Vacuum tube synth oscillator.

[ELS122]

I'm quite new to synthesizers, and oscillator circuits, but I have worked on tube audio for years.
I was wondering if there's a descent VCO using vacuum tubes that wouldn't drift a ton, I found a circuit using a neon lamp oscillator and changing the bias of the tube to adjust the frequency, but I think that would be unbearably drifty.
I'm not afraid of using sand, but I just want to make the audio chain mostly out of tubes for fun.
maybe having some kind of phase shift oscillator with a BJT changing the phase shift of the RC chain?
what would you suggest?

also, I don't quite get how synthesizers change the wave shape? do they change the operation of the oscillator itself? swap over to different oscillators entirely? or have some filter circuitry that just uses the frequency from the oscillator to create a desired waveform?

[BeBuLamar]

I really don't know but I think (of course I can be very wrong)
1. Analog synth has osillator the generate sine wave of different frequencies then either combine them to build the harmonic of filter them to change the wave shape.
2. Digital synth I think they just generate the wave form mathemaically then feed to digital to analog converter for the final stage.

[RJSV]

   You should do OK, with various concepts, such as RC filter are similar (to solid state).  I would start with a bit of study, of a product called:
   'The Analog Thing'

   That is a flat, writing tablet size, providing various sub-units, like X2 ac voltage amp, low pass filter, and a few separate VCOs, that is 'Voltage Controlled Oscillator'.
Plus the play surface has really easy terminals for doing your desired project, and also has a couple potentiometers, etc.  (The potentiometers maybe much higher resistances in TUBE circuits).
   I owned and 'took apart' my Fender Super-Reverb a couple times, made Leslie (tm) speaker copies (research only, no sales).

[TimFox]

A simple vacuum-tube circuit for an audio VFO would be based on an astable multivibrator, producing a square wave.
This page  https://www.schmitzbits.de/vtvco.html  has an op-amp circuit at the left to generate an exponential control voltage to match modern synthesizers, but is a tube circuit that varies the charging current into a capacitor, much as in a 555 IC.
This seems to be a better link, from a well-regarded vacuum-tube author:  http://www.synthpanel.com/tube/vco.html

[floobydust]

For low drift, I recall they used two RF oscillators around 500kHz and the demodulated beat frequency between them was the actual audio tone.
Because both oscillators drift the same amount, that gets cancelled out. Same principle as Theremins. I'll check what other tube synths used.
I found musicians dislike having to constantly re-tune every few minutes, that was the deal breaker with tube VCO's. Hammond went with tone wheels to stop that. Interestingly the tone wheel motor is synchronous, so mains frequency being low could not be fixed - my friend said Woodstock recordings with Hammond organs are flat due to generators powering everything, being off frequency (low).

For the time and hassle, I think a keyboard using a solid-state VCO is the way to go. Waveshaping can be done with tubes downstream if you wish.
The actual tone is filtered and VCF etc. so the oscillator's actual character does get muddled and harder to notice/hear. I've seen many synth designers waste all their time and energy trying to get a low drift oscillator - instead of working on the other circuits.

Sound Semiconductor is remaking vintage synth ICs that were originally by SSM.

[ELS122]

For low drift, I recall they used two RF oscillators around 500kHz and the demodulated beat frequency between them was the actual audio tone.
Because both oscillators drift the same amount, that gets cancelled out. Same principle as Theremins. I'll check what other tube synths used.
I found musicians dislike having to constantly re-tune every few minutes, that was the deal breaker with tube VCO's. Hammond went with tone wheels to stop that. Interestingly the tone wheel motor is synchronous, so mains frequency being low could not be fixed - my friend said Woodstock recordings with Hammond organs are flat due to generators powering everything, being off frequency (low).

For the time and hassle, I think a keyboard using a solid-state VCO is the way to go. Waveshaping can be done with tubes downstream if you wish.
The actual tone is filtered and VCF etc. so the oscillator's actual character does get muddled and harder to notice/hear. I've seen many synth designers waste all their time and energy trying to get a low drift oscillator - instead of working on the other circuits.

Sound Semiconductor is remaking vintage synth ICs that were originally by SSM.

well at that point I'd just rather use a neon bulb instead of the thyratron, probably some clever circuitry can be used to make it into a square wave, and too into a triangle ofc

[RJSV]

That was an interesting bit of history, that Woodstock Hammond Organ recordings were (sounding) flat, as per 'low speed' generators!
   Please do keep aware, usual Tube circuits have lethal voltages, (of course you know, I'm just saying).
I think, sincerely, that kind of vintage is beautiful, if that's the venue.  Analog isn't dead, just another avenue, to Technological success.
There has to be some profound advantage, like input hi-impedence, that got left in the dust ( under the general stampede to digital advantages.
   You could even, get silly, and claim;
   'Silicon is, maybe, going to RUN OUT in coming years ...while a VACUME is not in short supply, being nothing.

[Wallace Gasiewicz]

Mt uncle, a Navy Catholic Chaplain and Captain, was responsible for buying all the organs for the Navy Chapels. He bought Hammond because they were the most stable in all environments.
Apparently the Navy had good 120 HZ
I think some of the organs used in rock and roll were Navy Surplus.

Some of the old freq generators like the RCA ones, were not too bad in drifting. I think rebuilding the oscillator with new, better parts helps with the drift. Especially the old resistors, which are very prone to drift with changing temp. You can pick one up for maybe $5 and see if you can stabilize it.

[T3sl4co1l]

I mean, you can make whatever you like in semiconductors -- including multipliers, with difficulty*; it's just harder, I mean in general, because you don't have P-types.  So you need more level shifting shenanigans, on top of buffering and just generally needing more gain because triodes suck, pentodes aren't much better, and all of them are bulky, hot and high impedance.  So you tend to need more of them to do the same things, but can't spare even as many, so have to make do with heavily compromised circuits -- certainly compared to what's possible with ICs.

*Because BJTs have an essentially ideal exp(Vbe) characteristic, while anything else has less.  Tubes do still have an exponential cutoff, but it's so far down that it's hard to use (close to leakage current).

So, if you want to do a multivibrator -- anything will do, and you do have the unique option of doing a phantastron with pentode or other.  Neon lights, the threshold isn't controllable and varies with age and illumination, nasty things.  Thyratrons might do better, but you might also need a comparator of some sort to set the turn-on threshold (rather than relying on plate effect; which is stronger with triode than tetrode types I think, so, maybe those are better without a comparator to trigger it).  Diff pair is fine, but you do need another CCS to drive the capacitor, and probably a buffer or level shift or something to make it go up and down.  You do get triangle to saw waves from any of these with controlled edges -- so, not neon/thy (sawtooth only), but the others sure.

Frequency mixing is probably not a good way to go, as VCOs, even over narrow ranges, still aren't great; but maybe it can be done well, not sure.  Easier as far as wider range and easier control, with a varactor, if you don't mind that -- arguably semiconductor junctions were contemporary say in the 50s-60s so this might not be bad.  The other way is any variety of saturable reactor, a cored inductor with a perpendicular bias winding influencing it; which is practical all the way down to audio frequencies if you like (which would be good enough for an octave or two range, not going to do everything but alright for blipping out a melody track, say).

LC oscillators are nice and stable, or can be when tuned alright.  Not much range though.  So as awful as it sounds, it's not actually the worst possible option to just have one oscillator per note -- some organs did it this way as well.

Tim

[TimFox]

One possibility for voltage control is to use a reactance modulator (pentode, probably) to tune one oscillator and beat it against another oscillator (Theremin-like).
I am playing with a 6AS6 reactance modulator connected to a 6AS6 transitron (negative resistance LC) oscillator, but it's on the back burner here.

[ELS122]

well I thought of something like this, this is just an idea ofc, not a finished module.



I'm still not a fan of the tube oscillator in this configuration, it just seems that it would be really drifty, but I'll build it and see for myself. also the non-linearity of the tube would probably cause problems with low notes especially, but I need to balance the tube plate load resistor with whatever bias current I choose so it doesnt oscillate with "0" volts on the grid. (btw that resistor I put there so the plate wouldn't be super high impedance and wouldn't be too noisy)

the cathodyne phase inverter (using my favorite component values  and the tube diode make a sawtooth to triangle wave shaper, then I get the inverted signal from the plate and connect it to a switch between that and the plate of the oscillator plate, so I can either add or subtract the signal from the triangle wave output (or change the phase before mixing it)

then there's some cursed filthy soul-less sand-composed comparators for the square wave shaper and then a buffer for the other signal (plus a phase reversing switch).
so this should allow for a good amount of funky waves, for just 2 tubes... and an op amp

wouldn't pentodes as you suggested be even more drifty tho? but I'll look into it anyway, on the surface it sounds a bit better than a neon oscillator 
you mean like what was used for FM AFC tuners at the end of the tube radio era? I guess those never had issues with it drifting so much that AFC never engaged

[ELS122]

I was also thinking of using something like the magnatone amplifier vibrato circuit for a janky FM-modulated kind of sound, it wouldn't be an actual FM modulator but it may work good enough for an "FM Synth"

I just realized I could just add precision bridge rectifiers on the triangle wave output to double the frequency, so I won't need more oscillators (at least for octaves)

[TimFox]

Back in the day, the less-drifty tube oscillators relied on L-C resonance and were relatively independent of tube parameters (except for tube capacitance that changed with warmup).
General Radio liked to use beat-frequency units, with two higher frequency oscillators, especially since that allowed a full audio frequency range with one dial.
Reactance modulators were used in afc circuits, as well as FM generators, to affect the L-C resonant frequency from a voltage control.
Most of this can be found in  https://worldradiohistory.com/BOOKSHELF-ARH/Handbooks/Radiotron-Designer%27s-Handbook-4th-Edition.pdf  (a huge download, you might want to look for a used hardcopy).

[ELS122]

Back in the day, the less-drifty tube oscillators relied on L-C resonance and were relatively independent of tube parameters (except for tube capacitance that changed with warmup).
General Radio liked to use beat-frequency units, with two higher frequency oscillators, especially since that allowed a full audio frequency range with one dial.
Reactance modulators were used in afc circuits, as well as FM generators, to affect the L-C resonant frequency from a voltage control.
Most of this can be found in  https://worldradiohistory.com/BOOKSHELF-ARH/Handbooks/Radiotron-Designer%27s-Handbook-4th-Edition.pdf  (a huge download, you might want to look for a used hardcopy).

I already have thay book! 

[TimFox]

Another classic book, well worth owning in hardcopy, in another huge pdf:  Vacuum Tube Oscillators by W A Edson (Wiley, 1953)
http://www.tubebooks.org/Books/vto.pdf

[floobydust]

Hammond Organ Novachord (debut 1939–1942) first polyphonic synth, 72-note 163 tubes, 400W 500lbs.
I thought it uses dual RF oscillators to cancel out drift, one tuned inductor per key, and tube frequency dividers. Not possible to make a lone audio RC or LC oscillator with low enough temperature/B+ drift, to work. Have to do the math on cents tuning verses ppm.
http://www.novachord.co.uk/
https://www.discretesynthesizers.com/nova/intro.htm