fast risetime hv circuit

No, different processes versus the original.

Oh, maybe you mean newer parts in general. Could be. The good avalanchers are older, slower, low Ft diffused parts. Sometimes people make faster epitaxials and sell them as the same part number (probably they get a lot more per wafer) but they tend to not avalanche.

John

Anything will avalanche. At least once ;-)

What's your favorite transistor for avalanching?

I've used the Zetex parts to make hundreds-of-volts, very clean pulse generators, but they're slow, a couple of ns edges. I only have a little fooling-around experience with faster parts, so I couldn't really say what's good. I did enough experimenting to conclude that older, slower diffused parts avalanche best.

There are so many really fast parts around now - varieties of ECL, mmics, gaasfets, SRDs, shock lines, even cmos gates - that avalanche isn't as appealing as it once was.

We did have a fascinating, but money-losing, excursion into making fast 2KV pulses with water-cooled drift step-recovery diodes. Think about forward biasing a PN junction at +48 volts to pump in carriers, then reversing it from a -400 supply...

John

John,

Since it sounds like that stuff isn't a valuable trade secret anymore, sould you consider giving us the details? Using a rectifier to make fast kilovolt pulses is pretty neat.

Cheers,

Phil Hobbs

Step 1. Connect diode to AC powerline (13.8kV should do). Step 2. Capture pulse. Step 3. Replace charred diode remains. Repeat. ;-)

(Which probably wouldn't work very well, since turn off depends on extinguishing the arc, which probably takes a few miliseconds, if it dissipates at all.)

Hey John, were you going to send me some SRDs? I think your e-mail fell through a while ago, or something.

Tim

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

Did that not happen? Sorry if I dropped the ball. Send me another email and I'll get some diodes out.

John

in

Well, I suppose that as long as nobody seems to be buying these things...

This uses a couple of fast power mosfets to pull up to +48, and a couple more to pull down to -400, transformer coupled into the gates. The up-down node connects to the grounded power diode through an inductor, just a few turns of solid wire in mid-air. We forward bias the diode for around 100 ns, full 48 volts. It eventually realizes it's supposed to be on and starts to conduct. After 100 ns or so, it's up to 15 amps maybe, and we switch to -400. It looks like a short for another 80 ns roughly, reverse current builds up in the inductor, and at about -25 amps it suddenly snaps off. We go through a couple of SiC diodes into a 50 ohm resistor, with a little peaking L as I recall, and out a coax. At the end is a small capacitive load, the extraction aperature of a 3D imaging atom probe, where the pulse doubles to about

-2400, FWHM of around 3 ns but with a nice spikey shape, ideal for ripping atoms off things.

This is one of the Grehkov power-diode effects, the slow one in fact. The rest is just details, like where we get the diodes from.

It's interesting that dopant diffusion often accidentally produces the ideal carrier profile to make step-recovery diodes, as Boff discovered. Power diodes are typically not ideal, and recover very slowly - microseconds - when used as SRDs. Grehkov noted that a brief forward bias, as opposed to DC, makes them work much better. This was not a purely original discovery, as HP used pulsed forward biasing to improve the speed of their classic 1430 sampling head, using diffused SRDs, to get the bandwidth up to 12 GHz.

John

True. That's why nobody would even think about avalanching anymore when having to drive a diode sample. There you only need a few volts and a few tens of milliamps. But generating large pulses is another matter. The market for that so small though and that may be the reason for the scarcity of SRDs and such.

Tough market. Mostly universities I guess. Maybe you could let some of the rigid coaxes run on the outside of the T220, like the exhaust pipes on a Cobra. Looks more muscular.

ancient..

in

Interesting, thanks. I'll have to try that out.

Cheers,

Phil Hobbs

probably

ancient..

in

I could probably send you a couple of "diodes" to play with.

John

in

probably

ancient..

PN2369 in

If you have spares, that would be great. All this device work is making my soldering-iron finger itchy.

Thanks,

Phil Hobbs

IBM T. J. Watson Research Center

John Larkin a écrit :

in

probably

ancient..

PN2369 in

Now that is interesting info. Seriously...