The test is simple to perform and difficult to interpret in isolation. A single-phase alternating voltage is applied to one winding while the others are left open, and the resulting magnetising current is measured at each phase. The current reflects the magnetic circuit: how much flux the core can carry for a given applied voltage, and how much loss the magnetic path introduces.

What excitation current reveals at the core

Excitation current is the current required to establish flux in the core. Its magnitude depends on the core material, the effective cross-sectional area, the number of turns energised, the length of the magnetic path and the presence of any short circuit that diverts flux or creates a circulating current. Anything that changes one of those five quantities changes the measured current.

That list is the reason the test is useful during commissioning and after a fault. A shorted turn in the energised winding reduces the number of effective turns, which requires more current for the same flux. A core limb that has lost its insulation coating produces circulating eddy currents that increase loss. A core that has shifted, or a clamping structure that has loosened, changes the magnetic path. All of these appear as a change in the current pattern rather than as a change in resistance or ratio.

The test is also used as a receiving check after transport. A core that took a mechanical shock during shipping may not show a ratio deviation and may pass an insulation resistance test, because the insulation between laminations is not part of the winding insulation system being measured. Excitation current can reveal the change before the unit is energised at full voltage.

Reading the pattern rather than the number

The result that matters is the relationship between the three phases, not the absolute value on any one of them. Connected to a three-phase supply and measured phase by phase, a healthy transformer produces a current pattern that is set by its core construction. That pattern is repeatable and characteristic, and the deviation from it is the diagnostic signal.

Core construction determines the expected pattern. A three-limb core gives the centre limb a shorter return path than the outer limbs, so the centre phase typically draws less current than the two outer phases. A five-limb core distributes the return path differently. A shell-form construction behaves differently again. None of these patterns is abnormal, and none should be corrected in the report.

What matters is whether the pattern has changed. For a unit with a previous record, the comparison is direct: if the ratio between the highest and lowest phase current has widened, or if one phase has moved relative to the other two, something in the magnetic circuit has changed. For a unit with no record, the sister unit or the design family provides the reference, and the comparison is weaker because manufacturing variation is included in it.

Shifts that mean shorted turns versus normal variation

A shorted turn in the energised winding produces a clear signature: the current in that phase rises substantially, typically by tens of percent rather than by a few percent, and it stays high on repeat measurement. The rise is large because the shorted turn behaves as a loaded secondary, circulating a current that demands additional primary current.

Normal variation looks different. Core construction, the magnetisation state left behind by a preceding direct current test, and small temperature effects produce differences of a few percent that change between measurements and do not persist in one direction. A shift of a few percent that appears once and does not reproduce is measurement variation; a persistent shift of tens of percent is not.

A core fault produces a third pattern. Loss-related changes, such as a core that has become electrically shorted between laminations, tend to raise the current and the loss across all phases at once, or across a group of phases that share a magnetic path. That pattern points at the core rather than at a winding, and it is the case where the test adds information that ration and resistance testing cannot provide.

Effect of test voltage and supply quality

The current value has no meaning without the voltage that produced it. Excitation current is strongly non-linear with voltage, because the core saturates: a ten percent increase in applied voltage above the knee of the magnetisation curve can raise the current by far more than ten percent. Comparing a measurement taken at one voltage with a record taken at another produces a difference that is entirely artificial.

Supply quality matters for the same reason. Magnetising current is sensitive to waveform, so a source with significant harmonic distortion produces a different current reading than a clean source, even at the same root mean square voltage. Sources with poor regulation cause the voltage to sag as the core draws more current, which flattens the top of the waveform and changes the reading again.

Practical measures follow. Use the same voltage as the previous record, or correct explicitly and say so. Record whether the supply was mains, a generator or an electronic source, and note the waveform quality where the instrument reports it. Where a site generator is the only source available, taking the measurement twice at a short interval reveals how much of any change is supply instability.

Comparison with previous results and sister units

A previous result on the same transformer is the strongest reference available. It removes the design, the core construction, the winding geometry and the material from the comparison, leaving only the change. Given two measurements at the same voltage and the same tap, the difference between them is the physical signal.

Sister unit comparison is weaker but still useful, particularly during commissioning when no earlier record exists. Because nominally identical units share the design and the core construction, they share the expected pattern. A unit whose pattern differs markedly from its sisters is worth investigating. A unit that differs slightly is within the manufacturing spread, and treating that spread as a fault is a common commissioning mistake.

Records need to be comparable for either comparison to work. The applied voltage, the phase sequence used, the tap position where the measurement was taken, the winding that was energised, the supply type and the instrument should all appear in the record, and the units should be stated for every current and voltage value.

Combining with ratio and SFRA

Excitation current is most valuable when it is read alongside the tests that answer a different part of the same question. Ratio testing establishes whether the turns relationship is correct, which separates a winding problem from a core problem. A high excitation current on a phase with a correct ratio points at the magnetic circuit; a high current on a phase with a wrong ratio points at the winding.

Frequency response adds the mechanical dimension. Winding displacement changes the frequency response in the geometry-dominated band while leaving excitation current largely intact, and core faults change the low frequency behaviour of the sweep. Reading the two together turns a set of suspicions into a shortlist.

Resistance testing completes the set by ruling out connection and joint problems, which can mimic a winding fault in the current measurement when the excitation path includes a tap changer or a bushing connection that is not making properly. A ratio meter used in the same pass, such as those in the transformer ratio test equipment range, keeps the connection set unchanged between tests and removes setup variation from the comparison.

Limits of the measurement

Excitation current is a comparative test with a shallow diagnostic depth. It detects that something has changed and narrows the region, but it does not locate the change or size it. A current increase of forty percent in one phase tells an engineer that the excitation path has lost effective turns or gained loss; it does not say whether the cause is one turn or twelve, or whether it sits in the winding, at a tap connection or against the core.

The test is also relatively insensitive to small changes. A single shorted turn in a large winding with many turns produces a change that may fall inside the measurement repeatability. The same limitation applies to incipient core faults, which may take several thermal cycles to develop enough loss to show.

Instrument capability follows from the purpose. A tester intended for excitation and no-load measurement needs a stable, low-distortion source and a current measurement range that covers the low currents drawn at reduced voltage, because accuracy at the low end of the range is what makes a small deviation visible.

Reporting language that avoids overclaiming

Excitation results are frequently written up in language the test cannot support. A report that states a winding is shorted, on the basis of a current deviation alone, invites an outage and a rewind decision that the evidence does not justify. A report that describes the deviation, its repeatability, the conditions and the candidate causes is both defensible and more useful to the person who has to decide.

A workable structure is to state the applied voltage and the measured current per phase, then the comparison against the previous record at the same conditions, then the pattern observed, then the candidate causes in order of plausibility, then the recommended confirmation test. The conclusion should say what the test rules out as well as what it suggests.

Where the result will be used in a warranty discussion or a failure investigation, the record should also note the instrument and its calibration status, the connection configuration, and whether the supply conditions were consistent with the reference measurement. Those details are what allow a reviewer to reproduce the result rather than accept it.

Test voltage selection and its effect on the pattern

Voltage selection is a trade-off between sensitivity and safety. A low applied voltage keeps the core in its linear region, where the current is small, the measurement is quiet and differences between phases reflect the winding turns rather than the magnetic saturation behaviour. A higher voltage takes the core into the non-linear region, where a small change in effective turns produces a much larger change in current, at the cost of a less stable measurement and greater stress on the winding insulation.

The practical implication is that the voltage has to be chosen for the question being asked, and then held constant. When the question is whether anything has changed since the last outage, the same voltage as the last test is the right choice regardless of where it sits on the curve. When the question is whether a unit is fit to energise after transport damage, a voltage high enough to reach the non-linear region gives better sensitivity to a shorted turn.

Recording the voltage in the test record is not bookkeeping. It is the single field that decides whether two excitation measurements can be compared at all.

Faults excitation current cannot distinguish

The test cannot separate a shorted turn from a winding-to-core short on its own, because both reduce the effective turns in the excitation path. It cannot tell a shorted turn from a tap connection fault that bypasses part of the winding. It cannot distinguish an insulation fault inside the winding that has not yet produced a circulating current. And it cannot see a fault in a winding that is not part of the tested magnetic circuit at the time of the measurement.

Those limitations are not defects of the method; they define what it is for. The definition of the transformer characteristics being verified, including the winding configuration the test assumes, is set out in IEC 60076-1, and the dielectric test framework that governs how the unit is stressed afterwards is in IEC 60076-3.

For method detail on how excitation and no-load measurements are configured on modern testers, the technical documentation published by instrument suppliers such as OMICRON is a practical reference. For the wider diagnostic context, including how excitation results are combined with frequency response and dissolved gas data in a condition assessment, the study committee material published by CIGRE and the transformer research programme of EPRI both address the same decision. Applied field interpretation of these patterns is also documented by Transformer Consultants.