sparc5 said:
What about using a SCR instead of mosfets for discharging? They are easier to cool and will stay on even after the gate power turns off as long as the current is drain is sufficient. That might help out if the chip ever does reset. A separate board with MOSFETs means your current loop will be large and you'll get ringing with the MOSFETs.
I think, IIRC, the voltage drop across a SCR is a lot higher than a MOSFET.
And once the SCR turns on, there's nothing that can turn it off except waiting for the current flow to stop some other way (0% throttle?). If that's the case, I say skip the SCR since it's not really being used for anything. Also, back-to-back FETs allow for bidirectional current flow so you can have independent charge and discharge flow control in one wire. If using SCR's, you'd need to use separate charge and discharge wires (or use them in just the charge wire, but then it never turns off). SCR's are really best for use with AC current where you have an opportunity (typically) 50-60 times a second to turn them off.
IMHO, MOSFETS are the best choice here. With careful planning, you can keep the inductance of the gate wiring to the FET to a minimum. I have a 3" gate lead run without problems. The FETs can turn off in microseconds, even milliseconds, and still protect the MOSFETs from a short circuit so the speed of the driver can be fairly slow (compared to motor FET drivers). And even if the gate does ring a bit, that's OK as long as max Vgs isn't exceeded (and you can protect against that) or the FETs bounces on/off.
See the Design Guidelines for a description of the FET drive circuit. You can balance the drive's turn-off speed against its quiescent current draw as needed.
If you're set on using the EVM, it's best to get one, fire up the scope, and just start playing! It will be easy to see when the number of FETs you're using, or their wiring/arrangement, start to be a problem.