GDT keeps on killing IGBTs

[ritaismyconscience]

Looks like I didn't set it correctly.

Here's a simulation. I added a bit of inductance and resistance between the square wave output and the inductor to simulate the parasitic inductance of the wire:


Seems to have some phase lead.

[davekni]

Probing there makes it difficult to know phase accurately, as the scope signal depends mostly on parasitic resistance and inductance (20m and 200nH).  Probing at the high-voltage node, between the 22uH and 100m-ohm, will give you a more accurate measure of mmc voltage, which will be more accurately 90 degrees from current.

[ritaismyconscience]

I can't seem to get it work with your method. The controller does a really good job keeping the current constant, so waveform just looks like sine wave:



I'm measuring it at the highest voltage, which should also correspond to the highest current.

Edit: Here's a zoomed in version of the previous picture. There does seem to be a small bump right before the peak. Not sure if this is where the transistors are switching.


Edit 2: Tried a new way. I took the sine wave and subtracted the H bridge output, and it looks like this:


I'm only testing it at low voltage, but it's enough to trip the overcurrent.

[davekni]

You could probe the H-Bridge-to-MMC connection to verify that there's still a square-wave there.  Just in case one side of the H-Bridge fried, or the bus voltage is collapsing during the burst, then there wouldn't be the step on the MMC-to-primary connection.  If the MMC side of the H-Bridge has a good square-wave signal, then those same steps will show up on the high-voltage (primary winding) end of the MMC on top of it's higher-voltage sine-wave component.  (Check in your simulation too, probe the 50nF to 22uH node.  I think your simulation likely has less loss than real life, so you may see over 12kV, but there will still be the step on top.)

[ritaismyconscience]

I tried probing it and there is a square wave. I think the capacitor method doesn't work very well for the H bridge for some reason. I also tried measuring through a 10M resistor, and I got the same waveform (a sine wave with some ridges). I onl y got a square wave when I connected the probe directly to the output.

[davekni]

At low bus voltage, the H-Bridge output square-wave will be low amplitude.  With bursts long enough to ramp current to your 500A limit, then the MMC-to-primary node will still have max voltage.  Perhaps the H-Bridge output steps are just too small to see on the full MMC voltage.