SFRA Basic Analysis Volume 02 Version 1.0
SFRA Basic Analysis Volume 02 Version 1.0
Version 1.0
1 Introduction
This document addresses some of the variations that may appear in SFRA test results due to
different test procedures – looking at testing of cables, reasons for cable arrangements and the
impact of stray impedance. Short circuit tests and the effect of shorted turns are compared.
Variation at Neutral tap due to reversing switch position is demonstrated.
Table of Contents
1 Introduction................................................................................................................. 1
2 List of Figures ............................................................................................................. 2
3 List of Equations ......................................................................................................... 2
4 Modeling SFRA via a Two Port Network .................................................................. 3
4.1 Two Port Network Basics ....................................................................................... 3
4.2 Two Port Network Transfer Functions ................................................................... 4
4.3 Two Port Networks and SFRA Measurements....................................................... 4
4.4 Two Port Network Conclusions.............................................................................. 5
5 Testing Cable Integrity ............................................................................................... 5
5.1 Short Circuit Cable Testing .................................................................................... 5
5.2 Short Circuit Cable Testing Conclusions................................................................ 7
6 Cable Construction...................................................................................................... 8
6.1 Introduction............................................................................................................. 8
6.2 Noise Interference................................................................................................... 8
6.3 Ground Connections ............................................................................................. 10
7 Impedance in Measurement Circuit .......................................................................... 10
7.1 Practical Response of a Resistor ........................................................................... 11
7.2 SFRA Response of Coil and Resistor in Series .................................................... 12
7.3 SFRA Response of Coil and Resistor in Parallel.................................................. 13
7.4 Comparison of Red Coil and 50 Ω Resistor in Series and in Parallel .................. 14
7.5 Stray Impedances in Measurement Circuit ........................................................... 14
8 Impedance in Ground Circuit.................................................................................... 15
9 Coils and Core........................................................................................................... 17
10 Short Circuit Tests .................................................................................................... 19
11 Shorted Turns............................................................................................................ 24
12 Winding Arrangements: Tap Changers .................................................................... 28
13 Conclusions............................................................................................................... 30
14 References................................................................................................................. 30
2 List of Figures
Figure 1 Two-Port Network ........................................................................................................................... 3
Figure 2 SFRA Two Port Network................................................................................................................. 4
Figure 3 Doble Leads Arranged for Shorted Test........................................................................................... 6
Figure 4 SFRA Responses for Shorted Doble Cables .................................................................................... 6
Figure 5 Doble Leads with Extra Ground Connection ................................................................................... 7
Figure 6 SFRA Responses for Doble Leads with Extra Ground Connection ................................................ 7
Figure 7 Separate Coaxial Cables................................................................................................................... 8
Figure 8 Shield Connected to Reduce Noise .................................................................................................. 9
Figure 9 Noise Immunity Comparison ........................................................................................................... 9
Figure 10 Poor Noise Immunity from an Impulse Based System................................................................. 10
Figure 11 Measuring a Resistor.................................................................................................................... 11
Figure 12 SFRA for a 50 Ω resistor.............................................................................................................. 11
Figure 13 SFRA Responses of 50 Ω Resistor compared to Red Coil........................................................... 12
Figure 14 Red Coil in Series with 50 Ω Resistor ......................................................................................... 12
Figure 15 SFRA Response of Red Coil in Series with 50 Ω Resistor .......................................................... 13
Figure 16 Red Coil in Parallel with 50 Ω Resistor ....................................................................................... 13
Figure 17 SFRA Response of Red Coil in Parallel with 50 Ω Resistor........................................................ 14
Figure 18 Series and Parallel Comparisons for Red Coil and 50 Ω Resistor ............................................... 14
Figure 19 Variation in Short Circuit Results due to Grounding ................................................................... 15
Figure 20 poor Ground on One of Three Single Phase Transformers .......................................................... 16
Figure 21 Linear Scale Results of Three Single Phase Transformers........................................................... 16
Figure 22 Variation in Responses with Grounding ...................................................................................... 17
Figure 23 Red Coil and Core Parts............................................................................................................... 18
Figure 24 Core in position on Red Coil........................................................................................................ 18
Figure 25 SFRA Responses for Red Coil with and without core ................................................................. 18
Figure 26 Simple Model of a Transformer ................................................................................................... 19
Figure 27 Short Circuit Transformer Model................................................................................................. 20
Figure 28 Model autotransformer – set to measure Red Coil ....................................................................... 20
Figure 29 Open Circuit SFRA Responses .................................................................................................... 21
Figure 30 Open Circuit and Short Circuit Responses of Red Coil ............................................................... 21
Figure 31 Shifting of Resonance to a Lower Frequency .............................................................................. 22
Figure 32 Open Circuit and Short Circuit Responses of Brown Coil........................................................... 22
Figure 33 Comparison of SFRA Open and Short Circuit Results for a Transformer ................................... 23
Figure 34 Open Circuit HV Results ............................................................................................................. 24
Figure 35 Short Circuit HV Results – Indicating a Bad Transformer .......................................................... 24
Figure 36 Shorted turn on Red Coil of Autotransformer.............................................................................. 25
Figure 37 SFRA Response of Brown Coil – different Shorts Applied......................................................... 26
Figure 38 Brown Coil Responses with and without Red Coil Short – Main Resonance.............................. 26
Figure 39 Brown Coil Responses with and without Red Coil Short – Higher Frequency............................ 27
Figure 40 Red Coil Responses with and without Red Coil Shorted Turn .................................................... 27
Figure 41 Original Arrangement of Red and Brown Coils........................................................................... 28
Figure 42 Reversed Red Coil Connection .................................................................................................... 28
Figure 43 Variation in Brown Coil Response with Red Coil Connection .................................................... 29
Figure 44 Different SFRA Response for a single LV winding at Neutral .................................................... 29
3 List of Equations
Equation 1 Two Port Network Impedance Parameters................................................................................... 3
Equation 2 Two Port Network Boundary Conditions..................................................................................... 3
Equation 3 Descriptive Equations for Two Port Network .............................................................................. 3
Generally, any pair of terminals where a signal may enter or leave an electrical network is
described as a port.
When performing SFRA we have an input signal, referenced to ground, and a measured signal,
also referenced to ground. A transformer undergoing SFRA can thus be modeled by a two-port
network. There are many descriptions of such networks in the electrical engineering literature
and on the web [1,2]. Figure 1 illustrates a basic two-port network.
z11, z22, z12, and z21 are the open-circuit impedance parameters, and can be determined by solving
Equation 1 using the boundary conditions specified in Equation 2, where Vin, Iin, Vout and Iout are
given the specific values V1, I1, V2 and I2 under those conditions [1,2].
V1 V1 V2 V2
z11 = ⇒ z12 = ⇒ z 21 = ⇒ z 22 =
I1 I 2 =0
I2 I1 = 0
I1 I 2 =0
I2 I1 = 0
V1 = Z11 I1 + Z12 I2
V2 = Z21 I1 + Z22 I2
The impedances, z11, z22, z12, and z21, are formed by the complex RLC network of the specimen.
They include capacitances, inductances, mutual inductances and resistances and are related to the
construction and materials of the transformer.
The network illustrated in Figure 1 has a number of parameters that are associated with it; these
are given here as functions of jω to reflect their complex nature:
In an SFRA measurement, the port input and output voltages, Vin and Vout , are referenced to
ground, as indicated in Figure 2.
Vin Vout
It should be clear that if we vary the values of Z11 and Z22, we change the network, and the
results for SFRA may vary. This is the reason why good grounding is important, and why we try
to be as consistent as possible in applying grounds to the base of the bushings.
In an ideal situation, there is no substantial impedance between the two bushing flange
connections and Z21 approaches zero. Any stray impedance may affect the results, as described in
section 8
Z12 represent not just the winding impedance but also any other impedance paths between th
input and output signals. Due to the generally small size of Z12 compared to any other paths, it
dominates and Z12 is a close approximation to the winding impedance under most circumstances.
These practices are in an effort to remove stray impedances and to provide repeatability.
It has become common practice among many engineers to perform a test on the cables to check
their integrity.
By laying separate coaxial cables side by side on the ground and connecting the measurement
clips together and the ground clips together, we have a two-port network where Z12 and Z21, in
Figure 2 are approximately zero. The only impedances left in the neywro are Z11 and Z22. If they
are also zero, a ‘flat’ response up to 2 MHz may be seen. Any variations could indicate damage
within the coaxial cables.
That flat response changes when we add in a length of ground connecting cable for each separate
coaxial test lead. This addition reflects what happens in practice, as real bushings have some
height and so this ground lead is required, and we have some impedance in Z11 and Z22. The test
system affects the flat response and we see a small roll off in results at higher frequencies. This
is expected, predictable and not a problem.
Doble test leads have a pre-fitted ground connection within the test cables. When tested by
shorting the measurement clips together, as shown in Figure 3, we get the response shown in
Figure 4, which exhibits the roll off.
The results in Figure 4 are normal and do not indicate damaged cables. We can confirm this by
cutting through to the coax shields and adding a shorting link close to the measurement clips, as
shown in Figure 5. This is not recommended in practice and is done herely solely for
demonstration purposes.
Figure 6 SFRA Responses for Doble Leads with Extra Ground Connection
The leads provided by Doble have a clear roll off during a shorted cable test as they have the pre-
fitted ground. This is to be expected as they are prepared for testing a transformer.
In practice when any leads are used on a transformer, a ground lead must be attached and the 0
dB response across the frequency range is no longer relevant.
Consequently, the only benefit of seeing a 0 dB response across the frequency range is to check
cable integrity. It is just as valid to see the small roll off associated with Doble cables as that is
the expected result.
6 Cable Construction
6.1 Introduction
Cables used for SFRA would, ideally, not have any affect on the results.
Unfortunately, we do not live in an ideal world and the cables we use interact with each other,
with the test object and with the ground plane. What we must do is use cables that produce a
minimal and repeatable affect on results so that their effect is not misdiagnosed as an artifact of
the test object.
Doble test leads are optimized to produce minimum variation in results but also maximize
repeatability in practice. This has been achieved by considering two key areas: noise interference
and ground connections.
To make an SFRA test of a transformer winding we must connect the test set via coaxial cables
to two bushings – and ground those leads at the base of the bushings.
In a substation, the three leads act like antennae and pick up noise. Figure 7 shows three separate
coaxial cable leads in a ‘standard’ arrangement on a simple diagram of a transformer. The
ground connections to the bushings have been omitted for clarity.
Test Set
We can improve the situation by connecting the three coaxial cable shields together at some
convenient point along the cable length. This reduces the antenna effect and improves immunity
to noise. Figure 8 gives an illustration of how this is achieved in practice. The three separate
coaxial cables are still independent but have shields connected about halfway along their length.
Test Set
Comparative tests on the two systems have yielded the results shown in Figure 9. The test was
performed on a power transformer in a transmission substation, and we are looking at typical
results for a short circuit test. The result marked as ‘standard’, in white, is for a simple three lead
system; the result marked as ‘improved’, in yellow, are with the same test set on the same
transformer under the same conditions using leads with shields connected as shown Figure 8.
The test transformer was adjacent to an overhead transmission line and subject to a large amount
of interference.
Clearly both sets of results are affected by noise – it is impossible to completely remove noise in
the test object; what we have done is reduce pick up of noise in the test leads.
It is instructive to compare the performance of the M5100 against some competing systems at
low frequencies. Getting a good repeatable results at over 70 dBs down requires some complex
engineering. Figure 10 shows poor responses from an impulse system for comparison.
The transformer under test was not particularly close to any overhead lines, but the impulse
system is clearly highly affected by noise. The results should have shown a smooth response
including the low frequency dips associated with a HV delta winding. This susceptibility to noise
compromises repeatability and thus may lead to misdiagnosis.
To achieve a two-port network (see section 4.1) it is necessary to connect the shield of the coax
test cables to the base of the bushings. The best way to do this is to have an appropriate length of
ground cable for each bushing size and type. In practice, any handy length of ground connector
gets used and results, especially at higher frequencies (see section 5.1), are subject to variations
that are not caused by the test object.
To remove this uncertainty it is possible to provide an in-built ground connection set distance
back from the measurement clip. This is what the Doble leads do. As a result we introduce a very
small variation in test results at higher frequencies – but this is done in a consistent and
repeatable manner. What we gain is repeatable results that do not change with ground connection
cables.
Stray impedance may be introduced at the ground connections or at the measurement clip
connections. These may vary the responses of the transformer under test.
The SFRA response of a 50 Ω resistor, as shown in Figure 11, is given in Figure 12. We can
clearly see the effect of the cable roll off at higher frequencies, due to cable grounding, which is
expected.
Figure 13 shows the responses of a 50 Ω resistor compared to that for the Red Coil used in
previous examples. The coil clearly shows inductive roll off and capacitive climb back.
When the 50 Ω resistor is arranged in series with the Red Coil, as shown in Figure 14, the
response, in yellow in Figure 15, shows the expected characteristics, with the higher impedance
dominating.
With the Red Coil and 50 Ω resistor in parallel, as shown in Figure 16, the response is quite
different, as shown in Figure 17.
The combined response, in yellow, is dominated by the lower impedance component. At low
frequencies it more closely follows the Red Coil, and as the inductance of the coil introduces the
roll off, the response follows the resistor as the frequency increases.
Figure 18 shows the results for the two combinations of the Red Coil and the 50 Ω Resistor:
series and parallel for comparison.
Figure 18 Series and Parallel Comparisons for Red Coil and 50 Ω Resistor
From Figure 18 it should be clear that in series combinations, the higher impedance tends to
dominate the results, while in parallel, the lower impedance dominates.
In practice, it is unlikely that stray impedance will add in parallel between the two measurement
terminals.
The effect of stray impedance is to attenuate the response; the effect depends on the size and
nature of the impedance.
The ground connections are usually made via bolts or connectors on bushing flanges. These may
be poorly grounded and affect results. This is difficult to reproduce clearly in a laboratory or
classroom environment as the stray impedance may not be purely resistive and will involve the
z11, z21 impedances identified in Figure 1.
In an effort to show that there is an effect, a simple variation of the shorted cable test is used.
During tests to provide results from small coils, the Doble test lead grounds have been connected
together. By adding in a 50-ohm resistance between the two ground connections, and then by
disconnecting them from each other, different results are obtained, as shown in Figure 19.
It can be seen, in Figure 19, that introducing impedance between the ground connections
attenuates the high frequency results.
In practice, the effect is not so simple and variations may be introduced from 200 kHz upward.
Figure 20 shows the high side results from three single-phase transformers on a log scale. There
are clear variations at higher frequencies for one phase that have been attributed to a poor ground
on one connection during the measurement. The lower frequency results look good for three
sister single phase units.
The same results on a linear scale make the variation clearer, as the linear scale does not
compress the high frequency results so much, as shown in Figure 21.
Figure 22 shows field results from one phase of a three phase transformer tested three times to
illustrate variation in results with ground connections.
The basic response is in white; the red trace gives the response after the ground clip on the red
lead was removed from the bushing and laid on the tank; the blue trace is that after the ground on
the black lead was also removed and laid on the tank. The results are a little surprising as
removing the black ground seems to improve the situation, by returning the results closer to their
original form. What may be happening is that the two impedances to ground, Z11 and Z22 in the
two port network, see Figure 2, are varied in such a way as to balance their effects on the signal
through the impedance of interest, Z12.
To summarize, impedance in the ground circuit, introduced by poor contacts at the base of
bushings, or by attaching the ground clips to ungrounded bolts or flange parts, may cause
substantial variation in the high frequency results. In practice, as results are being collected, they
should be checked for consistency and any variations used to identify possible poor grounds.
Figure 23 shows the red Coil under test and the C-Core halves with retaining clip; Figure 24
shows the clip in position on the Red Coil. Note that the core is relatively small compared to the
Red Coil.
The results for the two arrangements in Figure 23 and in Figure 24 are given in Figure 25
Figure 25 SFRA Responses for Red Coil with and without core
The effect of the core is to increase the inductance of the coil, and hence it’s impedance. This
gives the more rapid roll off for the blue curve at lower frequencies in Figure 25. The increased
inductance means the LC constant for the coil is increased and the resonant frequency is reduced
as a consequence.
The amount of attenuation is also reduced, implying a reduced impedance at the new resonance
point compared to the previous resonance. However the level of attenuation at the new resonant
point is similar to that on the original trace at the same frequency.
The aim of a short circuit SFRA test is to allow phase-to-phase comparison on a three phase
transformer without the influence of the core.
An SFRA short circuit test on a transformer is performed by shorting one winding and testing the
other winding on the same phase. In the case of a three-phase transformer, it is usual to short
together all of the low voltage bushings, not including any neutral, and testing on the high side.
This is very similar to a Leakage Reactance test, but performed at lower voltage and at much
lower currents.
By testing the HV winding with the LV winding shorted, all three phases should look identical.
A simple model of a transformer is given in Figure 26. This is an illustration of the main
elements of a transformer when looking at low frequencies.
When testing the HV winding, the low frequency results are dominated by the effects of the core;
the main test circuit is shown by the red circle in Figure 26.
HV Winding LV Winding
R small R small
Core
R high
By putting a short across the LV’s, the high impedance core is removed from the test circuit, as
shown in Figure 27
HV Winding LV Winding
R small R small
LV
Core
Short
R high
A simple autotransformer may be produced from small coils. The model, as used in previous
discussions, is shown in Figure 28.
Open circuit SFRA responses from the Red Coil and the Brown Coil are shown in Figure 29
By shorting the Brown Coil, we get a different response from the Red Coil, as shown compared
to the open circuit response in Figure 30.
By adding the short to the Brown Coil, we have changed the Red Coil response in a number of
ways:
• the resonances around 200 kHz have disappeared – these are associated with the Brown
Coil
• the inductive roll off happens at a higher frequency
• the impedance in the roll off is reduced, implying a reduced inductance
• the main resonance at about 500 kHz is shifted slightly to a lower frequency, as shown in
Figure 31
The reduction in the resonant frequency implies that the LC combination has increased for the
components producing that resonance.
By shorting the Red Coil, we get a different response from the Brown Coil, as shown compared
to the open circuit response in Figure 32.
The addition of the short to the Red Coil has produced a number of changes:
• the resonances at the 400 kHz region, associated with the Red Coil, have been removed
© 2003 Doble Engineering Co. All Rights Reserved 22
SFRA Basic Analysis Volume 2
Version 1.0
The increase in the resonant frequency implies that the LC combination producing that resonance
has reduced. This is the opposite effect to that seen in the Red Coil with the Brown Coil shorted.
It is clear from the results in Figure 30 and Figure 32 that the effect of the short circuit is to:
• remove the resonances associated with the shorted coil
• alter the main resonant frequency of the ‘good’ coil
In practical terms, the effect is much clearer. Figure 33 shows the results comparing open and
short circuit results for one phase of a three-phase transformer.
Figure 33 Comparison of SFRA Open and Short Circuit Results for a Transformer
The open circuit results, in white in Figure 33, are clearly very different to the short circuit
results, in blue, at the low frequencies. The high impedance circuit involving the core has been
shorted out and the results now reflect the low impedance path through the windings. As all three
phases on a transformer should be very similar, all three short circuit results should look very
similar at low frequencies: they should overlay to within 0.2 dB, based on experience.
Three separate results for three HV phases of a transformer are given in Figure 34; these results
differ between the outer and center phases, and show some noise interference. They have the
characteristic shape of a HV delta winding. However, it is difficult to make an interpretation on
the basis of these results without reference results.
The same three results with the LVs shorted are given in Figure 35.
The effect of the core at low frequencies is clearly removed from the short circuit results. Each
phase should be identical at low frequencies – any variation is a cause for concern. In the three
traces in Figure 35, it is the yellow and blue traces that overlay almost exactly up to several kHz.
The white trace, for the A phase, is at a substantially higher impedance. In this transformer there
was a severe fault on the A phase tap changer.
11 Shorted Turns
By shorting one or more turns on a winding we are moving partway toward shorting the whole
winding. The effect could be expected to lie part way between the open circuit results and the
short circuit results discussed in section 10. Figure 36 shows a shorted turn introduced into the
Red Coil of the model autotransformer. This has been achieved by baring the conductor on two
nearby turns of the Red Coil and placing a short metal jumper between the bared sections.
Shorted Turn
Connection
Clearly it is possible that the short covers more than one turn. The intention is to short out a
small section of the winding. It should be noted that, in practice, a short between adjacent turns
in a practical transformer may include several turns of actual winding, depending on construction
(disk, layer) and position of the short.
The SFRA results for the Brown Coil in three states are shown in Figure 37 with:
• no short on the Red Coil
• a ‘shorted turn’ on the Red Coil
• the whole Red Coil shorted
The responses are similar but bear some closer examination. Figure 38 shows the effect on the
main resonance. The original Brown Coil response, in white, is shifted to a higher frequency, in
green, when a ‘shorted turn’ is applied to the Red Coil. This is shifted higher again, in blue,
when the whole of the Red Coil is shorted.
Figure 38 Brown Coil Responses with and without Red Coil Short – Main Resonance
The higher frequency responses, as shown in Figure 39, show the response of the Brown Coil, in
white, having a series of resonance associated with the resonant frequencies of the Red Coil
between 400 kHz and 900 kHz. The response in blue is with the Red Coil shorted, and the
resonance associated with the Red Coil is removed.
Figure 39 Brown Coil Responses with and without Red Coil Short – Higher Frequency
The green trace, in Figure 39, is for a ‘shorted turn’ and clearly lies between the Red Coil being
open circuit, in white, and the Red Coil being completely short circuited, in blue.
The response of the Red Coil when a short is applied to it is compared to its ‘undamaged’ state in
Figure 40.
Figure 40 Red Coil Responses with and without Red Coil Shorted Turn
• Shift the Red Coil main resonance, at about 500 kHz to higher frequencies
Note: the resonances associated with the Brown Coil, around 200 kHz, are only marginally
affected – they are shifted slightly to higher frequencies.
The shift to higher frequencies of the main Red Coil resonance implies that the LC combination
associated with it has been reduced.
When a transformer undergoes a shorted turn, it produces a ‘loop’ of conductor around the core.
It is possible that the short itself may cause that loop to break, resulting in a lost turn, or a ‘blown
turn’.
The arrangement of coils or windings not in the measurement circuit affects the results of the
actual circuit being tested. In Figure 41, the Red and Brown Coils are arranged as an
autotransformer.
In Figure 42, the Red Coil is left in the same position but the other end is now connected to the
Brown Coil.
The Brown Coil is tested with the two different connections for the Red Coil. The results, on a
linear scale, are given in Figure 43. Clearly the difference is not great in this case, but there is a
shift in two of the resonances.
Though the change is small with the demonstration Red and Brown Coils, the variation in
practice may be much greater. Figure 44 shows two sets of SFRA results for the X3-X0 phase of
a transformer, both taken at neutral position a few minutes apart. The only difference is that one
trace is having arrived at Neutral from a lower tap position, in white, the other having arrived at
Neutral from a raised tap position, in green.
The differences are substantial and need to be understood in terms of winding arrangements as
the tap changer has a reversing switch. At Neutral tap, the tap winding is not in the transformer
circuit but is connected to it by one end. By coming to Neutral from different directions, one end
© 2003 Doble Engineering Co. All Rights Reserved 29
SFRA Basic Analysis Volume 2
Version 1.0
or the other of the tap winding may be connected to the LV winding. Repeating the SFRA at
Neutral having come from below both times would yield identical results.
13 Conclusions
Modeling an SFRA test object using a Two Port Network gives insight as to why good
grounding and connections are required.
Doble test leads are designed to reduce noise interference and provide consistency and
repeatability in SFRA measurements.
Testing of shorted cables is useful, but a flat response to 2 MHz is not necessary.
Stray impedances may affect results through intrusion at the measurement clip or at the ground
clip.
Short Circuit tests change winding impedances and mutual impedances and are useful for phase
to phase comparison; a shorted turn gives results intermediate between an open circuit test and a
short circuit test.
Tap changers position may affect results at Neutral – depending on whether the previous tap was
raised or lowered; the resulting position of the reversing switch results in different winding
geometry.
14 References
1. http://www.ece.pdx.edu/~ece2xx/ECE222/Slides/TwoPortsx4.pdf
2. Lawrence P. Huelsman, 1984. Basic Circuit Theory, New Jersey, Prentice-Hall, Inc.