Transformer Winding Resistance Test: Method and Limits

Inject DC, measure the voltage drop, divide. That’s the whole test on paper.

In practice the reading climbs for ten minutes before it settles on a large unit. The value you write down is meaningless unless you also wrote down the winding temperature. The number your meter shows is not the resistance of the winding you think you measured — not on a delta connection. And when you finish, the core is left magnetized, so whoever runs the TTR after you gets garbage.

None of that makes the test hard. It makes it procedure-sensitive. Get the details right and winding resistance catches loose connections, broken strands, worn tap changer contacts, and slow winding deterioration — faults that stay invisible to most other terminal-accessible tests. Get them wrong and you’ve spent an afternoon producing a number nobody can use.

This covers the math, the practical decisions, and the interpretation.

What the number actually contains

The test measures DC resistance from terminal to terminal. Several things sit inside that single value:

Conductor resistance. Hundreds or thousands of meters of copper or aluminum wound into the coil. On a healthy winding this dominates everything else.

Joint and contact resistance. Brazed joints inside the winding, bolted connections at the terminals, tap changer contacts in the current path, bushing connections. Individually small. Collectively significant when one of them degrades.

Broken strands. Modern windings use stranded conductor, often continuously transposed cable. A few broken strands raise resistance slightly without causing an obvious failure. Nothing else you can run from the terminals will find them.

What the test does not measure: insulation, capacitance, inductance, or any AC behavior. It interrogates one thing — the conductive path — and that narrow focus is exactly why it finds defects that AC-based tests miss.

The four-wire measurement

Use a Kelvin four-wire connection. Always.

Two current leads carry the injected DC. Two separate sense leads measure the voltage drop across the winding itself. Because almost no current flows in the sense leads, their own resistance drops out of the measurement.

The reason this matters becomes obvious on an LV winding. A large transformer’s LV winding might read 3 mΩ. Your test leads, clamps, and contact points easily add 10 to 50 mΩ. A two-wire measurement on that winding is measuring your leads with a small transformer correction. It isn’t a marginal error — the reading is wrong by an order of magnitude.

Sense leads go inboard of the current clamps, contacting the terminal directly. If the sense connection picks up part of the current path, the lead resistance comes back into your reading.

Converting what the meter reads into phase resistance

Here’s where field values get compared against factory values incorrectly.

Your meter reads what’s between the two terminals you clamped. That is not the same as one phase winding. What the relationship is depends on the connection.

Star (wye) with neutral brought out

Measure line to neutral. What you read is the phase winding directly.

Star without accessible neutral

Measure line to line. You’re reading two phase windings in series.

Delta

Measure line to line. You’re reading one phase winding in parallel with the other two in series.

Which inverts to:

That factor of 1.5 is the one people forget. Skip it and every delta winding you test appears to have lost a third of its resistance against the factory report.

Worked example

A 33 kV delta HV winding. Measured between H1 and H2 at 28 °C: 4.50 Ω.

Phase resistance:

Corrected to 75 °C (copper):

Factory test report gives 7.85 Ω at 75 °C.

Deviation = (7.96 − 7.85) / 7.85 = 1.4%

Within tolerance. Note how much had to happen before that comparison meant anything: the connection had to be identified, the delta factor applied, the temperature recorded, and the correction run. Miss any one of them and the same measurement produces a deviation of 15%, 30%, or more — all of it phantom.

A practical note: some test sets output phase values directly if you tell them the vector group. Some output raw terminal values. Know which yours does before you compare anything.

Why DC, and why it takes so long

Why DC. A winding is mostly inductance from an AC perspective. At power frequency the reactance swamps the resistive component, so an AC measurement can’t give you a clean resistance. At steady-state DC the inductor is just a length of wire, and what remains is resistance.

Why the wait. Same inductance. Apply DC and the current doesn’t jump to its final value — it rises on the winding’s L/R time constant. Large inductance and low resistance means a long constant. Seconds on a distribution unit. Minutes on a large power transformer.

The voltage drop only means something once current has stabilized. Read early and there’s still an L(di/dt) term contaminating it, which reads as higher resistance than the winding actually has.

Modern sets manage this with high compliance voltage to speed the rise, and they monitor di/dt, only reporting when the rate of change falls below threshold. The physics still rules, though. Wait for the stability indicator. Don’t read at a fixed time.

Choosing the test current

The standards frame test current as a percentage of rated winding current. The usable band runs roughly 0.1% to 10% of rated, and both ends have real constraints.

Too low and the voltage drop is too small to resolve. Noise dominates. The reading wanders.

Too high and two things go wrong. I²R heating warms the winding during the measurement, so resistance is drifting while you read it — trending becomes impossible. And sustained high DC saturates the core more heavily than necessary, which compounds the demagnetization problem afterward.

Working guidance:

Winding	Typical test current
HV winding (high R, low rated current)	1–3 A, minimum 1 A for signal
LV winding (low R, high rated current)	Up to ~5% of rated; 20–50 A common
Winding above 100 mΩ	10 A or less is sufficient
Winding below 100 mΩ	20–50 A for acceptable accuracy

Most modern instruments range from 10 mA to 50 A and select automatically. Choosing manually, lean toward the upper end for accuracy but stay well clear of 10% of rated.

Core saturation on large units

On a large power transformer the inductance can be high enough that current settling becomes genuinely painful — fifteen minutes for a single reading is not unusual.

The traditional fix is to saturate the core. Once saturated, inductance collapses to a fraction of its unsaturated value and the time constant collapses with it. Current settles fast.

The problem is that saturating a large core can take 100 A or more, and most field kits can’t deliver that.

The workaround is dual-winding excitation: route the test current through both the primary and secondary windings in series. The primary-side current contributes ampere-turns toward saturating the core even though you’re measuring the secondary. Total ampere-turns go up, the core saturates, and settling time drops sharply.

Many transformer resistance kits build this in as a standard mode. If a reading simply refuses to settle, check whether yours supports it. It can turn a fifteen-minute measurement into under a minute.

Demagnetization

The DC you injected leaves the core magnetized. Two consequences.

It corrupts the next test. Excitation current measurements — inside TTR, in no-load loss testing, in dedicated excitation tests — are sensitive to residual magnetism. Run winding resistance and go straight to TTR and the TTR results are unreliable.

It affects energization. A transformer energized with significant residual flux can draw far higher inrush than design on the first cycle. Differential protection may operate. The DC offset in that inrush can cause unexpected trips.

Modern sets include a demagnetization routine that applies a decaying alternating current until residual flux is near zero. Run it after every resistance test. It costs a few minutes.

Test sequence: put winding resistance after TTR and excitation testing. If it has to come first, demagnetize before you move on.

Temperature correction

Copper resistance changes about 0.39% per °C near room temperature. Aluminum is similar. Between 20 °C and 75 °C, that’s a difference of roughly 22%.

So a field reading at 25 °C compared against a factory value at 75 °C shows a 20%-plus “deviation” that is entirely temperature. Every acceptance and trending criterion assumes both values sit at the same reference.

The reference is normally 75 °C for oil-immersed transformers. Some classes use 85 °C. Check what the factory report used and correct to the same figure.

The formula

For copper windings:

R₇₅ = R_measured × (234.5 + 75) / (234.5 + T)

T is the actual winding temperature in °C. The constant 234.5 is the inferred zero-resistance temperature for copper — older references round it to 235. For aluminum, substitute 225.

This is not the same correction used for insulation resistance. Different physical mechanism, different formula. Don’t reach for the IR temperature tables here.

Getting T

You can’t put a thermometer in the winding. The conventions:

• Oil-immersed, de-energized long enough to equalize — use average oil temperature. Top and bottom oil should be within about 5 °C of each other. Reaching that typically takes 3 to 8 hours depending on size.
• Dry or drained — use ambient air temperature.
• WTI reading meaningfully — use it.

The shortcut everyone takes is “just use the oil temperature,” and it works only once thermal equilibrium is reached. A transformer pulled out of service an hour ago has hot windings sitting in cooler oil. Use the oil temperature in that state and your correction is wrong in a direction that looks like a healthy winding.

Record the temperature with every reading. A resistance value without a recorded temperature can only be compared to other values at the same unknown temperature — which means it can’t be trended at all.

What values to expect

There’s no lookup table for this. Winding resistance depends on rating, voltage class, conductor cross-section, and winding geometry. Any table claiming to give “typical values” is wrong more often than right.

What is useful is order of magnitude, purely as a sanity check that you’re on the right range and the connection is sound:

Winding	Expected order
HV winding, power transformer	Ohms to tens of ohms
LV winding, power transformer	Milliohms to tens of milliohms
LV winding, distribution transformer	Single-digit milliohms
Tertiary winding	Between the two, closer to LV

If your HV reading comes back in milliohms, you have a connection problem or you’re on the wrong terminals. If your LV reading comes back in ohms, same conclusion.

The real reference is always the factory test report for that specific serial number. Get it before you go to site.

Acceptance criteria

Three comparisons, used differently depending on whether the unit is new or in service.

Phase-to-phase agreement

All three phases within about 1% of each other, at the same tap and same temperature. Wider spread points to a fault on the deviating phase.

This is the most useful field check because it needs no historical record. All three phases are right in front of you. And if you measure them within minutes of each other, they’re at the same temperature — so this comparison survives even without correction.

Against the factory value

After correction to the same reference, field values should land within about 5% of the original factory value. That tolerance absorbs instrument differences, lead effects, and temperature measurement error.

• 2% — unremarkable
• 5% — investigate
• 10% — clear flag

Trending across successive field tests

Change beyond about 2–3% between successive tests, same tap and same corrected temperature, is significant. Even if every individual value still sits inside 5% of factory.

On an in-service unit the trend is usually the sharper diagnostic. A winding creeping upward over several years of testing is telling you something is developing, regardless of whether the absolute value still passes.

The 5% and 2% figures above are the widely applied field values. Confirm against the acceptance table in the standard your job is contracted to — IEEE C57.152 for maintenance testing, IEC 60076-1 for the measurement itself. IEC specifies how to measure; it does not publish a blanket pass/fail deviation. That distinction catches people out on international projects.

Both the factory comparison and the trend collapse without temperature correction. Only the phase-to-phase check survives it.

Testing every tap

If there’s a tap changer, test every position. Off-circuit and on-load both, through the full range.

Resistance change per tap step should be smooth and roughly uniform. The progression tells you whether the selector is making proper contact everywhere. A tap reading markedly higher than its neighbors has a contact problem. A tap that won’t pass current at all isn’t making contact.

Static DC testing catches steady-state contact problems. It misses transients — a contact that reads fine standing still can still break momentarily during the actual operation. Dynamic resistance measurement, where resistance is monitored continuously through a tap change, catches those. Worth running whenever OLTC condition is the open question.

Fault signatures

What you see	What it usually means
One phase high	Loose connection or developing fault on that phase. Check terminals first, then suspect internal joint or strand damage. Often shows as slight ratio error and elevated excitation current on the same phase in TTR.
All three uniformly high vs factory	Almost always a temperature correction error. Recheck the temperature assumption and the delta/star conversion before suspecting the transformer.
One tap markedly different	Tap changer contact on that position. High but stable means worn or oxidized. Unstable means intermittent.
Rising across successive tests	Deteriorating connection or progressive strand damage. Correlate with DGA — localized heating at a degrading joint often shows in CO and CO₂ trends.
Settles slower than usual	Check setup before suspecting the winding. Connections, leads, grounding.
Won’t settle at all	Test setup, not the transformer. Kelvin clamps, lead integrity, grounding.

Setup points that change the number

Clean the contacts. Bushing terminals oxidize. Clamp jaws get dirty. A poor instrument-to-bushing contact reads as high winding resistance and there’s nothing in the number to tell you which it was.

Sense leads inboard. If the voltage-sense connection sits in the current path, lead resistance comes back into the reading.

Wait for stability. The instrument’s indicator, not the clock.

Hold conditions constant for trending. Same instrument, same test current, same tap, same corrected temperature. Every variable you change contaminates the trend.

Discharge before disconnecting. The voltage is modest but the stored energy in a large inductor is not. Modern sets discharge automatically. Wait for the indicator anyway.

FAQ