Sweep frequency response analysis (SFRA) compares the frequency response of a transformer winding with a reference fingerprint to detect mechanical and magnetic circuit changes such as winding deformation, core movement and faulty connections. The reading is a trace, not a single number: correct interpretation compares traces under identical conditions, reads deviations band by band, separates setup effects from real changes, and escalates only what the evidence supports.
Overclaiming from a trace is the most common interpretation error.
What an SFRA Trace Actually Shows
An SFRA measurement injects a swept-frequency low-voltage signal into a winding and records the response over a range that typically extends from a few hertz to several megahertz. The trace is a plot of the response against frequency, and different frequency regions are dominated by different parts of the transformer: the core at low frequencies, the winding geometry in the middle range, and the internal connections and structure at the higher end. Because the response depends on capacitances, inductances and the physical arrangement of the windings, any mechanical change that alters that arrangement changes the trace.
The word “fingerprint” is accurate: SFRA is a comparison method. A trace has diagnostic value only when it is compared with a reference obtained from the same transformer under the same conditions, or with traces from a sister unit of identical design. The reference may come from the factory, from an earlier site test, or from a measurement taken immediately after commissioning before the unit entered service.
Building and Reusing a Reliable Reference Fingerprint
The quality of the reference decides the quality of the interpretation. A factory fingerprint is valuable only if the site measurement can reproduce the same connection arrangement, lead lengths and measurement settings; otherwise differences in setup can be mistaken for differences in the transformer. When a reliable factory fingerprint is not available, the first site measurement after commissioning becomes the baseline for all later tests, which is one reason SFRA is worth performing before a new unit enters service.
Record everything needed to repeat the measurement: which winding and terminal pairs were used, the lead configuration, the grounding arrangement, the temperature and the instrument settings. Some test sets store this information with the trace. If the setup cannot be reproduced, the comparison is weakened no matter how carefully the trace is analysed.
Reading the Frequency Bands: Core, Windings and Structure
Interpretation begins by looking at the trace in frequency bands rather than at isolated points. The table below shows the typical relationship between frequency range and the part of the transformer that dominates the response. It is a reading guide, not a diagnostic law: the exact boundaries vary with transformer size, design and test setup.
| Frequency region | What dominates the response | What a deviation may relate to |
|---|---|---|
| Low, up to roughly 2 kHz | Magnetizing inductance of the core | Core movement, residual magnetism, changes in the magnetic circuit |
| Mid, roughly 2 kHz to 20 kHz | Winding inductances and inter-winding capacitances | Axial or radial winding movement, clamping changes |
| Higher, above roughly 20 kHz | Internal lead geometry and structural capacitances | Lead displacement, connection and structural changes |
Read the bands in order. A deviation confined to the low-frequency region points toward the magnetic circuit; a deviation in the mid band points toward the winding geometry; a high-frequency change suggests a structural or connection issue. When several bands move together, consider whether a single cause, such as transport shock, explains the whole pattern.
Quantifying Deviation Without Overclaiming
Several numerical indices have been proposed to compare SFRA traces, including correlation coefficients and band-specific deviation measures. These indices are useful screening tools, but they do not replace engineering judgement, and they are not interchangeable: an index that works for one transformer family may mislead on another. A trace that looks similar to the reference but has a small band-specific shift can matter more than a trace that looks different overall but changes only in a benign region.
When numbers are used, state what was calculated, over which frequency range, and against which reference. Do not convert an index into a pass or fail verdict unless the threshold has been validated for the transformer type and test method in question. The honest statement is often: “the deviation is concentrated in the mid band and exceeds the variation seen in repeated measurements of this unit, so further investigation is recommended.”
Common Setup Errors That Change the Trace
Setup errors are the largest source of false SFRA findings. A different lead length or routing changes the high-frequency response. A different grounding arrangement changes the low-frequency behaviour. Connecting to a different bushing or terminal, or leaving a connection loose, changes the whole trace. Even the position of the test leads relative to earthed metalwork can move the higher-frequency response.
Before interpreting a deviation, ask whether the measurement was repeatable. Make the same measurement twice without changing anything; if the two traces differ, the setup is not stable and the comparison is invalid. Only when repeated measurements agree should the trace be compared with the reference and the deviation treated as evidence.
When SFRA Results Require Escalation
SFRA is most valuable when it is used in a decision context: after transport, after a through-fault or short circuit, after a suspected mechanical event, or as part of a condition assessment on an aging unit. A clear band-specific deviation that is repeatable, that cannot be explained by setup, and that corresponds to the expected consequence of the event is a strong reason to escalate to a fuller investigation, including internal inspection where access and safety allow.
The decision framework for when these tests are needed is covered in What Transformer Tests Are Needed After Transport or a Short Circuit. SFRA findings should be combined with turns ratio, winding resistance and excitation data before a conclusion is reached, because a single test rarely identifies the full extent of damage.
Comparison with a sister unit is a useful second reference when no baseline exists. Two transformers of identical design, made in the same batch and tested with the same setup, should produce similar traces; a consistent difference between them in one frequency band is evidence worth investigating even when the absolute shape is unfamiliar. The comparison is only valid if the sister unit is genuinely identical, including the tap position and the connection arrangement, and if both units are tested with the same leads and settings. When repeated measurements of the same transformer scatter in the low-frequency region, check grounding and residual magnetism before reading core movement into the trace; when the scatter appears in the high-frequency region, suspect the lead arrangement rather than the internal structure. Only a deviation that is repeatable, band-specific and consistent with the event history should be reported as a likely mechanical change, and even then the report should name the supporting evidence and the recommended confirmation test rather than declaring a verdict.
Two practical habits improve trace comparison more than any index. The first is to overlay traces rather than compare them side by side from memory: plot the reference and the new measurement on the same axes, with the same frequency scale, and look at where they separate. The second is to state the comparison in writing before drawing a conclusion, using the same structure every time: which reference was used, whether the setup was reproduced, which frequency bands show separation, whether the separation is consistent with the event history, and what confirmation test is proposed. Reports written this way are easier for a second engineer to review and easier to defend if the finding later drives a repair decision. Where a trace is being used to support an insurance or warranty claim, the discipline matters even more, because the record itself becomes evidence and every missing setup detail weakens it.
Frequently Asked Questions
What do the frequency bands mean in an SFRA trace?
In a typical transformer trace, the low-frequency region is dominated by the core, the mid region by the winding inductances and capacitances, and the higher region by internal leads and structure. A deviation concentrated in one band points toward the component family that dominates that band, although the exact boundaries vary with transformer design.
Why do I need a reference fingerprint?
SFRA is a comparison method. A trace has meaning only against a reference measured under the same conditions, whether from the factory, from commissioning or from a sister unit. Without a reproducible reference, most apparent deviations cannot be distinguished from setup differences.
Can a numerical index tell me if the transformer is damaged?
Index values are screening tools, not verdicts. They summarize how far a trace deviates from a reference, but their thresholds are only valid for the transformer family and method for which they were developed. Use them to prioritize traces for review, then interpret the band pattern and setup repeatability before escalating.
For the instrumentation used in transformer diagnostics, see the transformer maintenance test equipment range, and for the complete test programme read What Tests Are Required for a Power Transformer.