It's about time for another installment... I have lots of pictures saved up so I'll spread 'em around finally. :3
This is the setup as of the last few months (over Christmas '06 and into January '07). On the left is my scope, the black box at the bottom is the bench supply (±15-18V at a few amperes), above it, the breadboard containing the power supply, oscillator and control circuits (see installment 6 for drawings, give or take some modification). Wires lead to the inverter, which is rigidly attached to the high voltage cap bank (two 8 x 470μF 200V packs in series for 1880μF 400V, center tapped), which is powered by a bridge rectifier (on heatsink) and MOT-come-isolation transformer. Above the inverter heatsink is the work coil, with kaowool insulation in place. To the right, the blue thing is the tank capacitor, barely visible behind the transformer, milk jug and radiator.
Rear view, a much better angle of the output and water system. A submersion pump in the milk jug sends water through the capacitors and tank coil. (Unfortunately the pump I have at the moment is crap for pressure and little flow gets through the 1/4" tubing of the coil. A positive displacement pump could be nice.) Return flow goes through a former automotive air conditioning core, equipped with a fan, and drains back into the resivoir.
Click to magnify. This is the special new manufacture part, a 30-80μH (depending on air gap), 80A capacity inductor, for Lmatch. I started by annealing a length of copper pipe, slitting it lengthwise like a sardine can, hammering it flat and cutting out a relatively straight strip 1/2" wide, 0.040" thick (give or take a few thousandths) and about 7 feet long. This I wrapped around a cardboard form, insulating as I went with a strip of 0.010" thick paper. Since 1/2" wide strip is rather hard to bend sideways, I opted to wind two pancake style coils and connect them in series for 12 turns. The core comes from four identical flyback transformers, glued together. Also in this picture, the BNC connector and toroid (covered in masking tape ;) are my 1:100 current transformer, registering 0.01V/A (i.e., 10 mili trans-ohms, so to speak). Well, it's actually 220:1, but it has a 2.2 ohm resistor (and RC snubber) so it looks 100:1 on screen.
Closeup of the inverter. (This picture was taken on the carpet, before I moved down to the Bench.) The power supply rails come in, loaded with film capacitors (2.2 and 0.47uF polyester, probably inductive wind). 12AWG wire turns to copper strip which the IGBTs are connected to. Some noninductive capacitors (stolen from the tank cap, so it's 19.8uF, BFD :-p) keep the rails locally under control. Overall, the trash from commutation is quite controlled, peak-to-peak about 5% of the supply voltage at 10 or 20A. Also notably, the gate drives are soldered and local to the inverter heatsink, not all the way over on the control board (also freeing up some real estate there). The same coupling capacitor, Lmatch and parallel resonant tank follows from here. I recently had the inverter up to about 80A peak with no ill results. Don't know if it'll take that constantly though.
These are the tank and inverter current waveforms when heating a steel crucible. Voltage is pretty low, reflecting the steel's hysteresis loss load on the tank -- the steel is still magnetic.
An odd load this time, potassium chloride tabs. A not unreasonable assignment, as it melts at 1422°F, right about the curie temperature of iron. Coincidentially, I found it begin to melt right about the time the voltage and frequency started to rise.
You can't see it well but this is completely transparent, water clear, and as mobile as alcohol. And I'd wager that if alcohol fumed in air, it would behave exactly like this, sans orange glow. Chloride salts have low vapor pressures when molten, so all the while there was a light whispiness coming off here, and I am probably more radioactive for it. (That is to say, slightly richer in potassium, which is naturally slightly radioactive. . .)
One final treat, a video: