Motor Insulation and Bearing Failures on a Long Unshielded Inverter Cable
A 400 V, 50 Hz, 37 kW, 4-pole motor is fed from a PWM voltage-source inverter on a 400 V supply, giving a nominal dc bus of 565 V and 620 V at +10% line. The motor cable is 130 m of unshielded four-core PVC cable in a tray, with the frame earthed only through the tray. Switching frequency is 4 kHz, the IGBT rise time is about 100 ns, and there is no output reactor. In eight months three motors have failed with turn-to-turn shorts in the first turns of the line-end winding, and one of them also had pitted bearings. A 100 MHz probe shows 1180 V peak at the motor terminals with heavy ringing. The motor is a standard 400 V machine with a peak withstand of 1350 V quoted at a rise time of 0.8 µs. The plant proposes to cut the carrier to 2 kHz and to fit a 4% input line reactor that is already in stores; a fitter adds that a longer cable would spread the stress out.
1. Explain the overvoltage mechanism, calculate the critical cable length, and explain why the failures cluster at the line end of the winding.
Read the worked answer
Mechanism. The inverter output is a fast step. The cable behaves as a transmission line with a surge impedance of roughly 40–50 Ω; the motor's surge impedance is much higher, so the reflection coefficient at the motor terminals is close to +1 and the incident wave doubles. For a step of amplitude Vdc, the motor terminal therefore sees approaching 2Vdc once the reflection returns.
Critical length. Propagation velocity in PVC-insulated cable (εr ≈ 4) is about 150 m/µs. Full doubling occurs when the cable is long enough for the reflected wave to return within the rise time: l_crit = v·t_r/2 = 150 m/µs × 0.1 µs / 2 = 7.5 m. At 130 m the cable is more than fifteen times that, so the reflection is fully developed and the peak is set by the dc bus, not by the cable length.
Numbers. 2 × 565 = 1130 V at nominal supply and 2 × 620 = 1240 V at +10%, with ringing giving the measured 1180 V. The 1350 V withstand of the motor is quoted at 0.8 µs rise time; at 0.1 µs the permissible peak is lower, and more importantly the surge distribution across the winding becomes capacitive and highly non-uniform - the first coil can take 30–50% of the total. That is why the failures are turn-to-turn shorts in the first turns at the line end, not general insulation breakdown.
The fitter is wrong: lengthening the cable does not spread the stress, it adds loss and common-mode coupling without lowering the peak, which is already at the 2 pu limit.
2. Evaluate the two proposed fixes and state what should actually be installed.
Read the worked answer
Cutting the carrier from 4 kHz to 2 kHz: this reduces the number of impulses per second and the drive's switching loss, but it changes neither the reflection coefficient nor the peak voltage. The damage per impulse is the same; only the repetition rate falls. It also increases current ripple and audible noise and slightly worsens low-speed torque quality. It is not a remedy.
A 4% input line reactor: it is on the wrong side of the drive. It reduces supply-side current harmonics and protects the rectifier, and has no effect whatever on the inverter output dv/dt or on the motor terminal peak. Fitting it and declaring the problem solved would leave the failure mode in place.
What should be installed, in order:
- An output dv/dt filter or reactor at the inverter terminals, sized for the drive current and carrier, to stretch the rise time to roughly 1 µs and hold the peak below about 1000 V. A sine filter is better still and also removes the common-mode step, but costs more and adds losses.
- A shielded, symmetrical three-core cable with a full-size PE, glanded 360° at both the drive and the motor, replacing the tray return path. This controls both the radiated EMI and the common-mode current that is causing the bearing damage.
- An inverter-duty motor if the existing standard machines must be replaced anyway; specify the insulation withstand against IEC 60034-25 rather than assuming a standard 400 V machine is adequate.
The priority is the winding failures: fix the overvoltage first, then the earthing and bearing measures.
3. Separate the bearing failure from the winding failure. What caused the pitting, and what would you do about it?
Read the worked answer
Different mechanism, same root cause. The inverter output has a common-mode voltage that steps between bus rails at the carrier frequency. That voltage drives a capacitive current through the stator-to-rotor capacitance, building a shaft voltage. The lubricating film is insulating until the shaft voltage exceeds its breakdown value, typically around 10–30 V, at which point a discharge current flows through the bearing - the classic EDM pitting. Eearthing the frame only through a tray raises the frame's high-frequency potential and makes the shaft voltage worse, because the intended return path is long and inductive.
Why the carrier change does not solve it: lowering to 2 kHz lengthens the interval between discharges but does not reduce the shaft voltage magnitude, so the discharge energy per event is comparable. Long cables and unshielded runs increase the common-mode current, and 130 m is well into the range where bearing damage becomes common.
Measures, in order:
- Shielded motor cable glanded 360° at both ends with a separate full-size PE conductor run with the phases - this gives the common-mode current a short, low-impedance return and is the single most effective step.
- A shaft earthing ring or brush at the drive end to hold the shaft voltage low.
- An insulated bearing at the non-drive end to break the circulating path where a brush cannot be fitted.
- A common-mode choke at the inverter output if the drive has one available for the frame size.
Note that fitting an insulated bearing without the earthing ring can shift the discharge to the driven machine through the coupling, which is a worse outcome.