Not "officially" part of the TW Theremin project - I just wanted to move some of my clutter from the "Goals" thread, and have an unobtrusive place that I can dump what might be junk (others are of course welcome to do likewise)
Following Uncle Howies inspiration, I have had another good look at the RCA data, and play with Armstrong oscillator simulations - This is what I came up with as a quick throw-together.. I have also posted this to the TW Theremin Discussion at E-14.
WARNING! The above is an ARMSTRONG oscillator, which has been erroneously identified as the oscillator used in the RCA theremin.
The actual oscillator invented by Lev is NOT an Armstrong oscillator - The above circuit, whilst it works, is GREATLY INFERIOR to Lev's Oscillator for Theremin use.
YOU ARE ADVISED NOT TO USE THE ABOVE CIRCUIT FOR THEREMINS!- The real Lev oscillator circuit is shown further in this thread.
The 42IF106 IF transformer has the required inductances for the tank, and pins relating to the simulated inductors are shown.
The above simulation waveforms - Red Sine = voltage on L1:A, Black Sine = Voltage on Q1:D, Blue square wave is current through L1:B and Q1 (10mA, 5mA average)
It should be noted that this current is through the "tickler" winding which has low R and L (100mR and 39uH max) so thermal and other problems should be minimal.
Not physically built or tested in this form, but FET model is a good manufacturer supplied one which I have found reliable. Best to look at this schematic together with data on the RCA.
(I have used similar circuits at much higher frequency - the above is designed to cover the RCA's 172kHz frequency)
Continuation of the above:
Simulated only, I have built similar fet mixers, but never this one.
Waveforms from first filters - Note, differential outputs have been genersted, but this is not strictly needed - one could take the lower [R4(2)] signal and use this directly:
And finally - differentially combining the signals, and sharpening the filtering at the opamp, one gets clean audio (the above is shown with 170kHz and 172kHz input signals, giving 2kHz audio out:
Filters are designed to give flat (1db) response from 20Hz to 3kHz, drops to -3db at 6.4kHz. 172kHz is -55db, 280kHz is -63db.
Some information on the 42IF106 transformer
FULL MEASUREMENTS ON 42 IF 106:
PINS: Min L(uH) Max L(uH) DCR(Ohms)
4-6 9.8 29.2 0.1
1-2 221.8 517.7 3.6
2-3 43.6 114.6 3.3
1-3 489.1 1164 5.8
One can also add the 4-6 winding to any others giving additional range (for example, add 4-6 in series with 2-3 to extend somewhere around 50uH to this range available.
It can be seen that tis cheap adjustable transformer can be used to omplement a wide range of adjustable inductances.
When used as a transformer, pins 4+6 are a seperate small winding, and pin 2 is a tap off the main winding.
There is no internal capacitor on this part, unlike some other IFTs.
great work!.... but it appears the your tickler has a much lower inductance- whereas in RCA they are about the same... 170-180(ish) mH.... and the RCA osc coils windings are in two layers with one on top of the other. The diagram for the 421F106 leads me to conclude that is is a single wind with multiple tap points. Fred, Im no engineer (should be my signature!) but couldnt these differences have an effect, albeit a subtle one, on the tone?
I dont think that I saved it,(and I may be wrong) but I recall another, or possibly followup email where Uncle Howie recommended against changing the coils at all.
If I may make a request? ... another sketch but with the original RCA pitch osc coils-
76 turns of #26 magnet wire on a 1.5" form X 2 layers (with about .03" layer of fabric seperating layers. ) w inductance @ about 170mH for inner layer , 178mH for outer and resistance of about....(wait let me check....) 1.6 ohms
Not too hard for the home builder. The wire can even be purchased @ Radioshack
The tickler is a separate winding normally used as the output of a 455khz IF amp. Fred's simulation does not use the tap in the primary.
Rewinding it is, in theory, possible, but these things are TINY. The wire is very thin and the ferrite is brittle. Messing with it just invites disaster. I've been winding inductors for almost half a century, and I don't want to fiddle with these.
Dual gate MOSFET's are available, and cheap, but so far I've only found surface mount types.
Also, there is a lot of conversion gain, probably too much. Two JFET's I think are better, and do the same thing.
You are absolutely correct - My Osc schematic bears little resemblance to the RCA - particularly with regard to the charactaristics of "L1". I simply knocked this simulation together as a quick first thought for evaluating the Armstrong oscillator.
As I see it, the "tickler" winding (L1:B) should have little impact - but I could be completely wrong about this.. my view is that, as long as it is capable of maintaining the required feedback to keep the oscillator oscillating, anything would work.
As I see it, the critical requirement is to get the value of L1:A and the tank capacitance C2 the same as the RCA if one wants the circuit to respond to antenna capacitance (via the antenna EQ circuit) the same way as an RCA.. It is the respective values of the tank resonator and the antenna resonator which should determine this - so I have kept these the same as the RCA..
This oscillator could be changed to work at EW frequencies by changing the values of L1:A (using pins 2-3 rather than pins 1-3 for L1:A) and C2, and replacing the antenna EQ values to those on the EW, for example.
It is only because Uncle Howie seems to think the oscillator design is critical (as in, Armstrong + the same LC arrangement) that I bothered to explore this - And the results are interesting.. This oscillator is simple and seems quite reliable ON SIMULATION.. But I really cannot see anything which particularly places it in any "special class" except perhaps the following:
What my circuit does not emulate is the somewhat strange configuration at the tank capacitors of the RCA - Referring to the above diagrams, it can be seen that my tank capacitors (C2) are taken to 0V, whereas the RCA tank capacitors (C1+C2) sit between the plate and the grid - I do not understand this (someone knowlegable on tubes could perhaps help me out and explain how this is working?).
The other factor I notice about the RCA circuit, which is different to most (if not all) transistor (certainly BJT) circuits I have seen, is that it looks like impedances will be a lot higher.. One has the antenna EQ connected to the tank resonator as expected, the "strange" tank capacitor configuration which, as I see it, has these capacitors connected not to a fixed ground but to an active driven signal which is contra-phase to the signal on the grid, and one has the oscillator connected to the high-z mixer grid..
Apart from the tank capacitors and "losses" to T8(3-4) the whole tank side looks like it could be much higher z than the tank circuits in most transistor designs.
Looking at C1+C2 and the connection to L1:B, (and not fully understanding whats going on) - I can see that the L1:B may be performing a function more than just providing required feedback - and that, in this case, the inductance / ratios etc WRT L1:A could be important.
My Osc schematic does nothing special - it produces a clean sine wave, thats all - IMO, it does not warrant further evaluation as it stands.. The Moog EW Osc is, IMO, a better choice..
The only area which may warrant further investigation is whether my different wiring of the tank capacitors changes anything substantially, and whether it is possible to "copy" the RCA configuration, and whether this has any advantages.
w0ttm.. Thanks for posting that simple RCA Osc schematic - it really helps discussion!
ok - I have just re-wired my oscillator simulation and connected C2 between G and D of the fet (copying the placement of G-P on original)..
I tried this configuration with L1:A = 598uH L1:B = 15uH, and therec was little change.
Then made L1:B = 598uH, and there was significant differences - more distortion on the waveform at L1:A.. this is becoming much more like what I expect to see to produce the tone I have heard from the RCA..
However - to get the simulation to run I did need to ignore simulator warnings - I have not looked at the levels etc in terms of the fets charactaristics, current through the fet increased to +/- 25mA (if I believe the simulation), and the frequency was not what I expected (much lower).. So big doubts.
I then connected C2 as in my original (to 0V) but left L1:A and L1:B at 598uH .. The results (again, ignoring simulator warnings) surprised me.. Waveform on L1:A was the right frequency and skewed the way I would expect to produce the RCA tone - Levels were sensible, currents (averaged) < 10mA.
So, it seems that the tickler coupling has more importance for this application.. Thinking about it, I suppose that if one wants a sine wave oscillator, then one designs so that the feed-back is just a little more than one needs to sustain oscillation... A tiny coupling is sufficient to do this, and in a controlled feedback loop, good clean sines come naturally..
When one wants to distort the waveform, overdriving the oscillator is a way to do this. Once again it seems like perhaps our dear Uncle has it right!
Now to hone the circuit, check tolerances etc, and (if it all looks ok) connect up a (simulated) mixer to two of these (simulated) oscillators to produce audio .. compare this audio to RCA samples, and fret because I cannot actually build a physical circuit at this time.. 8-(
I could not resist sim-ing it too :)
Results close enough to yours as to be trivial.
I'm off work Thursday, so I'm going to breadboard what ever you post. All I have for transistors are some MPF 102's.
Also, I only have some older IF transformers. The same inductance, but physically larger.
This may not tell us much, but I'm at the gotta solder something before I go nuts phase.
I have some serious confusion over the tank values..
My calculations say that, for a normal grounded tank capacitor, to get 172kHz I need 778 uH..
This value is almost right when I run the simulation with grounded tank capacitor (1100pF)
.. However, when I put the tank capacitor between gate and drain, I need to reduce its value to less than 300pF to get 172kHz.#
So, either the simulated fet circuit is not doing anything like the RCA (tube) circuit, or the values for the inductors (L1 in above schematic) need quite drastic reduction.
I have managed to get the simulated circuit to operate within component (particularly the fet) tolerance ranges by inserting some attenuation at the gate, and got rid of simulation warnings, convergence problems and errors . have run FFTs on the ouptut waveforms which produce a nice spectrum containing strong fundamental and both even and odd harmonics with a good profile (no nasty dominant harmonics, no other artifacts)..
But I cannot visualise what is happening across the tank capacitor, and this really bugs me - for me its possibly a bit like having gone blind.. the moment I saw this configuration I had the feeling that I was in new territory. What is also strange is that I have searched through every document I can find about the RCA, including every circuit evaluation, and seen no mention of this (to me) weird configuration... Is this because its obvious to everyone else? or is it just that no-one has noticed it...