I am making the board using a fullbridge driver with 4 fets or igbt's (same
TO-247/TO-3P footprints) - ie. FDH44N50 or similiar. I drive the
transformer primaries in parallel and put the secondaries in series. Since
the fullbridge effectively gives twice the voltage range compared to the
halfbridge I'm not using an input voltage doubler.
Stefan Heinzmann suggested this choice of MOSFET may not be the best. I
agree with his sentiment. Although it is tempting to believe that the
larger the MOSFET the lower the loss by punching in some basic I^2*R
calculations, but this is not always the case. Switching loss is quite
significant in switch mode power supplies. Larger MOSFETs switch slower and
have larger output capacitances. The output capacitance will produce
switching loss that is independent of the current the device is actually
switching. As such you cannot reduce this loss by merely using an extremely
beefy gate driver. Excessively fast switching is also bad from an EMI
standpoint. In order to get the EMI down often one must install big snubber
components which may in some cases introduce more loss than simply slowing
down the switching transition speed.
Your target output power is 20A*24V IIRC. That is 480W. Assuming 80%
efficiency then the input power would be 480W/0.8 = 600W. Since you are not
using a voltage doubler and this is a full bridge topology I assume the
input voltage will be about 160V DC. This is of course assuming you live in
North America or other location that uses about 120V AC. Presumably the
effective DC input current would then be something like 600W/160V = 3.8A.
In the full bridge topology the primary side MOSFETs need to be rated for a
minimum breakdown voltage at least as high as the input voltage. Since the
input voltage is around 160V DC, this would mean you need MOSFETs rated for
say 200V. Perhaps a little bit more margin is desirable in case of mains
transient events, but a 250V rated MOSFET should be adequate.
As you have probably noticed as MOSFET breakdown voltage goes up, so does
their on resistance. Therefore it is usually most desirable to use the
minimum breakdown voltage rated MOSFET that is needed for any given job.
The 500V device you mention is a poor choice for this item alone. Besides
the potentially lower efficiency for using an oversized MOSFET, this beast
or similar is certain to be extremely expensive. When you need four of them
this can run into real money if you design your supply such that you explode
a great number of them.
I would strongly suggest using a MOSFET such as the IRF644N instead:
http://www.irf.com/product-info/datasheets/data/irf644n.pdf
This is a 250V 14A 0.24 Ohm Rds(on) TO-220 part that sells for US $0.84 in
ten unit quantities from Digikey. Using this device will likely yield
somewhat better or same efficiency compared with the FDH44N50 and be much
easier on the pocketbook. You will likely need heatsinks on your diodes and
your MOSFETs no matter what choice you make in this application.
If you persist at working on this project, and you make sure to do lots of
research (reading/searching the web/libraries) after every explosion event,
you will be an SMPS pro in relatively short order. If after every explosion
event you simply solder in new devices and try again without figuring out
what caused the failure and how to prevent similar events in the future, you
will likely never be very successful in this project. That said. Good
luck.