Re: SSTC questions

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Jan Wagner by way of Terry Fritz <teslalist-at-qwest-dot-net>" <jwagner-at-cc.hut.fi>

Hi,

On Tue, 11 Mar 2003, Tesla list wrote:
 > Original poster: "Sean Taylor by way of Terry Fritz 
<teslalist-at-qwest-dot-net>" <seantaylor-at-attbi-dot-com>
 >
 > Hey all you SSTCers, I have a few questions . ..
 >
 > I'm working on building a high power SSTC with a full H bridge of IGBTs that
 > happen to have body diodes.  It will run off fully rectified & filtered 120
 > V mains.  The IGBTs are rated at 70 A, 600 V, and have a Tr of 350 nS, and a
 > Tf of 300 nS.  Pretty slow, I know, but I plan on making a fairly low
 > frequency coil, and with current being at zero, or close to during
 > switching, it shouldn't be much of a problem.
 >
 > So the questions I have are:
 >
 > 1) Why do people "disable" the body diodes of the FETs?  With a full bridge,
 > there shouldn't be a problem with too much current through the diodes, next
 > to none if the switching of the FETs is done correctly and there is at least
 > one rectifying diode on the input.

They have too slow reverse recovery. Newer mosfets have now intrinsic
diodes that are ultrafast recovery, but these mosfets aren't available at
budget price yet AFAIK.

If you run a very low freq SSTC and slow down the gate drive waveforms
then the "slow" 200..300ns intrinsic diodes should still work out fine.

a conducting slow-recovery diode & other mosfet turns on => short circuit
accross the rails, there goes...


 > 2) How do anticipate how much current I'll be drawing?

That's more a matter of decision, not anticipation.

Just take any secondary coil and slap around "enough" pri turns and you
get a low current draw. Then reduce the turns, check new current draw,
reduce turns, etc. And restrain yourself from removing one turn too
much in hope for better sparks... ;-)


 > 3) What should I use for the primary?  I was planning on using 8 or 10 AWG
 > fine stranded (cause thats what I have laying around) wire in a solenoid,
 > but should I use something fairly "coarse" stranded for lower RF resistance,
 > and how many turns approximately?

Helical, good coupling, well insulated from the secondary. Maybe 20 turns
to start with, and then reduce from that. Even mains cable works ok. Or
some cheap speaker cable.


 > 4) The inductance of just the primary is fairly low, so why is it that the
 > current draw isn't that great, despite the output being loaded down,
 > especially if an arc is drawn to ground?

An arc detunes the coil, so unless the driver adjusts the frequency to
keep track of this, the current draw goes down.

You'll have to check this but IIRC the impedance of the 1/2 wave mode (arc
to ground) is higher than that of the 1/4 wave mode, so once again, the
current draw goes down a bit even if the driver tracks the resonant
frequency correctly.


 > 5) For gate drive transformers, what will the waveform look like if I start
 > getting too few turns?

You'll start seeing a resonance between the primary coupling capacitor
and the inductance of the gate drive xfmr. It looks a bit like a
rounded-off square wave. Or a square wave consisting of large chunks
and bits of a sine wave.

The other way to tell is that the driver ICs have melt... ;-))


 > I have about 15 turns right now, and get a decent
 > amount of ringing on the output (when connected to a capacitive load).

You could use a small resistancce series resistor on the output to snub
out the ringing almost completely. Say 3.3 ohm or higher (you'll see what
the proper value is when you try out a few different resistors :-)


 > I'm
 > using TC4420 and TC4421 FET driver chips since the gate capacitance of the
 > IGBTs is relatively high (~13 nF).  Also, should I allow Vge to be +/-Vg, or
 > use a cap and clamp it to -0.7 V at the lowest?  I know letting Vge get
 > lower will turn off the IGBT faster, but I'd rather have more voltage in
 > order to turn the IGBT on better.

Well, using -0.7V +Vg halfs the current draw from the gate drive circuit
(vs +/- Vg) so that's sort of a benefit. IGBTs are slow anyway, not much
improvement in using a +/- Vg negative bias drive IMO.

With -0.7V +Vg You could use a reverse diode (1n4148?) accross the gate
resistor to speed up the turn-off, if you don't mind some small turn-off
ringing on the gate.

I think another thing to consider is the current tailing of the IGBTs.
I've a small resonant inverter welder working with an old three phase
IGBT bridge from Siemens (BSM 25 GD 1000, or something like that), with
very similar time specs to those of your IGBTs. Current tailing into a
resistive load is ~120ns. Running at 60..100 kHz. Just using a +/- Vgs
drive didn't work - there had to be additional dead time (earlier
turn-off at nonzero current, in addition to turn-on into zero volts
accross the IGBTs) so the module didn't get hot. Despite resonant
switching and all that, which already should keep losses low, those mods
were still necessary.

I'm not sure if the same applies to your setup but adding some dead time
to the drive signal wouldn't do any (large) harm IHMO...

cheers,
  - jfw

--
*************************************************
  high voltage at http://www.hut.fi/~jwagner/tesla
  Jan OH2GHR