SFRA Transformer Testing: Reading Winding Response Traces
A transformer can pass its routine electrical tests and still hide a mechanical injury. A winding that shifted a few millimetres during a short circuit, a clamping that loosened after years of vibration, or a core that moved during transport may show almost no change in winding resistance or turns ratio — yet it is a failure waiting for the next fault. Sweep frequency response analysis, universally known as SFRA, is the test built to find exactly that kind of damage. It looks at the mechanical condition of the windings and core through their electrical frequency response, and it does so without opening the tank.
This guide explains SFRA in plain language: what the test actually measures, how a trace is produced, how to read one, why the baseline comparison matters more than any single number, and when the test should be run. As with everything on this site, always work to your project specification, the manufacturer's instructions and the applicable standards for your region; the general guidance here never overrides those documents. If you are new to transformer testing as a whole, start with our power transformer testing and commissioning guide, which puts SFRA in context alongside the routine electrical tests.
What SFRA actually measures
Every transformer winding, together with its core, bushings and tank, forms a complex network of resistances, inductances and capacitances. That network has a unique frequency response: if you inject a small sinusoidal signal at one end of a winding and measure what comes out at the other end, the ratio of output to input varies with frequency in a pattern full of peaks and dips. Those peaks and dips are resonances, and their positions and heights are set by the physical geometry of the winding — the spacing between turns, the distance to the core and tank, and the clamping that holds everything in place.
SFRA records that pattern over a wide frequency range, typically from tens of hertz up to a few megahertz. The test signal is small, usually only a few volts, so the test is safe for the winding and does not stress the insulation. What matters is not the absolute shape of any one trace but how it compares with a healthy reference. If the winding has moved, deformed or loosened, its inductances and capacitances change, and the resonances shift. A trace that no longer matches its baseline is telling you that something inside the tank has physically changed, even if the transformer still carries load normally.
What SFRA can detect — and what it cannot
SFRA is remarkably sensitive to mechanical problems. The classic findings include axial displacement of a winding, radial deformation or buckling after a short circuit, loosened or collapsed clamping, core movement, displaced leads, and a tap changer left in an unexpected position. Because different frequency bands respond to different parts of the winding, experienced analysts can often narrow down not just whether damage exists but roughly where it is — for example, changes at lower frequencies tend to relate to the core and bulk winding, while changes at higher frequencies relate to individual turns and local geometry.
It is equally important to know what SFRA does not do. It does not measure insulation condition, moisture, or oil quality — those are the jobs of tests such as the insulation resistance and PI test, tan delta testing and oil testing. It does not directly detect partial discharge, which is covered by partial discharge testing. And a trace that matches its baseline does not prove the transformer is perfect; it proves that nothing mechanical has changed since the baseline was taken. SFRA is one tool in a diagnostic set, and its power comes from being combined with the others.
How the test is performed
SFRA is normally performed with the transformer de-energised, isolated, earthed and discharged, following the site safety rules for high-voltage work. The test set injects a swept sine signal, usually a few volts peak, into one winding terminal while measuring the response at another terminal, and repeats this for several winding configurations — commonly end-to-end on each phase, and sometimes between windings. Good practice also records the tap changer position, the oil temperature, and which terminals were earthed, because all of these affect the trace and must be reproduced for a valid comparison.
Connections matter more in SFRA than in most other tests. The test leads must be short, the same lead layout should be used every time, and the earthing arrangement must be consistent. A trace taken with different lead routing or a different earthing point can differ from the baseline for purely procedural reasons, and chasing a phantom fault caused by sloppy connections wastes time and can lead to wrong decisions. Photograph the setup, record the configuration, and use the same test set settings — frequency range, number of points, and signal level — as the baseline test wherever possible.
Reading a trace: the frequency bands
A typical SFRA plot shows the response ratio in decibels against frequency on a logarithmic scale. Analysts usually divide the plot into bands, and each band tells a different story. At the lowest frequencies, the response is dominated by the core and the magnetising inductance, so changes here can indicate core movement or a problem with the magnetic circuit. In the middle band, the winding inductance and the bulk capacitance dominate, and shifts here often point to winding displacement or deformation. At the highest frequencies, the response is set by local turn-to-turn and turn-to-earth capacitances, so changes here can indicate localised winding damage or displaced leads.
The practical reading technique is comparison, not absolute judgement. Overlay the new trace on the baseline and look for regions where the two traces separate: resonances that have moved in frequency, peaks that have changed in height, or new dips that were not there before. Small differences at a few isolated frequencies can be measurement noise or a slightly different connection, and should be checked by repeating the test. Broad, systematic differences across a band are the signature of a real mechanical change. Where the difference appears tells you where to look; how large it is gives a first sense of severity, but severity should always be confirmed with other tests and, where justified, an internal inspection.
Baseline comparison: the heart of SFRA
A baseline, often called a fingerprint, is a trace recorded when the transformer is known to be healthy — ideally at the factory before delivery, or at commissioning before energisation. Every later test is judged against it, which is why taking a fingerprint at commissioning is one of the highest-value actions in a transformer's life. It costs a few hours once, and it makes every future SFRA test dramatically more meaningful.
When no fingerprint exists, comparisons are still possible but need more care. A trace from a sister unit of the same design and rating can serve as a reference, since identical windings should give very similar responses; differences between phases of the same transformer can also be informative, because the three phases of a healthy unit usually track each other closely, and one phase standing apart from the other two is suspicious. A previous test on the same unit, even if it was not a formal fingerprint, can also show whether the condition has changed since then. In all cases, document what the reference is and how confident you are in it, because the conclusion is only as strong as the comparison.
When to run SFRA
SFRA earns its keep at specific moments in a transformer's life. The most important is after a short circuit or through fault: the electromagnetic forces in a fault can deform windings even when the protection operated correctly and the transformer appears to have survived. A post-fault SFRA, compared with the fingerprint, is often the deciding evidence for whether the unit can safely return to service or needs internal inspection. The same logic applies after transport or relocation, after a protection trip whose cause is unclear, and after any event that could have subjected the transformer to mechanical shock.
Beyond events, SFRA is used as a condition-assessment tool on older or critical transformers, sometimes on a periodic basis, to watch for slow mechanical deterioration such as loosening clamping. It is also valuable before and after major work — for example, before a unit is moved for overhaul and after it returns — so that any change caused by the work itself is visible. If you are planning an overhaul or major maintenance programme, building SFRA into the before-and-after test set is a straightforward way to prove the work did no harm and to catch problems early.
Limitations and common mistakes
SFRA has real limitations that a good engineer respects. It is a comparative test, so without any reference it can only tell you about differences between phases, not about absolute health. It is sensitive to test setup, so inconsistent connections or tap positions can create false differences. It can indicate that something has changed and roughly where, but it cannot by itself identify the exact fault — that usually needs corroboration from other tests or an internal inspection. And interpreting traces well takes experience; automated comparison software helps, but the final judgement still belongs to someone who understands both the test and the transformer.
The common mistakes are almost all procedural: forgetting to record the tap position, changing the lead layout between tests, testing with different earthing arrangements, or comparing traces taken with different frequency ranges. A disciplined test record — tap position, oil temperature, terminal connections, lead layout, test settings, and photographs — prevents most of these. If a trace looks wrong, repeat it before concluding anything; a surprising result that disappears on a careful repeat was almost certainly a setup problem.
Reporting SFRA results
A useful SFRA report contains the test conditions, the configurations tested, the reference used and its provenance, overlaid plots of new versus baseline traces, a band-by-band description of any differences, and a clear conclusion with a recommendation. The conclusion should be honest about confidence: "no significant change from fingerprint, unit suitable to return to service subject to the remaining tests" is a strong, useful statement, while "differences in the mid-frequency band on phase B suggest winding displacement; recommend internal inspection before re-energisation" is the kind of clear, actionable finding that justifies the test. Vague conclusions help nobody.
For engineers building a testing programme, the practical routine is simple: fingerprint every new or overhauled transformer at commissioning, repeat SFRA after any fault or move, and keep every trace with its full test record in one file per transformer. Combined with the routine tests in our transformer testing and commissioning guide, and with numbers checked using our free transformer calculators, SFRA turns the invisible mechanical condition of a winding into evidence you can act on — and that is exactly what a good diagnostic test should do.
Frequently asked questions
Can SFRA be done on an energised transformer?
No. SFRA is an offline test performed with the transformer de-energised, isolated, earthed and discharged. The small test signal would be meaningless and unsafe against system voltage, and the test requires access to the winding terminals.
How long does an SFRA test take?
A full set of configurations on a three-phase transformer typically takes a few hours including setup, with the actual sweeps taking minutes per configuration. Most of the time goes to safe isolation, connections and careful documentation.
Does SFRA replace other transformer tests?
No. SFRA checks mechanical condition; it does not replace winding resistance, turns ratio, insulation resistance, tan delta, oil analysis or partial discharge testing. Each test answers a different question, and a sound assessment uses them together.
← Back to the transformer testing guide · Next: tan delta testing → · Partial discharge testing → · CT and PT testing →