 cos (9
a+60
(*a+60
a
2 cos (9 + 60)/9d9
gives
'
\a+60
)
f j/a 9 -j/-a - 0/a
\
60
\<5+60
/
- ( j/b - 9 -j/-b - 9 )
/<'a+60
\
\5 60
+ 60
20 + 60
20 - 60 )
+ ( 0/ 60 - 0.5.//0/60 - 0.5//
/a+60
'/a 60 \ J--
jrm
492
Substitution of the limits gives
Substitution of the conjugate of Eq. (B5) for I* and of Vd0nj3 6 for
j/-60 +J/-2ol - 60 -j/60 -j/-2a- +60

j/ 60 28 - 60 +jr/HF +j/-2<5 + 60
-(5 - 60 - a + 60)/60 + (5 +60 - a - 60)/- 60
-0.5//- 2Q5 - 60) - 60 + 0.5j/ 2(a - 60) - 60
+ -5j/~2(d + 60) + 60 ~ 0-5y/-2(a + 60) + 60
2a 60 // 2a + 60 -j/-25 - 60 +.//-2 -f 60
=jI
- u(/60 / 60) - 0.5//- 25 + 60 + 0.5//- 2a + 60
+ 0.5//- 25 - 60 - 0.57/- 2a - 60
= /(/2a / 2<5)(/60 / 60) - m(/60 ~/-60)
+ 0.5/(/ 2a / 2<5)(/60 -/-60)
= (/60_ / 60)[ 0.5/(/ 2a -/-25) - ]
_
p' _|_
= tV-2
25 ~72m)
4tt -"7"
'
. /~2<x ~/~2
cos a cos <5
V3EII* = 3ELNi;
where P = active (average) power, W

Q = reactive power, var
Ell = rms line-to-line voltage, V
ELjV = rms line-to-neutral voltage, V
ij = conjugate of complex fundamental current, A
s*n
wn 2
cos
+7"
2
sin X sin u
P' = - 2
u cos X sin u
--
per unit
Q' =
(B13)
{B14)
(B15)
Per-unit apparent power is

(B7)
Power. The complex power is

P +JQ =
(B12)
j(u sin u/X)
(B6)
In terms of the direct current, /, = /s2(cos a cos 5),

4tc
(Bll)
sin 2a sin 25 + 2u
'
, / 2a / 2<5 /2m
ll=
7"d
-Q'
(B5>
Now, in the first per-unit system, the base alternating current is Is2 6/71, and
hence the per-unit fundamental alternating current is
!i_
--
Alternative expressions in terms of u and X a. + 5 = 2a + u, which may be

found by use of trigonometric identities, are
1,2
cos 2a cos 25
and the imaginary part gives the per-unit reactive power,
V2 = -(/-2 -/-25 -y2)

*s2
(B9)
The real part of Eq. (B10) gives the per-unit active power,
/}
ELNgives
Since base power in the first per-unit system is Vd0Is2 , the per-unit complex
power is
/2a /2<5 +Jjlu
P' -}- jQ = 4r
(B10)
4
Substitute /60 / 60 =j\j3:

7!
yMJ--/-+i2u
p+jQ =
493
(B8)
s' =
V(Oz + (Q'f = u2 + sin2 u~2u sin U cosA
(B16)
The per-unit scalar fundamental alternating current is likewise given by

Eq. (B16). This equation can be interpreted geometrically as a triangle having
angle X enclosed by two sides, one of length u, the other of length sin u; the
third side is 2I[ or 2S' , according to the law of cosines.
Displacement Angle. The angle 4> by which the fundamental current lags
behind the line-to-neutral voltage is given by
tan
r
Q 2u + sin 2a sin 2d
= u csc u csc X cot X
P = cos
2a -cos 2<5
-
(B17)
}
v
SB
INCLUSION OF DIRECT VOLTAGE DROPS
loop. We then put 5 = n y, and hence cos 6 = cos y, and multiply the
resulting equation by 1, obtaining:
Appendix C
(C3)
-Vt= Vd0 cos y-(Rc~ R,)Id + Farc
In Eqs. (CI) and (C3), Vd is negative for inversion (however, Vd0 is always
positive). The sign convention for Vd is sometimes reversed.
Inclusion of Direct Voltage Drops

Due to Resistance and Arcs in
Converter
Equations
j
if
A part, usually a small part, of the internal voltage drop of a converter is

caused by internal resistance, especially that of the transformer windings (both
primary and secondary). Another small part, usually negligible in hv con
verters, is the arc drop of ionic valves or the forward voltage drop of thyristors. The arc drop in hv mercury-arc valves is about 50 V per vhlve and is
almost independent of the current.
The following equations are given for completeness and for use in calcula
tions where high accuracy is required. The terminal voltage of any converter,
which may be either rectifier or inverter is :
Vd = Vd0 cos a - (Rc + Rr)Id - Farc
(CI)*
where Rr = true resistance of converter

Farc = voltage across all conducting valves in series
and the other symbols have their previous meanings :
Vd0 = ideal no-load voltage

Rc = equivalent commutating resistance
Id direct current
a ~ ignition delay angle
In inverter calculations it is convenient to put Eq. (CI) into a form using

cos y instead of cos a. For doing so, we first substitute into Eq. (CI)
Vd0 cos a = Vd0 cos <5 + 2RJd
(C2)
This equation pertains to the commutating loop, in which flows alternating

current whose amplitude is determined by the reactance Xc (n/3)Rc; the true
resistance has a completely negligible effect on this current, and the arc drops
in the two commutating valves are equal and in opposite sense around the
* This becomes Eq. (24) in Chapter 3 if Rr = 0 and Farc = 0.

494
495
The direct current passes through two phases of the ac circuit in series
(Y connection assumed) except during commutation, when it passes through
three phases in series-parallel. The resistance of the path, then, varies from
2Rp at u = 0 to 1.5Z?P at u = 60, where Rp is the resistance per phase. For
u = 0 we can readily verify that the same copper loss can be calculated either
from dc quantities as
Pca = 2RpId2
(C3)
Pou = 3JJ// = 3JJp(VI2 = 2J?//
(C4)
or from ac quantities as
,1
where Ip is the effective value of the alternating current, which consists of

positive and negative rectangular pulses of height Id and length 2n/3 rad.
Actually, the currents in both primary and secondary transformer windings
are nonsinusoidal alternating currents, and the copper loss in the foregoing
equations includes the losses in both sets of windings.
INDEX
INDEX
Ac line, inductive reactance, 26

Active gap, 288
Ac transmission, advantages, 19
half-wave, 27
limitations, 19
A.E.G., 7
Ainsworth, John D., 255, 320, 346, 363

Alborg (see Konti-Skan dc link)
Alternating current, ascendancy of, 2
Analog computer for C.E.A. control,
169, 170
Anisotropic earth, 405-406
Anisotropy, coefficient of, 405
Anode, 50, 424 {see also Electrode)
for cathodic protection, 435
Anode dampers, 260-269
Anode efficiency, 426
Anode reactor, 261, 262
Anodes in multiple {see Valve, multianode)
Apparent power of converter, 493
Arcback, 67, 157, 199-220, 241, 247
analysis, uncontrolled, 211-219
with blocking and bypassing, 207-210
caused by rapid commutation, 266-269
causes, 206-207, 266-269
consequences, 207
consequential, 232-233
crest currents, 216, 218-219
cure, 219
definition, 198
detection, 219
in inverter, 219
valves in series, 220
Arcback counter, 219
Arc chopping {see Quenching)
Arc drops, 50, 494-495
Arc quenching {see Quenching)
Arcthrough, 199, 227-228

causes, 227
consequential, 232-233
definition, 198
in inverter, -227
in rectifier, 227
protection against, 231
wave forms, 228
ASEA, 6, 7, 8, 16
Askarels, 365
Asynchronous tie, Eel River, 18
English Channel crossing, 14
Konti-Skan, 16
Atmospheric arc converter, 4
Auer, J. H., Jr., 440
Auroral displays, 420
Auto phase control, 104
Average direct current of bridge converter,
overlap greater than 60
96 (Eq. 72), 97
overlap less than 60, 83, 106-107
Average direct voltage of bridge converter,
no-load ideal, 77
no overlap, 77
overlap greater than 60, 97, 102
overlap less than 60, 83, 87, 106-107
Backfire {see Arcback)

Balanced currents, 335
Balance factor, 337
Beddington {see Kingsnorth dc link)
Bell Telephone System, 327, 328, 329, 331,
334, 336
Benmore {see also New Zealand dc link)
ground electrode, 450
uncharacteristic harmonic, 319
Bipolar dc link, 11
Blocking of fault clearing, 275

Blocking of valves, 15 1
Body current, 456
Body resistance, 456
Boundary conditions, 451
Break-even distance, 33
Breakthrough {see Commutation failure)
Bridge converter, analysis, 71-123, 198-233
during faults, 198-233
no overlap, 73-80
overlap greater than 60, 92-103
overlap less than 60, 26-92
relations between ac and dc quantities,
90,91,92,116
single-phase, 54
three-phase, 58, 71-123
Bridge inverter, analysis, 105-112, 203, 204,
219, 222-228, 230-231
Bridge rectifier, auto phase control, 104
complete characteristics of, 103
equivalent circuit, 88
wave forms, 76, 82, 86, 87, 101, 208,
217, 232
Bridges in parallel, 114
Bridges in series {see Cascade)
Brown Boveri, 7
Bundle conductors, 26
Bypassing, of bridges, 259
use of main valves for, 205-206
Bypass switch, 199
Bypass valve, 199-206
auxiliaries, 205
current divider, 205
damper, 261
discharge of dc line through, 244-247
grid-pulse supply, 205
transfer of current to and from, 201-204
in inverter, 203-205
in rectifier, 201-202, 204
CAB {see Consequential arc back)
Cable, submarine, compass error, 440-443

Cables, 20, 25,30, 35
reversal of voltage, 154
superconducting, 36, 39-41
Cable systems, harmonic impedances of, 351
Calverley, J. E., 4
Capacitors, 365-366
as cause of detuning, 356
497
Capacitors (continued)
cost, 345
heating from harmonics, 324
overvoltage from harmonic resonance, 324
protection of, 385-386
rating of, 369, 373, 383
reactive power of, 324
series, 438 {see also Compensation of .ac
lines)
shunt, 352, 354

temperature coefficient of capacitance,
365-366
tolerance, 367
Capillarity of soil, 450
Carson, John R., 337, 338
Cascade of single-phase converters, 65
Cascade of three-phase bridges, 67
Cascade of three-phase converters, 62, 63
Cascaded valve groups, 12
Cathode, ground {see Ground cathode,
Ground electrode)
Cathode spot, extinction of, 260
Cathodic protection, 423, 434-438
C.C.I.T.T., 327, 328, 330, 331
Directives, 338
C.E.A. control (see Constant-extinctionangle control)
Celilo {see Pacific NW-SW dc link)
Central Electricity Generating Board, 14
Channel Crossing (see English Channel dc
link)
Chart, converter (see Converter charts)
Circuit breakers, HV dc, 28, 36, 280-282
Clearing faults on dc line, 272-275, 279-280
C-message weighted voltage, 331
C-message weighting, 327-331
table, 329
Codrongianos (see Sardinia dc link)
Coke, 394, 398, 435, 443, 446, 447, 459,
462, 464
Commutating EMF, 81
Commutating reactance, per-unit, 89
Commutating resistance, 88, 102, 107
per unit, 8889
Commutation, 73 (see also Overlap and
Bridge converter, analysis)
currents during, waveforms, 82
Commutation failure, 105, 157, 158, 172,
198, 222-227
causes, 222
498
INDEX
INDEX
Commutation failure (continued)

consequent, 236-240
cure, 226
definition, 198
double, 226
symptoms, 225
wave forms, 223
Commutation margin, 110-112, 173
Commutation margin angle, 110-112, 174
Commutation resistance (see Commutating
resistance)
Commutators, 2
Compaction of soil, 450
Compass error, 440-443
Compensation of ac lines, 10, 23, 24, 27, 35
Complex power of converter, 492
Consecutive control (see Constant-extinction-angle control)
Consequent commutation failure, prevention
of, 236-240
Consequential arcback, 232-233

Consequential arcthrough, 232-233
Constant-current control, 163, 165-167
Constant-current transmission, 1, 3, 7,
156-157, 186
Constant-extinction-angle control, 158, 163,
167-174
errors in, 237
Constant-ignition-angle control, 163, 164-5
Constant-voltage transmission, 1, 7, 156-157,
184-185
Constitution of ac and dc links, 9
Contact resistance, 457
Conti-Skan (see Konti-Skan dc link)
Control, 148-192
actual characteristics, 158-164
basic means, 152
constant-current (see Constant-current

control)
constant-extinction-angle (see Constantextinction-angle control)
constant-ignition-angle (see Constant-ignition-angle control)
desired features, 157

frequency, 31, 182
instability of, 319
isochronous, 173
manual, limitations, 154-155
multiterminal lines (see Multiterminal dc
lines)
Control (continued)
power, 31, 180
stability of, 174-179
tap changer, 179-180
Converter, apparent power of, 493
bridge circuit, single-phase, 54
bridge circuit, three-phase, 58, 71-146
development of, 4
Graetz circuit (see Converter, bridge circuit,
three-phase)
Marx, 4
misoperation, 198-233
['
j
multibridge, 115-123
power vectors, 79
pulse number, 61
single-phase, 51-55
bridge, 54
center-tap, 52
full-wave,52, 54
half-wave, 51
six-pulse, 58-65
displacement angle, 493
effective alternating current, 484-489
fundamental current, 490-492
power as function of converter angles,
492-493
rms alternating current, 61, 484-489
undesirability, 332, 372
twelve-pulse, 67, 115, 332, 370
voltage drop due to resistance, 494
voltage drop in arcs, 494-495
Y-Y -interphase connection, 62, 63
Converter angles, 91, 105-106
notation for, 105-106, 110
Converter chart 1, 129-138
Converter chart 2, 138-146
error in reactive power by approximate
method, 145
Converter charts, 129-146
loci of constant active power, 137,
138, 140
loci of constant alternating current, 138
loci of constant apparent power, 137, 138,
139,140
loci of constant converter angles, 130-135,
139-143
loci of constant direct current, 129, 137,
140, 143, 144
loci of constant direct voltage, 129, 137,
140, 143, 144
*j
i
Converter charts (continued)

loci of constant displacement factor, 136,
139, 140
loci of constant reactive power, 137,
138,140
relations between charts 1 and 2, 146
Converter circuits, 49-70
comparative table, 66
Cook Strait (see New Zealand dc link)
Copper loss in converter, 495
Core of earth, 393
Corrosion, 18, 392, 423-438, 444, 465
Corrosion efficiency, 426
Cosine wave for C.E.A. Control, 170-172
Countervoltage of inverter, 105
Coupling, 334, 336
Coupling factors, 328
Crest currents, arcback, 218-219
short circuit on rectifier, 222
Cross Channel dc link (see English Channel
dc link)
Crust of earth, 393

Cryogenics, 36, 39-41
Current chopping (see Quenching)
Current command, 159
change of, 162
Current density, leakage (see Leakage
current density)
Current divider, 113, 114
bypass valve, 205
Current field, analogy with heat-flow field,
451
between electrodes, 417-419
buried spherical electrode, 399-402
in hemispherical volume surrounding a
hemispherical electrode, 397-399
in three-layer earth, 414-417
in two-layer earth, 406-414
in uniform earth, 393-397
log-log plots, 398, 401, 408, 413, 415,
416,418
natural, 419-422
compared to that of dc transmission,
422-423
near electrode, 393-417
of shore electrode, 402-404

reflections in, 410, 412
Current limit, minimum, 241, 267, 286
of conductors, 19
Current limits, converter control, 180-181
499
Current margin, 161, 164

Current order (see Current command)
Current oscillations in valves, 260-269
effects of, 269
Current ray, 410-412
Current regulator, 166, 167
gain, 179
Current setting (see Current command)
Current transformer, dc, 187-191
Dalles, The, 422 (see also Pacific NW-SW

dc link)
Dalziel, Charles F., 457

Damper windings, 323
Dampers, anode, 260-269
line, 270-272
valve, 247-269
Damping of harmonic resonance by loads,
350-351
Damping of oscillations of dc line, 175,
270-271
Dc current transformer, 187-191
Dc distribution, vestiges of, 2
Dc generators, Edison bipolar, 1
in Thury system, 3
limitations of, 2
series wound, 1, 3
Dc line, telephone noise from, 339-341
Dc link (see also Eel River, English Channel,
Gotland, Kingsnorth, Konti-Skan,
Nelson River, New Zealand, Pacific
Northwest-Southwest, Sardinia,
Vancouver Island, Volgograd-Donbass)
bipolar, 11, 391-392
constitution of, 9
construction in stages, 392
definition, 9
homopolar, 11, 392
kinds, 11
'//
x
monopolar, 11, 391,
39
reliability, 392
table, 13
Dc motors, limitations of, 2
Dc power measurement, 191
Dc reactor, 51, 235-247, 315
as buffer of overvoltage surge, 288
effect on crest discharge current of line,
246-247
effect on ripple in direct current, 240-244

WW****
500
INDEX
Dc reactor (continued)
effect on selectivity between dc line faults
and other faults, 275

for limitation of discharge current through
bypass valve, 244-247
for prevention of consequent commutation
failures, 236-240
purposes, 235-236
Dc smoothing reactor (see Dc reactor)
Dc transmission, advantages, 19, 32
applications, 32
British Columbia (see Vancouver Island
dc link)
Charlottenberg-Moabit, 7
constant-current, 7, 156
constant-voltage, 156
control (see Control)
economics, 33
Elbe-Marienfelde, 7
English Channel (see English Channel
dc link)
future of, 35, 39-41
Gotland.(see Gotland dc link)
history, 1, 37
Kingsnorth (see Kingsnorth
dc link)
Konti-Skan (see Konti-Skan dc link)
limitations, 19, 32
losses, 156
Manitobe (see Nelson River dc link)
Mechanicville-Schenectady, 6
Mellerud-Trollhattan, 8
Moscow-Kashira, 8
Moutiers-Lyons, 3
New Zealand (see New Zealand dc link)
Pacific NW-SW Intertie (see Pacific NW-SW
dc link)
Sardinia (see Sardinia dc link)
status during ascendancy of ac, 2
Sweden, 7, 8 (see also Gotland dc link)
Thury, 3
U.S.S.R., 8, 14-15
Vancouver Island (see Vancouver Island
dc link)
Volgograd-Donbass (see Volgograd-Donbass
dc link)
Wettingen-Zurich, 7
Dc transmission link, 3 (see also Dc link)
"Dc year," 15
Dead band of tap-changer control, 1.79
INDEX
Deblocking (see Unblocking)

DeForest, Lee, 5
Deionization of fault arc, 274
Deionization time of valve, 237
Dell, D. G., 465
Dent, 111, 119, 120, 173
Deterioration, coefficient of, 426
Deviation from sine wave, 326-327
Diametrical connection, 64
Direct current, average (see Average direct
current)
discontinuous, 240, 283
ripple in, 240-244
Direct Current Institute, Leningrad, 8
Proceedings, 37 (ref. A3)
Direct voltage, average (see Average direct
voltage)
drops due to arcs and resistance, 494-495
measurement of, 191
Disconnecting switches, 199
Displacement, transmission by, 26, 27
Displacement angle, 493
Displacement factor, 79, 91
Distortion of alternating voltages of multibridge converter, 118-119
Distortion of wave shape, measures of,
325-331
Disturbing current, 331
Disturbing voltage, 331
Donets Basin (see Volgograd-Donbass dc link)
Dorsey (see Nelson River dc link)
Double commutation failure, 226
Double overlap, in multibridge converter,
120-121
in single bridge, 93
Durichlor, 474
Earth return (see Ground return)

Earth-return circuits (see Ground-return
circuits)
Easton, V., 324

Echinghen (see also English Channel
dc link)
double-tuned filter, 363-364
Economics, ac versus dc, 33
Edgar Thompson mill, 7
Edison, Thomas A., 1
Edison Electric Institute, 327, 328, 329,
331,334,336
Eel River asynchronous tie, 18
Effective value of alternating current of sixpulse converter, 484-489

Elder, George R., 466
Electricite' de France, 14
Electrochemical equivalent, 425
Electrode, buried, 399-402, 455, 458 (see
also Land electrode)
cathodic protection, 435
fictitious, 408-412 (see also Electrode,
image)
ground (see Ground electrode)
hemispherical, 394
image, 399, 458
land (see Land electrode)
ocean (see Sea electrode)
sea (see Sea electrode)
shore (see Shore electrode)
Electrode line, 444, 466, 474

protection of, 285, 386
Electrode well, 467-469
Electrolysis of sea water, 467
Electromotive series, 425, 427

Electroosmosis, 436, 450, 460
EMFs of metals, 425
End effect, 418
English Channel dc link, 13, 14, 41-42
compass error, 442
grounding of, 445
variation of firing angles, 318
English Electric Co., 18
Equipotentials, 394, 395, 406, 407, 451, 462
Equivalent circuit, of dc link, 116-117,
153, 176
of inverter, 108, 116-117
of rectifier, 88, 116-117
Equivalent commutating resistance (see
Commutating resistance)
Excitron, 6
Extinction advance angle, 107, 110
normal, 168
Extinction angle, 81, 105, 107, 110
constant, 158
Extinction of cathode spot, 260
Faraday's constant, 425
Faraday's law, 424-425
Fault detectors, 274-275
Faults of valves (see Malfunctions of valves)
Faults on dc line, clearing of, 272-275,
279-280
501
Fence around ground electrode, 455, 470

Ferranti effect, 22, 23
Fibrillation, ventricular, 456-457
Field of ground current (see Current field)
Filter, 31, 333, 343-385
adequacy, 346
admittance of, 356, 360, 378-380
automatically tuned, 372-374
branch out of service, 367-368
components, 365-367
cost of, 344
damped, 344, 352, 354, 375-379

second-order, 345, 375-379
third-order, 375-379
dc, 383-384
design of (see Filter design)
double-tuned, 363-364
location, 343
high-pass (see Filter, damped)
impedance of, 355-358, 360, 364, 375-377
minimum, 333, 345, 381
minimum-cost, 345, 368-372
protection of, 385-386
purposes, 343
reactive power, 343, 372, 379-381
series, 343-344
sharpness of tuning, 344
shunt, 343, 344
single-tuned, 345
size, 354, 368
subdivided, 381
switchable, 381
tuned, 344, 352
tuning of, 367
types, 343
Filter design, 352, 354-363, 373, 375-384
damped or high-pass, 375-379
single-tuned, 355-363
Firethrough (see Arcthrough)
Fish, 465, 466, 470
Fleming valve, 5
Fourier series, complex form, 490
of current on network side of Y A trans
former bank, 301-303
of line current of 12-pulse converter, at no
overlap, 302-304
of train of rectangular pulses, 298-301
trigonometric form, 299
Frequency changer (see Frequency
conversion)
f1
|
|
|
5
1
j
1
502
INDEX
Frequency control, 182

Frequency conversion, 7
in steel mills, 7
Mechanicville-Schenectady link, 6
Sakuma, 16
Frequency deviation, 355-363
measurement, 191-192
Frequency discriminator, 191-192
Frischknecht, F. C., 405
Full-wave single-phase rectifier, 52
Fundamental alternating current of converier, 78, 490-492
Fusing of capacitors, 385
INDEX
Ground electrode (continued)
safety, 444, 455-459
Ground return, 11, 30, 391-476
advantages, 391-392
compatibility with other services, 423-443
equivalent depth of, 335, 336
problems of, 423-443
Ground-return circuits, coupling between,
334,335
Gamma (unit of magnetic field intensity),

420
Gardiner, G. E., 324
General Electric Company, 6
Generator, harmonic impedance of, 324
representation in harmonic studies, 349
Generators feeding dc link, 31
Germany, 7
Gotenborg (see Konti-Skan dc link)
Gotland dc link, 7, 13, 41, 392
compass error, 442
Grading electrode, 6
Greber, Henry, 35
Grid, control, 5
Grid bias, 151
transmission of, 151
Grid control, 73, 148-151, 153
Grid pulse, 149-151
transmission of, 151
Grid voltage, critical, 149
Ground cathode, 424, 443, 466 (see also
Ground electrode, Land electrode,
Sea electrode)
Ground electrode, 392, 443-476
accessibility, 444
corrosion of, 445
current-carrying capacity, 443
design, general, 443-445
duty, 445
heating of soil, 445
land (see Land electrode)
location, 444
maintenance cost, 443
requirements, 443
resistance to remote earth, 397, 398, 402,
403, 443
mutual impedance between, 337

telegraph, 1, 335
telephone, 335
Ground wires, effect on balance, 337
Half-wave rectifier, 5 1
Half-wave transmission, 27
Harmonic filters (see Filter)
Harmonic impedance of generator, 324
Harmonic instability, 319
Harmonic voltage, minimization of, 357
Harmonics, 31, 115, 295-326
ac characteristic, 297-313, 354
at no overlap, 297-304
at overlap, 305-313
effect of finite smoothing reactor, 3 15 ,
317-318
equations, 305-307
graphs, 308-313
with zero and infinite smoothing
reactor, 317
anticipated (see Harmonics, characteristic)
characteristic, 295-318
direct-voltage characteristic, equations,
307-313
graphs, 314-316
even, 320-321
noncharacteristic (see Harmonics,
uncharacteristic)
nontheoretical (see Harmonics, uncharac
teristic)
orders of, 295, 296,297
phase sequence of, 297, 313-314

reduction of, 332-333
theoretical (see Harmonics, characteristic)
total rms, 325-326
triple, 297,321-322
troubles caused by, 323-325
unanticipated (see Harmonics, uncharac
teristic)
Harmonics (continued)
uncharacteristic, 173, 296, 304, 318-322

amplification of, 319
causes, 318-319
consequences of, 319
diminution of, 320
field tests, 322
inverter vs. rectifier, 318-319
relation to errors in ignition angles,
320-322
Hayes, Thomas B., 440
Haywards (see New Zealand dc link)
Heating of machines by harmonics, 323
Heating of soil, 45 1-454
Hewitt, Peter Cooper, 5
Highfleld, W. E., 4
Hingorani, N. G., 191
History of dc transmission, 1, 37
Homopolar dc link, 11
Howard, P. E., 466, 476
Hunting of tap-changer control, 179
Hygroscopy of soil, 450
Ideal no-load direct voltage, 77

Igniter, 5
Ignition advance angle, 107
Ignition angle, 75, 107
measurement of, 192
minimum, 102-103, 113
production of uncharacteristic harmonics
by variation of, 320-322
Ignition delay angle (see Ignition angle)
Ignitron, 5,7
Iliceto, Francesco, 346
Image conductors, 334
Image electrode, 399, 458
Induction motor, 2
effect of harmonics, 323
Inductive coordination (see Telephone
interference)
Inductors for filters, 366-367
protection of, 385
Instability, harmonic, 319
; thermal, 451
Instantaneous currents in bridge converter,
during arcback, 209-210, 215-216, 232
overlap greater than 60, equations, 93-97
table, 98, 99
overlap less than 60, equations, 80-83
table, 84
503
Instantaneous voltages in bridge converter,

overlap greater than 60, equations, 97
table, 100
overlap less than 60, 83
table, 85
Integrator, 171
Interphase transformer, 62, 63
Inversion, 77, 105-112
Inverter, 10 (see also Converter and Inversion)
Inverter faults, detection, 231
generalization, 230-231
persistent, 231
Isochronous control, 173, 320
I-T product, 331-346
Jumps in direct voltage (see Voltage jumps)
Keller, George V., 405, 422

Kendall, H.C., 440
Kettle Rapids (see Nelson River dc link)
Kingsnorth dc link, 13, 18, 46, 183
assumed limits of harmonic impedance of
. network, 351, 353
criterion on harmonics, 346
Konti-Skan dc link, 13, 16, 44-45, 392
calculated and measured harmonics, 383
shore electrodes, 466-470
Krakowski, M., 337
Krishnayya, P. C. S., 318
kV -T product, 331
Lamm, Uno, 6
Land electrode, 445-464
Benmore, 461-463
construction, 447
cost, 462, 464
form, 447448
general, 445459
irrigated, 450
layout, 447448
life, 461463
location, 446, 461, 463
materials, 446447
resistance, 448449, 459, 461, 462, 464
Rice Flats, 463464
ring, 448, 449, 464
Rusck's design, 459461
safe voltage, 453
'
v.;v:v"
~"X,:
504
INDEX
Land electrode (continued)

safety, 455-459
siting, 446, 461, 463
soil, 450,451,455,461,463
star, 448, 449, 459
straight, 448
thermal time constant, 454-455, 461,

462,464
voltage, 453, 463
voltage gradient near, computation of,
458459
safe value, 458
Leakage current density, 426, 428-434, 470
cable, 426, 470

compact object, 428
permissible, 426
pipe, 428434,470
Let-go current, 456
Light beam, 151
Lightning arrester, 282-285, 287, 290
active gap, 288
dc, 287-290
for protection of filter inductor, 385
locations in converter station, 290
multiple-stack, 289, 290
on electrode line, 386
Lightning protection, 285-286
Line dampers, 175, 270-272
Line-drop compensator, 179
Link, transmission, 9 (see also Dc link)
Livestock, 455458,470
Loads in harmonic studies, 349-350
Longitudinal voltage, 330, 331, 335
Loss of load on rectifier, overvoltage caused
by, 287
Lydd (see also English Channel dc
INDEX
Measuring devices, 187-192, 197

dc power, 191
direct current, 187-191
direct voltage, 191
frequency deviation, 191-192
ignition delay angle, 192
Metallic circuits, coupling between, 334
Micropulsations of earth's magnetic field,
420-421
Minimum filter, 333, 345
Misfire, 198, 199, 228-229, 260, 283
definition, 198
protection, 231
Misoperation of converter, 198-233
Modulation, 319
Moisture in soil, 450, 461, 463
Monocyclic square, 7, 193
Monopolar link, 11,13
Moutiers hydro plant, 3
Moutiers-Lyons transmission, 3
Multianode valves, 6, 112-113
Multibridge converters, 115-123
Multibridge inverter, effect of mutual
reactance, 117-123
Multiterminal dc lines, constant-current
series scheme, 186
constant-voltage parallel scheme, 184-185
control of, 183-186, 197
fault clearing on, 279
Mutator (see Converter)
Mutual impedance between ground-return
circuits, 337
Mutual reactance between bridges, compen
sation of, 123
effect on multibridge inverter, 117-123
elimination of, 122
link)
measured scalar harmonic impedance of

ac network, 351, 353
valve dampers at, 248, 252

uncharacteristic harmonics, 319, 322
Magma, 393
Magnetic storms, 420, 438
Magnetization curve, 187
Malfunctions of valves, 198-199, 206-220,
227-233
Mantle, 393
Marx,E., 4
Materials, resistivities 0394
Natural frequencies, of valve current, 260,
263, 264,265
ov voltage across valve, 248, 251, 253
Natural load, 20, 21

Nelson River dc link, 13, 18, 47, 392
harmonics in generators, 323-324
Network analyzer, 349
Network harmonic admittance, region in
complex plane, 353, 360, 361, 382
Network harmonic impedance, 348-353,
358-362
calculation, 349
effect on filter design, 358-363
Network harmonic impedance (continued)

effect on filtering, 347-348
loci in complex plane, 349-352
measurement, 349
region in complex plane, 353
New Brunswick (see Eel River)
New Zealand dc link, 13, 15, 43-44, 392,
461,470
Noise voltage, calculation of, 338
Ocean electrode (see Sea electrode)
Oil-blast circuit breaker, 282
Oscillations on dc line, 270-272
Overhead lines, 10, 13, 19-30, 34, 237,
275, 285,349
Overlap, 73, 74, 80
double (see Double overlap)
greater than 60, 92-103
insufficient, 241, 284, 286
less than 60, 26-92
Overlap angle, 80 (see also Overlap)
Overshoot, of current during dc line fault,
276-279
of dc line voltage, 276
of rectifier voltage, 283, 286
of voltage across valve, 248, 253, 254
Overvoltage protection, 282-291
of converter transformer, 287
principles of, 282
summary of, 291
Overvoltages, external, 283
internal, 283-284
Pacific NW-SW dc link, 13, 17, 45-46
land electrode, 463-464

ocean electrode, 474-476
Parallel arrangements of anodes, bridges, or
valves, 112-115
Passband of single-tuned filter, 356-357
505
Phase sequence of harmonics, ac, 297
dc, 313-314
Phase-shift circuits, 171
Piecewise-linear circuit, 72
Pipe, cathodic protection of, 434-438
leakage conductance of, 436
leakage current density (see Leakage

current density)
Pipe coating, 427, 428, 435-437
I
li
Pipe-to-soil potential, 436-437

Power capability of overhead lines, 29
Power control, 180
Power factor, 157 (see also Displacement
/;
;
factor) .
Power limit of ac lines, cables, 25
overhead, 22, 23
Power reversal, 153-154
Power vectors of converter, 79
Protection, 235-291
j;
l\
ac and dc systems compared, 235
jj
of capacitors, 385-386
of converter transformers, 287
of filters, 385-386
Psophometrically weighted voltage, 331
Psophometric EMF, 338
Psophometric voltage, 331
calculation of, 338
Psophometric weighting, 327, 330, 331
table, 330
Pulse number, 61, 295, 296, 318, 332, 372
Pulse transformer, 151, 196
Q (see Quality factor and Sharpness of

tuning)
Quality factor of inductor, 355
Quenching, 150, 198, 199, 229, 283, 286
definition, 198
Quinn, P., 466, 476
\\
j i
ii
j |
|j
Radio interference, 20, 260
Pearl Street Station, 1
Radisspn (see Nelson River dc link)
'
Persistent arcthrough, 283

Persistent-arcthrough relay, 23 1
Per-unit quantities in converter equations,
Railway signals, 438-440

Reactive power, of ac line, 20, 21
of converter, 91, 92, 492-493
of delink, 24, 31
Reactor, dc smoothing (see Dc reactor) .
for filters (see Inductors for filters)
saturable, 187
Recovery voltage across valve, 253
88,89
Peterson, Harold A., 276
Phadke, Arun G., 276
Phase-angle meter, 367
Phase-locked oscillator, 320
j
i
j I
j I
i j
i j
506
INDEX
Rectifier, 10 (see also Converter)

short circuit on, 221-222
Re-energization of dc line, 275-276
Reeve, John, 318
Reflection coefficient, 4 12
Reitan, Daniel K., 276
Residual current, 335-336
Resistivity, apparent, 404-405
beach material, 471
sea water, 470, 472
soil, 450-451,461,463
table, 394
Resonance between filters and network,
348, 349
Reversal of power, 77, 153-154, 163
Reversal of voltage, 77
Ring connection, 332
Riordan, John, 337
Ripple, definitions, of, 325-331
in direct current, average value of, 242-243
calculation of, 241-243
effect of dc reactor on, 240-244
in direct voltage, 297
peak to peak, 52, 57, 61, 66, 326
Ripple control systems, 325
Rms alternating current of six-pulse con
verter, 484-489
Rusck, Sune, 427, 432, 448, 459
Safety, 444, 455-459

Sakuma frequency changer, 16
San Dalmazio (see also Sardinia dc link)
loci of network harmonic impedance,
349-352
Sardinia dc link, 13, 16,45
Saturable reactor, 187
Screened rod gap, 284
Screening, 336
Screening factor, 336
Sea electrode (see also Shore electrode)
general, 465-466
life, 474
Los Angeles, 474-476
Sea return (see Ground return)
Sea-return circuits (see Ground-return
circuits)
Series arrangements of valves or bridges,
112-115
INDEXSeries compensation of ac lines, 17

Series transmission, 3, 186
Seventh valve (see Bypass valve)
Sharpness of tuning, 344
of single-tuned filter, 355
optimum, 358-362, 381
Shielding, 336
Shoot-through (see Arcthrough)
Shore electrode, advantages, 465
construction, 467, 473
cost, 473
Danish, 466-470
efficiency, 404
general, 465-466
life, 473
materials, 465, 467, 471 , 474
mud flats, 476
North Island, 470-473
problems, 465
resistance, 469, 473, 476
voltage gradient, 470, 473, 476
Short circuit, monopolar, on bipolar dc
line, 283
on rectifier, crest currents, 222
uncontrolled, 221-222
with blocking and bypassing, 221
phase to ground, 275, 283
Short-circuit current, 29, 276-279
Shunt reactors, 10
Siemens, 7
Silicon controlled rectifier (see Thyristor)
Single-phase converters, 51
Single-tuned filter, 355-363
Skin depth, 335,417
Smoothing reactor (see Dc reactor)
Soil, capillarity, 450
compaction, 450
heating, 451-454
moisture, 450, 461, 463
parameters (electrical and thermal), 455,
461,463
resistivity, 450, 45 1, 461, 463
specific heat, 454, 455, 463
temperature, 451, 463
thermal conductivity, 455, 461, 463
S0ra, 466
Spark gap, 284, 285
Stability, of ac transmission, 17, 25
of control, 174-179
Stalingrad (see Volgograd-Donbass dc link)
Standard earth, 416

Start-control unit, 27 1
Step voltage, 392, 399, 401, 451, 458, 462
Stumpf, F. M., 346
Substitute networks, 269
Sunde, Erling D., 337
Sunspot activity, 420
Superconductivity, 36, 39-41
Surge capacitor, 285, 290
Surge diverter (see Lightning arrester)
Surge impedance, 20, 23
Surge-impedance loading, 20, 21
Sweden, 7
Switching surges, 19, 283
Sylmar (see Pacific NW-SW dc link)
Synchronous condenser, 26, 381
Synchronous machines, heating from
harmonics, 323-324
tan <p, 92
Tap changer, control of, 179-180

Tap range, 259
Telephone cable, 426, 427, 433
Telephone harmonic form factor, 328,
330, 346
Telephone influence factor, 328, 331
single-frequency values, table, 329
Telephone interference, 333-342
history, 334
reduction of, 341-342
policy on, 342
Telephone noise (see also Telephone
interference)
acceptable level, 338
from overhead dc line, 339-341
from three-phase line, 341

Telluric currents, 419, 438
Tertiary winding, 343, 381
Three-layer earth, 414-416
Three-phase converter, one-way, 56
two-way, 58
Thury system, 3
Thyratron, 5
Thyristor, 6
TIF (see Telephone influence factor)
Time constant, of current regulator,
166, 179
thermal, of ground electrode, 454-455,
461,462,464
Tolerances, capacitors and inductors, 367
-507-
Touch voltage, 392

Track circuit, 438-440
Transductor, 187
Transfer function of current regulator, 166
Transformer, converter, 51, 53, 54, 56, 59,
63-68
tertiary winding, 343, 381
dc, 36
favoring ac transmission, 2
interphase, 62, 63
overvoltage protection, 287
representation in harmonic studies, 349
saturation caused by direct current,
56,438
Transformer connections, 53, 54, 56, 59,
61-65,68,332
effect on phase of ac harmonics, 301-303
effect on wave shape of line current of
converter, 298
for increased pulse number, 332
Transformer ratio, 91, 116

Transpositions, 337
Transverter, 4
Trawlers, 14, 441
Triple harmonics, 297, 300, 321-322
Two-layer earth, 406-414, 449
Uhlmann, Erich, 274, 383

Unbalance, 335, 336,337
Unblocking of bridge, 271
Uniform earth, 393-397
Valve, 5
bypass (see Bypass valve)
characteristic of current vs. voltage, 49,
148-149
Fleming, 5
malfunctions of, 198-199, 206-220,
227-233
mercury arc, 5
deionization time, 237

multianode, 6, 102-103, 113
bypass, 205
Russian, 6
single-anode, 6
Valve dampers, 247-260
power loss in, 255-260
Valve faults, 198-233
consequential, 231
protection, 233
508
INDEX
Valves in series, 114

Vancouver Island dc link, 13, 17, 45
Vastervik (see Gotland dc link)
Vector power factor (see Displacement
factor)
Ventricular fibrillation, 456-457

Visby (see Gotland dc link)
Volgograd, harmonics in generators, 323
Volgograd-Donbass dc link, 8, 13, 14, 42-43
Voltage-dependent current limit, 181
Voltage-derivative unit, 274
Voltage divider, for measurement of direct
voltage, 191,287
Voltage drop, due to overlap, greater than
60, 102
less than 60, 83, 87
due to resistance, 494-495
Voltage jumps, across valve, 248
in direct voltage of bridge, 241
Voltage-level unit, 274
Voltage limits of transmission lines, 19
Voltage profiles of ac line, 21
Voltage regulation of ac line, 20
Wasserab, Theodor, 326

Wave distortion, definitions of, 325-331
Wave forms, ac line currents of 12-pulse
converter, 304
bridge inverter, 109, 110
bridge rectifier, currents during commuta
tion, 82
overlap greater than 60, 101
overlap less than 60, 86, 110
with no smoothing reactor, 317
dc current transformer, 190
in six-pulse bridge, 298

valve currents during arcback, 208, 217
voltage across valve, 110, 247-254
Wave shapes (see Wave forms)
Welligkeil, 326
Wcttingen, 7
Whitney, D. B., 466
Willesden (see Kingsnorth dc link)
Zigzag connection, 56, 332

Zurich, 7

Direct Current Transmission

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Direct Current Transmission

Hochgeladen von

Copyright:

Verfügbare Formate

Other Books by the Author

POWER SYSTEM STABILITY

,Volume i Elements of Stability Calculations, 1948, John Wiley & Sons,

Power Circuit Breakers and Protective Relays, 1950, John

1956, John Wiley & Sons, Inc.,

EDWARD WILSON KIMBARK, Sc.

Bonneville Power AdministrationFellow I.E.E.E.

ELECTRICAL TRANSMISSION OF POWER AND SIGNALS, 1949,

1971 by John Wiley & Sons, Inc.

All rights reserved. Published simultaneously

Reproduction or translation of any part of this work beyond that

Library of Congress Catalog Card Number: 72-142717

The most exciting new technical development in electric power systems in

1. GENERAL ASPECTS OF DC TRANSMISSION AND COMPARISON

Edward Wilson Kimbark

3. ANALYSIS OF THE BRIDGE CONVERTER

Analysis with Grid Control but no Overlap

Chart 1with Rectangular Co-ordinates of Direct Current and

5-16 Measuring Devices

Malfunctions of Mercury-Arc Valves

8. HARMONICS AND FILTERS

8-6 Means of Reducing Harmonics

TENTATIVE CONTENTS OF VOLUME II

Advantages and Problems

10. OVERHEAD LINES

14. HIGH-POWER VALVES

A. Effective Value of Alternating Current of a Six-pulse Converter 484

17. ASYNCHRONOUS TIES

Comite Consultatif International T616phonique

Comite Consultatif International T61egraphique

Elektricheskie Stantsii (Electric Powerplants),

Conference Paper (A.I.E.E. or I.E.E.E.)

univalent positive hydrogen ion

pounds force per square inch

silicon controlled rectifier

telephone harmonic form factor

Toka, Izvestiya (Proceedings of the Direct

peak inverse voltage

Power Apparatus and Systems

United States of America

DIRECT CURRENT TRANSMISSION

General Aspects of DC Transmission

1-1 HISTORICAL SKETCH817181

Early Discoveries and Applications

Both electrical science and the practical applications of electricity began

1-1 HISTORICAL SKETCH

GENERAL ASPECTS OF DC TRANSMISSION

Later Ascendancy of Alternating Current

Despite the general acceptance of ac transmission, some engineers never

They proposed, however, not to replace ac but to supplement it with dc.

The Thury System318

A system of hv dc transmission designed by a French engineer, Ren6

GENERAL ASPECTS OF DC TRANSMISSION

A converter is basically nothing more than an assemblage of controlled

The transverter, patented in 1920 by two British enineers, W. E. Highfield

main electrodes is ignited by a high-frequency spark getween auxiliary

The synchronously controlled switches heretofore described for use as

in the circuit. Generally, unidirectional conduction is desired. Devices

GENERAL ASPECTS OF DC TRANSMISSION

Present-day mercury-arc valves for high-voltage transmission, known as

The initiative in exploring the use of mercury-arc valves for dc transmission

LC bridge circuit called the monocyclic square. Constant-current operation

Secretariat intervened in favor of a hv dc cable line, which, it felt, would be

GENERAL ASPECTS OF DC TRANSMISSION