Need BIG high voltage PWM driver

Well, that's a real killer combination. I doubt there is any such chip. The ones with lots of drive are aimed at big motor drives, where the PWM frequency will be low, the ones designed for high-freq switching, if there are any, would be for smaller FETs.

I think you will have to roll your own. Making the driver for the low-side transistor won't be that tough, but you'll have to search a bit for the component to send signals over to the high-side driver without badly distorting the waveform. You still will be burning a LOT of power to charge and discharge the 2nF Cgs at MHz rates!

Jon

It's not that bad, done it before at lower voltages. For example with the LTC3810 which has really staunch 1ohm drivers. So it is possible. It's just that the isolation well in the LTC3810 can't do 200V.

It would be no problem to roll my own but this becomes a very busy and thus large circuit because I'd have to cook up the delay times, safety interlocks, UVLO and all.

Folks,

Looking up and down the various mfgs I can only find drivers with fairly wimpy muscles, nothing much below 10ohms Rdson on the drivers. I need:

PWM input with programmable dead-time

Example:

But ... it's wimpy. Above 200-300kHz it becomes seriously hot and it's outputs don't have much power. Do you know something similar that won't heat up too much below 1MHz and can send several amps into the gates for a zippy switch-over?

A buck driver such as the LTC3810 could be pressed into service. Unfortunately they don't go past 100V.

That's just a simple low side driver. What I need is high plus low side with deadtime control and the whole nine yards. Ideally already with a PWM input.

Den onsdag den 18. juni 2014 23.21.59 UTC+2 skrev Joerg:

put something like a ZXGD3006E6TA on the outputs of the irs20124s ?

-Lasse

What little I've been involved with at that sort of voltages (and it's always been looking over someone's shoulder, not doing it myself), the high-side gate drive circuits don't get handled by a single IC.

It's either a logic signal into an isolator driving a gate drive with power rails that are referenced to the FET source, or it's some sort of driver referenced to 0V into a transformer, with network on the secondary to clean up the waveform into the gate.

I assume you have this in your back pocket in case your one-chip desire fails.

Good idea. Maybe use the ZXGB3004E6 with 8A peak that Harry suggested. It won't solve the problem that the big driver chip heats up just on its own but maybe that becomes more manageable if it doesn't also have to drive the FETs directly. Prop delay looks nice and low so it should not mess up the dead-times too much.

Yes, it's in my back pocket. Next to where the poop bag for our dog walks is :-)

I can always resort to doing it roll-my-own style but with dead-time control, UVLO and all that it gets old.

Looks like you need a custom chip >:-} ...Jim Thompson

Good idea. Maybe use the ZXGB3004E6 with 8A peak that Harry suggested. It won't solve the problem that the big driver chip heats up just on its own but maybe that becomes more manageable if it doesn't also have to drive the FETs directly. Prop delay looks nice and low so it should not mess up the dead-times too much.

I have that problem now, trying to drive some big kilovolt mosfets in a couple of ns. Looks like I'll need 100 amps or so of peak gate current.

I always wonder at what point a gate path goes *PHUT* since the datasheets are usually silent about that.

Lasse's suggestion of using a fast low side driver "riding shotgun" on the high side is really interesting. That driver can seriously spiff up an otherwise slouchy drive. The ZXGB3004E6 looks nice even though it won't pull all the way to zero so turn-off will suffer a bit.

Why on earth does hardly anyone make gate drivers that also go negative?

Not enough qties this time, unfortunately. But no kidding, I think there is a good market for a good driver. Ideally one that can yank a gate into serious negative turf for a sledgehammer-style turn-off. Anything I've looked at, and it was around 100 chips, made me yawn.

Many marketing types think that one more pin or the need for an extra negative supply would not be cool. But they are wrong. High-speed power designers would almost kill for that.

I like the LM5112, which does let you pull the gate negative. I've used it for non-gate-drive apps, like small switchers.

But, like the ZXGB thing, you might toast it at high rep-rates.

IXYS used to make some brutal screaming gate drivers, but discontinued them. IXDN602 is nice, but nothing like the old IXDD415 beast.

Yeah, I need to get to at least 700kHz or so.

IXYS seems to do a lot of cool stuff and then drops it shortly afterwards. I got one consulting gig out of that so it's not all a bad thing 8-)

But it's nothing compared to the number of my Maxim design-outs.

Den torsdag den 19. juni 2014 01.44.14 UTC+2 skrev John Larkin:

this is rather beefy

-Lasse

M***m!!! That's like saying Vold****t! The Fab That Must Not Be Named.

The local rep said that there would be a replacement next year maybe. That could be it. 30 amps in 3 ns. It won't be cheap!

Available: