Miller Effect spikes in DRSSTCs
Steve Conner, Thu Jul 19 2007, 09:56AMHi all,
A while back, Richie Burnett and I were discussing gate drive in DRSSTCs. He told me about a strange consequence of the Miller effect that could possibly lead to bad things happening. He figured it out from theory, but I set up an experiment and was able to confirm that it really does happen.
I wrote up the experiment and results here
but to sum up:We know that once the burst is over, the anti-parallel diodes in your IGBTs act as a bridge rectifier and recycle the unused tank circuit energy back into your DC bus. But, while they are rectifying, the diodes switch extremely fast (they are fast recovery after all...) The result is a very high rate of change of voltage, that couples through the Miller capacitance to the gate, and actually spikes the IGBT on.
My DRSSTC (and everyone else's I guess) still works fine, but I thought I'd better share the results anyway
If anything, it's another reason to use local gate drivers: they would hold the gate down at -12V after the burst was done, instead of 0V.Thanks to Richie for making this interesting discovery.
Re: Miller Effect spikes in DRSSTCs
Marko, Thu Jul 19 2007, 11:59AM
Hi steve
I noticed exactly the same gate spikes but they never got dangerously big and I ignored them.
It seems that even 'crappy' gate drive is enough to prevent those getting over threshold voltage. It's still only few ohms of impedance between GS after all.
Marko, Thu Jul 19 2007, 11:59AM
Hi steve
I noticed exactly the same gate spikes but they never got dangerously big and I ignored them.
It seems that even 'crappy' gate drive is enough to prevent those getting over threshold voltage. It's still only few ohms of impedance between GS after all.
Re: Miller Effect spikes in DRSSTCs
Terry Fritz, Sat Jul 21 2007, 10:03PM
Hi Steve,
Interesting thing for sure!!
The high energy stored in the primary and the very high dV/dT should easily turn the IGBT back on.
The capacitance across the device should slow the dV/dT some, but probably not enough.
There would be a danger if the gate voltage exceeded the gate oxide breakdown of about 50 to 55V. But I think most use gate TVSs that should clamp that, and the gate resistance is usually pretty stiff. I guess you could ad external gate capacitance too if your really had to.
I would think cross conduction would not be an issue here. The IGBTs spiking on would just "help" to diodes a little
. Heating should not be an issue.
I can't think of anything really bad it would do as long as none of the obvious limits are exceed. But something to remember if one gets creative with the gate drive. FETs don't seem to mind "uber" switching much and I "think" IGBTs are pretty tough too. SCRs are not a factor anyway for our stuff.
IGBT diodes are meant to do very fast current, so the makers might "play" with die level things to mitigate any bad effects too. Larger bricks might have so much junk in the way that the spike never makes it back to where the IGBT is in the package too.
It's always a bit disturbing when circuits do things we don't tell them to do, but I can't see any bad effects here we need to worry about other than keeping this effect in mind.
Interesting stuff there
Cheers,
Terry
Terry Fritz, Sat Jul 21 2007, 10:03PM
Hi Steve,
Interesting thing for sure!!
The high energy stored in the primary and the very high dV/dT should easily turn the IGBT back on.
The capacitance across the device should slow the dV/dT some, but probably not enough. There would be a danger if the gate voltage exceeded the gate oxide breakdown of about 50 to 55V. But I think most use gate TVSs that should clamp that, and the gate resistance is usually pretty stiff. I guess you could ad external gate capacitance too if your really had to.
I would think cross conduction would not be an issue here. The IGBTs spiking on would just "help" to diodes a little
. Heating should not be an issue.I can't think of anything really bad it would do as long as none of the obvious limits are exceed. But something to remember if one gets creative with the gate drive. FETs don't seem to mind "uber" switching much and I "think" IGBTs are pretty tough too. SCRs are not a factor anyway for our stuff.
IGBT diodes are meant to do very fast current, so the makers might "play" with die level things to mitigate any bad effects too. Larger bricks might have so much junk in the way that the spike never makes it back to where the IGBT is in the package too.
It's always a bit disturbing when circuits do things we don't tell them to do, but I can't see any bad effects here we need to worry about other than keeping this effect in mind.
Interesting stuff there
Cheers,
Terry
Re: Miller Effect spikes in DRSSTCs
Sulaiman, Sun Jul 22 2007, 02:03PM
There's a nice application note here
that discusses this AND paralleling igbts.
Sulaiman, Sun Jul 22 2007, 02:03PM
There's a nice application note here
that discusses this AND paralleling igbts.
Print this page