Engineering practice, not a certification. This page states established practice and names the public standards to obtain. It does not certify a product, a factory, or a supplier, and it is not a test report. Use the edition named in the contract. Questions: hello@transformerpath.com.
Frequency-response analysis applies a low-voltage signal to a transformer terminal and records how the output changes across a range of frequencies. The shape of the trace is set by the inductances and capacitances of the windings, the core, and the connections. IEC 60076-18 is the measurement standard. It defines how to make a trace that someone else can compare later: connections, taps, frequency range, and the record of the setup. IEEE C57.149 is the IEEE guide to the application and interpretation of that kind of measurement. The two documents belong together in a procedure. One is how to measure. The other is how to think about the comparison. Using a different connection from the baseline makes a difference that looks like damage and is only a setup error.
The trace is not a pass-fail voltage. There is no single decibel limit that means "the winding has moved" for every design. Interpretation compares a trace with a baseline on the same unit, and sometimes with a sister phase or a sister transformer of the same design. IEC 60076-18 and IEEE C57.149 describe those comparisons and the frequency bands in which a difference is associated with the core, the windings, or the connections. Read the band in the guide you named. Do not treat every wiggle above a few kilohertz as a failed winding, and do not ignore a large difference in the winding band because the transformer still has a normal ratio.
The useful comparison is this unit against itself, before and after a stress. The stresses that justify a comparison are transport, a severe through-fault, a nearby short circuit, a bushing failure, and relocation. The baseline is the trace taken at the factory, or a trace taken on site before the unit is moved, on the taps and the connections the later test will repeat. If the factory trace was not required in the contract, the first site trace can only be a baseline for the future. It cannot prove that the road did nothing. The factory-acceptance guide on this site says to ask for the IEC 60076-18 record before shipment, in a format the site instrument can open.
Sister-unit comparison helps when the sister is the same design and the traces were taken the same way. It is weaker than a baseline on the same serial number. Two healthy units of the same design can differ a little. Two traces of the same unit, taken with the same leads, should not differ by a shift the guide calls significant. The report has to say which comparison was used. A conclusion that says "failed FRA" without the overlay is not a conclusion.
Ratio, winding resistance, excitation current, and the frequency-response trace answer different questions. A shorted turn changes ratio and often changes the low-frequency response and the excitation current. A broken strand or a poor contact changes resistance. Bulk movement of a winding can change the frequency response while the ratio, which depends on turns, stays inside tolerance. That is why a normal ratio does not clear a bad frequency-response comparison, and a normal frequency-response comparison does not clear a wrong ratio. After transport or a fault, do the set. IEEE C57.152 puts these measurements in the field-test context. The site-test guide on this site lists them in the order a commissioning crew can use.
The tap position, the oil level, and whether the unit is oil-filled all change the trace. IEC 60076-18 tells you to record them. A comparison between a drained factory trace and an oil-filled site trace, without the guide's allowance for that difference, is not a valid comparison. Leads draped differently, a different bushing short, or a grounded tank connection that was not in the baseline will move the high-frequency end. Repeat the setup before you repeat the alarm.
Impact recorders and a visual inspection come first. A recorder event above the manufacturer's limit, oil stains, a shifted bushing, or a radiator leak are reasons to test, and they are reasons to involve the manufacturer before energization even if the first trace looks ordinary. Frequency response is sensitive to some forms of movement and less sensitive to others. It is not a complete mechanical inspection. It does not see every loose blocking piece. It does not replace an internal inspection when the other evidence says the tank took a shock. It does give a comparison that an internal inspection cannot give once the unit is closed and filled, which is why the baseline matters.
After a through-fault, the same logic applies. Protection records say how large the current was and how long it lasted. IEC 60076-5 is about the ability to withstand short circuit at the design stage. It is not a field tool for deciding that this particular fault was harmless. If the current was within the nameplate withstand and the traces, the ratio, and the resistance match the baseline, the usual course is to return the unit to service with a plan for the next sample. If the traces have changed in the winding band, the usual course is to stop and to investigate. The words "usual course" are not a requirement. The owner's procedure and the manufacturer, with the traces in front of them, decide.
In the factory contract, require a measurement to IEC 60076-18 on each phase and on the tap positions the site procedure will use, with the connection diagram stored in the same file as the data. Require the electronic record, not only a plot. In the site procedure, require the same standard, the same taps, and a comparison using IEEE C57.149 or the manufacturer's written method, with the overlay kept. Say who is allowed to call a difference acceptable. A contractor and an owner who use different software scales will argue about a picture. Agree the scale and the baseline file before the unit ships.
Do not specify a decibel number as a universal rejection limit unless an engineer has taken that number from the edition of the guide you are willing to defend, for the frequency band it belongs to. A copied number from a training course is how sound units get rejected and damaged units get accepted. The guide's method is a comparison by a person who can see the setup record. The specification should buy that method, the baseline, and the time to do it. It should not buy a false precision the documents do not contain.
Frequency response does not measure dissolved gas, oil quality, or remaining insulation life. Those are IEC 60599 or IEEE C57.104, IEC 60422 or IEEE C57.106, and the loading guides IEC 60076-7 and IEEE C57.91. A transformer can have a perfect overlay and wet oil, or a stable gas history and a winding that moved. Keep the tests separate in the report so a later reader can see which question was asked. File the trace with the serial number, the date, the tap, the oil state, and the name of the person who compared it. The next comparison, after the next move or the next fault, is only as good as that file.
Keep the connection photograph or the connection schedule with the file. A later test crew that has the data and not the schedule will repeat the measurement in a convenient way and produce a trace that cannot be overlaid. IEC 60076-18 exists so that two crews, years apart, can make the same measurement. The schedule is part of the baseline, in the same way the serial number is part of the oil sample.
Other guides
- How to read a power transformer nameplate
- What belongs in a transformer inquiry
- Cooling codes and temperature rise
- Vector group, clock number, and paralleling