ping Win, SiC fets

You could easily get 10ns, faster if you work at it. I've slowed my 600V switch to 8ns to keep it smooth.

Also, 7kV and 50 ohms, with 50-ohm cable, means a current of 70A. Switching that much current is one thing, but trying to do it fast means inductive voltage drops, V = L di/dt, will be a killer. You can work to keep inductance under control with one MOSFET and a ground plane, etc., but when you put them in series, low-inductance paths go to hell.

Dunno, John, that's a very serious game to play. Behlke does it, but they stick to slower speeds.

Just between us chickens, I'm eyeballing this one for an upcoming task...

Cheers, James Arthur

SiC tends to have high gate resistance, 25 ohms for that one. That would make it hard to drive fast.

And I'm not a chicken. Quack.

Well, actually, 140 amps.

Switching that much current is one

Of course it will be hard to do! But it could be worth a lot.

I might have to use the fabulously expensive Ixys RF planar mosfets to keep the inductances down. Maybe even a couple stacks in parallel.

I'd have to transiently over-drive Vgs hard to turn it off quickly, which might kill it. Folks do that with MOSFETs to overcome gate inductance-- hard, high pulse, then back off the overdrive before the die ever sees the overvoltage.

The literature recommends negative Vgs to keep SiC FETs fully off, and

Honk! (They're flying over head these days, headed south)

Cheers, James Arthur

You need to source the (presumable short) 50-ohm coax cable with a 50-ohm back-terminating resistor. That means the switched load is 100 ohms, not 50 ohms. At the far end of coax you see full voltage, as the pulse arrives. The 2x reflected signal goes back to the source resistor and stops the high current draw. But to insure the fast risetime, you need to source the 70A current into 100 ohms, with your desired risetime. That's bad enough!!

I'm driving my SiC with +16 and -3.5 volts. This gives me an acceptable-low Rds(on) and is a tad bit faster than the higher spec'd voltages. Observed Vth = 10V.

I've seen those high numbers in the datasheets, but my measured performance shows they're wrong, at least for my parts.

I need 140 amps. 7KV into 50 ohms.

OK, John, if you're hoping for 10ns or better, I suggest you forgattabout it.

BTW, lower-case k for 7kV.

Since when ? multipliers are upper case... dividers are lower case.

K is already taken for SI temperature in Kelvin

(except in backwaters that still use F for Fahrenheit)

k = 10^3 is the exception that proves the rule.

Well, 'k' is an exception, as are 'h' and 'da'. Look up 'SI multiplier prefixes'.

Why, oh why, don't we see calculators, spreadsheet programs, numerical programs, etc., etc., accept multiplier prefixes? Spice does it, although it *still* gets 'M' wrong. Gnuplot has some clumsy support for it too, but that's about all I know of.

Jeroen Belleman

Nope; kilo, hecto, deka

Doh I've been doing it wrong for 50 years :(

If you are referring to LTSpice, I think it accepts Meg as a multiplier. That may have something to do with how it discriminates milli from mega disregarding single-letter capitalization. But I think it must be Meg rather than meg.

Blame the teachers and the standard of modern education ;}

Spice doesn't care about letter case. Meg or meg will work for

LTspice buffs pay attention now: A capacitor marked 1F is *not*

Jeroen Belleman

Win, do you cover triacs in your new edition? I was gobsmacked to find there was not even a mention of them in the 2nd for some unaccountable reason. A *very* rare oversight on your part, though, I freely admit.