Jimmy's DRSSTC 2

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

Original poster: "Jimmy Hynes" <JPHynes-at-gmail-dot-com> 

I have been working on a much bigger DRSSTC for a while, and it's
about time to say something about it to the group. Unfortunately,
progress has been rather slow, due to interest in more "normal people"
stuff  =\. I have access to more power than I could ever use
(something like two 3 phase, 440v, 200a lines), so I just have to take
advantage of that. Plus, Steve's getting too cocky; I need my record
back (just kidding, I had to say that :P)

All the stuff I have on it can be found here:
  http://hot-streamer-dot-com/chunkyboy86/drsstc2/

Estimated power draw is ~20kw, with ~200joules per bang. Sparks should
be in the range of 25 feet. It's really just gonna be whatever the
secondary can handle.

The toroid is going to be 24" by 96". I'm planning on using the normal
chicken wire/plywood deal.

The secondary is going to be 24" by 96" also. It's going to be 22 awg
wire space wound on a cardboard form.

The primary is going to be 4 turns 40" diameter helix. 1.25" by 0.021"
copper strip for the conductor.

MMC: up to 17 strings of 2. 2.55uF max
IGBTs: 2 CM300DY-24H IGBTs. Full bridge configuration.
Electrolytics: 16 12,000uF 350v caps in a voltage doubler.
Heatsink: 11"x9.75"x2.25" with two muffin fans for airflow.

Gate driver: http://hot-streamer-dot-com/chunkyboy86/drsstc2/gatedriver.jpg

Power for the gate drivers is supplied by a SMPS (tx2) and signal is
given by another transformer (tx1). A DC bias is put on the gate of
the lower MOSFET so that the IGBT is held off when there is no signal.
The DC bias is chosen so that it will come down slow enough to prevent
voltage spikes in the case that it has to shut off to stop shoot
through.

The TVS is just going to be 4 1.5kw 200v TVS in series. Each IGBT gets
it's own protection this time.

The whole thing will be controlled with a uC, like my last coil. This
time, the uC will have more to do. This time it monitors the DC bus
current (feedback and OC protection), secondary current, heatsink
temperature, instantaneous die temperature, and the voltage into the
SMPS for UVLO.

Using a uC also makes it easy to play music (heavily distorted bass
lines, or drum stuff), and keeps it very flexable. It will have an LCD
readout so I can monitor whatever I want.

Feedback will be done by using the counter timer to add up the time
difference between the current 0 crossings (pri or sec) and the output
signal. It isn't going to try to correct the frequency within a burst,
it will be a burst to burst deal. Open loop is still a possibility.

DC bus current will be measured through a current transformer,
however, some changes will have to be made to keep the core from
saturating (thanks Dan Strother!)

Here's my solution: http://hot-streamer-dot-com/chunkyboy86/drsstc2/ct.jpg

The diode and resistor directly across the transformer are needed to
keep flux from building up and saturating the core. The resistor
directly across the output is of higher resistance than the sense
resistor, and the L/R time constant of it and the winding is about one
cycle. The two transformers are cascaded to allow a huge step down in
current with a practical number of turns. The RC filter is just an
attempt to keep some really high F junk out (prolly gonna use 10n
too).

Sensing the die temperature is sorta tricky. My original idea was to
place a small thermistor on the die, and use an opamp differentiator
to try to get a higher bandwidth. The thermal time constant is too
long, and using the opamp amplifies the noise too much, so I decided
against that approach.

IR photodiodes/phototransistors can measure emitted IR with only 1uS
of delay, so I can shut it off if the die temperature gets too high
within one burst. The wavelength of the diodes is near 900nm, which is
a little short for the temperature I'm looking for (~100C or so), but
should work.
(http://hot-streamer-dot-com/chunkyboy86/drsstc2/irpic03.gif)

The heatsink temperature sensor is simply a thermistor stuck on there.

I have simulated everything, and everything checks out ok. These sims
show that everyone is being too easy on their bricks. I am going to
use the same bricks that Steve Ward is using for 'only' 11 foot sparks
(pansy :P).

Peak current sim: http://hot-streamer-dot-com/chunkyboy86/drsstc2/peakcurrent.sch
http://hot-streamer-dot-com/chunkyboy86/drsstc2/peakcurrent.jpg

Pspice happens to have a model of the exact IGBT that I'm using
(CM300DY-24H). I set the Vce drop at 10 volts, and found that the peak
current is huge even at fairly low gate voltages. Clearly not a
problem here.

Even if Pspice is inaccurate at currents so high, experimental data
has shown that peak current will not be a problem.

I also used Pspice to simulate the thermal situation. In this sim
(http://hot-streamer-dot-com/chunkyboy86/drsstc2/DRSSTCIIheat.sch) I set
up the simulation so that 4 of those IGBTs were driving a series LCR,
and found the loss. I multiplied the instantaneous current through the
device by the voltage across it, then integrated it. That gives me the
total loss per bang for each IGBT. The gain afterwards converts joules
to degrees C.

The answer in that sim was used in the thermal sims here:
http://hot-streamer-dot-com/chunkyboy86/drsstc2/DRSSTCIIheat(real).sch
http://hot-streamer-dot-com/chunkyboy86/drsstc2/DRSSTCheatslow.sch

All values for the IGBT came from the datasheet, except for the
thermal capacitance for the baseplate, which I had to calculate. The
thermal capacitance of the heatsink was just weighing it (10lbs) and
some simple math.

The thermal resistance came from a neat little calculator online. I
think I found my exact extrusion on it too.

The results are here:
http://hot-streamer-dot-com/chunkyboy86/drsstc2/heatlong.jpg
http://hot-streamer-dot-com/chunkyboy86/drsstc2/heatshort.jpg
(v are really degrees C)

The results of the thermal simulation make it look too easy. I'd like
someone to check my numbers. I really hope it's right.

Looking forward to your comments  ;)