You are on page 1of 26

SFRA App.

Note 2003/10/06-01
Draft Version 1.0

Transformer SFRA Application Note #2003/10/06-01

Use of Sister Units in SFRA Analysis

Tony McGrail
Doble Engineering
October 6th 2003

1 Summary
Sister transformer may be used when making a diagnosis of SFRA results.

Examples are given where sister units have been tested and comparisons made and
conclusions drawn.

It is important that differences between sister units are understood, and that nominally the
same design of transformer may actually have internal differences that are not apparent,
and thus provide variation in SFRA results. Variation in construction and materials may
affect results in different ways.

Table of Contents Page


1 Summary ...................................................................................................................... 1
2 List of Figures .............................................................................................................. 2
3 Introduction .................................................................................................................. 3
3.1 SFRA Analysis Strategies ..................................................................................... 3
3.2 Causes of Variation in an SFRA trace................................................................... 3
3.3 What is a ‘Sister Transformer’?............................................................................. 3
4 SFRA Results from Sister Units................................................................................... 4
4.1 Two Units in the Factory ....................................................................................... 4
4.1.1 HV Windings .................................................................................................. 4
4.1.2 LV Windings................................................................................................... 6
4.1.3 Interwindings .................................................................................................. 7
4.2 Three Units – Successive Serial Numbers............................................................. 9
4.3 Two units – Similar Serial Numbers – Small Differences .................................. 12
4.3.1 Comparison of Overall Results..................................................................... 12
4.3.2 Comparison Phase by Phase ......................................................................... 14
4.3.3 Comparison of Successive Results ............................................................... 16
4.3.4 Comparison of LV Results............................................................................ 17
4.4 Sister Unit Used for Comparison After Bushing Failure .................................... 18
4.5 Sister Unit Used as Reference after Alarm.......................................................... 22

Page 1 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

2 List of Figures
Figure 1 HV Windings Overall – Log Scale....................................................................... 5
Figure 2 HV Windings Overall – Linear Scale................................................................... 5
Figure 3 HV Windings Detail ............................................................................................. 6
Figure 4 LV Windings Overall – Log Scale ....................................................................... 6
Figure 5 LV Windings Overall – Linear Scale ................................................................... 7
Figure 6 LV Windings Detail ............................................................................................. 7
Figure 7 Interwindings Overall – Log Scale....................................................................... 8
Figure 8 Interwindings Overall – Linear Scale................................................................... 8
Figure 9 Interwindings Detail ............................................................................................. 9
Figure 10 Three Sisters 2 kHz Results.............................................................................. 10
Figure 11 Three Sisters 20 kHz Results............................................................................ 10
Figure 12 Three Sisters 200 kHz Results.......................................................................... 11
Figure 13 Three Sisters 1 MHz Results ............................................................................ 11
Figure 14 Overall Results – Suspect Unit –log scale........................................................ 12
Figure 15 Overall results – Suspect unit – linear scale..................................................... 13
Figure 16 Overall Results – Assumed Good Unit –log scale ........................................... 13
Figure 17 Overall Results – Assumed Good Unit –linear scale ....................................... 14
Figure 18 Sister Unit comparison – U Phase - Log Scale................................................. 14
Figure 19 Sister Unit comparison – U Phase - Linear Scale............................................. 15
Figure 20 Sister Unit comparison – V Phase - Log Scale................................................. 15
Figure 21 Sister Unit comparison – V Phase - Linear Scale............................................. 16
Figure 22 V Phase - Log Scale - Successive Measurements ............................................ 16
Figure 23 V Phase - Linear Scale - Successive Measurements ........................................ 17
Figure 24 LV Responses U phase Log Scale.................................................................... 18
Figure 25 LV Responses U phase Linear Scale................................................................ 18
Figure 26 HV Windings of Suspect Transformer – Log Scale......................................... 19
Figure 27 HV Windings of Suspect Transformer – Linear Scale..................................... 19
Figure 28 H3-H0 Log Scale.............................................................................................. 20
Figure 29 H3-H0 Linear Scale.......................................................................................... 20
Figure 30 Y Winding Center Phase – Log Scale .............................................................. 21
Figure 31 Y Winding Center Phase – Linear Scale .......................................................... 21
Figure 32 Y Winding Center Phase – Detail .................................................................... 22
Figure 33 Results from Suspect Unit ................................................................................ 22
Figure 34 Results form Sister Unit ................................................................................... 23
Figure 35 Debris under a winding that has shifted vertically (axially)............................. 24
Figure 36 Subsequent results from sister unit................................................................... 25
Figure 37 Comparison of results for good phase .............................................................. 25
Figure 38 Comparison of results for axially shifted phase ............................................... 26

Page 2 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

3 Introduction
This application note covers the role of ‘sister’ units when considering SFRA analysis.

The aim is to provide a rationale for variances between sister units, and an understanding
of why they exist. Several examples are given to reinforce the message.

3.1 SFRA Analysis Strategies

The most efficient way to analyze SFRA traces is with respect to a previous trace taken
on the same transformer. This is not always possible, as previous results may not be
available.

We can however compare results with those taken from a ‘sister’ transformer; the
question then is, ‘What makes a transformer a ‘sister transformer’? This is discussed in
section 3.3.

Given two ‘sister transformers, how well should results compare? Several examples are
given where results are similar, and some where results are dissimilar, in section 4.

3.2 Causes of Variation in an SFRA trace

An SFRA trace relates to the complex web of inductances, capacitances and resistances
that make up the internal construction of a transformer.

When making the same test on the same transformer – on different dates, for example –
the SFRA traces should be identical.

3.3 What is a ‘Sister Transformer’?

Ideally, sister transformers would be built of identical materials in the same manner to
produce identical transformers down to the minutest detail. They would probably have
successive serial numbers.

Transformers are a complex web of inductances, capacitances and resistances. In the case
of identical sister transformers, their internal:
• capacitances between windings
• capacitances to ground
• inductances
• mutual inductances

Page 3 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
If all of these elements were identical, then they would, as a consequence, have identical
SFRA results.

However, in practice, transformers do see some variation in construction: there may be


variation in lead placement internally, or in absolute position of windings with respect to
the core, and the core with respect to the tank. They also undergo stresses and strains in
service which may cause movement and/or distortion in internal lead arrangements, for
example. All of these variations may cause differences in the SFRA traces, as shown in
section 4.

4 SFRA Results from Sister Units


A number of results are presented here. They are traces taken over several years using
both HP and Doble M5100 SFRA test sets. These are proven and reliable field
instruments that allow traces to be compared in great detail.

Results may be presented in Log or Linear format – the former emphasizes lower
frequencies while the latter emphasizes higher frequencies. Note that it is possible with
the M5100 to zoom in on any area of interest.

4.1 Two Units in the Factory

In this example, two new transformers were tested using SFRA in the factory as part of
their commissioning tests.

The transformers were 45 MVA, 66/13.2 kV with successive serial numbers.

Small variations in responses may be attributed to variation in construction.

4.1.1 HV Windings

The overall results for the H1-H3 windings, as shown in Figure 1, look very similar, with:
• Low frequency variations consistent with variation in magnetization of the core
• Some higher frequency variations

Page 4 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 1 HV Windings Overall – Log Scale

The same results on a linear scale are given in Figure 2. It is clear that in this method of
viewing the overall data, the higher frequency results are more clearly represented and
the low frequency core response is almost lost. The variations between phases relate
almost entirely to amplitude, and is small. This implies a variation in resistance, rather
than in any element of an LC combination which produce the resonances. Consequently,
the results at high frequency are likely to have variation due to some minor impedance
introduced either at the bushing contact or at the ground contact, but may also be due to
minor variation in construction between the two transformers.

Figure 2 HV Windings Overall – Linear Scale

The linear scale display, in Figure 2, still shows some minor variations at frequencies
round 150 kHz. More detailed display of this data is given in Figure 3. It is clear based on
the evidence that there are indeed some variations between the two transformers.

Page 5 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 3 HV Windings Detail

However, since both of the transformers here are brand new and have passed all other
electrical tests, the HV winding SFRA results show only differences in construction and
manufacture and do not reveal any cause for concern in the mechanical integrity of the
two transformers.

4.1.2 LV Windings

The LV windings of the same two transformers also show some variations, as indicated
in Figure 4. The low frequency variations may be ascribed to variations in core
magnetism. Higher frequency variations need to be investigated further.

Figure 4 LV Windings Overall – Log Scale

The same results on a linear scale, as shown in Figure 5, accentuate the high frequency
responses. It is clear that the responses have the same general shape, with slight

Page 6 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
variations in amplitude at high frequencies. These are probably related to very small
differences in construction as variation due to contact impedance at a lead would be
expected to have an effect across the whole of the high frequency range.

Figure 5 LV Windings Overall – Linear Scale


The details at lower frequencies, as shown in Figure 6, are very good. It is clear that there
are small variations between the two transformers, but they are minor.

Figure 6 LV Windings Detail

Again, as both of the transformers shown here are brand new and have passed all other
electrical tests, the LV winding SFRA results show only differences in construction and
manufacture and do not reveal any cause for concern in the mechanical integrity of the
two transformers.

4.1.3 Interwindings

Page 7 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
Interwinding results are not usually considered to be as effective in identifying winding
deformation and/or movement as individual winding responses. However, they are useful
in providing simple phase to phase comparisons between transformers.

The interwinding results for one phase, as shown in Figure 7 and in Figure 8, clearly
show variations between the two transformers.

The low frequency variation is due to variations in core magnetism and core construction.

Higher frequency variations, above about 1 MHz, are probably due to minor variations in
internal construction between the two transformers; variations in impedance or contact
resistance would be expected to have a more substantial affect on the results.

Figure 7 Interwindings Overall – Log Scale

Figure 8 Interwindings Overall – Linear Scale

Page 8 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
The details at lower frequencies, up to about 300 kHz, as shown in Figure 9, show some
variation between the two transformers. The responses between 100 kHz and 150 kHz
show some significant variations – new resonances and shifts in resonances. If the two
sets of results were from the same transformer on two different occasions, there would be
strong evidence here of some winding movement in this phase.

As it is, the results are from two different transformers in the factory, and the variation
may be ascribed to differences in construction.

The results may be taken as reference results, as no significant variations were found in
any other electrical tests on these two sister units.

Figure 9 Interwindings Detail

4.2 Three Units – Successive Serial Numbers

Three 750 MVA transformers were tested on different dates. Typical results for the three
units are given for the A phase.

These results are given in ‘traditional’ bands used with SFRA that evolved during
development of the test procedure during the late 1980’s.

The 2 kHz results, as shown Figure 10, show very good correspondence even in this
highly inductive region where remnant magnetism has a strong effect.

Page 9 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
0

-10

-20
Amplitude in dB

-30

-40

-50

-60

-70

-80
0 500 1000 1500 2000

Figure 10 Three Sisters 2 kHz Results

The 2 kHz results show the expected form at low frequencies where the ‘double dip’
resonances between 200 and 400 Hz relate to variations in the inductive path provided by
the core.

The 20 kHz results, as shown in Figure 11, are also very similar. Clearly the construction
of the three transformers is very similar.

0
-10
-20
Amplitude in dB

-30
-40
-50
-60
-70
-80
0 5 10 15 20

Figure 11 Three Sisters 20 kHz Results

Page 10 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Similarly the 200 kHz bands have very little variation, as shown in Figure 12.

-10
Amplitude in dB

-20

-30

-40

-50

-60
0 50 100 150 200

Figure 12 Three Sisters 200 kHz Results

10
0
Amplitude in dB

-10
-20
-30
-40
-50
-60
0 0.5 1

Figure 13 Three Sisters 1 MHz Results

Variations between the three transformers begin to appear around 700 kHz, as shown in
Figure 13.

The results for the three transformers are very consistent. Variations at high frequency
may relate to some internal movement within the transformer or may relate to small
construction variations.

Page 11 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
Generally speaking, construction variation tends to affect smaller transformers more than
larger transformers.

4.3 Two units – Similar Serial Numbers – Small Differences

In this case we look at two sister 235/21 kV 500 MVA transformers in service at a power
station. A close in fault brought the integrity of one of the two transformers in to
question. SFRA was used to investigate the situation.

4.3.1 Comparison of Overall Results

The overall results for the suspect unit, as shown in Figure 14, are generally good. The
log scale accentuates the low frequencies where there is the expected variation between
phases.

Figure 14 Overall Results – Suspect Unit –log scale

The same results on a linear scale, in Figure 15, brings out the higher frequency variation
between phases. All three phases have clear variation, but the U phase seems most
different.

Page 12 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 15 Overall results – Suspect unit – linear scale

A sister transformer was also tested using SFRA; the serial number were not successive
but were very close. The results on a log scale, as shown in Figure 16, show the expected
variation at low frequency but also have clear variations at higher frequencies.

Figure 16 Overall Results – Assumed Good Unit –log scale

The same results for the assumed good unit, on a linear scale in Figure 17, show high
frequency variations not unlike those found in the suspect unit.

This is an indication that the variation may be related to construction.

Page 13 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 17 Overall Results – Assumed Good Unit –linear scale

4.3.2 Comparison Phase by Phase

A direct comparison of the results from the two U phases on both log and linear scales, in
Figure 18 and Figure 19, show an acceptable variation at low frequency which is
assumed to relate to the magnetic state of the core when the measurements were made.

Figure 18 Sister Unit comparison – U Phase - Log Scale

The linear results, Figure 19, show very good correspondenece across the frequency
range, with only slight differences between the two transformers.

His implies that any variation between phases in each transformer is more likely due to
design and layout rather than random construction variation as it is unlikely that this
phase could be so similar between the two transformers when manufacturing variation
has such an effect within a transformer.

Page 14 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 19 Sister Unit comparison – U Phase - Linear Scale

There is a different story when looking at the V phase. The log scale results, as shown in
Figure 20, have the acceptable low frequency variation below 1 kHz due to remnant
magnetism, but also clearly have higher frequency variations.

Figure 20 Sister Unit comparison – V Phase - Log Scale

These high frequency variations are brought out in more detail when a linear scale is
used, as in Figure 21. The differences between the two V phases is clearly far more
substantial than those between the two U phases. The W phase was also very similar
between the two transformers.

The individual comparison of phases between the two sister transformers gives an
indication that there is a difference between the two V phases; this difference could be
the result of design and construction, but the similarity between the other phases means

Page 15 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
that this is unlikely. It is possible, however, that the two phases are different as a result of
some fault which has caused some minor lead movement, or as a result of shipping.

Figure 21 Sister Unit comparison – V Phase - Linear Scale

Further investigation of the assumed good unit indicated that it, too, had seen some fault
current during its time in service. This may be the cause of variation.

4.3.3 Comparison of Successive Results

Further measurements were made on the suspect transformer some months later to
ascertain whether there had been any deterioration in state of the transformer. The results
shown on a log scale, Figure 22, do show some interesting variations. There is a very
small second resonance at low frequencies, about 500 Hz, which indicates an imbalance
in the magnetic state of the core during testing. This is likely due to previous electrical
tests performed or to the switching process that brought the transformer out of service.

Figure 22 V Phase - Log Scale - Successive Measurements

Page 16 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

The same results on a linear scale, as shown in Figure 23, do have some variation at high
frequencies.

It is important to differentiate between variation in resonant frequency (a shift to left or


right) that relates to a change in LC combination, and a variation in amplitude (shift up or
down) that relates to impedance magnitude in the circuit. The variations in Figure 23 are
of the magnitude type which implies some small stray impedance in the measurement
circuit. The final conclusion is that the results are acceptable.

Figure 23 V Phase - Linear Scale - Successive Measurements

4.3.4 Comparison of LV Results

Overall, the LV results for the two transformers were very similar. Typical results are
given here for comparison.

Figure 24 shows the LV results for the u-v phase on a log scale. There are clear variations
at low frequency which relate to core magnetisation. The two responses are of the
expected form in the low frequency range.

Page 17 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 24 LV Responses U phase Log Scale

A linear scale for the same results shows very minor variations across the frequency
range, as shown in Figure 25.

Small variations are in magnitude, rather than in frequency and are likely to relate to
small impedance variations between the measurements.

Figure 25 LV Responses U phase Linear Scale

4.4 Sister Unit Used for Comparison After Bushing Failure

A bushing failure on a GSU gave cause for concern about the mechanical integrity of the
transformer windings. SFRA was used as part of a suite of electrical tests to assess the
electrical insulation condition and the mechanical integrity of the transformer.

Page 18 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
In this case, both the transformer under suspicion and a sister unit of the same design and
vintage were available for testing. Extensive electrical tests took place, with SFRA used
to confirm that no significant winding movement had occurred.

Figure 26, below, shows the overall results for the suspect transformer. There is good
correspondence between the phases.

Figure 26 HV Windings of Suspect Transformer – Log Scale

Figure 27 HV Windings of Suspect Transformer – Linear Scale

Small variations are evident at higher frequencies, above 1 MHz, in the trace for the H3-
H0 winding. These would normally be considered acceptable, but as this was the phase
where the bushing had failed, a more detailed examination of results in that range was
performed.

Page 19 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
To give support to the interpretation of the results, further SFRA measurements were
made on the sister transformer. These results are given in Figure 29 which contains the
H3-H0 results for the suspect transformer and for the sister unit.

Figure 28 H3-H0 Log Scale

Figure 29 H3-H0 Linear Scale

Clearly there is very good correspondence between the two units. In fact, variation
between phases in the suspect unit is mirrored in variation between phases in the known
good transformer. The variation in the H3-H0 phase in the suspect transformer is seen
again in the transformer known to be good.

Additional analysis of the LV windings added further support to design and construction
variation, as shown in Figure 30, Figure 31 and Figure 32.

Page 20 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 30 Y Winding Center Phase – Log Scale

On a log scale, there are good correspondences at low frequency, in Figure 30, while on a
linear scale, Figure 31, there are some clear differences between the two transformers.

Figure 31 Y Winding Center Phase – Linear Scale

Looking at the differences in detail, Figure 32, we see differences in magnitude rather
than frequency of a significant resonance. It is clear that there are differences between the
two transformers, but these are likely to be construction and design related as a fault
would tend to cause variations across a more significant frequency range and be more
consistent in shifting resonant frequencies.

Page 21 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 32 Y Winding Center Phase – Detail

Overall, SFRA measurements have not revealed any significant winding movement or
distortion. The suspect transformer was subsequently successfully returned to service.

4.5 Sister Unit Used as Reference after Alarm

In 1994 a 240 MVA Transmission transformer was taken out of service after a gas alarm.
SFRA responses were taken for both suspect unit and for a sister unit. They clearly
indicated a problem on one phase of the suspect transformer.

SFRA tests for the suspect transformer, as shown in Figure 33, showed variation between
phases, with a consistent but small right shift of the responses for the A phase.

-10

-20
Amplitude in dB

-30

-40

-50

-60
0 0.5 1

Frequency in MHz
A-Phase B-Phase C-Phase

Figure 33 Results from Suspect Unit

Page 22 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

A sister unit was tested, and the results for each phase looked very similar, as shown in
Figure 34.

¾ Results for good unit - 1994


Ratcliffe SGT3 T4681 AEI N to LV Tap 1 22/11/94
0

-10

-20
Amplitude in dB

-30

-40

-50

-60
0 0.5 1
Frequency in MHz

A phase B phase C phase

Figure 34 Results form Sister Unit

As a result of the comparison of SFRA traces the suspect transformer was diagnosed as
having winding movement. An internal inspection of the transformer showed axial
shifting of the A phase, a collapsed winding, as shown in Figure 35.

Page 23 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

Figure 35 Debris under a winding that has shifted vertically (axially)

The SFRA results, with a clear shift to the right, indicate a reduction in the LC values for
each resonance. This is consistent with an axial collapse.

In 2001 the sister unit, which had good SFRA results in 1994, was taken out of service
after an alarm. The results showed the same shift of SFRA results for one phase. Without
further testing it was possible to diagnose the same failure mode as with the unit which
had failed in 1994.

Figure 36 shows the overall results for 2001. In comparison with Figure 33 there are clear
similarities for the suspect phase. It seems to have shifted in the same way.

As reference results were available, phase by phase comparison was used to confirm the
diagnosis. Figure 37 shows very good agreement up to 2 MHz for the two B phase
windings. Slight variations at higher frequencies probably relate to some stray ground
impedance.

Page 24 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0

¾ Results of sister unit - 2001


Ratcliffe SGT3 T4681 AEI N to LV Tap 1 8/11/01
0

-10

-20
Amplitude in dB

-30

-40

-50

-60
0 0.5 1
Frequency in MHz

A phase B phase C phase

Figure 36 Subsequent results from sister unit

-10

-20
Amplitude in dB

-30

-40

-50

-60
0 0.5 1
Frequency in MHz

22/11/94 8/11/01

Figure 37 Comparison of results for good phase

Comparison of results for the affected phase is given in Figure 38. The resonances
between about 0.4 MHz and 1MHz are slightly right shifted in a way that corresponds to

Page 25 of 26
SFRA App. Note 2003/10/06-01
Draft Version 1.0
the failed unit in 1994. The shift is consistent across the frequency range and provides
strong evidednce that this winding has suffered the same fate as its sister.
0

-10

-20
Amplitude in dB

-30

-40

-50

-60
0 0.5 1
Frequency in MHz

22/11/94 8/11/01

Figure 38 Comparison of results for axially shifted phase

Page 26 of 26

You might also like