Controller for large buck regulator

>1000A that is a lot of current!    What is the input voltage and output voltage?

Maybe you could use a multiple  multiphase dc/dc controllers,

http://focus.ti.com/docs/prod/folders/print/tps40180.html

or

http://focus.ti.com/docs/prod/folders/print/tps40090.html

400v in

270v out

and it has been reduced to 800A.

Generally, buck controller can accept low input voltage like 60Vdc. 400Vin is not a range for buck controller.

this would be very challenging design, due to a few raw facts of physics:

if your IGBT's have conduction voltage of 2v7 the switch resistive losses go to 2kW range 

if you would use asynchronous rectification the diodes would dissipate 0.5-1KW

capacitors at 800A 15KHZ single phase synchronous buck would be huge unless the load can accept many volts of ripple

if you would split the design to use multiple parallel 120-200A modules (5-9 of them)  and use multiphase controller,

if you have 9 modules operating at 40 degree phase shift at 15KHz, the output ripple will be at 135KHz and that is way easier to filter than 15KHz

unfortunately the inductors of multiphase converter would still see the 15KHz and be of respectable size even for 150A current range.

I have designed  500A +/-200V output power supply (with 115V 3phase AC input) (it was actually a class D amplifier but could be used as PSU)and it was implemented using 24 parallel half bridge stages, and the output inductor alone weighted some 20Kilograms

and was made out of 100 or so TV flyback cores and 5mmx50mm copper band the

that design was done in 1985-1986 and the power switches were implemented with Fuji/Mitsubishi/Toshiba dual transistor modules of 100A 1200V specs

400A units were tested in 4 parallell units, but after a few explosions of transistors it was determined to be bad engineering to go that high in currents

the design/manufacturing/testing of 25 4KW output modules is easier but not necessarily safer than bigger units

also for multiple parallell units it becomes possible but not trivial to have redundancy and individual as well as global current and temperature protection circuitry

in this unit each of the 48 transistor modules operated at 5KHz  and the unit was installed in 2meter hight 80cm wide cabinet

I would say that semiconductors have improved a lot in 25 years, but laws of physics must still be obeyed.

p.s. don't kill yourself if you go ahead with this design.

Bill,

While TI does not offer a controller for a non-isolated 400V to 270V @ >1000A design, it may be possible to use some TI controllers for this application.

The TPS40091 controller is a 4-phase peak current mode control BUCK controller designed to use external drivers.  This would allow you to break your power stage into 4 250A power stages, but even that is likely more than you want to drive with a given controller.

The TPS40140 might be a better design.  This Dual Synchronous BUCK is designed to be stackable in upto 8 devices for 16 individual time phases.  62A is still a lot more than we normally see per phase.  I've never tested the TPS40140 at 15kHz, I'm not certain it will function properly that low frequency, but if it does:

You'll need external drivers for the 400V IGBTs with anti-cross conduction circuitry (the internal drivers drive MOSFETs with a maximum voltage of 40V and the external drivers will invalidate the internal anti-cross conduction circuitry.

While you could have 1 TPS40140 drive multiple IGBT drivers, I would recommend 1 TPS40140 controller per IGBT H-Bridge (So, if you wanted to divide the current over 32 31A Hi-Bridges, each with their own inductor with 2 H-Bridges running at each of 16 time divisions, I would recommend 16 TPS40140s rather than the "minimum" 8 controllers.  This will allow each controller to provide peak current mode control for each individual H-Bridge.

You'll also need to provide current feedback, likely through a current sense transformer.  The TPS40140 uses a high-bandwidth differential input current sense amplifier that can take the output of a current sense transformer (Though I would recommend a series bias diode on the ground side of the current sense resistor to raise the CS- pin voltage to 700mV as the current sense amplifier can have reduced accuracy below that expeected minimum voltage)

It's going to be a lot of components and circuitry, but the 16 phases will give you a lot of ripple cancellation on both input and ouptut.