8kW PFC Boost Converter

Updated : 5/30/2006

The project is a continuation of the 5kW Boost converter project.  After a few failed attempts, I realized that I was really after more power, perhaps 10kW or so.  I changed the title from 12kW to 8kW.  While the converter is designed for 12kW, its very unlikely that I will ever run at that power level.  So I expect the maximum output power to be 800VDC at 10ADC, so 8kW. 


Converter Type Discontinuous current, power factor correcting boost converter.  2 converters are interleaved, each operating at 6kW maximum input power.  There is an "active" snubber circuit that recycles snubber energy, this greatly reduces IGBT switching loss.
Switching Device IGBT: BSM200GA120DN2  200A 1200V. (warning PDF file)
Operating Frequency 17.86khz maximum.  21uS ON period, 56uS per cycle.
Inductor Parameters 68uH measured.  3 E ferrite cores. Cross sectional area : 21cm^2.  10 turns of 12awg machine/tool wire.
PWM controller UC3526 (warning PDF file)
Expected Efficiency 95%... we shall see.

Schematics for the PWM and low voltage section.

Schematics for the input power and high voltage section.

Revised power section schematic showing the active snubber design:

Explanation of the Active Snubber:

Because the converter operates in a discontinuous mode, this implies 2 particular conditions.  Firstly, the peak current will be about twice that of a continuous mode converter.  Secondly, because the IGBTs don't turn on with any current, we only care about the turn OFF losses, which will be quite high if not taken care of.  The Turn OFF loss is caused by the instantaneous overlap of the IGBT voltage and current (in this case 850VDC and 100A).  This overlap is very short, but can cause quite a bit of loss, which equates to hot IGBTs.  The trick is to then place a capacitor across the IGBT so that the instantaneous voltage at shut off, is 0V, that is, until the capacitor gets charged.  There is a diode in series with the capacitor, so that it could never discharge back through the IGBT when it turns ON.  The problem then becomes: what to do with the energy in this capacitor?  A typical snubber places a resistor in parallel with this capacitor to drain off the charge.  In this case, we are basically shifting the IGBT losses to power dissipated in this resistor.  The nice thing is that we can locate said resistor remotely, but still, we have to get rid of that heat.  My approach is to, instead, make use of this energy stored in the snubber capacitor.  Basically, this capacitor becomes the "power supply" for another small boost converter.  The capacitors energy is dumped into a small inductor (through another IGBT).  When the IGBT opens up, the energy in the inductor is forwarded to the main output capacitor of the converter.  My pspice simulations showed a dramatic drop in power loss.  With no snubber circuit employed, IGBT losses totaled around 850W.  With the active snubber, the total IGBT loss dropped to an amazing 200W, while delivering 12kW!!  Obviously, 200W is a lot easier to deal with than 850W, though will still require a considerable amount of cooling (consider a typical computer CPU dissipates 50-70W).

Over Current Limiter:

In order to keep the IGBTs safe, and to avoid saturation of the main inductors, I needed a way to monitor the current through them.  Richie Burnett set me straight on how to use CTs to monitor this current (which has a DC component!).  If I was using a single converter, the trick would be to use a diode in series with the CT output, so that the core would "reset" on every negative current pulse.  But, since I am using 2 converters, I can feed them both through the CT, but in *opposing* direction.  This keeps the v*s balanced.  I then need to use a full-wave rectifier on the output of the CT.  If the CT voltage exceeds the preset reference voltage, the controller ends the pulse right there, shutting off the IGBT.  The controller (nicely) makes sure that the pulse width sent out from A is the same length that B will put out, and vice-versa.  So if the current trips on pulse A, then pulse B will be that same truncated length.

High Power Testing:

Finally I have reached the most exciting part of the project, actually seeing if the darn thing works!  A few weeks back I tested the converter with a bucket of slightly salty water as a dummy load.  I ran it up to 4900W output (that's 780V at 6.3A).  I only ran for about 25 seconds since I was producing what appeared to be a good amount of hydrogen through electrolysis. 

The most recent test was on 5/28/06 where I actually powered up my DRSSTC-2 with the power supply.  It worked exactly as hoped.  The output voltage was very stable, and there was plenty of power to spare.  I only ran it up to 6400W that night due to an IGBT failure on the coil (a TVS string had a bad solder joint and fell apart... only a single IGBT failed and it was the one which had the broken TVS).  At 6400W output, the inductors stayed cool to the touch, and my main heatsinks were only warm.  Seems pretty efficient so far!  The converter survived something I didn't plan for: a dead short on the output, i.e. a dead TC.

Preliminary construction photos:

Converter set up for testing on my bench.

Home made PWM board.

Large cooling fan to move a fair amount of heat.

The boost inductors, stacked E cores.

Large IGBTs and diodes.  The active snubbers are partially hidden by the small yellow capacitors.

This picture shows how the CT is wired.

The smaller 700uH inductors.

A glimpse inside.  This thing is not much fun to work on!

Main control relays.  The yellow cap on the side of the relay is the "charging" cap.  White caps in the foreground are the DC filtering and must support 40ARMS!

All boxed up now.  A Plexiglas sheet holds an array of switches and the voltage control pot.  Up top there is a multimeter (for voltage) and an ammeter.

Volt and amp meters.  Amp meter uses a shunt to read 0-10A rather than 0-1mA.

Using a common receptacle as the 850VDC output, hopefully it doesn't mind the voltage!