Stator Winding Inter-Turn Short: How We Found One and How the Surge Test Catches Them

By the Baoluo Motor Engineering Team · Published 2026-09-20

A customer sent a 48 V 10 kW AC asynchronous traction motor back to us with one sentence on the paperwork: "starts, runs for a second, stops, and the shaft feels notchy when you turn it by hand." They had already replaced the inverter. That did not fix it, which told us the drive was probably innocent and the motor was worth opening.

We went through the two obvious suspects first — the controller and encoder pairing, then the shaft and bearings. Both came back clean. The fault was inside the stator: a handful of turns in one phase had shorted to each other. This is the failure that a multimeter, a megger and a hipot test can all miss, and it is why every stator we build goes through a surge comparison test before it is wound into a housing.

Rule of thumb: if an AC traction motor trips the drive immediately but the inverter checks out, suspect the winding. If the winding measures fine on resistance and insulation to earth, the remaining suspect is the insulation between turns, and only a surge test will show it.

1. What came back from the field

The motor was an XYQ-10PC01: 48 V DC battery supply, 10 kW rated, 34 V AC rated voltage, 220 A rated current, 1560 rpm, a 10–200 Hz operating band, 60-minute duty, IP54, and a 64 pulse-per-revolution encoder on a standard B5 flange with a temperature sensor. Those figures are the published rating for the model.

The symptom pattern mattered more than the symptom itself. "Runs briefly, then stops" is what a drive does when it sees too much current. A winding with a shorted turn draws exactly that. The shorted turns form a closed loop with almost no resistance and almost no inductance, so the current in that loop runs far above normal and the drive shuts down to protect itself. "Notchy when turned by hand" was the part that made us check the mechanics, because that feel usually means bearings or a bent shaft.

2. The two things we ruled out first

2.1 Controller and encoder matching

An AC asynchronous motor is useless without an inverter, and a mismatch between the two is the most common reason a new installation misbehaves. We checked the pairing in this order: encoder pulse count against the controller's setting (64 P/rev here), motor parameters entered into the drive against the nameplate, and the drive's fault log against the reported symptom.

Everything matched. The fault log showed overcurrent on acceleration, which is consistent with a bad winding, and the customer's replacement inverter produced the same log on the same motor. Two identical drives reporting the same fault on one motor points at the motor.

Our C750 series AC motor controller is sized in pairs with our motors for exactly this reason, and we cover the mismatch problem in detail in our AC traction motor failure guide.

2.2 Shaft and bearings

Next we put a dial indicator on the shaft extension and measured runout, turned the shaft by hand through a full revolution feeling for a tight spot, and checked axial play and bearing noise. Runout was inside tolerance, there was no tight spot, and the bearings turned smoothly. The mechanical side was healthy, which left the electrical side — and specifically the winding.

3. What we found inside the stator

We stripped the motor and looked at the winding. One phase had a discoloured band of enamel near the slot exit, and the copper underneath had started to darken. A few turns in that coil had lost the insulation between them and were behaving as one thick turn.

That is an inter-turn short. The shorted turns are a closed circuit sitting inside a much larger one, and because the loop has very little resistance and very little inductance, the current circulating in it can be many times the motor's rated current. The heat is concentrated in a few cubic centimetres of copper. The enamel carbonises, the carbonised path conducts, and the short spreads until the winding finally goes to earth and the motor stops for good.

Two things put a defect like this into a motor. One is damage during manufacture: a sharp lamination edge scraping the enamel, or winding tension pulling the film thin. The other is service stress: voltage spikes at the motor terminals from the inverter, and repeated heating and cooling cracking the film over time. Both leave a weak spot buried between layers where nobody can see it.

4. Why the usual tests miss it

This is the frustrating part for anyone diagnosing a motor on a bench. An inter-turn short can survive the three tests everybody reaches for first.

IEEE Std 43 gives the accepted minimum insulation resistance as roughly (rated kV + 1) MΩ, with a practical floor of 5 MΩ for random-wound low-voltage machines. That test is about the winding-to-earth path. It says nothing about the insulation between two neighbouring turns, which is where this fault lives.

5. The stator is a capacitor, an inductor and a resistor

To understand the surge test you have to stop thinking of a winding as a resistor. A stator winding is a resistor, an inductor and a capacitor all at once, and all three are distributed along the copper rather than sitting in one place.

The inductance is the important one here, because it scales with the square of the number of turns. Take turns out of the circuit and L falls sharply. That single fact is what makes the surge test work.

6. How the surge comparison test works

The idea is simple. You charge a capacitor, then discharge it into the winding through a thyristor. What you get is not a destructive high-energy surge. The stored energy is small enough that the test does no harm and can be repeated, so you can run it on every stator that leaves the factory. The winding responds by ringing, and the frequency of that ring is set by its own L and C:

f = 1 / (2π √(L C))

Then you do the same thing to a known-good reference winding. You put the two traces on top of each other. A healthy winding and a good reference produce nearly the same picture. A winding with a shorted turn produces a visibly different one.

6.1 Three stages of one pulse

One captured waveform has three distinct parts, and it helps to know which part carries the information.

From t0 to t1 the voltage climbs. This is the wavefront, and its steepness is set by the front resistance in the tester and by the distributed capacitance of the winding. Almost no current flows during this stage, so it does no damage to the coil. Its steepness matters for a different reason, which we come to in section 7.

From t1 to t2 the thyristor is conducting, and the pulse capacitor and the stator inductance ring together for roughly half a period. The frequency here depends on both the winding and the tester.

From t2 to t3 the circuit is on its own. The frequency now comes from the winding's own L and C, and the rate at which the ring dies away is governed by r / 2L. This is the part you compare.

6.2 Why a shorted turn shows up

A shorted turn removes turns from the coil. Fewer turns means lower inductance. Lower inductance in f = 1 / (2π √(L C)) means a higher frequency, which on screen is a shorter period. The shorted loop also carries extra current, and that current wastes energy, so the ring dies away faster too. You are looking for two things at once: a trace that swings faster and fades quicker than the reference.

7. Choosing the test parameters

Three settings decide whether the test finds anything: the peak voltage, the wavefront time, and how long you apply it.

Peak voltage. The working figure is Up' = K1 × K2 × Ug. K1 is √2, converting the RMS withstand value to a peak. Ug is the power-frequency withstand value for the stator or rotor, taken from the applicable standard — for a vehicle drive motor that is GB/T 18488. K2 is an application factor:

On the peak voltage tolerance, the national standard is fairly loose, but a supplier can hold it much tighter, and ±3% is achievable on a production tester. Worth asking for, because voltage that drifts between tests makes waveforms drift with it.

Wavefront time. The two standard values are 0.2 µs and 1.2 µs, and the definition is worth knowing: wavefront time equals 1.67 times the interval between the 30% and 90% points on the rising edge. This is a real trade-off. A short wavefront puts a steep voltage gradient across the winding. Most of the stress lands on the first few turns at the line end, and a defect deeper in the coil feels very little. A long wavefront reaches further into the winding, but it is less sensitive to what it finds there. The usual choice is a shorter wavefront for coils with few turns, and a longer one for coils with many.

IEC 60034-15 covers impulse withstand for form-wound coils and sets a wavefront of the order of 0.2 µs. It also notes the fact that catches people out: a three-level inverter can overshoot by a factor as high as 1.7 at the motor terminals. If your motor spends its life on a drive like that, the insulation sees far more than the nameplate suggests. The test level should reflect it.

Test duration. A manual test usually runs one to three seconds. Automated testers shorten that, and the test can be repeated as many times as you like without harming the winding, because the energy in the pulse is small.

8. Reading the waveform: pass or fail

There are two ways to put a number on the comparison.

The waveform area difference method compares the area under the two traces and expresses it as an overlap. Ninety percent overlap or better is the usual pass. The difference-area method divides the area of the difference by the area of the reference, and ten percent or less is the usual pass. Both are saying the same thing in different units, and you can also compare peak amplitude, oscillation period and phase directly if you prefer.

9. Why two good windings can still look different

Here is the part that causes arguments on the shop floor. Two windings that are both perfectly serviceable can still produce traces that differ by a few percent, and in awkward cases by several tens of percent. That is not a fault, it is manufacturing reality:

• Lamination stack height. A difference of a few tenths of a millimetre in the stack changes the magnetic path and therefore the inductance.
• Turn-to-turn and layer-to-layer capacitance. Small differences in how the wire lies, including a broken or shifted turn, change C.
• Winding tension. Wind the same coil at 0.7, 0.75 and 0.78 of the same setting and you get three slightly different resistances.
• Test voltage accuracy. If the tester holds ±5% instead of ±3%, the traces move accordingly.

The practical answer is to set your pass band against a reference built the same way, on the same machine. Treat a single borderline trace as a reason to re-test, not to scrap. What you are hunting for is the clear signature: a shorter period, plus a faster decay.

10. What we changed on the line

Finding one of these is only useful if it changes something. After this return we did three things.

1. We added a surge comparison test at an earlier stage, on the wound stator before it goes into the housing, so a coil with a weak spot never gets built into a finished motor.
2. We tightened the accepted wavefront steepness so defects deeper into the coil still get stressed, rather than only the first few turns at the line end.
3. We reviewed the lamination edge finish and the winding tension window, because those two are where manufacturing damage comes from.

The third point is the one that matters most. A surge test catches a weak spot. It does not stop one being made. Clean lamination edges and a controlled tension window are what stop it.

11. What to ask your motor supplier

1. Do you surge-test every stator, or only a sample? Ask for the pass criterion in numbers.
2. What peak voltage and wavefront time do you test at, and do they account for inverter overshoot?
3. What is the tolerance on the test voltage?
4. Do you test the wound stator before assembly, or only the finished motor?
5. Can you show the trace, or at least the overlap figure, for the unit I am buying?

Those five questions separate a supplier who tests from a supplier who owns a tester. The same thinking applies when you are comparing motor types in the first place, which we set out in our DC vs AC traction motor comparison.

Frequently asked questions

What is an inter-turn short?

Two or more turns of a coil losing the insulation between them and becoming one electrical turn. The shorted turns form a closed loop of very low resistance and inductance, so a large current circulates in them, and the local heat destroys the surrounding enamel.

Why did the insulation resistance test still read healthy?

Because a megger measures the path from winding to frame. A turn-to-turn short is a path between two turns, not to earth, so it can show a perfectly good reading. IEEE Std 43 gives the minimum acceptable value as roughly (rated kV + 1) MΩ, with 5 MΩ as a practical floor for random-wound low-voltage machines.

Can a multimeter find it?

Usually not. Losing a few turns out of dozens changes the phase resistance by a fraction of an ohm, which is inside the scatter of the measurement and inside the spread between two healthy motors.

Is the surge test destructive?

No. The energy stored in the pulse capacitor is small, the test lasts one to three seconds on a manual tester, and it can be repeated as often as you like without ageing the winding.

How much difference between two windings is acceptable?

Ninety percent waveform overlap is the common pass criterion, or ten percent difference area. Keep in mind that two healthy windings can still differ by a few percent. Blame the stack height, the spread in capacitance, and the winding tension. So a borderline result is a reason to re-test, not to reject.

Does a short wavefront or a long one find more faults?

A short wavefront stresses the first few turns hard and barely reaches deeper into the coil. A long wavefront reaches further but is less sensitive. The usual choice is short for coils with few turns, longer for coils with many.

Can an inverter cause an inter-turn short?

Yes. Voltage overshoot at the motor terminals from inverter switching is one of the two main sources of the defect, and IEC 60034-15 notes an overshoot factor as high as 1.7 for three-level inverters. Matching the AC motor controller to the motor is the first defence.

How we build and test

Every AC traction motor we ship gets machine winding for an even slot fill, then vacuum pressure impregnation of the winding. Then comes a full electrical check, which includes the surge comparison test described above. After that each unit is run under full load, balanced, and given an electrical safety inspection before it is packed.

We have been building traction motors for special vehicles since 2004. We hold ISO9001 and ship more than 500,000 motors a year across 800+ specifications. The XYQ series is our three-phase AC asynchronous range, built for heavy forklifts, tow tractors and electric vehicles, including the XYQ-10PC01 in this story. If you are diagnosing a DC machine instead, the equivalent walkthrough is our top 5 causes of DC motor failure, and there is a real teardown in our DC motor failure case study.

Sources

• GB/T 755 — Rotating electrical machines: rating and performance.
• GB/T 18488 (current edition 2024) — Drive motor system for electric vehicles; source of the power-frequency withstand value Ug used to set the surge peak.
• IEC 60034-15 (Ed. 4.0, 2025) — Impulse voltage withstand levels of form-wound stator coils; wavefront of the order of 0.2 µs; inverter overshoot factor up to 1.7 for three-level drives.
• IEC 60060-1 — Definitions used for impulse wavefront time (1.67 × the 30%–90% interval).
• IEEE Std 43 — Recommended practice for testing insulation resistance; minimum value approximately (rated kV + 1) MΩ, 5 MΩ practical floor for random-wound low-voltage machines.
• IEEE 522 — Guide for testing turn insulation of form-wound stator coils; reduced level applied to machines already in service.
• IEEE 60085 / IEC 60085 — Thermal classification of electrical insulation (Class F = 155 °C).
• Baoluo Motor Enterprise Knowledge Base — factory test items including 100% load testing, dynamic balancing and electrical safety inspection (Section 1).

Send us the symptom and we will tell you where to look

Tell us the battery voltage, the inverter you are running, and what the motor does when you power it up. A photo of the drive's fault code helps more than anything else.

• Describe a fault or request a quote
• Browse the AC traction motor range
• Browse the AC motor controller range
• Read the AC traction motor failure guide

About the author: the Baoluo Motor engineering team designs and tests three-phase AC asynchronous traction motors and DC traction motors for forklifts, tow tractors, AGVs and electric vehicles. These guides come from the test bench and from motors customers send back, not from theory alone.