Interupting xenon flash current ?

I do have plans for reducing the fanout - actually having either 1 driver chip for each IGBT or possibly one driver chip for every two IGBTs. However I have reduced the priority of doing that move as it is starting to appear that the problem I currently has nothing to do with gate drive.

My priority is now how to get a clean turnoff. Originally I was blaming poor design of the gate drive, but I have seen it work fine with a resistive load rather than the flashtube so I will now be looking at ways to 'tame' the flashtube.

I have now added a 'freewheeling' diode between the colectors and the positive supply (of the capacitor). This has greatly reduced the positive overshoot to about only 80 volts above the capacitor voltage. I quite naively hoped this would solve the problem so I tried firing the tube at 700 volts. Worked a few times, but only just the few. After about the tenth discharge one of the IGBTs blew - full short circuit between all three terminals.

It was suggetsed that it could be something as simple as the inductance of the tube so I tried to make an experimnent to see whether that was the case. I built a resistor out of 200 (!) small resistors, giving it a shape as similar as possible to that of the xenon tube . The idea was that the inductance would be very similar and the resistance would be equal to that of the tube at the time of turnoff. Unfortunately the resistor was no match for the current I tried to put through it and just went up in a puff of smoke and a loud bang. Turns out I was a bit over optimistic assuming that keeping the pulse short enough would allow the resistors to survive - and I also made a mistake setting the pulse length to 200uS instead of 20uS as I had meant to.

I've now decided to take an opposite (and counter intuitive) approach. I will be instead adding inductance in series with the tube. One thing I think this might do is provide a high frequency blocking between the tube and the IGBTs. ALthough this will not necessarily solve the problem I am hoping it will allow me to figure out better where the oscillations are originating. I will be able to check the voltage on each side of the inductor. I should see much larger oscillations on one side than the other. If it is on the IGBT side then I will know there must be some very significant collector to gate feedback. If on the other side it would confirm the suspicion that the tube is causing the oscillation - due to its negative resistance or some other weird plasma physics thing.

As i mentioned, the flashtube plasma is the source of oscillation (and cannot be tamed). I also suggested the use of an artificial transmission line for determining the on time; that could be programmable in steps as needed.

I will make you a deal. I think that the problem is some weird plasma physics, thus a resistor pack does *_NOT_* constitute an equivalent load. Accordingly i will offer you a bet, cut the driver to IGBT ratio to no more than 1 driver to 3 IGBT and see what happens. If it does not help i will apologize profusely, if it helps you apologize profusely.

Robert, Thanks, I have not ignored your suggestion. I have been trying to figure out what the artificial transmission line would involve but I haven't got very far.

With what little I know about it I have two concerns:

1. Something tells me that it will involve some rather largish inductors and capacitors. I am already 'over budget' in terms of space requirements so that can be a problem
2. I need the timing to be arbitrarily determined by software on the PIC, one of the objectives being also a returned light exposure control, as used in ordinary automatic flashguns.

Would it be possible to guide me with a few words in the right direction as to what it would involve and whether these two concerns are valid?

Joseph, as I said, I do intend to increased the driver to IGBT ratio but that is going to take me quite a while. Presently I have the IGBTs mounted on copper bars and the drive consists of components soldered to copper strips and bars directly on the IGBTs. To have the multiple drivers this system is no longer practicable so I'll have to design and order a PCB. I was going to postpone this but maybe I'll put some more energy back into it. Put it this way, I cannot lose either way. Either I'll be rewarded by a solution to my problem or a profuse apology :) A winner any way. :)

I hate to get picky, Robert, but real or artificial transmission lines at kilovolt and kiloampere ranges are not small nor simple but interestingly difficult.

Pick all you want; you are correct. However, do *you* have another workable suggestion?

A full solution, no. Some ideas, some tried with mixed results. Some not tried yet, some very similar to others and OP ideas. OP does let us know when trying something posted here.

Yes; excellent feedback!

My experiment with the series inductor went all wrong, as in loud bangs and smoke :(

Still not sure what exactly went wrong. I would have expected the inductor to just discharge properly through the 'freewheeling diode'. Somehow however it demolished the diode together with an identical diode in the snubber and three of the IGBTs. Most of the noise and smoke came from some resistors that blew up and a 1sq mm wire that was used to connect the freewheeling diode that completely vaporised. The diodes are rated 1.2KV 150A continuous and 3000 amps for 6mS so they should have easily taken the strain. That vaporised wire seems to be telling me that a much larger current flowed, though where that could have come from remains a bit of a mystery.

Anyway, I will just file that under 'bad ideas' and move on.

I'm struggling to design a PCB with a better driver to IGBT ratio. Here's a jpeg of the schematic:

Not sure if it will help, but I am hoping it will at least remove any doubts about the suitability of the gate drive. Any comments and suggestions about the schematic would be appreciated.

The resistors in the supply lines to the gate drivers are intended to provide decoupling both between the drivers as well as from grounbd bounce and such ills that the very large current flowing in the immediate vicinity are likley to cause.

The values of those resistors as well as the gate and emitter resistors are at best a good guess (perhaps not even a good one). For the decoupling resistors I'm looking for sufficient decoupling while avoiding excesive voltage drop.

The gate resistors are probably too low as the gate driver can source a maximum of 2 amps, so I may be changing them with 33 ohm or thereabouts instead but I'm not sure yet.