So, kind of close to RF. Returning to attempt reuse of some hardware I built long ago to drive a Tesla coil, as I want a big, hard-to-fry, high bandwidth signal source to play some games with bunching, gradients, and recirculation in the fusor. Didn't seem to me like a new build in solid state stuff would be as robust, and would surely require more step-up turns ratio than this old thing (1kv or so plate supply). This uses a modification on an already odd grounded grid amateur radio linear amplifier. In this case, I'd ungrounded just the control grid, and grounded the cathodes of the tubes instead (the usual small anti-parasitic R's are there of course; I think of them as fuses at these speeds and feeds.).
The result has a rather amazing plate current vs grid bias curve, or at least, I've never seen anything quite like this before. In doing a little curve plotting, it was rock stable and never hinted at oscillating, even though the plates output lead does go under-chassis with the grid stuff. Bill and I didn't run this up to high power where there'd be big dissipation, we just wanted to find the lower end of the curve on this, so I could design a driver to go between an arbitrary waveform generator and this thing (which in turn will drive a homebrew wideband stepup transformer to the main grid, with separate DC bias for the grid via the secondary winding).
Having built this oh, 40 years or so ago..there was a bit of trepidation firing it up again, and it looked like I used good parts for the time, but was also in a hurry...at least I used a copper foil ground plane around the tube sockets. As luck would have it...it all flew. Looks like this under the chassis. There is an adjustable supply to the left to run the curves with, and various meters out of picture on the right to get the data.
And that data, again not to full power or even close, created this unusual curve - I didn't plot every point once I realized we got to straight line land.
Wow! I can drive this with modern fast opamps! I bought ADEL 2020's, one per tube as the input C is kind of high for one opamp to do them all. I don't think I'll need more than 20 MHz.
I can DC couple this all the way to the output if I want to and am careful. The supply is around 1kV at around 1 amp continuous, but will do a LOT more (Navy destroyer radio supply stuff) for awhile, and the tubes peak at around 1.1 amps - each - cathode current, and will easily take 10kv or so peaks on the plates.
This is why I resisted all the comments I should use some crappy old skool ham amp tubes. The're not even close to this kind of (well, ICAS) performance - opamps vs 100w driver, I'm in!
Here's what was in the box, more or less.
I don't guess I'll need the bias supply, but I will beef up the half-wave rectified +/- 28v supplies (from spare filament windings) to run the opamps, so making that power supply and figuring out how to fit that in are next. Also, for reasons that will be obvious to anyone with some experience here, I want to move the plate connections fully above-chassis and make a new bottom plate to shield the input side of things.
And people on the web are hitting on my "why isn't this done yesterday"...yeah, I know you guys (members) know this isn't something you do before breakfast after writing remote control code and sysadmin-ing replicated databases for the rest of this.
This one has the potential to make a lot of pretty expensive smoke. That's one thing when you can be there in the room with it and slap the off switch at the first sign of trouble, before the tubes get "bent" or transformers stink. It's quite another when you hope you notice something is wrong in time over a time-lagged video feed that is one of many at the operating position.
So, this stuff is slower to do than it might be otherwise.