Capacitor bank pretty well complete

impressive. And what is it designed to deliver ? How many Farad at how many volt ?

Rene

Thanks. Oh yes, the specs! :-[

20uF total. 200 x 495-1333-ND from Digikey: (15mm, 0.1uF, 630VDC/250VAC rated.)

I've extrapolated the graph here:

If that is indeed RMS amps, then the bank is rated for 460A RMS (twice what my welder is rated for!). At 250VAC max, I can get a maximum 115kVA at 15kHz. At 445VAC max (sinewave peak = DC rating) I can get a whopping 205kVA!

Speaking which, does anyone have any comment as to which limit I should obey, DC peak or AC rated volts?

Anyway, not bad for $80 of parts and a few man-hours of thinking, bending and soldering.

Tim

-- Deep Fryer: a very philosophical monk. Website:

Yeah, probably five, maybe more, but the kind of hours where you spend one or two each day and work goes real slowly, you just take your time.

Nah, nothing so crude. ;) Although... The capacitors are rated for 400V/us peak, so I can safely, repeatedly pull

8kA peak out of the bank.

There's probably 0.1uH or more there in the wiring, which constitutes a resonant frequency of 110kHz. For a lossless connection and 630VDC initial, that would constitute 445Vrms across the inductance for a current of 6300A, within ratings.

Or better yet, short the terminals through a coil. 5kA through a few turns will make some generous magnetic field, maybe I can shoot an aluminum plate into the air a few feet (not much since it's only four joules stored).

But anyway, enough of the fantasy, I'll tell you how it works.

Know anything at all about radio? Tuned circuits and whatnot? (You should anyway- when you're switching 200A on and off, a lot of pieces of wire suddenly become inductors. Bypass capacitors can ring with these bits and potentially make things worse!)

The basic idea is, to heat metal with an AC magnetic field, you need a pretty strong field. A good way to generate this is an iron-cored transformer, but that needs big scales to work on, so ain't gonna cut it here. So we use a coreless inductor.

When you run amps through a coil, it makes a magnetic field, which induces voltage in the work, which has resistance, P = V^2/R and you get heating. In the coil, it takes a lot of amps to build that field, though. If you figure 1" insulation around a 4" dia slug of iron, that's a 6" coil. The cross-sectional area is 16 to 36, or 44% of the primary is coupled to the secondary (coupling factor 0.44, roughly speaking). It'll be a bit lower in practice due to stray field and whatnot. So if we run 100kVA (such as, say,

316V * 316A) through the primary, 44kVA is coupled to the work, which reflects back some of this as back EMF (silver, copper, gold and aluminum being the worst load metals for this reason), and dissipates perhaps 20kW of that. Overall, you have a 5:1 ratio between volt-amps in the primary and actual power consumed.

If you run transistors (or SCRs, or tubes) into that primary alone, you have to handle all 100kVA as if it were real power. If your inverter itself is

90% efficient, you'll be burning (100k / 0.9) - 100k = 11kW in the process! That's 66% overall (31kW input, 20kW output, 11kW dissipated) for what should be a 90% efficient system. There's a better way.

If you cancel the inductive component of the primary, you are left with real current that does real work. A capacitor is the opposite of an inductor. You can connect them in parallel or series to accomplish this. In series, currents must be equal; in parallel, voltages must be equal. In either case, the reactive components cancel at resonance (you may remember an inductor's reactance rises as frequency goes up, while a capacitor's falls; when equal, they resonate).

That leaves the problem of driving it. With 20uF anchoring the tank circuit, you aren't going to make the voltage change much faster than the sinewave on it. It's perfectly okay to drive with a sine wave, but it needs either a much faster class D amplifier, or an inefficient linear amplifier. Class D amps just aren't cheap or easy to make in the 200+kHz, 10kW range, and aren't very efficient either. (Class A through C means linear amplification with a varying amount of conduction time; class D is with the transistors always either ON (low voltage, high current) or OFF (low current, high voltage). Since voltage or current is low at any given time, power loss is low, except during switching.)

Inductors are class D's friend, so what I did is connect an inductor ("Lmatch") in series with the tank. The inductor provides "springiness" so I can swing 300V in a microsecond without shattering my transistors, allowing me to get from one side to the other with a minimum of power loss. The only problem is it causes a series-resonant peak, where current can rise to a hundred amperes or more (and it's worse when the coil is lightly loaded: Q multiplication effect).

So in summary to your question, it cancels the reactance of the work coil, allowing me to drive it with 10-20kVA instead of a full 100-200kVA.

Tim

-- Deep Fryer: a very philosophical monk. Website:

Looks extremely nicely done. No cut corners. Congratulations. Do you care to share the parameters of this bank (voltage, capacitance, etc).

i

Fuck yeah ;-)

Hopefully a sheet steel case around the coil and whatnot will keep submarines from tracking the signal into my back yard...

(I'm in the ELF area, that subs use for comm.)

Tim

-- Deep Fryer: a very philosophical monk. Website:

It looks very nice, but was it really only a few hours?

So, forgive me for an ignorant question, but what are you planning to do with your caps? Charge them with line voltage and discharge into your heating coil? Or what? I do not have a mental picture of what you will do, and am very ignorant about electronics.

i i

Beautiful! Ed

Very nice explanation. I now understand why you needs the caps. Sounds like a very interesting project actually. Will it generate any interference?

i