SiC Devices Behave Differently

Silicon carbide devices switch faster than silicon and, in the case of the diode, store essentially no charge. That makes them easier to design in for efficiency and EMI, but it also changes the failure modes: a SiC converter rarely fails because of recovery, and instead fails because of overshoot from loop inductance, a thermal problem, a surge event or a paralleling imbalance. This article presents a systematic method for diagnosing the common issues in Cree SiC MOSFETs and Schottky diodes inside high-frequency converters.

Overshoot and Ringing

The most common SiC problem is excessive voltage overshoot at the switching edge, and it almost always traces to the commutation loop inductance and the gate drive. Because SiC switches fast, the same loop inductance that a silicon design tolerates produces a much higher spike. Keep the DC-link capacitor close to the device, minimize the loop area, use the four-lead package with the driver-source connection to remove source inductance from the gate loop, and tune the gate resistor to slow the edge within your overshoot limit. Measure the overshoot at the device terminals, because the bus measurement hides the spike the device sees.

False Turn-On

In a bridge, the fast dv/dt can couple through the gate-drain capacitance and turn a device on when it should be off, causing shoot-through. The negative off-state gate supply and a low-impedance gate loop resist this, and a small gate-source capacitor can help. Check the gate waveform at the device during the complementary switch's transition.

Surge and Inrush

A frequent cause of SiC diode failure is a surge or inrush current that exceeds the diode's rating. At power-up, a large capacitor charges through the rectifier and draws a current pulse far above the steady-state current. Check the inrush against the datasheet surge rating, and if it is close, add a soft-start, an NTC or a pre-charge circuit. Measure the surge with a clamped current probe during start-up to confirm the magnitude.

Thermal Problems

Conduction loss in a SiC MOSFET scales with the on-resistance hot value, which rises with temperature, so the thermal design must account for it. If a device runs hot, check three things: the actual current, the hot on-resistance and forward voltage, and the interface and heatsink. A thick, uneven or contaminated thermal interface raises the thermal resistance enough to overheat the device at rated current. Use a thin, uniform interface and verify the case temperature under load. The high junction temperature rating gives headroom, but it is not a substitute for a sound thermal path.

Paralleling and Sharing

SiC diodes have a positive temperature coefficient of forward voltage, so current sharing improves as one diode heats, which makes paralleling practical without thermal runaway. SiC MOSFETs can be paralleled too, but the gate and source inductances must match so the devices switch together. Keep the layout symmetric, match the devices, and verify sharing under load.

A Systematic Method

Work from the simple to the complex: check the overshoot and the gate waveform first, then the thermal path, then the surge and inrush, then the paralleling. Keep a reference board that is known good, and record the overshoot and temperature of each new design on the bench, so a later change is immediately visible. BeiLuo supplies genuine Cree devices with import declaration, certificate of origin and RoHS documents, and our FAE team can help you diagnose a problem and choose a design change that resolves the root cause.