UCC21710: Gate driver tripping on over current fault when it is already bypassed.

Hi Ankit,

This behavior is related to the way the output die communicates faults with the input die. There is an on-off keying circuit on the input side that that counts for 3 rising edges from the fault detection circuit on the output die. However, an OUT falling edge can cause a cross-channel noise similar enough to incorrectly triggers nFLT. One way to limit this problem is to slow the OUT falling edge to eliminate undershoot. The problem is worst at cold temperature, which reduces the effectiveness of internal filters.

Yes, you can re-set the nFLT by sending a  negative pulse on the EN/nRST pin.

If you want greater simplicity, you can use a gate driver such as UCC5390 that does not have this nFLT as a potential failure mechanism.

Best regards,

Sean

Hi Ankit,

My colleague has tested a high-voltage ~100pF capacitor across the isolation barrier to shunt this noise with some success. This can work on the low-side, since there will not be any charge cycling. 

Higher HV Bus voltage usually means higher noise during switching.

Best regards,

Sean

Hi Ankit,

She added this capacitor across the grounds. Hopefully you have a gound plane on the input side that can absorb the noise injection without creating a ground-bound issue.

Best regards,

Sean

Hi Ankit,

Adding a capacitor between the high-side ground (Switch node) and the logic ground is NOT an good idea. The capacitor will heavily load the switch node, and add lots of losses to the FETs, In addition, it will inject high AC current into the logic ground.

It is not really a good idea to add a capacitor across the low-side gate driver either, but it was found to mitigate the UCC21710 false triggering of nFLT in my colleague's testing.

Best regards,

Sean