Hi,
I’ve had three EPC2304 failures over the past few months. Each time, the entire high or low side of one phase failed (not always the same phase), and the FET furthest from the gate driver was consistently the most damaged. It is also closest one to motor.
Setup details:
I mostly copied EPC9186 phase schematic with difference of using 3 pairs instead of 4 and EPC2304 instead of EPC2302, beefier Schottky flyback diodes.
-
3 paralleled FETs per half-bridge (EPC2304)
-
Layout derived from eval board EPC9186
-
Gate driver: STDRIVEG600 (previously another driver 2EDF7275K, same result)
-
50ns deadtime (i also tried 300 and 600ns, still had failures), 30kHz switching frequency
-
Failures only occurred during >100 V operation, not at ≤ 80 V
-
Test conditions where failures occurred:
-
Double pulse test (high load, two pulses at 40us + 5us idle + 10us pulse)
-
Motor detection (medium load, maybe couple amps requested to spin the motor)
-
RL measurement (VESC procedure) (low load, just energizing motor coils)
-
That is my 3rd pcb and I’m starting to loose hope. I changed almost everything (including config ie deadtime) and exact same result with low inductance motor (big outrunner 8kW). I’m nowhere near the theoretical limit of 3x EPC2304. Scoping gates look clean and only interesting event is that during second pulse, I see sub 10ns current spike on 5Mhz bandwidth current sensor (when low side turns off, high side turns on).
The only consistent things are failure at > 120V and the fact furthest ganfet was most damaged. Is there any explanation what is going on?
My current hypothesis: local common-source inductance di/dt causing false turn-on or current redistribution between paralleled FETs but honestly im not sure how to even attempt to fix it.
