Sie sind auf Seite 1von 6

THE CALIBRATION OF PARTIAL DISCHARGE (PD)

MEASURING CIRCUITS
ACCURACY AND ESTIMATION OF ERRORS

HAE FE LY Dr.Th.Praehauser

ABSTRACT
With increasing electrical stresses and further "tightening" of the construction of high
voltage apparatus nondestructive tests like the partial discharge measurement are an im-
portant help for producer and user.
Limits for pd-values are based mainly on experience during the last 30 years with radio
influence voltage (RIV) measurements in pV and widebandmeasurements of the apparent
charge in PC. Even when following standards, possible errors with RIV'measurements
are higher than with wideband measurements of apparent charge. Errors can be kept
in both cases within + 3 dB, and should even for an earlier stage of pd measurements be
smaller than + 10dB.
Exact knowledge of the data of comparable measurements is essential, therefore calibra-
tion and calibration standards are necessary.

RESUME
Avec l'augnientation des contraintes electriques ainsi que la rationalisation des cons-
tructions des apparells utilises en H.T., les essais non destructifs comme les mesures
de decharges partielles sont d'un important secours pour l'utilisateur comme pour le
fabricant.
F Les limites de d&charges partielles sont surtout basees , apr& 30 annees d'experience,
sur les mesures de perturbations radiodlectriques et sur les mesures en bande large
en PC. Les erreurs possibles peuvent Stre plus importantes avec les mesures de per-
turbations radioelectriques - m6me en appliquant les sp&iflcations donndes - qu'avec
les erreurs en PC. Les erreurs peuvent Btre comprises dans les deux cas dans une tolb-
rance de 2 3dB et devralent neanmoins i%tre plus petites que + 1OdB pour des mesures
antdrieures.
Les connaissances exactes des donnbes comparables sont essentielles, done l'etalonnage
et l'etalon sont n6cessaire.

ZUSAMMENFASSUNG
Mit zunehmenden elektrischen Beanspruchungen und kompakteren Konstruktionen der Hoch-
spannungsapparate ist eine zerstorungsfreie Priifung wie die Teilentladungsmessung eine
wichtlge Hilfe fur Hersteller und Beniltzer.
Die zullssigen Grenzwerte fiir Teilentladungswerte gehen haupts&chlich auf Erfahrungsn
der vergangenen 30 Jahre mit Starspannungsmessungen in pV und Breftbandmessungen in pC
zuriick. Fehler kdnnen bei der )1V-Messung hijher sein - selbst be1 Anwendung vorhandener
Vorschrlften - als bei der pc-Messung. Fehler kijnnen in beiden FSllen innerhalb + 3dB
gehalten werden, sie sollten aber selbst fiir weiter zuriickliegende Messungen kleiner
als + 10 dB sein.
Genaue Kenntnir aller Daten von zu vergleichenden Messungen ist wesentlich, deshalb
rind Eichung und Elchnormale unbedingt ntitfg.

International High Voltage Symposium Ziirich 9.- 13.Sept. 1975 El-33


3.2-07 p 265-270
1. INTRODUCTION measurements in UV and in pC for every
type of test sample, thus determining
For a long time, the safest means for its individual conversion factor. Con-
checking the elctrical Insulation of sidering the variety of existing mea-
high voltage equipment before putting it suring circuits and calibrating me-
into service was to apply severe test thods, and the accuracy of RIV-measure-
methods using test voltages which were ments, the latter method appears to be
high compared with the operating volta- more reliable. But as the method of pa-
ges, and of long duration. Weak points rallel measurements in pV and in pC is
in the insulation cause in most cases a applicable only as an exception, the
complete breakdown. Sometimes however accuracy of different pd-measuring me-
the quality of the insulation is reduced thods is discussed in the following pa-
by the test, so that breakdown occurs ragraphs.
perhaps after only several years of ope-
ration. In spite of the low electrical Generally it must be kept in mind,
stresses and the large safety margins that the actual discharge quantity neu-
in those days, this already presented tralised in a void or gap can be measu-
difficulties for the designer, but now- red only in very special cases; in fact
adays the problem is much more severe. partial discharges in a test object re-
The need to economise in electrical sult for the test object as a whole in
equipment and systems and to reduce a voltage drop and change of charge
costs for reasons of competition com- (=the apparent charge) causing tran-
pels the producer to increase electri- sients in the measuring circuit,
cal stresses and tighten construction which are detected by the measuring
to the point, where the safety factor Instruments. The measured quantity
is small but still adequate to guarantee is much smaller than the actual dis-
satisfactory operation. This raises charge quantity in most cases and may
not only the problem which tests should be different for different terminals
be applied, but furthermore by what of the test sample. This latter pro-
methods damage caused by a test can be perty is used for locating partial dis-
revealed. Such a nondestructive check charges in high voltage equipment (2),
of the insulation is the partial dis- but is not discussed here.
charge (pd) measurement, when applied
before and after a test voltage.
2. THE RADIO INFLUENCE VOLTAGE (RIV)
The amounts of partial discharges, MEASUREMENT (uV)
the changes of these amounts during
tests that can be tolerated for safe 2.1. RIV-METERS
operation, and the values which shall When partial discharge measurements
be established as pd-limits at operating were started on a wide scale in the
voltage must be based on the experience 1940's, only sensitive radio noise meters
in partial discharge behaviour of high were generally available. As there was
voltage equipment during the last thir- practically no alternative, they were
ty years. This experience includes un- used in USA for partial discharge measu-
certainties in the measuring methods rements also, according to the early NEMA
and of the measurement accuracies. In standard (3), which was revised in 1964
order to receive comparable results of (4).
pd-measurements for the various makes of Two principal types of RIV-meters were
high voltage equipment, it is necessary developed, namely the American and the
to standardize measuring circuits, mea- CISPR radio noise meter (5). They Indicate
suring quantities, measuring instru- a "weighted" voltage, which means, that
ments, and calibrating methods. In the the reading of the meter for an impulse
past thirty years all these parameters (remaining unchanged in size and shape)
have been changed or improved respec- increases with its repetition rate. But
tively and were settled finally 1968 in the relative increase of the meter reading
the IEC recommendations Publication No. is different according to the two stan-
270 "Partial discharge measurments"(1). dards: this means e.g. for an increase of
Up to that time the radio influence the number of discharges per half cycle of
voltage (RIV) in pV had been prescribed a 50 Hz sine wave from one to 10, that the
by most national standards as the mea- reading of an US RIV-meter increases
suring quantity. But IEC No. 270 re- c.30%, that of a CISPR RIV-meter c.70%.
commends beside some other measuring The CISPR reading dependance on the repe-
quantities the "apparent charge" in tition rate is adapted to the sensitivity
picocoulombs, and does not accept RIV of the human ear. This dependance on re-
anymore - except as a mere Indication. petition rate might be in some cases the
In order to make full use of all same one as for a damage done to electri-
pV-measurements up to now, pV-values cal Insulation, but rather unlikely in
must therefore be transferred to pica- general.
coulombs. This is done by either ap-
plying a factor of e.g. 1,O (see 4.) The accuracy of the RIV-meters varies
generally or by performing parallel between + 1,s dB (c. + 18%) and + 3 dB
(f 35 $1, the difference between readings the results of a NEMA-instrument could
at the lower or at the upper portion of arise, and so simulate a better circuit
the scale may be up to 2 dB, when changing RIV-factor: the NEMA instrument for it-
from one range to another. Comparison self has a high impedance, so that the
measurements with various instruments sho- current distribution in the measuring
wed in general rather good agreement, when circuit is not affected, if the instru-
PC-pulses or sine waves were applied (6,7). ment is connected either to the coaxial
The dependance of RIV-meters on the pul- cable leading to its measuring impedance
se repetition rate can be instead of max. Zm or to the object under test, whereas a
+ 6 dB (according to CISPR) 8 to 12 dB (6) CISPR-instrument with its low impedance
for some instruments (related to 1 pulse shifts the current distribution when
per half cycle). With many different ins- being connected once to the measuring
truments in use (and out of use again) in point at high voltage and then to the ob-
the past thirty years, the maximum diffe- ject under test with the radio frequency
rence between Instrument readings for the generator. Therefore during the calibra-
same imput, just caused by the errors of tion of a NEMA circuit with a CISPR ins-
these instruments, thus could be c.12 dB trument: two Instruments should be used
(+ 3 makes 6, and additional dependance simultaneously, or an equivalent impedan-
on repetition rate 12-6 = 6), correspon- ce inserted alternately.
ding to a factor of 4 between the extrem It is often supposed, that the amount
readings - the maximum error being c.9 dB of the measuring impedance influences di-
(factor of c.3 for too high values). rectly the PV-measurement. This is true
for the reading of the instrument. But
2.2. CALIBRATION OF THE MEASURING CIRCUIT with calibration according to NEMA with
BY VOLTAGE COMPARISON voltage comparison this reading is cor-
The NEMA standard 1940 (3) prescribes rected by the circuit RIV-factor, which
a selfcontained calibrating signal-source is the ratio of the reading of the ins-
for the calibration of the measuring ins- trument to the voltage input at the test
trument, but not necessarily a calibra- sample. Therefore the calculated "true"
tion of the measuring circuit as a whole. value is independent of the amount of
The transferfunction between high fre- measuring impedance.
quency voltage input at the high voltage AS the reading of RIV-meters is depen-
apparatus under test and the reading of dant on the repetition rate of partial
the instrument is checked only with the discharges, the frequency of the test
compensated circuit, but not with the or- voltage Is of influence too. With higher
dinary circuit: two curves show the error frequency the number of discharge/s is
due to stray capacitances of the test directly increased, assuming the same
object. This procedure seemed to be jus- number of discharges/cycle.
tified by the high capacitance of the
coupling capacitor with not less than The following table gives the approxi-
2500 pF. In the compensated circuit . mate factors for some samples- of pulses
however a radio frequency source Is ap- per half period referring to 50 Hz as
plied across the test object and so the l,o:
ratio determined by which the readings are Hz Pulses per 1
to be multiplied. The 1964 NEMA standard I half oeriod 2 4 8
substituted the 2500 pF of the coupling
capacitor by 1000 pF, the 600 ohm of the 50 l,o 1,o l,o 1,o
measuring Impedance by 150 ohm and chan- 100 1,25 1,2 1,ll 1,l
ged the time constant of the detector 150 1,38 1,26 1,17 1,14
(charging/discharging time) from 10/6OOms
to 1/600ms. (CISPR RIV-meter: 1/160ms). As the table shows, the higher the
As Harrold and Dakin point out, It was pulse repetlton rate per half cycle, the
hoped when introducing the new standard, lower is the influence of the frequency
that these changes would compensate each of the test voltage.
other roughly, and that therefore exis-
ting corona test limits would not have 2.3. CALIBRATION OF THE MEASURING CIR-
to be changed: which Is about valid for CUIT BY COMPARISON OF CURRENTS
apparatus with a capacitance between 100 Instead of calibration by voltage com-
and 500 pF, as the values remained si- parison at the test sample and at the mea-
milar, but not for transformers with suring impedance according to NEMA (see
capacitances between 300 and 5000 pF, Z.Z.), the German VDE Standards (8) use
where the values approximately doubled a radio frequency current of constant a-
at low pulse repetition rate, coming to mount as partial discharge calibrator.
better agreement with Increasing pulse This current is produced either by a
repetition rate (7). series connection of a sine wave gene-
At the application of NEMA standards, rator with c.50 pF or by a pulse callbra-
but with CISPR Instruments and error of up tor and is applied to the measuring cir-
to e.g. 6 dB (factor of c.2) compared with cuit in parallel to the test sample. It
is measured once directly by inserting a
3
measuring impedance in series to the cur- circuits, as its application is very sim-
rent source and once by inserting the ple. It leads to the same results as the
same measuring impedance in series to the correct application of a radio frequency
coupling capacitor (or to the test sample sine wave generator, but leaves less
instead). The circuit RIV factor may vary chances for errors. For this reason the
with the kind of radio frequency genera- partial discharge standard for instrument
tor (-if it gives the same results groun- transformers (11) of the Swiss Electro-
ded or not grounded-), the length of the technical Association recommends Only the
coaxial cable (if used at all), the way pulse calibrator for RIV-measurements as
of connecting this coaxial cable to the well as for PC-measurements. To facilita-
series capacitor, to the measuring impe- te comparisons of RIV-measurements fur-
dance and finally to the test object it- theron only one measuring impedance for
self: a spread of 6 dB (factor of 2 bet- RIV-measurements, namely 150 ohm is per-
ween minimum and maximum) seems to be not mitted (see 2.3. last section). Besides,
unusual, although c.10 dB can be found the direct comparison of PV with pC-va-
too. But e.g. 8 or 16 uV as result of the lues is established.
partial discharge test means for an in-
strument transformer the essential diffe- 2.5. SURVEY OF POSSIBLE ERRORS AND DE-
rence of being accepted or of beeing re- VIATIONS FOR PD-MEASUREMENTS WHEN
jected respectively. Compared with a ra- USING RIV-MEASURING CIRCUITS.
dio frequency generator the pulse call-
brator seems to be less problematical In see :
application (see 2.4.). Accuracy of modern RIV-
In contrast to the calibration of the instruments $ 5:; (2.1.)
measuring circuit by voltage comparison former ones
(see 2.2.), the amount of measuring im- Influence of repetition
pedance influences the result directly, rate on former instruments
although not exactly proportionally. As up to + 6dB (2.1.)
long as the whole current of the radio
frequency generator (or pulse generator) NEMA 1964 standards
passes the measuring impedance (direct against 1940 up to + 6dB (2.2.)
series connection), the Instrument rea- NEMA standards with CISPR
ding corresponds directly to the change instruments up to cy+ 6dB (2.2.)
of the measuring impedance. But if the
measuring impedance is in series to the Testfrequency of 150 Hz
coupling capacitor (or test sample ln- instead of 50 Hz + 3dB (2.2.)
stead), the whole current distribution VDE standards with sine
of the measuring circuit is shifted by wave generator by unlucky
a change of the measuring Impedance and choice of connections c. 2 3dB (2.3.)
therefore the current through the measu-
ring impedance itself too. The deviation VDE 60 ohm against NEMA
150 ohm - 8dB (2.3.)
will be relatively small though, and the
circuit RIV-factor will change only Certainly these errors were partly
little. But the instrument reading in- compensated by not calibrating at all
creases with increasing measuring impe- (in former times), or by choosing mea-
dance, so that the calculated "true" va- surinq frequencies with low response to
lue increases too. Therefore RIV-mcasu- @'SE, or by a more or less lucky choice
rement with calibration by current com- of radio frequency generator arrangement
parison Is dependant on the amount of (e.g. standing waves).
measuring impedance.
So the spread of pd-measurements during
The circuit RIV factor itself is fre- the last thirty years when applying
quency dependant, but in a different way standards in good trust should be less
as the circuit RIV factor found by cali- than approximately + 1OdB of the true
bration by voltage comparison, as once values. Certainly this range is much
the current is kept constant and once the smaller in all those cases, in which pd-
voltage respectively in these two me- measurements were performed continuously
thods. This means, that only a specific through the years with long series of
amount of measuring impedance at a specl- high voltage apparatus by means of one
fit radio frequency gives the same pd- stable method or by several methods, the
results as received by measurements with relations of which to each other were
calibration by voltage comparison and known.
150 ohm.

2.4. CALIBRATION OF THE MEASURING CIR- 3. THE MEASUREMENT OF APPARENT CHARGE


CUIT BY A PULSE CALIBRATOR. (PC)
The pulse calibrator (9, 10) recom- 3.1. PARTIAL DISCHARGE DETECTORS.
mended by CIGRE facilitates the calibra-
As already mentioned (see 1.)) only
ting procedure also for RIV measuring
in special cases it is possible to mea-

4
sure the actual charges being dischar- be checked especially In cases, where the
ged and neutralized In a partial dis- test object has a high capacitance compa-
charge. In fact only the resultant red with the other elements, because dls-
changes of voltage or charge on the test charges from a low capacitance are recor-
object as a whole can be determined. ded much more sensitively than those from
RIV-instruments measure the voltage the testobject (6). After the calibration
drop caused by a pd just within a narrow across the test sample, the transfer be-
frequency band transforming it by a haviour of pd's coming from small capaci-
weighting circuit according to the sen- tances in the circuit should be checked.
sitivity of the human ear. Broad band
partial discharge detectors measure the The accuracy of instruments and pulse
maximum peak values of apparent charges calibrators is + 1dB each; if the pulse
within one half cycle of the industrial calibrator is applied at Its - 1dB to the
frequency. This gives results more in- instrument at its + 1dB point, but the
dependent of the frequency behaviour of latter changes during measurement to its
test object and measuring circuit. Such - 1dB point, then the whole error is
instruments are available on the market - 3dB. The accuracy of a correctly cali-
for the measurement of the apparent brated circuit then is + 3dB.
charqe according to IEC 270 and have an
accuracy of c. + 1dB. As the reading of
most instruments can be adjusted to the 3.3. SURVEY OF POSSIBLE ERRORS FOR DIS-
PC-input value at the test object during CHARGE MEASUREMENTS WHEN MEASURING
calibration, a separate calculation of THE APPARENT CHARGE WITH WIDEBAND
the transfer factor is not necessary in INSTRUMENTS.
most cases. see :
3.2. CALIBRATION OF THE MEASURING CIR- Accuracy of wideband
CUIT BY A PULSE CALIBRATOR. partial discharge de-
tectors + 1dB (3.1.)
The shape of discharge pulses Is far
from a pure sinewave of any frequency, Accuracy of pulse
althouth discharges may cause travel- calibrator + 1dB (3.2.)
ling waves similar to sinewaves. There- Resultant accuracy + 3dB (3.2.)
fore CIGRE recommends pulse generators
(see 2.4.) for calibration of pd-measu- Measurements from the past without ca-
ring circuits. The application of these libration across the test object and just
calibrators is less problematical for calculated transfer function may have an
everyday measurements in a test plant, error of e.g. + 10dB.
than radio frequency generators. Despi-
te that straycapacitances cannot be dis-
regarded and the connections should the- 4. RELATIONSHIP BETWEEN pV and PC.
refore be as short as possible. At the For the conversion of pV to pC it is
biginning of " apparent charge " mea- necessary to know: the type of high vol-
surements only measuring instruments tage apparatus under test, the measuring
with built in calibrators were used. instrument, the measuring circuit and the
These calibrators inserted the pulses applied standard, the frequency of the
just at the input of the coaxial cable tert'voltage, the radio frequency for the
(coming from the testobject) into the RIV-measurement and the method of cali-
measuring instrument. The transferfac- bration (radio frequency generator or
tor was calculated. However checking pulse calibrator, voltage- or current
these calculated amounts suggested then comparison). The amount of measuring im-
a direct calibration by inserting the pedance is especially important, if the
calibrating pulses across the terminals current comparison method was used (see
Of the test sample. Therefore it is 2.3.).
essential for the comparison of pd-
testresults to be aware of the method IEC 270 (1) recommends a conversion
of calibration. If only the internal factor for pC/)~v of 1 (using a measuring
calibrator is applied - possibly even impedance of 150 ohm), which was adopted
without "calculating" a transfer bsha- also by the Swiss standards for instru-
viour -, an error of + 1OdB (factor of ment transformers (11). (7) gives for
3) is not unusual, which means that the transformers a margin dependent of the
recorded values are only l/3 of those pd-location and the capacitive network
found after direct calibration. The of the transformer from about 0,2 to 60,
pulse calibrators themselves available on with an average from 0,25 to 5, and for
the market should be accurate within other high voltage apparatus like
+ 1dB. switchgear, lightning arresters,
bushings c. 1 (measuring impedance 150
Generally it must always be ascertained, ohm) , (12) gives 2,6 pC/pV for 100 pul-
if the test circuit is discharge free, so ses/s and a measuring impedance of 60
that measured discharges do not come from ohm. For the actual reading in UV (wi-
other elements of the circuit. This must thout correction by the circuit RIV faC-
5
References:
tor and including a measuring impedance 1 IEC Publication 270, partial dischar-
of 60 ohm) (6) gives a margin between 5 ge measurements, 1968
and 100 for the relation pC/pV. (All ins- 2 T. Praehauser: Lokalisierung von Tell-
truments narrowband either ASA or CISPR entladungen in Hochspannungsapparaten
standard.) Bull. SEV 63(1972)16, p.893-905.
Results from a measuring circuit with (Translated paper available at HAEFELY
a 60 ohm impedance and calibrated by the Basel)
current comparison method (see 2.3.) must 3 Methods of measuring radio noise,
be calculated for a 150 ohm impedance by 1940, Edison Electric Institute, NEMA,
multiplying with a factor of max. 2,5 Radio manufacturers association
(- If the calibration was done by voltage
comparison (see 2.2.) the amounts remain 4 NEMA Publ. No. 107-1964, Methods of
unchanged (see 2.2.)). measurement of radio influence vol-
tage (RIV) of high voltage apparatus
5. CONCLUSIONS. 5 CISPR Publ. 1 1961, Specification for
Limits for permissible pd-values at ope- CISPR radio interference measuring
rating voltages of high voltage apparatus apparatus for the frequency range
are agreed upon partly on theoretical O,lSMc/s to 30Mc/s
reasons but mainly on experimental data. 6 T. Praehauser: comparison of mea-
As these data were collected during the suring instruments and circuits
last thirty years, it is essential to CIGRE report 12-10 Part B 1968
know exactly the type of measuring lnstru-
ment and its electrical characteristics, 7 R.T. Harrold, T.W. Dakin: the
the amount of measuring impedance, the me- relationship between the plcocou-
thod of calibration (radio frequency or lomb and microvolt for corona mea-
pulse generator, voltage or current compa- surements on HV transformers and
rison, connections), the industrial fre- other apparatus, Paper T72086-2 IEEE
quency of the test voltage, the measure- Winter meeting NY 1972
ment radio frequency, the characteristics
of the coupling capacitor, the data of the 8 VDE 0434 Teil 1 u. 2, 1966, Richtli-
measuring circuit and the applied stan- nien fCir Teilentladungspr3fungen an
dards or recommendations. With this know- Betriebsmltteln mit Wechselspannun-
ledge It should be possible to estimate gen bis 500 Hz
the probable error of the measured values: 9 F.H. Kreuger: Recommended arrange-
if they stay within the limits of approx. ments for the standardisation and
+ 3dB given by the used instruments, or the testing of circuits for the
if wider tolerances have to be assumed. detection of discharges. Report
More and more limits for pd-values during CIGRE 21-01, Supplement III, 1968
tests are being specified officially by
technical committees for high voltage 10 G. Mole: Basic characteristics of
equipment. As these limits are essen- corona detector calibrators, Trans.
tial for acceptance or rejection of ap- IEEE, PAS 89 (1970)2, p.198-204
paratus, it is necessary to check mea- 11 SEV(Swiss Electrotechnical Associa-
suring instruments and calibrators from tion), TE-Regeln fiir Wandler 1974
time to time. Therefore it is desirable,
that the national bureaux of standards 12 E.M. Dembinski, J.L. Douglas: Cali-
(or equivalent offices) establish cali- bration and comparison of partial
bration stations for such instruments discharge and radio interference mea-
and calibrators. suring circuits, PRCC. IEE 115
(1968)9, p.1332-1340

HAE E F LY
EMIL HAEFELY & CIE AG BASEL
C H - 4 0 2 8 EASEL
SCHWEIZ SUISSE S W I T Z E R L A N D
TELEFON 061/4118 17
TELEX 62 469 ehb ch
1.1978 1000
E 1-33 2.Edition