Beruflich Dokumente
Kultur Dokumente
+ =
1
11
For the SHORT measurement:
SF
RF
DF meas
E
E
E S
+
=
1
11
Once the values (magnitude and phase) of the error terms are known at each frequency, the
computer within a modern VNA can correct them out and display the true value of S
11DUT
.
6.3.2 Sliding Loads
For the 50 Ohmload, any residual reflection will cause an error in the calculated value of the
directivity term. A solution to this is to use a SLIDING LOAD. With a sliding load, S11 is
measured at a number of different physical positions. It is then found that S11 forms a circle on the
Smith Chart: The centre is exactly at EDF, and the radius is the magnitude of the S11 of the 50
Ohmload. By measuring at seven or more positions, a wide frequency range is covered. Sliding
loads are expensive precision mechanical devices, and have mostly been replaced because of more
modern calibration techniques, described in following sections.
E
DF
residual
reflection
from 50 ohm
load, at 7
different
positions
centre of chart
Fig. A20 Technique for finding directivity term using a sliding load
(c) Ian Robertson 2008
6.3.3 Calibration standards
In practice, there are no ideal calibration standards. For example, the open circuit has some fringing
capacitance, which is frequency dependant. The short circuit has some inductance. In addition, there
is an important issue of REFERENCE PLANES. Because PHASE is now being measured, and we
are at very high frequency (SHORT WAVELENGTH), even a very simple adapter has significant
electrical length. So, the correct position of the SOL and T must be maintained throughout the
calibration. You cannot mix different connectors and adapters when calibrating.
The calibration kit consists of all the calibration standards, plus a file or files describing their
electrical characteristics. The whole lot is supplied in a nice wooden box, with a diskette or tape,
and there may be other components such as connector gauges and male/female adaptors. Connector
gauges might be used to check there is no unacceptable protrusion or recession of the centre pin of
any coax connectors. The cal kit file has to be loaded at some point into the VNA memory.
The data on the tape/disk describes the electrical characteristics of each calibration standard. Coax
open and short circuits are usually offset from the connector reference plane. A waveguide
calibration set may contain set lengths of guide and offset shorts. Also, in real life, the open, short
and load are not perfect. For the open and short circuit, a common solution is to have an equivalent
circuit for them:- for example, the open could be treated as having a certain value of fringing
capacitance, which is frequency dependant. The short can be modelled with an inductance. The cal
kit file in the VNA stores all this information, and is supplied by the manufacturer of the
calibration standards on tape or disk. For on-wafer measurements, Cascade Microtech supplies its
own software package which runs the VNA remotely during calibration.
Fig. A21 A calibration kit in a box; contains calibration standards, adapters, spanners,
gauges, and a data disk possibly.
(c) Ian Robertson 2008
6.3.4 Two-port S-parameter measurements
In order to measure two-ports without turning the DUT around, the VNA essentially requires the
same hardware as the one-port, duplicated on Port 2 as illustrated in Fig. A22. For two-port devices,
the error model has to be extended and have either 8, 12 or 16 error terms depending on what
accuracy is needed. The block diagramshows a VNA with S-parameter test set: both ports have
dual directional couplers, and the RF stimulus is continually switched fromport1 to port2, usually
with a solid state PIN switch. The frequency converter usually has one channel for each raw
parameter. Then, all four parameters can be measured nearly simultaneously, and the calibration
applied to the displayed data. The full details of the two-port calibration algorithms can be found in
Agilent Application Notes, such as AN 1287-3 Applying Error Correction to Network Analyzer
Measurements
Frequency Converter
A to D COMPUTER
RF Signal
Generator
PORT1
b2
PORT2
TEST SET
"Network
Analyser"
display
IF signals
Directional coupler
a2
Directional coupler
DUT
a1 b1
Generator
LO Signal
Splitter
Splitter
Non-reflective
SPDT switch
Fig. A22 Simplified block diagram of a VNA with S-parameter test set.
(c) Ian Robertson 2008
6.3.5 Calibration Example
The following screen dumps are taken froma 6GHz 8753 VNA, before and after calibration. The
results should be self-explanatory and powerfully demonstrate the importance of calibrating the
instrument.
Fig. A23 S21 (0.5dB per division) and S11 (10dB per division) for a thru line, BEFORE
calibration (the reference line is 0dB)
Fig. A24 S11 of a short circuit, BEFORE calibration
Because the VNA is not calibrated this is almost completely useless: although the magnitude of S11
is approximately ONE, the phase is going round and round the Smith Chart. You cant even tell if
this is an open or a short.
(c) Ian Robertson 2008
Fig. A25 S21 (0.5dB per division) and S11 (10dB per division) for a thru line, AFTER
calibration (the reference line is 0dB)
Fig. A26 S11 of the short circuit, AFTER calibration (there is a dot on the left hand side)
(c) Ian Robertson 2008
Fig. A27 S11 of the open circuit standard AFTER calibration
Fig. A28 50 Ohm load, AFTER calibration
An advanced topic is hiding in these measurements: The 50 Ohmload appears to be really excellent
(70-80dB return loss). But, in fact, because the load is a calibration standard, you are actually just
comparing it with itself. The 70-80dB return loss limit shown here is largely the connector
repeatability limit. Likewise, we can now clearly identify that the short circuit is a tiny dot on the
left hand side of the chart, but in reality it has some inductance and it appears perfect only because
you told the VNA that it was. To get more precise measurements, a knowledge of the
imperfections of the calibration standards is needed. This is the purpose of the calibration kit data.
(c) Ian Robertson 2008
6.4 Time domain measurements
VNAs measure in the frequency domain using a swept or stepped signal source. The reason for this
is that the receiver(s) can then use heterodyne detection, giving a low noise floor and thus better
dynamic range. However, many VNAs can performa Fourier transform of the frequency domain
data to present the measurement in the time domain. This is mostly useful for troubleshooting; for
example, if you had a long cable which had an unacceptable return loss measurement, in the time
domain you could clearly identify whereabouts along the cable the problem lay. S11, for example,
in the time domain will show individual reflections separated in time:-
calibration
reference plane
termination
(e.g. port 2)
good
connector
faulty
connector
time
mag
t
t =0
(cal. ref . plane)
good connector
faulty connector
Fig. A29 Time Domain Measurement Example
In this cable example, the two reflections are separated by a significant distance compared with the
wavelength at the highest frequency of the measurement. In fact, the time between pulses, t is
equal to twice the delay of the cable. The delay can be calculated fromthe dielectric constant of the
material in the cable (or effective dielectric constant for microstrip / CPW / etc.) as follows:-
C
l
delay
eff r
or
=
where l is the physical length and C is the speed of light.
Note in the VNA display that the two reflections each have a Gaussian shape; this is a function of
the limited bandwidth of the frequency domain measurement and the windowing function used in
the Fourier transform. The window function can be made a bit more narrow, at the expense of
increased sidelobes.
(c) Ian Robertson 2008
As the two reflections get closer and closer together, they become impossible to resolve. The only
way to get a major improvement in resolution is to increase the measurement bandwidth. With a 45
MHz to 110 GHz measurement, for example, it is just possible to identify the different air-bridges
on a GaAs MMIC CPW line if they have a separation of ~200 microns.
Many VNAs offer a choice of LOWPASS or BANDPASS transform; the bandpass is most often
used, and gives the sorts of graph already described. The lowpass type has the advantage that it
yields a SIGN of mag , so that inductive and capacitive discontinuities can be distinguished.
DIRECT time-domain measurements are sometimes used: A time domain reflectometer uses a
nonlinear source (e.g. step recovery diode) to create an impulse train, which is then observed in the
time domain via a sampler and oscilloscope. More modern systems exist which employ the ultra-
short pulses froma laser to create impulses on a chip with ~THz equivalent bandwidth. The
impulses are often created by illuminating photoconductive switches, and the pulses detected using
opto-electronic sampling. These research-lab systems have been used to measure devices up to a
few THz. However, the frequency resolution is not as good as a frequency-domain measurement,
and the technique would not give good results for a narrowband filter, for example.
In passing, we note that opto-electronic techniques are sometimes also used for in-circuit field
probing of GaAs ICs. Here, the electro-optic effect in GaAs means that the electric field changes the
refractive index, so a laser beamcan be used to probe around the chip. This technique is known as
internal or direct electro-optic sampling. For greater sensitivity, and for application to circuits on
substrates which do not exhibit the electro-optic effect, a small crystal probe can be moved around
the surface of the circuit, and the crystal itself is probed via a laser beam. LithiumTantalate is often
used for the crystal as it exhibits a strong electro-optic effect. This technique is known as external
electro-optic sampling. In electro-optic sampling the probing systemis measuring the fields, and the
(microwave) stimulus signal and port terminations must be applied separately.
7. Microstrip Test-Fixture Measurements
For test fixture measurements on microstrip (or CPW etc) circuits there are several ways of making
measurements with reference planes located in the microstrip. These methods can be classified into
two groups:-
1. A coaxial calibration is performed, with the network analyser's reference planes located
at the end of its coaxial cables, and then de-embedding (or "stripping") is used to shift
the measurement reference planes to the DUT.
2. Calibration standards are placed directly within the test fixture, using techniques such as
TRL (through-reflect-line)
or similar calibration.
De-embedding can be performed using an equivalent circuit model of the test fixture, or by
employing a secondary calibration routine (such as a TRL calibration) on the test fixture itself. Test
fixtures are still widely used for device measurements because they offer a great deal of flexibility,
such as allowing DC bias circuitry to be located next to the device. An illustration of a basic 2-port
test fixture for measuring an MMIC is shown. The MMIC is permanently attached to the chip
carrier with either conductive epoxy glue or solder. The metal housing is employed to hold the chip
carrier and the connector/launchers in place. Because of the fixed nature of this setup, it would be
difficult to use in-fixture cal standards, and deembedding would be easier. The parameters for
deembedding are obtained by measuring an MMIC thru line (if available) so that all the connector,
(c) Ian Robertson 2008
fixture and microstrip-MMIC transition effects can be modelled (using tuning and optimisation on a
CAD package).
In principle, time-domain gating could also be used for deembedding, but in practice the
limited network analyser bandwidth means that the spatial resolution is too coarse to make this
reliable.
MMIC
50 line
Coax-to-
microstrip
launcher
"Filtercon"
DC feedthrough
Wires to PSU
Multilayer chip
capacitor
(e.g. 10000 pF)
Single layer
chip capacitor
(e.g. 100 pF)
Ground
connection
Raised surface
of chip carrier
Alumina
substrate
Housing wall
Bond-wires
Wires to PSU
Fig. A30 A typical 2-port test fixture
(c) Ian Robertson 2008
7.1 In-Fixture TRL calibration
The SOLT method has serious limitations for microstrip measurements and there has been a lot of
work to develop other techniques. When calibrating in microstrip, or on a semiconductor wafer,
clearly sliding loads were not convenient. The TRL calibration technique is an important solution,
fromwhich a number of others are derived. The standards used are a THROUGH, a REFLECTION
standard, and a known length of LINE.
In-fixture TRL-type calibration gives much better performance than deembedding using equivalent
circuits. Typically, a special purpose test fixture is used which allows the speedy but repeatable
insertion of the required microstrip calibration standards. For TRL calibrations the calibration
standards consist of a number of through lines of different lengths, and a non-critical reflection
standard (e.g. a short or open circuit). In order to cover a useful frequency range, it is necessary to
employ a number of different delay line lengths to overcome phase ambiguity at all the
measurement frequencies. In order to accommodate these different lengths of calibration standard
the test fixture must either have coaxial-to-microstrip launchers fixed onto sliding carriages, or it
must use a split-block design with an interchangeable centre section. For example, the Anritsu test
fixture uses the sliding-carriage approach and allows accurate measurements to be made from0.01
to 60 GHz.
Fig. A31 Anritsu Microstrip Test Fixture. Copyright 2008 Anritsu
(c) Ian Robertson 2008
Microstrip TRL standards
50 line
Coax-to-
microstrip
launcher
Sliding carriage
Measurement reference plane
THRU
50 line
Coax-to-
microstrip
launcher
Measurement reference plane
REFLECT
50 line
Coax-to-
microstrip
launcher
Measurement reference plane
LINE
Fig. A32 TRL in-fixture calibration example
(c) Ian Robertson 2008
8. Probe station measurements
Accurate MMIC measurements can easily be made using a microwave probe station. A photograph
of the Agilent Technologies / Cascade Microtech 45MHz to 110GHz systemis shown. Repeatable
measurements can be made up to 110 GHz as a result of development in measurement equipment,
probes, and calibration techniques. The key to the high accuracy is the use of CPW probes, which
allow a smooth transition to be made fromthe network analyser's coaxial connectors (or waveguide
flanges) down to the tiny probe pads of the MMIC, with 50 ohm characteristic impedance
maintained throughout by tapering the CPW centre conductor width and ground-plane spacing.
Fig. A33 Photograph of the Agilent Technologies / Cascade Microtech
8510XF RFOW System
Fig. A34 Close up view of the probes above the chuck
(these are low loss typeprobes where the coaxial connector is in-line with the probes,
rather than coming in vertically fromthe top)
(c) Ian Robertson 2008
coax outer
CPW tips
pitch
Fig. A35 Probe tip details
8.1 Waveguide Test Sets
Coaxial connectors are still restricted to 110 GHz, and even below that it is difficult to realise high
performance components in microstrip and coax. So, rectangular waveguide still plays an important
part in mm-wave measurement systems. Test port combining refers to the process whereby a
coaxial S-parameter test set is operated in tandemwith a waveguide test set, in order to achieve a
higher frequency range with a single connection to the DUT.
Test port combining can be external, using a diplexer to separate the coax and waveguide
test set signals:-
DUT
waveguide
test set
coax test set
RWG
65 GHz
coax
110 GHz
coax
diplexer
Fig. A 36 External test-set combining (one port shown)
This approach has the advantage that the coax and waveguide test sets can be operated
independently when required, making the system more flexible. However, the diplexer after the test
sets (and the adaptors etc) significantly degrade the raw performance of the VNA (especially the
directivity). Calibration still works, but the user may find that the calibration drifts off more rapidly
due to temperature changes etc.
(c) Ian Robertson 2008
To improve directivity, Agilent use internal combining in the (old) 8510XF system. In this system, a
single ultrawideband coaxial-connector directional coupler is used over the whole measurement
rage (45 MHz to 110 GHz). AFTER the incident and reflected signals have been separated, the
signals are separated into coax and waveguide bands (45MHz-50GHz goes into coax, 50-75 GHz
and 75-110 GHz go into separate waveguide mixers):
DUT
coax
frequency
converter
RWG
110 GHz
coax
45MHz - 110 GHz
coupl er
m
u
l
t
i
p
l
e
x
e
r
75-110 GHz
waveguide
tfrequency
converter
110 GHz coax
stimulus input
50-75 GHz
waveguide
tfrequency
converter
50 GHz coax
Fig. A37 Internal Coax/RWG combining (8510XF)
Above 110 GHz, waveguide test sets must be used in isolation. Currently, Oleson Microwave Labs
produce suitable modules for use with both Anritsu and Agilent network analysers. Waveguide
modules covering 90-140, 140-220 and 220-325 GHz are available. The mm-wave module
comprises predominantly directional couplers, frequency multipliers to multiply the ~50 GHz
source signal up, and harmonic mixers. The 140-220 GHz RFOW system, using a pair of
transmission-reflection modules is shown.
(c) Ian Robertson 2008
Fig. A38 220 GHz Measurement System
Fig. A39 220 GHz Measurement System: Close up of waveguide-fed Picoprobes
TM
(c) Ian Robertson 2008
8.2 Prober Calibration Techniques
Probe stations can use any of the following calibration procedures
:
1. SOLT
2. TRL / LRL
3. LRM
4. LRRM
The calibration standards can be on a wafer or on a precision impedance standard substrate
(ISS). On-wafer standards can characterise the probe-to-MMIC transition more accurately, whereas
ISS standards are not subject to manufacturing variations and can be more traceable.
Generally, the SOLT has poor accuracy above 50 GHz because of the difficulty in realising good
quality standards (especially the open circuit). The LRL and TRL techniques require a minimum of
two delay lines and a non-critical reflection standard (usually an open or short circuit), like the
microstrip in-fixture calibration already described. The reference impedance is taken fromthe
characteristic impedance of the delay lines. In order to cover a wide frequency range, a number of
different delay line lengths are required because each delays electrical length is limited to a range
between 20 and 160 degrees, approximately. This means that the probe separation has to be
adjusted during the calibration procedure. For many applications, such as automated test systems,
this is a major limitation and for these applications the Line-Reflect-Match (LRM) calibration,
developed by Cascade Microtech, is preferred to LRL/TRL. The multiple delay lines required with
the LRL/TRL calibration are replaced by matched loads, representing an infinitely long delay line.
LRRM (Line-Reflect-Reflect-Match) is an enhancement to LRM which uses both the short and
open standards to improve accuracy.
Fig. A40 WinCal Screen Dump showing main window and LRM cal submenu choice
(c) Ian Robertson 2008
The software package WinCal from Cascade Microtech is dedicated to on-wafer calibration and
measurement. For TRL/LRL calibrations, the Multical programme fromNIST is more powerful
but less user friendly. The steps required for calibrating with the LRM technique using an
impedance standard substrate (ISS) are as follows:-
Fig. A41 Probe set up and calibration:
(1) Planarise the probes (2) Align them to alignment marks
- make sure all three tips make equal with the probes not quite touching
marks on the contact substrate
(3) Set the overtravel (4) Measure the thru line standard
- the tips are moved down until
they skate the exact prescribed distance
(c) Ian Robertson 2008
(5) Measure the 50 Ohm loads standard (6) Measure the open standard
(lift the probes in this case)
(7) The software calculates the calibration constants; download them into the VNA
(8) Measure a test component for calibration verification
Here, a 25ps delay line, open at one end, gives a severe test
Note the black line along the edges is resistive material to help prevent unwanted modes
0.2 0.4 0.6 0.8 1 1.5 2 3 4 5 10 2050
0.2
0.4
0.6
0.8
1
1.5
2
3
4
5
10
20
50
-0.2
-0.4
-0.6
-0.8
-1
-1.5
-2
-3
-4
-5
-10
-20
-50
0.0 GHz
120.0 GHz
S11
HP 8510
04-22-199912:11:05
Fig. A42 This is the S11 of the open-circuited 25ps delay line
(c) Ian Robertson 2008
8.3 Microstrip MMIC On-Wafer Measurements
Most MMICs use microstrip, not CPW, and an arrangement for measuring a MIM capacitor for
modelling purposes is shown. It should be noted, of course, that the CPW-to-microstrip transitions
must be accounted for to give accurate measurements for the capacitor in a microstrip environment.
The best method of removing their effects is to performa calibration with microstrip standards, as
shown. Be warned, however, that microstrip standards are problematical because the characteristic
impedance is affected by dispersion and is critically dependant on the wafer thickness.
CAPACITOR UNDER TEST
FEED LINE
CPW-to-MICROSTRIP
TRANSITION
C
O
A
X
I
A
L
P
A
R
T
A
I
R
-
C
O
P
L
A
N
A
R
P
A
R
T
REQUIRED REFERENCE PLANES
C
O
A
X
I
A
L
P
A
R
T
A
I
R
-
C
O
P
L
A
N
A
R
P
A
R
T
Fig. A43 RFOW measurement of a capacitor
Some points to be aware of are:-
1. The line Zo depends on substrate thickness; the wafer thickness is surprisingly poorly
controlled (e.g. +/- 10%).
2. The measured data may be referenced to the line Zo, not 50 Ohms (depending on
software used).
3. The Thru must have sufficient length to ensure evanescent modes at the probe-
microstrip transition have decayed away.
4. The opens must be far enough apart that no coupling occurs.
5. Multiple LINE lengths will be required for wideband measurements, and their lengths
must be carefully chosen. Multical carefully smoothes over the band edges between the
useful range of each line standard.
6. At >> 100 GHz the LINE standard length is comparable to the overtravel skate
distance!!! Probe placement errors become a severe limiting factor.
7. Substrate modes (TM ) can be excited above ~100 GHz, depending on the substrate
height.
(c) Ian Robertson 2008
REFERENCE PLANES
THRU
REFERENCE PLANES
REFLECT
REFERENCE PLANES
LINE
Fig. A44 Microstrip on-wafer LRL calibration standards
(c) Ian Robertson 2008
8.4 Measuring Ordinary Microstrip Circuits on a Probestation
Such is the convenience of probestations, it is likely that one might try to measure an ordinary
microstrip circuit with one. The best method is to have laser-drilled via-hole grounds, to realise
wideband CPW-to-microstrip transitions, and to then use on-substrate calibration pieces like the
ones already described for MMICs. However, laser-drilling may not be available and is fairly
expensive to have done. So, several researchers have studied CPW-to-microstrip transitions that do
not need via-holes. One simple technique is to use radial stubs as (narrowband) grounds:-
Fig. A45 Narrowband CPW-to-microstrip transition without via holes
(c) Ian Robertson 2008
Further Reading
1. LUCYSZYN, S. : Measurement Techniques, Chapt 12, RFIC & MMIC Design and
Technology,, eds. I. D. Robertson and S. Lucyszyn, pub. IEE, 2001
2. S. Lucyszyn, "RFIC and MMIC Measurement Techniques", Chapter 11 fromMicrowave
Measurements, R. Collier and D. Skinner (Editors), Published by the IET, pp. 217-262,
London, Oct. 2007
3. LANE, R., 'De-embedding device scattering parameters', Microwave Journal, pp.149-156,
Aug. 1984
4. LUCYSZYN, S., MAGNIER, V., READER, H. C., and ROBERTSON, I. D.,
'Ultrawideband measurement of multiple-port MMICs using non-ideal test fixtures and a 2-
port ANA', IEE Proc. Part A, vol. 139, no. 5, pp. 235-242, Sep. 1992
5. HEUERMANN, H., and SCHIEK, B., 'The in-fixture calibration procedure line-network-
network-LNN', 23rd European Microwave Conf. Proc., pp. 500-503, 1993
6. GODSHALK, E. M., 'A W-band wafer probe', IEEE Int. Microwave Symp. Dig., pp. 171-
174, 1993
7. BAHL, I., LEWIS, G., and J ORGENSON, J ., 'Automatic testing of MMIC wafers', Int.
Journal of Microwave and Millimeter-wave Computer-aided Engineering, vol. 1, no. 1, pp.
77-89, 1991
8. EUL, H. J ., and SCHIEK, B.,' Thru-match-reflect: one result of a rigorous theory for de-
embedding and network analyzer calibration', 18th European Microwave Conf. Proc., 1988
9. PRADELL, L., CACERES, M., and PURROY, F., 'Development of self-calibration
techniques for on-wafer and fixtured measurements: A novel approach', 22nd European
Microwave Conf. Proc., pp. 919-924, 1992
10. FERRERO, A., and PISANI, U., 'Two-port network analyzer calibration using an unknown
"thru"', IEEE Microwave and Guided Wave Letters, vol. 2., no. 12. pp. 505-507, Dec. 1992
11. PURROY, F., and PRADELL, L., 'Comparison of on-wafer calibrations using the concept
of reference impedance', 23rd European Microwave Conf. Proc., pp. 857-859, 1993
12. PENCE, J . E., 'Technique verifies LRRM calibrations on GaAs Substrates', Microwaves &
RF, pp. 69-76, J an. 1994
13. MARKS, R. B.: A multiline method of network analyser calibration, IEEE Transactions,
MTT-39, J uly 1991, pp. 1205-1215
(c) Ian Robertson 2008
Decibel Units
The decibel (dB) is a unit based on the logarithmic ratio of powers. dB units are able to
conveniently represent a huge range of signal levels and high gain/attenuation ratio: e.g. from pW to
100's of Watts is often encountered in radio. In the case of an amplifier, for example, we can write:-
Gain in dB =10 log OUTPUT POWER
INPUT POWER
Note that this is a ratio and has no absolute units.
Since power =V
2
/R, it is important to be careful when working with voltages. For the same
amplifier, the squaring function to get power from voltage means that 20 log must be used:-
Gain in dB =20 log OUTPUT VOLTAGE
INPUT VOLTAGE
The following table shows that a 10dB step refers to a factor of ten in power ratio:-
dB Power ratio
60 1,000,000
50 100,000
40 10,000
30 1000
20 100
10 10
0 1
-10 0.1
-20 0.01
-30 0.001
-40 0.0001
-50 0.00001
-60 0.000001
A special case of the dB unit is 3 dB which closely approximates to DOUBLE the signal power,
or -3dB which is HALF the signal power.
For absolute units, for example to express the output power of an amplifier, it is common to use a
suffix on the dB unit:-
Watts) in 10log(P dBW in Power =
) mW in 10log(P dBm in Power =
In spectrum analyser measurements, it is common to encounter the unit dBm/Hz, which expresses
the fact that noise and randomdata signals have continuous spectra and the power is spread out over
a range of frequencies.
(c) Ian Robertson 2008
Appendix
The History of 50 Ohms
http://www.rfcafe.com/references/electrical/history_of_50_ohms.htm
This fromHarmon Banning of W.L. Gore & Associates, Inc. cable:
There are probably lots of stories about how 50 Ohms came to be. The one I am most familiar goes
like this. In the early days of microwaves - around World War II, impedances were chosen
depending on the application. For maximum power handling, somewhere between 30 and 44 Ohms
was used. On the other hand, lowest attenuation for an air filled line was around 93 Ohms. In those
days, there were no flexible cables, at least for higher frequencies, only rigid tubes with air
dielectric. Semi-rigid cable came about in the early 50's, while real microwave flex cable was
approximately 10 years later.
Somewhere along the way it was decided to standardize on a given impedance so that economy and
convenience could be brought into the equation. In the US, 50 Ohms was chosen as a compromise.
There was a group known as J AN, which stood for J oint Army and Navy who took on these
matters. They later became DESC, for Defense Electronic Supply Center, where the MIL specs
evolved. Europe chose 60 Ohms. In reality, in the US, since most of the "tubes" were actually
existing materials consisting of standard rods and water pipes, 51.5 Ohms was quite common. It
was amazing to see and use adapter/converters to go from 50 to 51.5 Ohms. Eventually, 50 won out,
and special tubing was created (or maybe the plumbers allowed their pipes to change dimension
slightly).
Further along, the Europeans were forced to change because of the influence of companies such as
Hewlett-Packard which dominated the world scene. 75 Ohms is the telecommunications standard,
because in a dielectric filled line, somewhere around 77 Ohms gives the lowest loss. (Cable TV) 93
Ohms is still used for short runs such as the connection between computers and their monitors
because of low capacitance per foot which would reduce the loading on circuits and allow longer
cable runs.
Volume 9 of the MIT Rad Lab Series has some greater details of this for those interested. It has
been reprinted by Artech House and is available.