UCC21732: How to reset the fault latch of UCC21732 without risk of fireworks

Kjell, 

Kjell Ljøkelsøy said:

* Do you have any suggestions about how to clear the fault latch without having to turn on the power transistors?

I am a little confused about this question. Clearing the fault latch will not turn on the power transistors unless the PWM inputs are already high. 

If this is a concern, why not just ensure your controller keeps PWM inputs LOW after DESAT is detected?

Kjell Ljøkelsøy said:

A crude workaround could be to short the 5V supply. The fault handling logic could do that by turning on a large crowbar transistor briefly, so the undervoltage/powering up logic in the driver IC would clear its fault latch.

Kjell Ljøkelsøy said:

* Is this something you would recommend, does shorting the 5V rail to clear the fault latches make sense?

This does not make sense to me, as I do not see what this accomplish differently than toggling RST/EN apart from adding complexity and cost to the design. You don't really solve any problems with RST/EN by doing that. And I am not quite sure about the problem you described with RST in your earlier point.

In the worst case, if you had accidentally turned the FET ON back into a short circuit, DESAT detection would be tripped again, and fireworks would be unlikely to happen.

Kjell Ljøkelsøy said:

A crude workaround could be to short the 5V supply. The fault handling logic could do that by turning on a large crowbar transistor briefly,

Best

Dimitri

In our case the PWM inputs can be at any state at any time, as these signals come straight from an analog'ish control circuit, where the two PWM inputs gives shoot through protection. Adding logic gates for enable at the PWM inputs could give the function we want, but would prefer to avoid that as these gates would add a little latency to a latency budget that is quite strained.

The En/Rst' signal is common for several drivers so the other drivers that not in fault state will be turned on when we try to clear the fault latch.
Desat protection would normally save the transistors, but the large current pulse can destroy something else.
This is what we want to avoid.  The converter has already signaled that there is something wrong somewhere, so we can assume that the system is more fragile than usual.  

Shorting the 5V supply came up as an idea for a quick and dirty workaround. It would do the same as manually cycling the power supply as we do now for clearing faults. 

The En/Rst' pin would have done exactly what we need if the reset latch was cleared by low level and not by rising edge.

Kjell, 

Perhaps you could use a buffer in front of EN/RST. 

If that can't work, I suppose you can use the cut power method if thats the only thing that can work in your system. 

Best

Dimitri

A buffer at EN/RST' will not have any effect  here, as driving this pin is not a problem.  The way it works is the issue here. 

Apart for putting functional change of the EN/RST pin on the wish-list for the next chip revision it seems as power cut /crowbar is the least distasteful solution.