Super duper hype fast FET driver?

The real "wall in the sky" is at 100 picoseconds. That's where things get seriously difficult.

John

It's easy if you're not looking for sub-ns edges. We use

FDV302 SOT23 logic-level drive

BSS83 SOT23 gp gumdrop

NDT2955 SOT223 for more power

These all run around 25 cents each, as opposed to a 2N7002 at 3 cents.

The n-ch FDV301, complement to the 302, is only 10 cents. Holes must cost more than electrons.

I'll test them for switching speed, eventually.

John

Did I say "insanely good" ?

The point is that we're talking about fet drivers, and you are arguing.

John

witch-burning?

stone-age

I have heard they are simply issued some sort of registration pass.

Yes. I have lived in countries with bans and without. With bans there were less people who fried their brains via drugs and less durg-related funerals. I prefer that.

But that one is unobtanium :-(

How did you keep the PNP out of saturation? Or did you still have some of those rare doped ones?

It might be surprisingly easy to get that going again, depending on the ADC sampling rate needed it could be as simple as one of the cheap uC eval kits. Those comes with USB nowadays.

ote:

and

hanks!

I remember seing tp0601 listed as complement to 2n7002, haven't checked the datasheet though

-Lasse

Thanks. Although storage and other times look really ghastly. I'll have to try it but I don't quite see how to get that down to a nanosecond or less.

The 2369 held true to it's data. When driving the living daylights out of it it'll switch much faster but there was always that small dreaded storage time of 10nsec or so.

Careful, the BSS83P (the P-channel version) has and end-of-life notice on Mouser. They have about 17,000 left, maybe time for hoarding those?

reath.

the

I don't think you'll get it under 1nS--I chimed in here where you were asking for a gold-doped PNP in general. I didn't me to imply it's going to meet your "sub-nS" in both directions requirement-its capacitances are too high. I don't remember getting it that fast either.

Sub-nS edges, measured. That's too slow?

The fastest way is to use an emitter current sink that doesn't allow the collector to swing the load far enough to saturate the BJT.

Slightly slower is a good schottky, c-b. I think I used microwave diodes. Another incremental improvement: the drive waveform had a big initial current pulse, then starved the BJT d.c. so it'd turn off faster.

Change diodes! Or bypass them with good caps. A schottky c-b and a slow trr diode in series could be interesting too, using the diode's Qrr to yank the transistor off.

Capacitance c-b has to be extraordinarily low--Miller really kills it.

-- Cheers, James Arthur

Yeah, I'll have to sledge-hammer a capacitance around anyhow. Requires at least an amp to do that.

I meant loaded down with the gate capacitance of a FET and such. That's when AC runs out of steam.

I'll have to look into the emitter sink. The current spike is something I did a lot but sometimes when I overdid it ... *PHUT* ... gone.

In my case Miller might not be too bad because I only need to swing 12V or so.

A 74ACT octal buffer makes a damned fine high-speed output driver or fet gate driver. Use 4 or better all 8 sections in parallel.

John

breath.

I was thinking, some years ago, about stacking cmos gates to get more voltage swing. Imagine, say, one 74ACT240 type buffer, all 8 sections in parallel, driving the ground pin of a second one.

John

breath.

Gold doping kills minority carrier life-time in P-type material (the base of an NPN) but has very little effect in N-type material, as in PNP's. 2N2369 is the most famous of gold-doped NPN switching transistors. ...Jim Thompson

Can it rival the NL37 series? The advantage would be that those come in octals. In the old days I have sometimes soldered several on top of each other but only in experiments. I know that was naughty but it did drive the big pulser (the super expensive lab grade driver box had croaked). It was 74AC though.

breath.

I was seriously thinking about doing such a bootstrap ride. But it'll add too much inductance to the whole game, more so than good magnetics.

Thinking? Works great. Try it some time.

John

There was a hiss, a kaboom and then a clean-up on Otis Street.

Oh wait, wrong city :-)

But seriously, that stuff does work. You need a higher voltage source and a TL431 style shunt for the one riding up. And of course a gate drive transformer to drive it, probably would have worked with a cap, too. Back in the early ultrasound days we tried a lot of such tricks because we had to cram umpteen pulsers onto a board, cost was an issue, and FETs were fairly poor and simply didn't work well with 6V or so in drive amplitude.

Later someone tried it with non-CMOS logic. Forgot which one, ABT or so. The lid on one of the chips flew off, probably latched up. I was surprised the substrate paths on the CMOS parts held up.

This was all 8-10nsec stuff and it was just about drive oomph and amplitude. Really, really slow compared to what I need now.

They are almost as fast as NL37s, sub-ns, but have the advantage that they are "ttl" compatible. If you power an NL from, say, 6.5 volts and drive the input from 3.3 or even 5 volt logic, they can really get hot.

Plus, you get 8 in a can.

John