Voltage losses in iGBT trippler

Got a schematic?

On a sunny day (Wed, 18 Nov 2015 16:48:44 +0000) it happened TTman wrote in :

For me y'r text is cryptic. How 'bout a cicuit diagram?

BTW I gave up on resisters and use taps on the voltage multiplier, less power, more stable. Did you scope it for waveforms? And some HV diodes drop more than a few volt. So diagram and part list needed.

IGBTs are really slow, 500 ns may not be long enough to turn them on fully. Your subject line says IGBT, your text says BJT. Which is it?

Jon

I'll try and get a circuit and post a link. They are IGBTs...

IGBTs are slow, very slow.. that 500nS aint doing it as far as I can see.

Jamie

Well, not the fast ones. The more majority carrier conduction (MOSFET character), the faster they are, down to 50ns or thereabouts.

But it is hard to get speed in the high voltage models. Partly due to junction length and recombination, and partly due to economics (>1200V devices are geared towards motive power, where up to a few microseconds just isn't a big deal). So more in series might be needed. If they end up 600V or below, MOSFETs might even be worth considering, unless the pulsed current also needs to be quite high.

Back in the day, induction heaters would use stacks of 200V thyristors, because they were available in fast grades (t_q < 10us). Same thing applies here, but recombination at low hFE is better (t_stg ~ 500ns?), so it's not *that* terrible.

For more ideas, check out some of the references that turn up, concerning IGBTs, pulsed operation and drive, magnetic compression and whatnot -- particle accelerators need a variety of pulse drivers for deflecting and focusing beams. e.g.,

Basic idea for magnetic compression is, you put a saturable reactor in series with the load (usually consisting of a big enough core made of square ferrite, square permalloy, orthonol, or nanocrystalline material, and enough turns to get the desired saturation flux and therefore time delay), then turn on the relatively slow switch, leaving enough time for it to fully saturate, before the reactor saturates much more suddenly and

Hmmmm... I wonder if there is a charge delay effect in IGBTs. That is, does the collector have an inductive characteristic, analogous to forward recovery in diodes? The physics being, the junction is only as conductive as the free charge carriers (mostly minority carriers in junction devices -- diodes, BJTs, dominant in IGBTs), the concentration of which is due to the history of applied current. Apply an upward current step, and the voltage rises a little until more charge carriers are freed. (Apply a downward current step, and it remains rather conductive until charge storage and recombination have cleared out, hence reverse recovery and storage time.)

This would be relevant to a magnetic pulse compressor, where the current spikes up suddenly at saturation, which could yank the collector voltage up, even though the transistor is otherwise apparently "fully on". It might be combatted with a relatively large capacitor strapped C-E, so that a large discharge current must flow, priming it for the peak current flow later. Or put another way, a C || (R+C) || ... network, or something like that, wired in parallel with the load, so that collector current remains relatively constant during turn-on.

Concerning drive, there are several papers out there talking about squeezing extra speed from commercial "monsters" (circa 1200V, 2400A, ~2us), which involves goosing the gate terminal to +30 or more for a few hundred ns, dropping to a modest simmer of 15-20V during the pulse. The overdrive is dangerous (they aren't usually rated beyond +/- 20 or 30Vge), but combats the gate spreading resistance (usually 1-2 ohms, even for the big ones).

All these methods of course concentrate a lot of effort on the transition region, where huge power flows, so you probably don't want to resort to these methods for high duty cycles, high switching frequencies, or commercial production... (For lab equipment where grad students are always handy to replace exploded parts, no worries!)

Tim

We stock 5K amp thyistors, hooky pucks rated at 1500V. But I would not be using those to steer a beam.

We had some older equipment that used magnetic amplifiers, I am glad they are gone :)

Some times a slower component is benefiical. Fewer external components needed.

We replacement a scan amplifier that used HV NPN in quasi state with a class D amp using IGBT module pack. That was stable to be used for both a 200hz scan for the accelorator, steering and focus control.

The old system suffered great problems if there was an unexpected interruption in the supply. It would at times destroy the upper row of the HV NPN's. This is also with the use of 40 amp stud mounted diodes to by pass the event to the supply. Problem is, it depended on what angle the scan was on when this happens. The quenching diodes were just to slow and to high of R in that state. I suppose a bank of smaller diodes in parallel may have help that? It seems the uncontroled collapse of the magnetic coil, a large one, can really cause some damage. The IGBT design seems to handle it because it, along with using welding caps for high current and low ESR, seems to capture the event.

We put those across the main supply caps so we don't suffer from the inherent L.

Jamie