I've been perfecting this thing for this use, finding out that some scope probes are "consumable cheap" imports...and got all my test equipment agreeing with each other and reality. Which is a bigger deal than one might think when trying to tweak something up.
So, down to 4 tubes (easy, just unplugged a couple), same circuit - I could increase the screen volts a little from the current 35v and not get into quiescent draw, but...for now...
At 300 ma avg plate current (arbitrary limit set by a thermal circuit breaker as Bob recommends), I'm getting 1kv pp on the plates, and around 800v into a 500 ohm load when transformer coupled with a slight step-down. I could probably have kept the 6 tube arrangement if I used less inductance in the plate circuit. This setup has a fair amount of capacity plates<>ground.
For the RF choke/transformer I used a 4" piece of .5" diameter type 61 ferrite rod, Al = 66, 35 turns primary, 30 secondary, but on the outside of a 3/4" quartz pipe with the primary in the inside....because that's going to save me some grief driving something that also has ~ 20kv of the opposite polarity DC on it - that'd be a hard to get capacitor.
This is kind of a low Q output stage that works between 1.7 and 3 mhz at more or less full power with decent waveform in this kinda-strange class B/C mode. Tubes are "off" with no signal and around -22v bias, but barely, and it takes some drive to start the "on" part of the cycle - so I get class C like operating angle. The 2020 opamps drive the control grids around 5v positive, with no obvious rounding off due to loading. Could be worse. If I need more power, I'll have to come up with another screen supply. The current one is a diode/cap off the seriesed filament windings that have one side grounded. I could maybe double off some tap of that and get close - 10 more volts might be nice and should double the ~~ 145w rms I can get out of this without heating up anything but the dummy load.
I went for those numbers as without load the fusor pi network is around 470 ohms at resonance with no ions there. It'll go up in Z and down in F0 with more ions present, but that will also (I think) cut the rather enormous voltage stepup I get from the pi network as the C goes up if nothing else. It's around a 29 pf load now, averaging around 750k ohms with minimal ions in the tank. I'm using a 113uH pi coil and 1035pf to ground on the input side. I have a beautiful 2.5 mh choke with the best numbers I've ever seen to be the DC feed isolation, but it's getting near self resonance and I may have to get a smaller one. The one I have looks a bit like this (original part must have been their last one, and was a little nicer):
https://www.rfparts.com/rfc4.htmlLike perhaps:
https://www.rfparts.com/102-752.htmlPricey stuff. I wish there was a way to tell how good they were before buying...I need one in series with the DC bias for the main grid (~ 20kv) to avoid back feeding RF to the Spellman, I'd guess it wouldn't like nearly a kv of RF fed back into it's little ballast R. Which, as Cliff warned me, is potted and kinda hard to replace if you cook it.
So, about ready to carry this heavy deal back across the shop and try it...I think I'm beginning to understand why Hams use a separate box for the HV supply....split up the weight! This is all in out, and that's about half a cubic foot of iron/copper (the power supply is choke input!).
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.