1
\$\begingroup\$

enter image description here

I have been having a hard time finding out how to drive 32 gate drivers for SiC MOSFETs from a single function generator signal of 5 V.

My function generator is a Keysight 33500B. I'm using Genesic 1200 V SiC MOSFETs (G3R12MT12K) and cree gate drive boards (CGD15SG00D2). I'm designing a two series connected set of MOSFETs to boost voltage rating and paralleling 16 of them for a total of 32 MOSFETs (I will eventually use the 3.3 kV Genesic SiC MOSFETs (G2R50MT33k).

The MOSFET bank has to withstand and inductive voltage spike in a pulse inductive energy storage pulse generator. Really, I'm trying to split a trigger signal into 32 trigger signals that can still drive the gate driver boards. For some reason, this part is not explained well anywhere. I've read some about using various kinds of drivers or buffers to amplify the trigger signal back up in voltage, since the voltage will drop in splitting the trigger signal with traces or cables. Basically, I want to understand how this works and why we do it one way as opposed to another.

No one yet seems to address what I'm asking: how to get 32 trigger signals to the Si826 optocouplers. I've got plenty on the other stuff, just need to know how to build an amplifier system for a 5 V input trigger with pulsewidths from 50 ns to 2 us.

\$\endgroup\$
10
  • \$\begingroup\$ I remember SCRs and TRIACs to be triggered using signals/gate impulses - whatever, even flashes of light. Assuming one "bank" of 16 SiC-FETs "grounded", I'd be concerned about driving the other, floating bank, and supplying power to those drivers. \$\endgroup\$
    – greybeard
    Commented yesterday
  • 3
    \$\begingroup\$ Please draw a schematic of your arrangment, currently it's hard to understand how the mosfets are configured \$\endgroup\$ Commented yesterday
  • \$\begingroup\$ Tried to upload image but too big. \$\endgroup\$
    – Jay Eifler
    Commented yesterday
  • \$\begingroup\$ Reduce size and try again. \$\endgroup\$
    – winny
    Commented yesterday
  • \$\begingroup\$ (since the voltage will drop in splitting the trigger signal with traces or cables. It will? To any consequence? The drivers seem to have opto-coupler inputs, 4 mA are enough, apparently. I'd be concerned about timing more than drive strength here, and about delay differences between the individual driver modules as well as a systematic timing difference between "upper" and "lower" MOSFET.) \$\endgroup\$
    – greybeard
    Commented yesterday

2 Answers 2

1
\$\begingroup\$

This is a view of your Cree/Wolfspeed driver from here: -

enter image description here

It uses an Si8261 optical interface (U1) and that internally looks like as you would expect: -

enter image description here

The Cree module has R1 in series with one input and R2 in series with the other. Both appear to be 620 Ω resistors. However, because it is an optical input you can wire several inputs in series so that they share the same current thus limiting the loading they would otherwise take if you wired them in parallel.

Of course it would be highly beneficial to short-out the 620 Ω resistors on all but one of the devices so that the driving voltage could be lower.

Presently, you need at least 10 volts to activate the Cree module but, looking deeper into the Si8261 data sheet I can see that if the 620 Ω resistors on the module were shorted, you would only develop a maximum of 2.8 volts across each photodiode. If your drive voltage is 15 volts you would probably be able to get four modules working in series.

This reduces your problem by a factor of four. And, of course, you could use another (the same one) driver module to provide a strong output that can drive the 32 other inputs (8 strings of 4-in-series if you want). Or use the Si8261 on its own to drive the 8261s on the Cree drivers.


Regarding your choice of GeneSiC MOSFET, you may well be overestimating how much transient power this device can handle when switching high power loads. It is limited to 10 kW for 100 μs duration (see the SOA curve) whereas something like the ON semi NTHL020N120SC1 can handle switching transient powers of up to 50 kW (down to durations of 10 μs) and up to 100 kW at durations of around 1 μs. The GeneSiC part gets nowhere near this performance and, might be the reason why you have so many in parallel.

I must have blown maybe 4 or 5 up trying to get the GeneSiC MOSFETs to work when the data sheet was telling me all the time the bad news about those parts.

G3R12MT12K SOA: -

enter image description here

NTHL020N120SC1 SOA: -

enter image description here

\$\endgroup\$
5
  • 1
    \$\begingroup\$ The optocoupler is rf not optical actually. Not sure about the power ratings for the mosfets you gave. I've used NTHL020N120SC1 before. When I look at the SOA figures I don't seem to get the same answers. Maybe the series of optocouplers is a good idea. Really I just want to drive 32 of them, I can use an AWG to begin with. \$\endgroup\$
    – Jay Eifler
    Commented yesterday
  • \$\begingroup\$ Right it appears that genesic just didn't give shorter times of their SOA. \$\endgroup\$
    – Jay Eifler
    Commented yesterday
  • \$\begingroup\$ @JayEifler no, I have blown up maybe 4 or 5 trying to get them to work in the uncharacterized area and, I gave up on them big time. It's a massive flaw in every GeneSiC device specification I've read. Believe what the data sheet is telling you; don't use them outside the limits because they are crap. \$\endgroup\$
    – Andy aka
    Commented yesterday
  • \$\begingroup\$ Yeah it's 521 ohms, picture just fuzzy. \$\endgroup\$
    – Jay Eifler
    Commented yesterday
  • \$\begingroup\$ No, it's 620 ohms @JayEifler \$\endgroup\$
    – Andy aka
    Commented yesterday
0
\$\begingroup\$

Your schematic is unintelligible. If you are asking basic questions about how to drive these devices, then I suggest you STOP -- you will be dealing with extremely dangerous and deadly voltages and energy levels.

FYI -- minor differences in propagation delays and input thresholds will mean that the FETs won't all synchronize precisely enough to guarantee sharing. In addition, you cannot put 2 1200V banks in series and expect to get ~2x the V rating -- they will not share the applied voltage equally.

\$\endgroup\$
1
  • \$\begingroup\$ Already dealing with the voltage and energy levels all the time. It's true series connection of mosfets is dangerous but it has been done a lot. Mostly I'm interested in getting trigger signals to the cree gate driver boards. Sorry about the hand-drawn layout, maybe you didn't see where the mosfets were. \$\endgroup\$
    – Jay Eifler
    Commented yesterday

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.