7

I'm working a circuit to send very short pulses to a load, with the requirement that the pulses have the fastest rise time possible (~10ns or less).

I've picked out a driver that has a fast rise time, the UCC27524P (http://www.ti.com/lit/ds/symlink/ucc27525.pdf).

I have a question about choosing a MOSFET, is it the driver that determines the rise time or the MOSFET's rise time in it's datasheet?

I couple of MOSFETs I'm looking at has rise times of around 20ns, are they simply unable to switch any faster?

I've attached a simple circuit to illustrate.Simple Switching Circuit

Brian
  • 73
  • 1
  • 1
  • 6
  • Why not a method that guarantees 1ns or 100ps? or even less then easy to slow down to 10ns with a filter? What a pulse without a load impedance ( R+X(f)) or source impedance or load current defined with 10ns? a design that never reaches it goal... reflect on this for a while. – Tony Stewart EE75 Jul 15 '17 at 21:29
  • 1
    I'll consider anything that can meet the rise time requirement @ 400 volts. If there is another method, I'm all ears. This is just what I came up with on my own. – Brian Jul 15 '17 at 21:33
  • then start adding specs above, also include inductance and I assume you know how to probe this rise time correctly and that you know that all wires, resistors have inductance and understand the result of On to Off. – Tony Stewart EE75 Jul 15 '17 at 21:36
  • 1
    Learn how to design by specs then choose parts. Sooner or later this a priori. – Tony Stewart EE75 Jul 15 '17 at 22:01
  • This is not your everyday power supply, I'd recommend at least looking into some avalanche breakdown transistors (i.e. the ones *specifically designed* to operate in the breakdown region). I remember seeing a paper from the 60's or so where they were switching even higher voltages at faster speeds so there's definitely ways to do this that presumably doesn't require super advances components (it's the 60's everything was handmade after all). You might be able to get what you want with a DC-block and a big GHz-class RF amp, just feed it a single pulse assuming it doesn't start oscillating. – Sam Jul 15 '17 at 22:47
  • 1
    Read this PDF on using GaN transistors, without oscillation. http://www.transphormusa.com/document/recommended-external-circuitry-transphorm-gan-fets/ – analogsystemsrf Jul 16 '17 at 03:46

4 Answers4

4

It's both the driver and the mosfet that determine the rise time. The net rise time is going to be about twice if both driver and mosfet are similar. You might also be interested in fall time and this can be harder to achieve especially if the gate isn't driven negatively when removing gate charge.

If the data sheets say 20 ns and there isn't some obvious improvement in the example circuit used to measure rise time then you are probably out of luck. Keep plugging away and look for better mosfets. I think there is a new type out referred to as something like silicon-carbide technology but don't quote me on it.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
4

Here's your solution. 400V, 4 ns.

https://www.avtechpulse.com/medium/avrf-4a/

The fact it costs $15k could hint at the fact that it is not a problem that can be easily solved with 2 active devices... hmmmmm...

Now, if you really insist...

Okay, you want to turn on a MOSFET really fast. The MOSFET gate is a capacitor. So you need to pump lots of current into it really fast. This means lead inductance is not your friend. This means leaded packages like TO-220 will only result in more pain. So, I searched DigiKey for a suitable part, and whaddya know, found one!

Tada!

(It is also available in TO220, if you want to take chances).

Now, this is not your average MOSFET. It's actually a cascoded GaN-FET. Ths cascode handles your other problem, Miller effect of Cgd capacitor, which at 400V is going to be truly humongous.

Notice how the gate charge is ludicrously low compared to your IRF840 (like 10x less). Also the rise/fall times are very fast (like 5ns).

And the TPH3206LSB version has the Source on the big fat cooling pad on the back, which means you can solder it to your ground plane to cool it. (If it had been the drain, extra capacitance on the output would have been introduced). Note the TO220 version also has the Source on the tab, so you can ground the heatsink and even remove the insulator.

I've never used a GaN-FET, but damn, this part looks like it means business!

With a solid layout and a bit of luck, it could work. Maybe... Well, you can always try... Although you should really work on your specs first, since we don't know anything about the load yet, so this is really a shot in the dark.

bobflux
  • 70,433
  • 3
  • 83
  • 203
1

The problem is not the gate charge (you chose a good gate driver) or intrinsically the MOSFET itself or even the stray inductances (assuming 10nH gate-source) but it is the drain-source charge combined with the load impedance.

A quick LTSpice simulation with the STP8NM60 (650V MOSFET that is built in) shows 50ns rise time with 600R load, but just under 10ns with a 60R load (agreeing with the datasheet 10ns figure). The output capacitance for the STP8NM60 is 100pF. A GaN transistor such as GS66502B (650V part available at Mouser for £8.19) has an output capacitance of 17pF, which means it could just about manage your required 10ns with a 600R load.

Matthew
  • 419
  • 4
  • 5
1

You can use ISL55110 driver with GS66502B GaN Mosfet, they should allow you < 2nSec fall time

Nikolay
  • 31
  • 1
  • 1
    Welcome to EE.SE. This may be a good solution, but a circuit diagram would help a lot since it's very short. – winny May 22 '18 at 13:31