Isolating oscillators via emitter followers

Hello, I've made a post before about building my own analog theremin, but this question is largely irrelevant, so I'm making a new thread.

I'm suspecting the two oscillators of the pitch generator influence each other too much. So I got the idea to isolate them from the mixer stage with emitter followers. Has anyone even tried this before? I haven't seen this approach used in any other schematics.

Anyways, most emitter follower designs I find are meant for higher power audio applications, such as driving speakers. I designed my amplifier more via trial and error than via calculations, but I know the basic theoretical principles of the amplifier and transistors generally. What I don't understand is the role quiescent current plays in the amplifier. But I found that my 250 kHz sine wave gets distorted with too little quiescent current. So I have to use a somewhat small emitter resistor, which results in the need of small base resistors. Which then means the signal from the oscillator has a low resistance path to ground, which again doesn't work well with the oscillators.

Is it possible that I won't be able to get good enough behaviour with an emitter follower at all? Should I try to use a FET transistor instead? Or should I rather abandon the entire endeavor?

With analog design absolutely couple your signals after using two emitter followers. The oscillators must not know the other one exists.  Have distance between them and shield the fixed oscillator from the pitch antenna as that could also cause unwanted waveshape distortion.

Use coupling distortion to create your waveshape for a wonderful sound. The authentic voice of the theremin is being lost to digital imitation and simpler designs..

I am saving my special knowledge so my MIT friends can reintroduce the authentic theremin phenomenon and its many natural voices. Even Moog Music came up short, genius as he was.

A beautiful theremin design is not an easy journey, I wish you the best.

Christophe S Becker
Old Temecula

"I'm suspecting the two oscillators of the pitch generator influence each other too much. So I got the idea to isolate them from the mixer stage with emitter followers. Has anyone even tried this before? I haven't seen this approach used in any other schematics."  - Shweik

Yes, the ESPE01 is a FET follower module that has been installed in many Moog Etherwaves.  And there's my DIY bipolar version of it, the YAEWSBM.  I suspect very similar circuitry was built-in to the ill-fated updated Etherwave, which had a trimmer you could adjust to vary the coupling.  There are other Theremin designs that incorporate a variable coupling control.

And there is the Aetherwellen mod, which uses a simple RC filter to accomplish largely the same effect.  For Etherwaves I would recommend this approach as it is super simple and much less invasive.

Why do some Theremins have a variable coupling control?  Because the coupling can serve two purposes: 
1. Expand the "null zone" around 0Hz, which can help to shut the instrument up when no one is near it.  
2. Provide a "raspy" low end timbre that can be musically useful.  

If you completely obliterate coupling then you gain a more linear and controllable low end, but you also lose the advantages.  It's a give and take, zero sum kinda situation where the hope is to find some sort of sweet spot in the strength of the coupling / decoupling.  Fully decoupled Theremins are easier to play on the low end, but they sound extra bland to me because there is very little timbre variation going on.

I thank you very much for the input. This is another piece to the theremin mosaic which I was completely missing until now. I feel like every time I think I have solved all problems, I just find another one very soon. But it has been a great learning experience so far.

I did some research and it seems JFETs will be somewhat more suitable for the task, but MOSFETs will be easier to work with. Is there a notable difference between the two from your experience?

"I did some research and it seems JFETs will be somewhat more suitable for the task, but MOSFETs will be easier to work with. Is there a notable difference between the two from your experience?"  - Shweik

I'd just go with bipolar.  They're easier to obtain, are more predictable, have higher gain, and the base presents a sufficiently high impedance load.  To really cover all the bases you might investigate the variability of parasitic capacitive loading, as it's buffering the LC tank.  For maximum oscillator isolation you could use separate 3-terminal regulators to eliminate coupling via the supply.

The Aetherwellen mod is quite fine though: a ton simpler, adjustable, and easily removable.  

"I'd just go with bipolar.  They're easier to obtain, are more predictable, have higher gain, and the base presents a sufficiently high impedance load.  To really cover all the bases you might investigate the variability of parasitic capacitive loading, as it's buffering the LC tank.  For maximum oscillator isolation you could use separate 3-terminal regulators to eliminate coupling via the supply.The Aetherwellen mod is quite fine though: a ton simpler, adjustable, and easily removable.

As I've pointed out in my original comment, I did try to build BJT emitter followers, but I cannot get the input impedance above 100kOhm, which doesn't seem to be sufficient. I might do some further testing though.

...but I cannot get the input impedance above 100kOhm, which doesn't seem to be sufficient. I might do some further testing though.

Why not taking for example a 470 kOhm resistor between collector to basis and a 2.2 kOhm from emitter to ground. The impedance is round beta*2.2 kOhm parallel 470 kOhm. Use a small coupling capacitor of 15 pF to the basis of the BJT.   

Why not taking for example a 470 kOhm resistor between collector to basis and a 2.2 kOhm from emitter to ground. The impedance is round beta*2.2 kOhm parallel 470 kOhm. Use a small coupling capacitor of 15 pF to the basis of the BJT.

I just tried this and the base voltage sits at around 3.3 volts. I use 12 Vcc, so I wanted the base voltage to be about 6 volts in order to reach maximum amplitude without distortion. This way I can get 4 Vpp, which isn't terrible I suppose. I'd have to go even lower with the resistance if I wanted higher amplitude, but I don't think I can afford that.

Yes, it is depending on the beta-value. With my interpretation of your data I calculate your transistor has a beta of 66, what is a bit low concerning the goal of high input impedance. With beta=240 the 470 kOhm resistor would match the 6V condition perfectly. 

To get 6 Volt at emitter, the current must be 6V/2.2k=2.7 mA. Basis current then is 2.7 mA/beta. The basis resistor must see 5.4 Volt, for this resistor resulting in R=5.4V/(2.7mA/beta). With beta=70 you would get R=130 kOhm. This is too low, the whole input impedance were only 70 kOhm. With beta=240 R increases to 480 kOhm. The whole impedance results in 480 kOhm parallel to beta*2.2 kOhm = 250 kOhm. 

For simplicity and better stability you can use a 1 Megaohm or 500 kOhm adjustable resistor between Vcc and ground. The center tap to basis. So you can adjust exactly the halve of Vcc at the emitter resistor. In addition use better a transistor with beta between 200 and 300. 

Hope the estimation is reasonable an I made no mistakes... 

Oh that's a good point, I didn't think about the beta value too much before. 250kOhm output impedance seems a bit more reasonable.

I tried setting the base voltage with a voltage divider before. But I figured out the Vcc-base resistor has to be at most 100 kOhm. The base-ground resistor can then be around 500 kOhm. But the series impedance of the 15 pF coupling capacitor then forms a voltage divider for the input, so only about half of the input amplitude makes it to the base.

The original schematic from which I took the oscillator design has about 400 kOhm output impedance. But it has no subsequent amplifiers, the oscillators are directly coupled through the resistors. So I'm already deviating quite a bit from the original design.

I'll try to look for higher beta BJTs. But it seems to me that FETs would still be the safest option, since the input impedance can be an order of magnitude higher. Dewster did mention they are used in some professional theremins afterall.

If you don't mind, I have a slightly off-topic question. The oscillator design is from this site
https://eleneasy.com/2019/08/20/theremin-v-1-pitch-reference-oscillator-analysis/
Is it possible that low beta has a negative effect on the oscillator as well? I used the same transistor for it.