Electrical Tester online
July 2018
Best practice for sweep frequency response analysis (SFRA) - Part 1

Best practice for sweep frequency response analysis (SFRA) - Part 1

02 August 2018

Robert Foster - Application Engineer, Megger USA 
Sanket Bolar - Application Engineer,  Megger USA


IEEE and IEC standards reference numerous tests that can be performed to evaluate the condition of a transformer. One of these is Sweep Frequency Response Analysis (SFRA) where low voltage, multi-frequency sweeps are performed on the transformer. The results are analyzed to reveal issues with the transformer’s internal components including the core, windings, tap leads and connections. SFRA tests can detect small physical changes in these components, but results can be affected by the way the test set is connected, specific transformer settings and even tests previously performed on the transformer. This two-part article explains how proper connections and methodical testing will help to make SFRA test results more accurate and more repeatable.

Basics of SFRA testing

During commissioning of a three-phase two-winding transformer, 15 tests should be performed: 

  • six open-circuit tests (one for each winding);
  • three short-circuit tests (performed on the three high side windings with X1, X2, X3 shorted on the low side);
  • three capacitive inter-winding tests;
  • three inductive inter-winding tests. 

All tests are conducted by injecting a low voltage, generally 10 Vp-p, on one terminal of a winding and measuring the response on the other end of the winding or, for inter-winding tests, on the corresponding terminal of the secondary winding. The terminals that aren’t being measured are left open, shorted, or grounded depending on the test. The frequency of the test voltage typically varies from 20 Hz up to 2 MHz and the ratio of the output to the input, expressed in dB, is plotted against frequency. Phase versus frequency is also plotted. 

Although each transformer has a unique fingerprint, a general pattern emerges based on transformer type and test performed. If the transformer is faulty, analysis of different frequency ranges indicates where the fault may lie. With further analysis and possibly further testing it might be possible to identify what type of  fault has occurred and its probable location within the transformer. 

There are three frequency ranges of interest for most transformers: the low, mid, and high frequencies. The actual frequencies corresponding to each range depend on the size, type and design of the transformer. However, from the typical sweep for an autotransformer, which is shown in Figure 1 (taken from IEC 60076-18), it can be seen that effects related to the core of the transformer dominate the low frequency range up to roughly 2 kHz. The sweep begins with a decreasing magnitude based on the magnetizing inductances of the core, extending down to a minimum that occurs at a resonance point between the bulk capacitances of the transformer and the magnetizing inductance of the core. It should be noted that this is a typical response where the A phase and C phase have two local minimums and nearly overlay each other whereas the B phase has a single local minimum resonance point and responds differently throughout the entire range. This is because the B phase core leg has symmetrical return paths for the flux unlike  A and C phases. 

Moving into the mid-frequency range (2 kHz to 20 kHz), the response is most influenced by the coupling between the windings, so the shape of the curve and resonant points will vary depending on the type of connection and arrangement of the windings. As the sweep progresses into the high frequency range of 20 kHz to 1 MHz the leakage inductances, along with the series and ground capacitances of the winding determine the overall shape. For both mid and high frequency ranges, the curves for all three phases almost exactly overlay since the response depends on the winding and, in general, all three windings will be nearly identical. 

Once the sweep extends above 1 MHz for transformers greater than 72.5 kV, or above 2 MHz for transformers 72.5 kV and below, the response depends more on the test setup and connections than on the transformer itself, although the internal tap leads will have some influence. At this point the response for the phases will start to diverge. These comments describe typical results; the actual frequency ranges and the effects of the various components on the sweeps will vary from transformer to transformer.


Figure 1: General relationship between frequency and transformer component (IEC 60076-18 Figure B.6)





Recommendations for consistent measurements

IEEE C57.149 states “the test configuration can have an impact on the test results. It may be difficult to determine if these minor variations are  due to differences in test configuration or some other physical change. Therefore, it is important to document the test configuration and connections for future test repeatability.” It also states “grounding techniques will have a significant effect on test results. Grounding techniques, including selection of ground conductors as well as their routings, should therefore be precise, repeatable, and documented.” CIGRE Brochure 342 and IEC 60076-18 also highlight the need for consistent connections and transformer settings. The general recommendations of these three documents can be summarized as follows:

  • Transformer shall be completely isolated from high voltage
  • Transformer tank shall be grounded „
  • Test instrument shall be grounded „
  • Transformer should be as close to “in service” condition as possible
    • Note any difference such as lack of oil, transportation bushings, etc.
  • All external bushing connections should be disconnected 
  • If applicable, the tap position for both DETC and LTC shall be recorded
  • DETC shall be in the “in service” or “as found” condition
  • LTC is recommended to be in the extreme raised position. If sweep is measured with the LTC set to neutral position it should reach this from the raised position and both tap and previous  tap position should be noted.
  • „Grounding leads should be as short as possible (without coiling leads) and of flat braid type 
  • Solid connections should be made when attaching test leads to the terminals of the transformer 
  • SFRA tests should be performed before the winding resistance test. If a  winding resistance is performed,  demagnetize afterwards. 

By following these rules, the technician will ensure that the measurements made are as accurate and repeatable as possible. When analyzing the results it will therefore be easier to determine if a difference in sweep response compared to fingerprint is due to an actual mechanical change inside the transformer or simply the result of using a different setting  or connection.

Test setup and transformer settings

To highlight some of the effects that various settings on the transformer as well as the test setup and the connections can have on the results, several measurements were carried out over a two-day period on the Dyn1 67 kV/12.47 kV 12 MVA 3Φ transformer shown in Figure 2. During this time, no internal abnormalities or physical changes occurred in the transformer. Only the test setup, connections, and tap settings were changed. Additionally, a winding resistance test was performed. The next section discusses the results in detail. 


Figure 2:  Dyn1 67 kV/12.47 kV 12 MVA 3Φ transformer

Effect of magnetization

The presence of residual magnetism in the core can influence results in the low frequency range. Residual magnetism may be the result of performing a winding resistance test. For such a test, direct current is injected into the winding; the core is magnetized and saturated so that the voltage drop due to the inductance of the winding is excluded from the measurement. If the core is not demagnetized after this test, significant differences could show up on open-circuit SFRA measurements in the low frequency range.

To demonstrate the effect of core magnetization, an open-circuit SFRA sweep was carried out on the phase B of the LV winding of the transformer with the core in the demagnetized state. Winding resistance tests were then performed at 10 A, 25 A and 50 A. After each of these tests, an open-circuit SFRA sweep was carried out. The SFRA sweeps done after the resistance tests produced the same curve, irrespective of the DC current used for the test. There was, however, a noticeable difference between the pre-magnetization and post-magnetization curves, as shown in Figure 3 and 4. 

Figure 3: Effect of magnetization on magnitude response (orange traces - pre-magnetization and after demag)





Figure 4: Effect of magnetization on phase response






Figure 5: Difference between results pre and post magnetization of transformer





In the low frequency area, the curve shifts upwards and to the right after magnetization. A shift in the first resonant frequency is observed at approximately 500Hz. A difference curve was plotted to highlight the difference between the pre-magnetization curve and post-magnetization curve.

As can be seen from the difference curve in Figure 5, there is a considerable difference in the low frequency range of 10 Hz – 3 kHz. Minor differences are also observed at higher frequencies.

Based on these observations, it is clear that care should be taken to ensure the core is not in a magnetized state during SFRA testing. Ideally, winding resistance tests should not be conducted prior to SFRA tests but if this is unavoidable – or if core magnetization is suspected – the core should be demagnetized before starting the SFRA test.

Effect of removing the core ground

Often, the core-ground insulation resistance measurement is the first electrical test done on a transformer. If the technician forgets to ground the core terminal after the test and begins SFRA testing, the ungrounded core terminal may lead to a deviation in the curve. To simulate this condition, the core was disconnected from the ground, as shown in Figure 6.

Figure 6: Core terminals disconnected from the ground




Figure 7: Magnitude response of transformer with (blue) and without (red) core grounded 





Figure 8: Phase response of transformer with (blue) and without (red) core grounded





Figure 9: Difference in magnitude response between grounded and ungrounded core

Figure 10: (left) Braids grounded at the  bushing flange

Figure 11: (right) Braids grounded at the edge of the grounding pad.






The curves in Figures 7 and 8 were obtained with open-circuit sweeps performed on the HV phase A of the transformer.

The magnitude difference between the curves shown in Figures 7 and 8 was plotted separately, and is shown as Figure 9.

As can be seen from Figure 9, differences are noticeable up to 9 kHz but then the curves essentially overlay up to 900 kHz, after which there are further noticeable differences. For valid fingerprints and comparisons, the core should therefore be kept grounded during the SFRA test.

Effect of the grounding braid length

Since SFRA measurements are sensitive, grounding plays an important role. The grounding loop must be as short as possible and the grounding  should be connected using braids at the flange of the bushing where the signal leads are connected. To demonstrate the effect of the length of the grounding connections, two measurements were made. The shortest possible length was used for one measurement, as shown in Figure 10. The whole length of the grounding braid was used for the other measurement, as shown in Figure 11.

The curves shown in Figures 12 and 13  were obtained. The difference curve is shown in Figure 14.

Figure 12: Magnitude response with short and long ground connections 

Figure 13: Phase response with short and long ground connections 

Figure 14: Difference plot of responses with long and short ground connections


There are essentially no differences in the two curves until frequencies of 900 kHz and higher are reached.

In the next set of open-circuit measurements, shown in Figures 15 and 16, the grounding braids were removed altogether. The difference curve is shown in Figure 17. Once again, there was no change in the low and mid frequencies but the deviation increased dramatically in the high frequency range. There was a small difference between the two curves at 100 kHz, which is lower than the frequency where differences produced by long and short grounding braids were noted. Above 200 kHz, the differences increased significantly.

Figure 15: Magnitude response with proper grounding (blue) and with  no grounding (red)

Figure 16: Phase response with proper grounding (blue) and with no grounding (red)

Figure 17: Difference between response with proper grounding and with no grounding

The grounding arrange-ments for the instrument were also changed. But no differences were observed whether the instrument was grounded to the transformer, to the substation grid, or even not grounded at all. The magnitude and phase responses are shown in Figures 18 and 19. Note that even though the grounding of the instrument does not affect the measurements, it must nevertheless be grounded at all times to ensure safe operation.  

Figure 18: Magnitude response with test equipment grounded and ungrounded 

Figure 19: Phase response with test equipment grounded and ungrounded 


Based on these observations, the grounding of the instrument may not affect measurements, but the position and effective length of the grounding braids does have an effect on the high frequency range. Standard grounding procedures should therefore be followed to allow accurate comparisons of SFRA results.

The second part of this article, which will appear in a future issue of Electrical Tester, will look at the effect of tap position, reversing switches, length of shorting leads and the influence of the test voltage used on SFRA testing. It will also cover delta stabilizing windings and how to verify the correct functioning of the test instrument before concluding with a summary of the main recommendations for dependable and repeatable SFRA testing.

Read part 2 here