Re: MOSFET arrays

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

Original poster: "Stephen Conner by way of Terry Fritz <teslalist-at-qwest-dot-net>" <steve-at-scopeboy-dot-com>

At 07:25 28/03/03 -0700, you wrote:
>Original poster: "Matthew Smith by way of Terry Fritz 
><teslalist-at-qwest-dot-net>" <matt-at-kbc-dot-net.au>
>
>Hi All
>
>I'm just looking for a little general guidance on the use of multiple 
>MOSFETs in SSTCs, OLTCs, flyback drivers, etc.
>
>
>
>1) Parallel operation.  I believe that the gate capacitance of a MOSFET 
>device can cause slow "off" switching unless the current can be sunk nice 
>and quickly, hence the use of a totem-pole driving stage.  I see these in 
>use both as discrete components and also in the output stages of various 
>PWM chips.
>
>Given sufficient current sourcing/sinking of (probably bipolar) driver 
>devices, what are the practical limits to the number of MOSFET devices 
>that they can drive in parallel?  With our potentially high ambient 
>temperatures, I'm working on a current handling capacity of only 3.5A per 
>device with the IRF620; would eight in parallel to switch 28A be practical?

The main problem with using devices in parallel is the parasitic inductance 
in the source circuit. High di/dt generates enough voltage across the 
inductance to partly cancel out the gate drive signal in some of the 
devices. These devices don't turn on fully, suffer very high losses and may 
blow out. An individual transformer-isolated gate drive for each MOSFET 
makes sure that all devices turn on fully, but still may not ensure equal 
sharing. These issues are explained in much more detail on Jan Wagner's 
SSTC pages:

http://www.hut.fi/~jwagner/tesla/

FWIW, an inverter that I designed switched 70A peak at 100kHz, using an 
H-bridge each leg of which had four MOSFETs in parallel. I connected all 
four gates in parallel and drove them with a push-pull pair of bipolar 
transistors that could source/sink 2A. The mosfets were carefully laid out 
to minimize/equalize parasitic inductance.

http://www.scopeboy-dot-com/elec/RE/inverter1.jpg
http://www.scopeboy-dot-com/elec/RE/inverter2.jpg



>2) Gate drive transformers.  I can see that these would be required when:
>         a) outputs are required out-of-phase (as in a bridge
>         configuration).
>         b) devices are employed in series for higher voltages
>Are there other circumstances where these should be employed?

See above. Their main use is for when the source of the MOSFET you're 
driving is not grounded, as in the top two legs of an H-bridge. You can use 
so-called "bootstrap" or "high-side driver" circuits instead, but the 
consensus seems to be that the less silicon you have around your Tesla 
Coil, the better.


>3) Series operation.  Since we only need a few (relative) volts on the 
>gate, I guess that this is a job for a transformer (see above); since the 
>secondaries will float, I would see the issue here of being one of 
>winding/winding and winding/core insulation, especially when switching 
>larger voltages.
>
>I seem to remember Terry looking at using optical drive on a large IGBT 
>array, which would give an alternative (with better isolation) to the gate 
>drive transformer.
>
>Again, what are the real practical limitations to the number of series'd 
>elements that we can use?  Could I switch, for instance, the rectified 
>output of an MOT, using a series array of, say, twenty IRF620's?  (200V 
>per device max.)

The challenge is getting all the MOSFETs to turn on/off at exactly the same 
time, even though the gate voltage and capacitance can vary widely from one 
MOSFET to the next. If one MOSFET lags behind the others, it's doomed to die.

There is a circuit where you only drive one MOSFET and it triggers all the 
others in a kind of domino effect. Dan McCauley is the TCML resident expert 
on this one and will hopefully be releasing his findings soon... The main 
limitation of this circuit is that the more MOSFETs you have in series, the 
slower it will switch (because the switching has to ripple through) and the 
greater the losses will be.

Optical drive is the technique that power transmission engineers use for 
their 100 megawatt thyristor stacks. Getting all the devices to turn on/off 
at the same time will still be a challenge, perhaps more so, because of the 
variability in LEDs/photodetectors/etc. You still need to provide an 
individual isolated power supply to each optical receiver. Solar cell-based 
receivers exist, they require no power source, but have very slow switching 
times due to the small photocurrent of only a few microamps.

Personally, for series switches, I'd favour the gate drive transformer 
approach. I would bifilar wind the GDT using pieces of RG-58 type coax, 
with the screen being the primary and the core the secondaries. AFAIK, 
RG-58 can stand about 16kV between core and screen. As for synchronisation, 
by fitting each MOSFET with a Miller capacitor (from drain to gate) the 
device's own capacitance would be swamped. It would also increase the gate 
drive requirements quite a lot and you would probably end up putting tens 
of amps peak into the GDT primary, but that's easily done with low-voltage 
MOSFETs.

Solid-state series switches (whether GDT or cascode driven) are probably 
going to be the future of Tesla coiling and I would strongly encourage you 
to experiment and push the boat out a bit. You can afford to blow a couple 
of MOSFETs after all :D

Steve C.