Why is our Bootstrap high side drive circuit blowing up?

[FvM]

“The law (for mains voltages and above) is that you must use a SiC bootstrap diode, even if that means having to use a TO220 SiC package so as to get a low Vf.”

Do you agree with this law?

I hold my experience against it, see post #2.

 

[treez]

A possible circuit problem could be large switching node undershoot due to bad layout.

thanks, I agree with the above from post#2, but surely this doesn't make the bootstrap diode's reverse recovery any worse?

 

[FvM]

No. The undershoot can increase pulse currents in forward biased state. If we have reason to assume that the resistor fails during reverse recovery, it won't matter.

Anyway, you know I wasn't referring to this statement (possible failure causes) but to the observation that US1M can work well as bootstrap diode.

 

[treez]

The undershoot can increase pulse currents in forward biased state.

Thankyou, this is a massive point, so you are saying that the series damping resistor could be getting blown up by the "forward" current pulse in these kind of cases of massive undershoot of the switching node?

 

[FvM]

I never experienced this (massive undershoot won't happen in a reasonable design), just considered that it might happen.

 

[FlapJack]

If the damage to the resistor is a tiny black spot and the resistor is OC then it is from high voltage. If the resistor is OC and no visible damage then it's likely fused from high current.
My guess is that the resistor damaged from high current and it needs to be increased to 10R.

What is OC.

 

[treez]

open circuit

 

[Easy peasy]

STTH1L06, or STTH1R06 assuming 600V is high enough...

- - - Updated - - -

US1M is horribly slow when it warms up....

 

[treez]

thanks, but this bootstrap is at the output of a three phase rectifier, so its 564V pk, but with tolerance it could go above 600v.
However, the boss was considering using lower voltage UF diodes (which as you know often have lower reverse recovery times) and putting them in series. He said he believed they would always share the voltage equally....(?

 

[Easy peasy]

a small snubber for 2 x 600V in series is a good idea, e.g. 33pF 1kV, 47 ohm, across each, should be checked with a scope at full power.

 

[Easy peasy]

A good guide to mosfet gate drivers: Application Note AN-937 from IR

 

[PatrickPribyl]

This is an interesting problem, with lots of answers, but I agree with Treez and Fvm early posts about reverse recovery transient.

As you know, the difference between SiC and Schottky is the zero reverse recovery time of the former - i.e. no current transient. Since it broke when you changed it, I would strongly suspect that is the reason for the problem. It sounds like you have concluded essentially this with "Rule 1".

Solving the problem without a replacement SiC diode is then somewhat harder. Adding some series inductance would be my first try, but then you need to make sure it does not introduce ringing and overshoot.

I wonder if another thing that might have the same effect is just dumb parasitic capacitance in the parts, and so trying a different Schottky diode might work? (My guess is no, since you probably have faster transients than other responders that have successfully used a schottky).

For a cluge, you could try stacking up 6 pcs 30 ohm 1206 resistors on the pads (one on top of the other).

Please let us know the resolution!