Original poster: Steve Conner <steve@xxxxxxxxxxxx>
Hi Steve, How did you keep hits to the primary from killing your bricks??
Well, why _should_ strikes kill them? There's 400A flowing in the primary, so why should it even notice a tickle of streamer current? I figured out a lot of nasty fiddly answers to that question that you may not even want to know:
1) The streamer current has a very high di/dt similar to a lightning strike. If the H-bridge layout was sloppy, L*di/dt could overvolt things. With a tight layout, the flywheel diodes will clamp these overvoltages down to the DC bus voltage.
2) The streamer current is common mode. It came effectively from a capacitor whose other plate is RF ground, and it wants to go back to RF ground. If there is no nice AC path from the primary circuit back to RF ground, the streamer voltage will go flying through everywhere untill it "Finds" RF ground. That can couple nasty things through your GDTs etc. back into the driver.
This is a problem in classic coils too: the high inductance of a NST (in both differential and common modes) effectively lets the primary circuit "Float" at RF. So a primary strike causes huge stress on the NST's insulation. We use safety gaps between the transformer bushings and the case to fix this, but a DRSSTC calls for a different approach.
I used Faraday-shielded GDTs wound with miniature coax, and some capacitance between DC bus negative and the green wire ground. I also try to choose RF grounds that are fairly well related to the green wire ground. (Indoor coiling is good here because any metal nearby should be connected to the green wire quite directly.)
3) The strike current may flow through the part of the primary that goes through the feedback CT. The high peak current of the streamer will create a dirty false signal with big spikes. I had trouble with this latching the feedback circuitry up, but adding some zeners cured it.
The PLL circuit just ignores anything that isn't within 10% of the expected operating frequency so there is no harm done by a bad feedback signal. Unless the IGBTs are being run so hard that an error in the switching timing will make them turn-off a current outside their safe operating area and go into latchup. But I choose to run my devices a little less hard than that, so even if the feedback goes totally wrong, they won't die.
Steve Conner http://www.scopeboy.com/