Most MOSFET Failures Trace to Four Causes

A power MOSFET that fails in a switching supply usually fails for one of four reasons: an avalanche event from a load dump, a gate-drive problem, a thermal problem, or a layout problem. The symptoms overlap, so a systematic diagnosis is faster than replacing parts. This article presents a method for diagnosing the common issues in NCE Power SGT and Super Junction MOSFETs inside switching supplies and inverters.

Avalanche Events

An inductive load stores energy that must go somewhere when the MOSFET turns off, and if the drain voltage rises above the rating, the device enters avalanche. The NCE SGT family is 100 percent avalanche tested, so it tolerates a limited energy dump, but only within the datasheet limits. If a device fails at power-up or at load removal, check the peak drain voltage and the avalanche energy. The first fix is to tighten the layout and reduce loop inductance, which lowers the overshoot; the second is to add a clamp or a snubber if the event cannot be avoided.

Measuring the Peak

Always measure the peak drain voltage at the device terminals, not at the bus, because the device sees the worst spike. Compare it with the rating and keep a margin of at least fifteen percent for the worst case. A bus measurement hides the spike and gives false confidence.

Gate-Drive Problems

Gate-drive problems look like thermal or EMI problems at the system level. Ringing on the gate waveform is parasitic oscillation between the gate capacitance and the gate-loop inductance, and the fixes are geometric: shorten the loop, increase the gate resistance in small steps, or add a ferrite bead close to the package. False turn-on in a half bridge comes from fast dv/dt coupling through the Miller capacitance, and the fixes are a negative gate supply, a lower-impedance pull-down, or a lower off-transition gate resistance. Measure the off-state gate during a switching event to confirm.

Drive Strength

A weak driver charges the gate slowly, which increases switching loss and heats the device. A driver that is too strong can cause ringing. The right answer is a driver with adequate current and a tuned gate resistor, with the driver placed close to the device.

Thermal Problems

Conduction loss scales with the on-resistance times the current squared, and the on-resistance rises with temperature, so a poorly cooled device heats, which raises its resistance and its loss again. Check the current, the on-resistance hot value, and the interface and heatsink. A thick, uneven or contaminated interface raises the thermal resistance enough to overheat the device at rated current. For a surface-mount device such as a DFN or TO-252, the PCB copper and thermal vias are part of the thermal path, so an undersized copper area is a common cause of overheating.

Interface and Mounting

Use a thin, uniform interface material and the specified mounting torque for a through-hole part. Re-measure case temperature after any rework, because a disturbed interface rarely goes back to its original quality.

Layout Problems

Layout decides both overshoot and EMI in a switching supply. The commutation loop formed by the DC-link capacitor and the switching devices is the dominant contributor, so keep it tight and place the capacitor close to the devices. A large switching-node area radiates EMI, and long gate traces add inductance. A disciplined layout is quieter and more reliable than any component change, so review it before the board is built.

Conclusion

An NCE Power MOSFET is a robust device when it is applied correctly. Most failures trace back to an avalanche event, a gate-drive problem, a thermal issue or a layout mistake. Check the peak voltage, verify the gate waveform, confirm the thermal path, and tighten the layout, and the device performs as designed.