Ping Tim Williams

Sure. If you use a Colpitts, you can use just any RFC that meets the freq and Q specs needed.

You'll be shooting for relatively high impedances, on the order of 10kohms. So for a Q of 10, a load of 10k ohms and reactances in the 1k range (parallel equivalents, or ESRs well under 100 ohms). And whatever capacitances and inductances that corresponds to.

You'll probably not be hand-widnding inductors because it's a pain to put down that many turns (100s). So also, anywhere you can deal with off-she-shelf inductors will save trouble.

Remember that reactances don't absorb energy, they only store and pass it on. There is resistance everywhere, and that determines the system impedance, frequency response and all that. Even if there is no explicit resistor in the circuit, losses (ESR, plate resistance, grid admittance?) will do the job for you.

Ideally you'd use a heptode for the mixer, but if you want to do it entirely with subminis, I'd suggest using a diffamp configuration -- you'll get better isolation between input ports, which means less need for buffering (amplifiers, voltage dividers, cathode followers*), which is important when your two oscillators are very close together in frequency and can't see any coupling.

*CFs aren't all that great for isolation, comparatively speaking. You've got a modest output impedance (1/gm, maybe 200 ohms for this class of device), and a direct capacitive path for reverse transfer (Cg1k, ~4pF for this class of device).

Whereas a common cathode amplifier has Cg1p for reverse transfer (fractional pF) -- because it's so small, it's better, despite the higher gain too. (That is to say, if forward gain isn't as high as 1/(reverse gain), you can get oscillation very easily (tuned grid oscillator, etc.), which is one of the problems with triode or BJT amplifiers, and why you need need neutralization on them.)

If you get "dual input" type pentodes, you can do that too (Cg3p doesn't matter, since you'll be shunting that RF with a C-input filter), but the g3 transconductance isn't so great.

But if you're getting a bunch of tubes, I'd suggest avoiding dual control types -- they aren't so good for general-purpose use, because g3 is intentionally finer, which makes the whole thing work worse (6HZ6 requires as much heater power as a 6V6, but only delivers as much as a 6AU6).

There are plenty of circuits out there using a handful of tubes, but the beat frequency probably stinks because of coupling (hey, if you don't need a clean tone below 1kHz, that's fine!) and may not even have volume control.

The volume control I used in the solid state version is simply a slope detector. Nothing fancy. A ratio detector would be fine too, and might be feasible at RF without conversion. You can use conversion to make the frequency range more detectable (say, dropping it to 20kHz so you only need a few kHz bandpass to do slope detection). Do mind the Q factor, which gets poor at low frequencies, so at some point, you have diminishing returns on sensitivity.

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

Thanks so much for the reply - it will take a little while to digest. I was thinking about doing something fun like giving this theremin a digital audio output for direct recording into a computer, or something. That could be cool.

Do you think those antennas you see on WiFi routers would work OK as pitch and volume controls? They're cheap, but maybe they're too short.

I think a "cooler" approach would be to use the "controls" to drive a digital synthesis -- perhaps adding a third channel (controlled by body proximity? height of left foot off floor? extent to which MOUTH is open??) to modify waveshapes/ADSR profiles...

Forget the tubes. You will get much better electronics with FETs.

--

-TV