If you are buying a transformer from an overseas factory, power transformer testing is the last part of the purchase you can still influence after the design is frozen — and it is where a specification turns into evidence. IEC 60076 sorts every test a transformer can undergo into three classes: routine, type and special. Which class a test belongs to decides whether it is inside the quoted price, whether it delays delivery, and whether a certificate from an earlier project can be accepted instead of a new test.
This guide is written for procurement managers, EPC contractors, MEP consultants and distributors who have to sign off a factory acceptance test (FAT) on a unit built in China or anywhere else. It covers what each test class proves, the acceptance limits in IEC 60076-1, what a complete FAT dossier must contain, and the readings that should stop a shipment.
Mixing up the three classes is the most common misunderstanding in a transformer test report. Routine tests confirm that the individual unit in front of you was built correctly. Type tests prove a design, once, on one representative unit. Special tests are everything the buyer asks for beyond that.
| Class | Performed on | Inside the price | What it proves |
|---|---|---|---|
| Routine | Every unit | Yes, standard | This serial number matches its rating plate and is free of manufacturing defects |
| Type | One unit of each design | Usually yes, once per design | The design itself meets the guaranteed performance |
| Special | Only when specified | No — extra cost | A characteristic beyond the standard requirement |
A type-test certificate from an identical design built earlier is normally acceptable, provided it covers the same rating range and the same core and winding construction. Routine tests are never transferable — they belong to the specific unit you are buying. A FAT is not a fourth class of test either: it is the routine programme run with you, or your inspector, watching. Nothing extra is measured.
Routine tests are defined in IEC 60076-1 and cannot be waived by either side. They are non-destructive and repeatable, which is why they can be applied to a whole production run.
| Test | What it verifies | Typical acceptance check |
|---|---|---|
| Winding resistance, all windings and taps | Bad joints, wrong conductor cross-section, wrong number of turns | Phase-to-phase deviation within about 2% of the average |
| Voltage ratio and phase displacement | Turns ratio at every tap and the correct vector group | Within ±0.5% of the specified ratio at the principal tapping |
| Short-circuit impedance and load loss | Conductor losses and the impedance your fault study assumed | Inside the IEC 60076-1 tolerance of the guaranteed value |
| No-load loss and no-load current | Core steel grade and core assembly quality | Inside tolerance of the guarantee; no-load current within +30% of design |
| Separate-source AC withstand | Insulation from each winding to earth and between windings | No flashover for the full applied test duration |
| Induced overvoltage withstand | Turn-to-turn and interlayer insulation | No collapse, and partial discharge within limits where measured |
| Insulation resistance | Dryness and cleanliness of the insulation at the moment of test | Recorded as a baseline value, not as pass or fail |
| Oil tests, liquid-immersed units | Dielectric strength, moisture, acidity and a dissolved-gas baseline | Per IEC 60156 and the limits in your specification |
| Tank leak and pressure test | Weld, gasket and bushing-seal integrity | No leakage and no pressure drop over the hold period |
Losses and impedance are the commercial heart of the routine test, because they are measured on the unit you are actually buying rather than on a sibling. Load loss is measured at rated current and every result has to be corrected to the 75 °C reference temperature; if the report does not state the correction method, ask for it. No-load loss is measured at rated voltage and frequency with the secondary open, and it is the clearest signal of core quality and assembly.
One caution. A no-load loss well below the guarantee is not automatically good news. Check the test voltage and frequency waveform, the calibration record, and whether the tested unit matches the approved design before you treat the saving as real. An unexplained low reading is a question, not a bonus.
Dielectric routine tests follow IEC 60076-3. For the medium-voltage classes most export orders fall into, the standard separate-source AC test levels are:
| Highest voltage for equipment (Um) | Separate-source AC test | Standard lightning impulse level |
|---|---|---|
| 12 kV | 28 kV, 60 s | 75 kV |
| 24 kV | 50 kV, 60 s | 125 kV |
| 36 kV | 70 kV, 60 s | 170 kV |
| 52 kV | 95 kV, 60 s | 250 kV |
| 72.5 kV | 140 kV, 60 s | 325 kV |
A report that records “applied voltage: PASS” with no voltage level and no duration is not evidence. Insist on numeric values, the test duration, and the connection diagram used.
Where partial discharge (PD) is measured, IEC 60076-3 sets a ceiling of 100 pC at 1.1 Um for equipment up to and including 72.5 kV. PD is a routine test on many higher-voltage orders and a special test on others, so confirm where it sits in your contract rather than assuming it is included. A quiet PD reading taken in a noisy bay at the wrong time of day is worth very little — ask when the measurement was taken and what the background noise level was.
On oil-immersed units, add oil dielectric strength (breakdown voltage, per IEC 60156), moisture content, acidity and interfacial tension, plus a dissolved gas analysis (DGA) baseline before shipment. That baseline is the reference your site team will compare against after transport and again after six months in service. If you attend the FAT, watch the sample being drawn — a DGA result with no chain of custody is hard to defend in a warranty discussion.
Type tests are performed on one unit representative of a design family. Once the design passes, the certificate normally covers later units of the same design, which is why the tests are not repeated for every order. Two of them carry most of the weight.
Dry-type units add partial discharge measurement and, where the site demands it, sound-level verification. The ratings a compliant dry-type unit has to state — voltage class, temperature class, PD level, clearances and enclosure — are set out in our reference on dry-type transformer specifications.
Three checks settle it. First, does the certificate cover the same design family and rating range, not merely the same manufacturer? Second, is it issued against the edition of the standard your contract names? Third, does the laboratory’s accreditation actually cover that test? Where a certificate is borderline, the cheapest protection is a clause stating that the accepted type-test evidence is attached to the purchase order as an annex.
Special tests are the ones agreed between buyer and supplier at contract stage. They are not required by the standard, and every one of them adds cost — and sometimes weeks.
| Special test | Reference | Specify it when |
|---|---|---|
| Short-circuit withstand | IEC 60076-5 | Utility, infrastructure or high-fault-level duty, where mechanical withstand must be demonstrated instead of calculated. Run at an independent high-power laboratory; the tested unit is generally not saleable afterwards |
| Switching impulse withstand | IEC 60076-3 | Transmission-level units where switching surges govern the insulation design |
| Sound level measurement | IEC 60076-10 | The substation sits near housing, a hospital or offices, or planning consent sets a noise limit |
| Partial discharge measurement | IEC 60270 | High-voltage, cast-resin, data centre, offshore and renewable duty — anywhere long insulation life under continuous stress matters |
| Zero-sequence impedance | IEC 60076-1 | Earth-fault protection settings must be calculated precisely, typically with star windings and solidly earthed neutrals |
| Harmonics of no-load current | IEC 60076-1 | The unit feeds or is fed by converters, variable-frequency drives or rectifiers |
| Frequency response analysis (FRA) | IEC 60076-18, IEEE C57.149 | Large units where you want a fingerprint to compare against after transport and installation |
| Capacitance and dissipation factor (tan δ) | IEC 60076-3 | Oil-immersed units where a condition-monitoring baseline is part of the asset strategy |
| DGA before and after testing | — | You want a gas baseline and confirmation that the tests themselves generated no fault gases |
Do not order the whole menu by reflex. The practical rule is to pick the two or three that your application actually drives. A data centre or an offshore platform justifies partial discharge and sound measurement in a way a rural distribution station does not, and unnecessary tests buy cost and time rather than confidence.
The standard sets a ceiling, not your target. These are the principal-tapping limits for a two-winding transformer:
| Parameter | IEC 60076-1 tolerance |
|---|---|
| Total losses | +10% of the total losses |
| Each loss component (no-load and load) | +15% of that component, provided the total-loss tolerance is not exceeded |
| Voltage ratio at the principal tapping | ±0.5% of the specified ratio, or one tenth of the actual percentage impedance, whichever is lower |
| Voltage ratio at other tappings | ±0.5% of the design turns ratio |
| Short-circuit impedance at the principal tapping | ±7.5% where the impedance is 10% or above; ±10% where it is below 10% |
| No-load current | +30% of the design value |
Notice the asymmetry. The loss tolerances are one-sided and positive, so a supplier can deliver a transformer whose losses are up to 15% worse than quoted, per component, and still be compliant with the standard. If an energy budget or a green-building target depends on those figures, write a tighter contractual tolerance — many buyers negotiate zero positive tolerance on no-load loss — and attach it to a liquidated-damages clause. Also confirm which edition of the standard your contract references, because older editions number the test clauses differently and the numbers move with the edition.
A FAT adds no new measurements. What it adds is your presence when the measurements are taken, and the right to challenge a reading before the unit is crated. On an export order the FAT is the last cheap moment to find a problem: correcting a defect at the factory costs a fraction of correcting it after a long sea voyage, a crane lift and a site crew on standby.
There are three options and they are not mutually exclusive. Attending in person gives the strongest evidence, because your engineer can watch the zero check on each instrument before any measurement is taken. Remote witnessing by live video works when it is structured: a fixed agenda, a camera that shows the serial number plate, the instrument displays and the test connections, and the same signed report package afterwards — plus a realistic time-zone plan, since dielectric tests on a quiet bay are often run at night. Third-party inspection through SGS, Bureau Veritas, TÜV or Intertek is the usual compromise when travel is not possible, and it also covers the packing check.
Whichever route you choose, fix it in the contract: notice period, test agenda, accepted remote method, report language, and a rule that shipment is not released until open points are closed.
| Reading | What it usually means |
|---|---|
| Winding resistance imbalance above 2% between phases | Wrong number of turns, a bad tap connection or a failed crimp |
| Load loss more than 5% above the guaranteed value | Undersized conductor or poor-quality copper, and higher operating temperature in service |
| Impedance more than 5% away from the specified value | Fault levels and voltage regulation fall outside your system study |
| Applied voltage test recorded as passed with no voltage level or duration | The test as described cannot be verified, so it is not evidence |
| No-load loss far below the guarantee with no explanation | Check test voltage, frequency and whether the tested unit matches the approved design |
| Any test run with out-of-calibration instruments | Every reading taken with that instrument is void |
Most FAT disputes are created at quotation stage, not on the test floor. Close these points before the order is confirmed:
For the wider picture of how testing evidence fits into supplier qualification, see our guide to choosing a power transformer manufacturer in China and the capability summary on our transformer manufacturer page.
A transformer can pass the FAT and still fail at site. Transport can shift a core, moisture can enter a poorly sealed unit, and a tap changer can move. The site acceptance test repeats the fast checks — insulation resistance, winding resistance and ratio, oil tests, and on larger units a capacitance and dissipation factor measurement — and compares them with the FAT numbers corrected for temperature. That comparison is only possible if you received the FAT data, which is the real reason to insist on the full dossier rather than a certificate of conformity.
Power transformer testing is not paperwork to be collected at the end of a project. It is the mechanism that converts a specification into verified performance, and the three-class structure in IEC 60076 tells you whether a given test is standard, once-per-design, or an extra worth paying for. Put the tolerance table in your contract, require numeric results rather than ticks, agree the witness arrangement in advance, and treat FAT approval as the gate that releases the balance payment. Buyers who do those four things rarely discover problems at commissioning.
Kampa Electric builds oil-immersed and dry-type transformers from 11 kV to 35 kV with in-house routine testing, type-test documentation and support for witnessed or third-party FAT, including dry-type power transformers for industrial, utility and commercial projects. If you have a specification sheet ready, send it to our engineering team for a factory-direct quotation with a proposed test plan.
Routine testing on a single unit typically takes a few hours to a day, so a full witnessed FAT usually fits into three to five days including visual inspection, dimensional checks and the accessory review. Add 24 to 48 hours if a temperature-rise test is running on the bay at the same time, and allow extra days if impulse or special tests are scheduled.
A routine test is performed on every unit and proves that this particular transformer was built correctly. A type test is performed once on a representative unit and proves that the design meets the specified performance. Routine results are never transferable between units; a type-test certificate normally is, provided the design family and rating range match.
No. A FAT is the routine test programme witnessed by you or your appointed inspector, and it measures nothing extra. A type test is a separate, more severe programme — temperature rise, impulse withstand — carried out once per design, often years before your order.
The routine set under IEC 60076-1: winding resistance, voltage ratio and phase displacement, short-circuit impedance and load loss, no-load loss and current, and the dielectric routine tests per IEC 60076-3. Insulation resistance and, on oil-immersed units, oil dielectric strength, moisture and a DGA baseline are standard practice even where they are not strictly mandatory.
Usually yes, if the certificate covers the same design family, the same rating range and the same core and winding construction, and if it was issued against the edition of the standard your contract references. Check the accreditation of the issuing laboratory as well, and attach the accepted certificate to the purchase order as an annex.
Total losses +10%, each individual loss component +15% provided the total is not exceeded, no-load current +30%, voltage ratio ±0.5% at the principal tapping, and short-circuit impedance ±7.5% where the impedance is 10% or above and ±10% where it is below. Tighter tolerances must be agreed in writing before the order is placed.
Not always. It is a routine requirement in many higher-voltage specifications and a special test in others. Specify it where insulation life under continuous stress matters — cast-resin dry-type units, data centres, offshore substations and renewable projects — and confirm in the contract whether it is routine or special for your order.
