Can't make xenon strobe fire!

[Oznog]

I have a xenon strobe tube I'm trying to fire. It's pretty big. I have a 300uF cap bank and the charger takes it up to 435v which should be plenty.
https://www.electro-tech-online.com/custompdfs/2005/09/093601.pdf

I've got a nice big trigger transformer. It's powered off the main cap bank, the primary is in series with a 0.022uF cap in parallel with a 47Mohm resistor.
https://www.electro-tech-online.com/custompdfs/2005/09/114265.pdf

The primary is pulled down by a high voltage FET which should be capable of 5 amps at that gate voltage. The FET's triggered by a pulse from a PIC. The tube current also goes through a FET so I can make it "twinkle" a bit with short flashes by cutting short the discharge cycle and pausing before retriggering.

Anyways, the cap charges fine and the PIC seems to be properly switching the tube onto the cap and then powering the trigger. But the tube won't fire. I measured the pulse with the caps only charged to 12v and the transformer secondard hit about +150v on the scope. As far as I know that's supposed to be positive. I can't measure the pulse with my scope directly since it's around 6kV.

I don't get it. Why doesn't this ******* fire? There's nothing going on in that tube that I can see or hear.

 

[williB]

hmmm .. when i was working with my little strobe project .. i seem to recall trying to get it to fire with a fet , but it wouldnt ..
so i used a relay , and then it worked...
hope this helps.

 

[audioguru]

This site has info about which end of a strobe light bulb is its cathode:
https://www.misty.com/~don/donflash.html

 

[Someone Electro]

Did you notice that the transformers rated output(6000V) is right on the minnimum triger voltage of the strobe(6kV).

 

[Oznog]

Hey, I got it to fire!
I'm not sure why- I first tried pulsing the trigger primary manually instead of with the FET by flicking the ground wire past it real fast. Then I tried adding to the capacitor so it would see a longer current pulse. Somewhere in the mucking about it fired. I'm not sure of the exact sequence of events now. In fact I'm not sure what changed at the moment it started firing. But, now it fires just fine!

I wonder if there's some significant break-in effect here? Like it needed a super-strong trigger the first time?

From start to finish voltage, it's taking 27 Joules. Big flash! I'm trying to figure out what I can reasonably expect from these tubes in terms of more flashes, maybe at reduced duration.

 

[Oznog]

OK it fires. The problem I have now is the commutation isn't happening. I knew the flash's power was consumed quickly but I had no idea how fast. I this case I'm hitting the trigger and shutting off the power in only 3 PIC instructions, with the internal osc@4MHz this is only 3 microseconds. The ending cap voltage is the same as making it 10x as long.

Maybe I should try adding some inductance. I already put in a high speed, high voltage flyback diode to accomodate the inductive spike from the cable run.

Actually I'm confused. If I can drain 400v out of 450v in 3uS, that's 133MV/sec dv/dt rate. That would require 40 kiloamps out of a 300uF capacitance. Can somebody tell me if my math is flawed? This is obviously impossible (duh!). The transistor alone only has 5v on vgs and the specs say that makes an Ids of around only 5 amps.

I must have something wrong with the oscillator freq (will check that right away) or the transistor is experiencing far greater turn-off times than usual. That is just an N-channel 600v MOSFET that has the source to ground, it pulls down the tube cathode to complete the current path. So I don't quite see why this would happen.

I need to rig up a high voltage probe here. Being able to scope this would tell me a lot. I just fear arcing resistors or whatnot, my scope is a nice one that I'd strongly prefer not be destroyed.

 

[Nigel Goodwin]

OK it fires. The problem I have now is the commutation isn't happening. I knew the flash's power was consumed quickly but I had no idea how fast. I this case I'm hitting the trigger and shutting off the power in only 3 PIC instructions, with the internal osc@4MHz this is only 3 microseconds. The ending cap voltage is the same as making it 10x as long.

The duration of the flash is goverened entirely by the capacitor, flash tube, and any resistance (internal or otherwise). Your trigger pulse is irrelevent - it only turns it ON, it turns OFF when the capacitor is discharged.

If you're trying to switch the flash tube out of circuit with an FET, I can see that's going to be fraught with problems? - you're switching a very high current. If you want smaller flashes, switch different value capacitors in circuit - the smaller the capacitor, the shorter the flash.

 

[Oznog]

Right, the trigger pulse doesn't determine the flash duration. And yes there is a second FET here controlling the pulse width. I would like to keep the PIC software in control of the exact pattern of flashes it blinks in. An ideal pattern I am initially aiming for is weak-weak-strong. The cap is basically sized to be close to the max rated flash energy. Also the little transformer needs about 6 sec to charge it currently from a full flash so quicker ones are not possible. Not only is there no room in the enclosure for multiple caps, each would require its own switching element anyways. It could then be done with an SCR in that configuration but other than that there's little gained and it's way outside the desirable spec.

Still, it's a mystery just what is happening. I tried to check the board over and over and found nothing amiss. This is the transistor:
https://www.toshiba.com/taec/components/Datasheet/2SK3562.pdf
and it's driven with 5Vgs which is around 7.5 amps Ids at room temp. (I'm going to replace that with one that does 15 amp @ 5vgs later) So even if the tube took that much, it should be nowhere near done with its pulse in 3 usec so it should have been cut short but wasn't. I wondered if the current could have been high enough to create an IR drop on the ground trace to somehow keep it on, but nothing in the board layout would make this likely that I could see. I wondered if the pulse could have shaken up the PIC. It has a 0.1uF bypass cap next to Vdd/Vss and its ground was not placed between the flashtube ground and flashcap -. Actually I'm wondering now because when I think about it, if the PIC reset itself it would restart its timing interval which is currently all it does, thus I would not have noticed it reset. I will check on that.

Circuit works fine otherwise. I had it popping out 10 flashes/min for awhile, transformer got hot, everything else pretty cool. So we can rule out having shorted out that FET, a single pulse would have toasted it.

Normally I would troubleshoot this hands-on much more, but without a HV probe or two for the scope I can't do a lot. Nor am I really comfortable doing a lot of handling on a hacked together circuit with 450v and enough energy and current capacity to kill me many times over. So it's going to require more though than random tinkering.

Any ideas? I mean, it should cut the flash short, I don't see how it the transistor could fail to do so.

 

[Oznog]

Oh wow the transients are something to behold.
I got the scope on the gate and it's a load of noise swinging all over the place.

I checked on the Vdd on the PIC and as it fires it jumps up to 5.5v, maybe 6v. Enough to say there's no certainty the PIC is executing code properly. It's not resetting though.

I was curious if maybe it's just induction in the scope leads alone so I clipped them together and left them nearby, and saw nothing at all on the scope.

That's with a 0.1uF ceramic on Vdd. Since the reg output is only 0.1" away this is also the reg's stabilizing capacitance.

The 5v rails are a couple of inches away from the cap and don't share traces with the high current path. I can shorten the wires connecting the 2 FET gates a bit to cut down on induction but only so much.

Or could it be capacitive coupling between the gate and source voltage? I'm having a hard time seeing that because it would try to drag down Vg when the gate transistions high and vice versa. But Vdd could only get boosted like this if a >5v voltage was induced when the pin is pulled high, unless it generates enough current to pull it up that high when the pin is driven low (this would require a LOT of induced current).

 

[audioguru]

Mosfets have a huge gate to source capacitance, and more is added by the Miller effect. It takes a high current to charge or discharge that capacitance quickly. Try a complementary pair of transistors as a totem-pole (emitter-followers with their emitters tied together as the output to the gate through a 10 ohm resistor to stop oscillation) current booster for the gate.

 

[Oznog]

g-s cap isn't really huge here. The PIC pins have strong driving capabilities and it's a sharp enough 0-5v pulse when the HV is disabled. So I'm thinking Miller effect.

It's showing very asymmetric. It takes about 50usec to rise to 7.5v (and this is with a 5v source voltage) and then I see an exponential decay like a RC time constant and it takes about 1800usec to decay to 1v.

I don't really get it. If this were Miller effect then the it would slow the charging of the gate less than trying to turn it off with the cap partially discharged. The turn-on times being so much quicker than turn-off is the opposite of what I expected. Nor would this explain the overshoot of the gate getting to 7.5v.

 

[audioguru]

Don't you have a Mosfet pulling-down to ground the cathode of the flash-tube? Camera flash circuits don't have anything to reduce the current. I think the Mosfet has a fairly high voltage across it during each flash because it can't pass the very high current.
Then maybe you have a capacitive voltage divider between the fairly high voltage at the drain of the Mosfet when it can't fully saturate due to the massive flash-tube current through it. The capacitive voltage divider passes some voltage to raise the output voltage (and therefore its supply voltage) of the PIC. It sounds backwards but maybe that's what is happening.

 

[Oznog]

Don't you have a Mosfet pulling-down to ground the cathode of the flash-tube? Camera flash circuits don't have anything to reduce the current. I think the Mosfet has a fairly high voltage across it during each flash because it can't pass the very high current.
Then maybe you have a capacitive voltage divider between the fairly high voltage at the drain of the Mosfet when it can't fully saturate due to the massive flash-tube current through it. The capacitive voltage divider passes some voltage to raise the output voltage (and therefore its supply voltage) of the PIC. It sounds backwards but maybe that's what is happening.

Oh you may be onto something here. If the MOSFET pulls the drain down to 0v, then the other FET fires the trigger, if the tube current exceeds the MOSFET's drive for 5Vgs the drain voltage will rise suddenly. Due to the capacitive coupling between drain and gate it might pull the gate voltage up with it. The gate's already high so this will be conducted to the positive rail inside the PIC. This is serious because the PIC is being powered by a 5v reg with little pull-down ability on its output (as opposed to a battery which would strongly resist this). Thus the latch-up and it looks like the pin just tristates after this and it won't conduct well to anywhere so it takes forever to get back down. I'm pretty sure that's it because the scope shows that pin voltage surging to 7.5v.

I put a 5.1v zener on the gate-to-source. That should conduct the surge to ground but I'm still not seeing it. I'm getting 6.5v or so here, like maybe the current switches too fast for the zener to clamp?

 

[Sebi]

Possible cause also the GND wiring...