Originally Posted by valerun
probably possible but:
1. 4000 amps of IGBTs (assuming 4000 amps is a combined continuous rating) will give you maybe 2000A peak. You will need to derate ~30-50% to allow for heating and then you need to derate for uneven current sharing (a LOT of derating if you use, say, 7 600A IGBTs). So in the end, you will probably have to use 10 600A IGBT modules to get reliable 2000A. Plus 10 similarly sized modules for freewheel diodes (unless, of course, you find dual IGBT modules - which are hard to come by used at that power rating).
2. Of course, at that point, you have several feet of busbars connecting things together (even more if you are using separate freewheel and switch modules). Every inch of a standalone conductor is ~20nH. at 2000A turned off in 1us, this will kick back with 40V inductive spike (assuming linear ramp-down which it ain't so likely 2x higher or so). 12 inches - with 480V, 2 feet - with 960V and so on. So proper bus designs etc are critical and even then you would probably be looking at 500V+ spikes at your desired current levels.
3. This means you *will* have a voltage limitation on your design. It will be tough to find modules with voltage ratings beyond 1200V. Also, other isolating components (Dc-DC for gate driver, isolator for drive signal, etc) can likely be taken to 1kV tops. Which gives you 500V max battery voltage input. But that's with zero margin so derate some more. Pretty soon you're at 300V pack voltage rating.
4. You can reduce the voltage stresses by slowing down your switching. Then you will need to reduce switching frequency to keep the switching losses under control. But then you will have another problem of input caps becoming too large (as they need to support the switching current). At 2000A output, 50% duty (when they work the hardest), you are looking at 2000A draw for, say, 50us (10kHz switching) and then 2000A fill for same time from battery. If you want to limit your battery current ripple to, say, 10% (or 100A in this case), you are looking at 5V max voltage ripple on caps (assuming 100 100AH CALB cells with 0.5 mOhm per cell). 5V at 2000A for 50us is 20,000uF input caps. Chances are, with components you can easily find, switching frequency has to be reduced further. At 5kHz, you will need 40,000uF, and so on. All that is in addition to your caps needing to support 2000A ripple current which is not trivial and sounds like 20-40 high-ripple smaller caps in parallel on a laminated bus to me...
Overall, according to my analysis (not really expert but after having read a LOT of stuff and taking my pseudo-controller to ~500A), it will be 'really' hard to get more than 500kW out of the controller build with reasonably accessible parts. And I might be optimistic here, too... Then again, you have a 300kW Soliton1 and 500kW WarpDrive already built for you...
If you want to go higher than that, it's into the exotic land of 1600A 1700V $1000 IGBT modules, custom film capacitors, active shaping of the switching curves, exotic snubbers, etc.
Not impossible but far from just mounting a few IGBT modules on a heatsink.
PS. Don't have any vested interest in Soliton1 (yet, anyway ;-) - just appreciating what these guys did after having tried a few of those things on my own..