Ignoramus965 said:
Don, what are your thoughts on using IGBT's for switching?
Oh, boy! Now, you're REALLY getting in deep water!
The FETs are a lot more forgiving of shorts and linear operation, and
at the voltages you are looking at, they work a lot better, too.
But, still, you need perhaps 2-3 A of gate current for EACH FET in
your circuit! If you put 5 in parallel, you need 10-15 A of gate
current to charge the gates. Now, that is only for 50 nS or so,
then the current in the gate drops to zero. But, if you don't
provide this kind of current, the transistor will burn up in the
linear region before it ever gets fully turned on.
With IGBTs, if they are ever allowed to operate in the linear region,
even for a hundred nS, they are destroyed! The FETs have what is
called negative temperature coefficient, when they get hot, the
current they will conduct drops. This allows them to distribute
current evenly across the transistor, and across multiple transistors,
even when in the linear region. IGBTs flatten out in the saturation
region, but have a strong POSITIVE temp coeff. in the linear region.
Current will "hog" to the hottest part of the transistor die in the
linear region, and the transistor will self-destruct.
So, you really want to avoid the IGBT.
And, you will not be able to use your signal generator because it
can't provide several amps. You could do a low-power test, maybe
lighting an LED with it, to see the rise-times, etc. on your
scope.
For an H bridge, you need 4 different gate drives. Two of them
can be ground-referenced, but two of them (for the "high-side"
transistors) will have to be floating. You don't want to use
complementary transistors, ie. P channel, as they have MUCH lower
performance than N channel. So, using all N channel, the high side
transistors have their source connected to the floating output
terminal. So, the gate driver must supply gate voltage referenced
to the floating output voltage.
The IR 2113 and related chips can handle much of this stuff for
you.
One final problem is the DC welder will have a huge inductor at the
output. Thus, the output will probably have large voltage excursions
when the electrode touches the workpiece and the arc starts. So, you
can't depend on getting a steady 40 V DC or whatever. When the arc
breaks or sputters, you may have hundreds of Volts on the electrode
cable. I think the standard practice is to move the inductor to the
electrode cable AFTER the switching circuit. But, the design of
the welder may make this hard to do. Certainly in AC "buzz box"
welders, the output inductance is just built into the transformer,
it is not a separately-connected coil. I guess they can't do that
in a DC welder, so you may be able to change the connection there.
Jon
