Mosfet with gate overvoltage dies with short cct Drain-source but not explode?

[ocset]

Hello,
Is it possible for a DPAK FET to die by going  short circuit from drain to source, (due to an overvoltage on its gate) but without actually exploding  or looking damaged?

We have a 40W offline LED driver which has a current clamp FET, M4,  in it which limits transient overcurrents. (as in the schematic  attached).

  We are worried that the zener diode D21,  is sometimes  soldered on but with a  dry jointed anode connection, and therefore when first powered on, the NFET , M4 (M4 is actually a DPAK) gets an overvoltage between Gate and Source, and then dies and goes short circuit from drain to source, and just stays short circuit, without exploding. 

Is this at all likely to happen?
If it does , then we have got no current clamp protection in our product.       
 

[Andy Watson]

I see no resistance in the gate! Is M4 stable?

[ocset]

Thanks, do you mean whilst its clamping?
We believe it is.
We scoped it after hitting the input  with a transient and we dont see the clamp NFET oscillating on and off.

[T3sl4co1l]

MOSFETs only explode if exposed to a large fault surge sufficient to propel the plastic case-bits as shrapnel.

The die itself melting occurs microscopically and silently.

Tim

[Andy Watson]

Yes. When it's clamping you're operating the device in linear mode. It's a fast switching fet - I would be surprised if drain-gate capacitance and gate lead inductance did not conspire to oscillate under the right (wrong!) conditions - given the lack of damping resistor in the gate.

We scoped it after hitting the input  with a transient and we dont see the clamp NFET oscillating on and off.

Sometimes you get lucky. I'm not saying this is definitely the problem but the addition of a grid-stopper would reduce your reliance on luck.

[dmills]

Agree about the gate stopper, but remember that especially in linear mode (which that part is not really characterised for) you will be fighting the Miller effect hard, you might want a diode across the stopper to make ramp down faster.

Regards, Dan.

[ocset]

Thanks, woops, sorry i wasnt expecting the advisory to go into the type of FET, it was originally the concern about the dry joints...

The FET we actually use here is actually a STD3NK90Z

http://www.st.com/content/ccc/resource/technical/document/datasheet/a8/09/96/48/1d/14/42/7d/CD00003170.pdf/files/CD00003170.pdf/jcr:content/translations/en.CD00003170.pdf

[capt bullshot]

If you exceed the Gate/Souce voltage of a MOSFET, the gate oxide will break down. This shorts the gate to the channel, result depends, but "MOSFET always turned on" may happen as a result. The package explodes if enough energy is available only, otherwise only the die fails.

[floobydust]

The FET we actually use here is actually a STD3NK90Z
http://www.st.com/content/ccc/resource/technical/document/datasheet/a8/09/96/48/1d/14/42/7d/CD00003170.pdf/files/CD00003170.pdf/jcr:content/translations/en.CD00003170.pdf

Providing pseudo-schematics with missing or different parts, you make it very hard for us 
There's no IP or anything special about this topology. Providing unfiltered schematics would be best, instead of LTSpice simplifications.

If this is part of the LED streetlight failure autopsy, at least give that thread some closure instead of starting "it failed-why?" threads, and then abandoning them so we never know what cause was found. I don't like to see the community used, and here we go with partial information, speculation and a colossal waste of our time IMHO.

[Andy Watson]

,it was originally the concern about the dry joints...

Yes, any excess voltage on the gate-source will damage the device - and until the current clamp operates there is nothing but the zener to restrict the gate-source voltage. Visible damage requires a relatively high level of unchecked energy availability.