Interupting xenon flash current ?

I do not think that inductance is the issue / problem source. Without very special precautions (gas pressure, current range, aging, addition of radioactivity, and i do not know what else), gas in a tube will have regions of negative resistance; in the right conditions, one can see a "crawling worm" of plasma in a gas tube. If you want to get "precise" data, then use a current-limited supply set to (say) 50V (no less than 32V) and slowly adjust that limit from (say) 1mA ("engineering zero") to the max of the supply, while monitoring the tube voltage and current with a scope. Use a trigger transformer to start the ionization.

Sorry Robert, this is not any ordinary flash. OP is talking kiloamperes and kilovolts.

Robert, OP is trying to deal with megajoule pulse at microsecond and shorter times. The instrumentation must necessarily be suitable for those conditions.

This makes me think that a different approach for commutation may be more appropriate. Perhaps something more similar to how SCRs are commutated. But at this scale normal approaches must be modified.

Just a quick note, the IGBTs are surviving with no problem turning off up to about half a megawatt ( 550 V @ 900 A). There is of course the very dirty turnoff but the IGBTs seem to handle it quite well at this voltage. It is only when I go above that voltage/current that the IGBTs start to die.

I'll probably be making the test resistor and doing some tests tomorrow or Friday. I'll let you all know of the results.

(PS: My original target was actually much higher - 1400V @ 3000A. If I eventually manage to get it working at 700V I'll probably move my target back up to that again )

ance.

rocess.

e

em. =A0At

o a

ng as

fine.

d e

Appears to me that you are frying the IGBT on turn-off. The situation you described with turn-on where you crashed your PC could be a fluke. I build quite a few flash lamp power supplies and the only times I fried IGBT's are turn-on with too high of a gate resistor and as such not fully turning the IGBT on with a thermal failure as a result. The other failur is to have a gate resistor that is too low and have the IGBT anode current try to exit through the gate.

What is the total gate capacitance from your parallelled IGBT's ? Is your driver capable of supplying the peak curent without any hick-ups ?

What you are describing is "after glow" in the flash lamp. Keep in mind that we are talking thermionic emission here. That cathode is hot from the pulse and will keep sending electrons out making ions along the way. This process is a multiple micro second process and function of lamp, drive current etc.

What flash tube are you using again ?

But time the cathode is hot is relevant as it will keep sending out electrons, making ions in the process. That's where the afterglow photons come from.

Are you trying to get micro second flash pulses at these operating conditions ? I'm not checking this list often, it is OK to use my emal address.

The shortest pulse I was ever able to drive a xenon flash lamp with is

2us. (laser pumping) At these pulse times your flashlamp doesn't even come fully on. No use of an IGBT was made to obtain these times.

Nah- fluorescent tubes have thermionic cathodes, heated by ion bombardment.

Mercury and xenon (low pressure) have the longest deionization times of most gas discharges, I think- typical thyratron ratings are around 1ms. Look up 2D21 for instance. High pressure puts a lot more ions and electrons around to recombine, so naturally it's going to be faster. It's essentially a thermal plasma, and being not too much matter, it cools down pretty quickly, recombines and settles back to the ground state.

Tim

-- Deep Fryer: A very philosophical monk. Website @

That doesn't sound right at all. I suspect your calculation is too simple! ;-)

Lemme see here. COIL.EXE (by Brian Beezley) says 5 turns on a 4" former,

0.5" length (i.e., a solenoid geometry) gives 4.86uH at 1MHz (and Q = 180, not too bad). This program does *everything*, from proximity effect to parasitic capacitance, and correctly predicts inductance, Q and resonances of any shape coil. (If you used the old N^2 / 9+10 somethings equation, you should know that only works nearly right for solenoids about as long as they are wide.)

Tweaking some values, it looks like a 1.5" diameter loop gives 0.1uH. In fact, a 100" loop gives 13.5uH, or about 43nH per inch circumference. I don't remember what the limiting value is, something like 20 or 30 or 50, as radius goes to infinity. That's one way of calculating inductance per length of a "free" wire. Incidentially, that 1.5" loop comes to

21.2nH/inch.

Tim

-- Deep Fryer: A very philosophical monk. Website @

you described with turn-on where you crashed your

Actually that is what I intended to say but maybe I didn;t phrase it clearly. I found out quite early on that the turn-on was no porblem and neither was maintaining the current, even if I kept the IGBTs long enough for the capacitor to discharge fully. The problem only seems to occur when I turn off the IGBTs while the capacitor voltage is still somehwat over 500V. The starting voltage does not matter, it is only the voltage at the time of turnoff that seems to matter. I can start a discharge at 700 volts and maintain the IGBTs on long enough for the voltage to go down to 500 or less and then turn off - that causes no problems and I have done that some 50 times without blowing any IGBT.

The instance I described when the PC crashed was to me showing a failure during turnoff, but I may have not described it clearly.

What I observed on the oscilloscope trace was that the turnoff did occur but it seems that one or more of the IGBTs either did not turn off fully or at one point started conducting again of its own accord, that is with no gate voltage applied. Perhaps it went into avalnche breakdown or something, but it was very clear that the turnoff was what went wrong.

your driver capable of supplying the peak curent without any hick- ups ? I don't have the info here with me but I'll check that this evening. Worth mentioning though that I added a 33nF capacitor across the Gate- Emitter of each group of six capacitors. Now that may seem an odd thing to do but it actually improved the turnoff. Without the capacitors I was blowing IGBTs even at lower voltages. What I believe was happening was the collector DV/DT being too high and feeding back into the gate. My hypothesis is that the added capacitors acts as a capacitive divider thereby reducing the damage caused by the collector/ gate capacitance.

Without the gate capacitors I had turnoff problems even into a purely resistive load, which I assumed was oscillation due to the collector- gate capacitance feedback.

The Drivers are two TC4422 with a 15 volt supply, each one driving 12 gates. There is a 1 ohm resistor in series with each gate. The TC4422 is rated for 9 amps and from the waveform I observe on its output it appears that it is indeed delivering that both on turn on and turnoff (though I can only reliably measure that without applied collector voltage)

conditions ? I'm not checking this list often, it is OK to use my emal address.

No, I am aiming for reasonably square(ish) pulses down to 200uS and down to about 40uS accepting that the light output will have a triangular shape rather than square. I have confirmed optically that the lamp can achieve a rise time from 0 to about 80% of its rated output (at 700V) within 20uS. There is also an additional delay of about 25uS until the light output actually starts to rise but that doesn't matter since what concerns me most is the actual duration of the light output rather than the delay between the trigger and the actual light output. Actually to improve the rise time (at the expense of added delay) I am pre-simmering the tube for some 40uS at 2 amps before applying the full voltage. (though that as most other things is software definable)

I do have a future project in mind for sub microsecond photography but that will be something entirely different and haven't really started thinking about it yet. I think I would be use spark gaps and exploding wires for that.

Correction to above post:

"I added a 33nF capacitor across the Gate- Emitter of each group of six capacitors. "

should of course read:

"I added a 33nF capacitor across the Gate- Emitter of each group of six IGBTs"

Someone also what flashtube I am using and I forgot to reply to that.

It is an IFK-2000, the 2000 representing its rating of 2000 Joules. Meant to work up to 1000 V but is OK with 1400 at lower energies. Makes a nice 'bang' discharging at that voltage. I thought the thing exploded the first time i tried it 8)

[url= of IFK-2000[/ url]

If/when I sort out the problem I'll actually be running two of them simultaneously.

Power level and timing do not "prevent" the use of an artificial transmission line; just the part specs and wiring layout.

Yes; "time of flight" is in nanoseconds or less - ion relaxation / buildup is in hundreds of microseconds or so as a guess based on how long it takes a tube plasma too die (seconds).

Some possibly dumb questions about what goes on in the tube:

If I calculated right (which I probably didn't) an electron would take a minimum of about 50nS to travel from the cathode to the anode if it didn't hit anything along the way. (assuming a 300mm path and 700 volts) The Ions I think would take much longer, being thousands of times more massive. Is that anywhere near being correct, and would the ion motion play a significant part in what appears on the terminals of the tube?

Would electrons and ions actually travel all the way or would most collide and recombine long before reaching the other end of the tube ?

If I understand correctly, most of the emitted light is due to the temperature of the gas rather than electron-ion recombination. If that is the case quite a significant time is needed for the gas to cool down, which I think accounts for the (relatively) slow decay of the light output. Would the presence of the hot gas in itself cause voltage/current through the electrodes? COuld that be causing the overshoot (assuming it is not just inductance)?

That sounds about right. I've made calculations of a couple ns for vacuum tubes (about 1/8" travel, near zero initial velocity, about 300eV final velocity).

Since the electrons are surrounded by gas, they take a whole lot longer to get there, ricocheting off atoms until they gain enough energy (heck, only a few eV for xenon) to ionize them, at which point the collisions go from elastic to inelastic and energy is transferred. Electrons spall off and a plasma discharge is born.

Of course, the electrons have to come from somewhere. There are a few just sitting around, but despite the unstable nature of a discharge, there aren't enough. A high voltage pulse is usually added to a flash tube to pull electrons from nearby surfaces with field emission. Once the electrodes are hot, thermionic emission also serves to keep things conducting.

Indeed, you can imagine the ions and atoms being in about the same place, bumping around inside the tube, while the electrons dance around them about, oh, 2 x 10^5 times faster. (That's slow, but it also doesn't take much velocity to have a good kinetic energy on an atom that size.)

If you assume the ions and electrons fully migrated to opposite ends of the tube, making assumptions about the number in the tube (pressure, internal volume, etc.), ionization level, etc., you could guess what the literal capacitance would be (since we're talking seperation of charge here). Such a capacitor would discharge quickly!

Yup. Don't forget that the electric field isn't constant- each time an electron has enough energy to ionize a neutral atom, it soon will. In low pressure, low current situations, you can actually see this as glowing bands above the cathode: electrons go so far, gaining kinetic energy in the field, then they smash it away, then go on again and so forth... Each ionized region has a nearly equal voltage across it, while the un-ionized [not union-ized, heh] regions have somewhat more field in return.

That should be correct. If you have a spectrometer on hand, you could see if it looks black body (at that temperature, probably centered around blue?) or has any spectral lines characteristic of xenon (or anything else for that matter!).

I don't think so. It's quite conductive before it cools down. It takes a lot of energy to seperate charges, and that energy is soon released if it can be!

Tim

-- Deep Fryer: A very philosophical monk. Website @

I suppose that OP might use something like what they drive the STP nitrogen lasers with.