Replacing the oscillators in the Etherwave Theremin?

"The most important question I have about the design right now, is how much trimming needs to be done on a regular basis?" - [i]CharmQuark[/i]

If you mean "Pitch" trimming (a knob on the front panel) the answer is "everytime". Everytime you place your theremin in new location, environment, for new player and even for specific piece of music.

"and replacing the obsolete AN5265 with another amp." - [i][i]CharmQuark[/i][/i]

In last design I use TDA7056A (it has 1.4V control voltage versus 12V of AN5265).
There is more tiny package -- TDA7052A (phone amp).

I feel I'm unnecessary roundabout regarding my questions. So, to put it simply: what I want to do is to move the trimming to a pot. I'm perfectly fine with having fixed inductors and fixed capactors in the design, but potentiometers are readily available, comes in many flavors and is comparable easy to replace if one cannot be sourced. It there is no front-panel variable caps or inductors that are easy to source in time and space, the pot is a more obvious answer than rolling your own I think. Yes, it is fun to build own components, but I want reproducibility and to be able to put a part number in the bill of materials that everyone can understand. Yes, I understand the benefit of having an L in the tank and I want to have a "COTS" inductor, but there must be some LC oscillator out there where the trimming is done by varying R (a "LCR oscillator"? ). (I suppose that another theoretical way to go is an conventional RC-oscillator driven by a ridiculous high voltage to get a good antenna swing.)

Thank you for the tip about TDA7056/7052A. They were non-stocked at mouser, but I should not be very hard to find a suitable amp.

"So, to put it simply: what I want to do is to move the trimming to a pot."  - CharmQuark

You and everyone else!   

Moog moved the "fine trimming" to a pot, and you can sometimes do this in other ways (e.g. by playing with the oscillator bias).  But you also need a way to tune the thing in a bulk manner, particularly if you have a linearizing or EQ series coil going to the antenna.  FredM was playing around with electrical tuning by saturating the EQ coil core via a DC current.  Varactor diodes don't provide a ton of range and may require higher voltages.  Variable caps can be difficult to source, variable inductors harder.

If I were designing and building an analog Theremin I'd probably try to manually pad the oscillators with fixed caps until the tuning was close enough, then have an internal coarse trim capacitor with panel fine trim capacitor (or electrical equivalent ala Moog).  Electrical trimming (ala Moog) might introduce temperature drift, this is something I would investigate up-front before going too far down that road.  Or I would bite the bullet and make something like one of ILYA's designs, or the Moog Melodia, where it's all variable caps / coils brought out to the front panel.  Roger (pitts8rh) would be a good resource for advice, though you can learn a lot just looking at his pictures: [LINK].

I'm not saying you're beating a dead horse, but many others who came before you were up against this same wall and knew it.  And there have been some pretty sharp people attacking this problem.  I don't want to discourage you, but I don't want you to get discouraged by this problem either.  If you don't have a working Theremin in your hands at the moment, I encourage you to get one, just to get a better feel for the whole thing.  I owned an EWS and a Theremini before I built the D-Lev, and the experience was invaluable to the design process.  (The EWS taught me how difficult it is to tune an EQ type Theremin and the limits of linearity for that approach; and the Theremini demonstrated the value of high gestural bandwidth and high antenna voltage swing - because it has neither!)

I don't mean to annoyingly plug my digital approach so much, but this issue was absolutely dead central to why I went with digital in the first place: if you do it right you can plug just about any decent (i.e. adequate Q at resonance) fixed coil, over a wide range of values, into the analog front end (AFE) and the thing just works with no tweaking.  It gives you the highest possible antenna swing by maintaining quadrature between drive and antenna, and it is simple (both in terms of math functions, and in terms of required user input) to linearize the field mathematically - and make it much more linear than an analog Theremin.  An analog Theremin oscillator could likely be designed as an analog PLL, but you would still need the heterodyning stuff, hum filtering, etc. which is more straightforward to do digitally, as is the linearization, so why leave the digital realm at all?  I suppose you could use the VCO voltage directly to drive an audio VCO, but accommodating various environmental C factors would require something variable analog somewhere.  It's a tough nut with no obvious dead-easy solutions, which is why I did the bare minimum in analog and ran screaming into digital space (where no one can hear you scream).

I've tried to play with pot + varistor based trimming in LTSpice, but didn't get acceptable results.
BTW, Open.Theremin uses very strange trimming. It has DAC connected via SPI to drive varistor

In order to not compromise the linearization coil resonance, I'd only provide means for internal tuning of the variable pitch oscillator. This would be done once after production and later only in service intervals when needed. 
External tuning control is simpler to do on the fixed oscillator since there is no impact on secondary resonances from the linearization circuit. An internal coarse tuning means (be it variable cap or ind) must also be provided for servicing (re-centering the external pitch knob).

First: Thank you all for putting up with me and give me such good advice!

@dewster: Yes, exactly. I know I'm not the first. And not the last, who want to trim via pot rather than inductor/capacitor. And at least some oscillators should be straight forward to replace. Look at the fixed pitch oscillator in the "Paradox" design. It uses a 470µ trimmer inductor. The signal is fed into a clock input of a counter, so the waveform isn't critical at all (save for jitter, if any) so I really don't understand why that one should be trimmed by expensive and hard to obtain parts. BOM optimization cannot be the reason.

And, if I understand the theory, the beat frequency is the shit, and the IF/RF from the oscillators is only means to get a beat frequency. So, one of the oscillators shouldn't need any trimming at all, provided that the trimmed oscillator had sufficient trim range. And in this case, where an oscillator very well may produce square wave, shouldn't is suffice with a fixed pitch oscillator (maybe simple trimmer cap on the PCB instead of a front panel one) and then do the trim with the fixed pitch oscillator, which can be replaced (the context here is the Paradox design) with just about anything, wine bridge, 555, neon bulb relaxation osc (ok, ok, the last one wasn't serious!) or whatever oscillator that can provide the frequency and may have a panel mounted potentiometer? And no, I don't mind the digital approach. Except I want to build from a proven design with proven sound. "Paradox" is my best bet I think, I like the design (and it do have a crucial digital step).

@Buggins: Me too. I'm playing in LTSpice under wine with different LC-oscillators and try to make them trimable by pots. No cigar yet. Trimming is possible, but with horrendous bad range or linearity.

@Thierry: Yes, I'm following you and this is precisely what I think is reasonable. Problem is, I'm not really confident when it comes to theremins. I mean, I'm just a run-of-the-mill engineer with focus on digital stuff and all my analog experience comes from measurement and calibration industry. Here is the most stupid thing: I've never been into audio. At all. Never built an amplifier for audio. So, if you guys have a hard time replacing those trimming inductors and -caps, there is something I don't understand.

It feels like I missing something. I *hope* theremin builders put variable caps and coils into your constructions because of the handiwork as a hobby, not out of necessity.

I'm following different paths here, and one is to come up with some design of variable caps. But I don't like it. Don't really have the tooling or time. But for the most part, I of course want to understand the concepts of theremin design.

What is your spontaneous reaction to my silly suggestion that one maybe can use an ordinary RC-oscillator for the var pitch osc, and having a high voltage to compensate for doing away with the inductor? Don't mind the problem with generation of higher voltages. A 70 V full-swing RC-osc, how about that? Doable? Idiotic?

(Totally off topic: Isn't it strange that Firefox's spellchecker knows "theremin", but not "inductor"? )

"And in this case, where an oscillator very well may produce square wave, shouldn't is suffice with a fixed pitch oscillator (maybe simple trimmer cap on the PCB instead of a front panel one) and then do the trim with the fixed pitch oscillator, which can be replaced (the context here is the Paradox design) with just about anything, wine bridge, 555, neon bulb relaxation osc (ok, ok, the last one wasn't serious!) or whatever oscillator that can provide the frequency and may have a panel mounted potentiometer?" --  CharmQuark

CharmQuark,
a key word is "stability".

Things to consider:

1. A Theremin that uses series EQ inductors actually has three LC resonances: one for the fixed oscillator, one for the variable oscillator, and one for the EQ inductor and antenna C.  The latter two interact, and they all have to "line up" in a certain way for you to experience good linearity and your desired playing note range, hence all of the variable L's running around inside of these things.  Tuning a Theremin such as the EW that uses series EQ coils isn't a trivial process (perhaps harder than doing IF alignment without tools).  In many ways you're tuning to the EQ inductor, as it tends to be large and fixed in value.

2. RC oscillators generally aren't stable enough for heterodyning use in a Theremin.  LC is stable enough but doesn't give a wide range with hand capacitance changes, hence the need for heterodyning.

3. Predicating your choice of oscillator topology on "sound samples" is overly restricting your search space.  If you can get a ~sine wave out of the mixer you can manipulate that many ways to get all kinds of sounds.

@dewster: (Starting from below). No, I think you misunderstand. I'm doing the opposite, I found a general concept I like and try to build it verbatim, but feel forced to tinkering in order to be able to source parts. In the "Paradox" design the oscillators topology has nothing to do with the sound, they could very well be producing square waves. The mixing is, so to speak, already accounted for, by ILYAs eminent design. The only way the waveform is important there, is in the interaction with the antenna. Waveform generation is disconnected from sound mixing by a digital counting stage (again, save for jitter and side channel noise, e.g. loading of power rails).

I would say that there is many stable oscillators without inductors. From the humble wien bridge to XTALs with DDS or PLL, ovenization, temperature compensation, using parallel topologies in both oscillators and such. My talk about neon lamps and 555s were not entirely serious and more connected to the discussion of waveform than all the other things.   

As I suggested and Thierry hinted at, there seems to be reasonable to have a conventional LC tank for the antenna osc., and adjust via the fixed pitch osc. So, ideally, the variable pitch oscillator can be verbatim from ILYAs schematic, only that the tuning cap is an ordinary trimmer cap and not connected to the panel, and instead trim the fixed oscillator via the front panel. The inductor, if 470 µ is too much, can also be trimmed on board in a 1,2,4,8-fashion of smaller inductors and jumpers or such. Then, as the actual sound generation is waveform-agnostic regarding the pitch oscillators, the fixed oscillators can be of any type, as long as it has a suitable trimming range and stability. It has some form of trimming done from the front panel. Does this sound reasonable? Or am I bonkers? (I haven't tried to attack the volume tuning yet)

Edit: Clarity.

OK, didn't mean to misread you CharmQuark!

The Wien bridge is RC based, so it's my speculation that it isn't stable enough for "serious" Theremin use.  Though my LC oscillator testing was in search of minimum phase noise (as well as longer term frequency stability). 

RC oscillators are the most pullable but the least stable; XTAL oscillators the most stable but the least pullable; LC oscillators are in the sweet spot for Theremin use being somewhat pullable and fairly stable (and the L can really jack up the antenna voltage swing).   Often Theremin antennas are padded with capacitance, which renders the oscillator even less pullable / sensitive (in an absolute sense, not in a musical sense).

I think most of your questions will be answered once you start farting around with real circuitry.