Transformer isolated gate drive can blow up the SMPS?

Hello,

Article [A] below says that a gate drive transformer with a DC restoration circuit can turn the MOSFET on when its not required to be on and blow up the SMPS.
-Is this correct?...surely not?.....i have never heard such thing before...........and just because a PWM controller drive output goes off doesnt mean it stays low...surely it goes tri-state (high Z) ?


[A] Page 16, figure 6 of the following tells of a great problem with gate drive transformer circuits with DC restoration circuitry.
Gate Drive Design Tips - Ridley


Figure 35, page 32 of the following gives the same circuit but doesnt mention this problem
**broken link removed**

Hmm - OK... I think I see what he's trying to say. And yes, his argument relies on the PWM controller drive staying low.
This sounds like a pretty low probability type of failure mode - but I guess it would be a turd to debug

*Perhaps* if the driver had a totem pole output stage, "off" might = tied low. His use of 'Steady state operation' is a little cryptic in the context of a PWM system but I took that to mean "when everything is continuously pulsing". I could envision the problem he describes then occurring in one of two situations - When the drive is cut off (and the input to the capacitor-transformer network is held at either rail) AND ...

1. The system's power supply rails are collapsing (as would happen at switch-off). If the output devices (likely) survived the (one-off) current pulse, we'd never know it even happened.
2. The rails stayed up (i.e. the converter was just powered down). Even then, I imagine you'd have to be pretty unlucky!

I've never had (or seen) a gate-drive transformer saturate, but... maybe this is giving me clues as to why my designs never seem to attain the utmost in efficiency..?!
Thanks for sharing the article grizedale - personally, I'll probably continue to use the restoration circuit, but maybe be a little wiser in my debugging when the MOSFETs mysteriously snuff it

Figure 35, page 32 of the following gives the same circuit but doesnt mention this problem

Click to expand...

It's not the same circuit, it has a pull-down resistor.

Aye, they've simply included the resistor alluded to in Ridley's comment: "(a gate load resistor is recommended)" but the 'DC restoration' via diodes concept appears the same.

The article [A] to me seems to make a lot of sense.....when the output goes onto no load, the PWM controller will stop switching and simply hold the bottom fet off......which will undoubtedly cause the problem Ridley tells.

This will undoubtedly put stress on the fet, and make its lifetime less.

I'd say Ridley has called this one exactly right.......i just wonder if there is any way of stopping it when the DC restoration circuit is used?

The alternative transformer gate drive without the DC restoration circuit is lousy.......the gate drive voltage varies with the duty cycle even if duty is 50% or less.....can you imagine the problems that causes for the power source of the PWM controller?

As FvM said in post #3, it is not the same circuit, see carefully. There is an additional pull down resistor.

Transformer coupled gate drivers without active waveform shaping on the secondary achieve at best fair performance, particularly the switching speed isn't mind blowing. For this reason, I'm not strongly motivated to reasearch on this circuit topology.

It's O.K. to say, there's a risk of undefined behaviour with diode level restorer, but I assume so far, that save operation can be achieved. A pull-down resistor is necessary in my opinion, other parameters, e.g transformer inductance and coupling capacitor value need to be considered for a detail analysis. The posted conclusions are too general in my view and more preconception than analysis.

Raza i understand what you are saying, but even with that pull-down resistor the problem will still be there.

Or rather, the pull-down resistor would have to be of a very low ohmic value (and therefore be very dissipative) if it was to mitigate the described problem.