a friend of mine and I are trying to build a theremin from scratch. For inspiration we have been studying various theremin designs. Among other things we have been looking at the schematics of the EM Theremin in the article by Robert Moog "Build the EM Theremin". The pitch variable and pitch reference oscillators look quite brilliant: they have less components than the most simple (Colpitts) oscillator that we were able to come up with. I am including a picture of the oscillator from the EM theremin at the bottom of this text. It would be great if someone could answer some questions about this oscillator:
1) What is the name of this type of oscillator?
I am new to electronics so it would be great if someone could tell me wether the following assumptions are correct:
2) As far as I understand the design incorporates the oscillator into a long tailed pair, am I right?
3) It seems to me there are no phase inversions which is probably why this circuit can be so simple.
4) Furthermore I see that usual Rc for a LTP is not present on the right and on the left it is replaced by the tank consisting of C1 and L5 which should then have a rather high impedance at the operation frequency of approx 260 kHz.
5) I believe (by looking at the resistors R3 and R4) that approx 1/48 of the signal's voltage is fed back to the input of the LTP so the LTP must have an amplification factor of a bit more than 48.
6) Finally I observe one component C4 which, I assume, is not vital but is probably there to suppress noise. Am I right?
7) Are there any drawbacks to this oscillator? How about stability? Does it have a good Q-factor? How about temperature dependence of the circuit etc...
Ok, I hope this is not all nonsense and thanxs in advance for your patience.
1. It's a differential pair oscillator.
2. It uses a differential pair of transistors, where the bases are the differential inputs, the collectors the outputs, with feedback through the common emitter resistor. I've never heard it called a long tailed pair but yes that's what it is.
3. Because of the differential pair there is 180 degree phase inversion (i.e. it is an inverter).
Looking only at Q1, it conducts because the base sees ground through R1 and the emitter sees -12V through R2. Q1 conducting causes a current to flow through the tank C1 | L5, which causes the collector voltage of Q1 to start heading downward. Conduction will decrease once the collector voltage nears ground, but may continue downward due to the "flywheel effect" of C1 | L5, but then will head in the positive after bottoming out. Once the Q1 collector voltage gets high enough, it will start to turn on Q2 through C3 + R3 + R4, pulling current through R2, which makes Q1 current starved so it starts to turn off.
4. I suppose.
5. The feedback gain is 1/48, but the differential gain at oscillation is slightly greater than 1 so that oscillation is sustained. It's basically providing just enough "pull" on each cycle to make up for losses in the tank. Think of a kid on a swing being pushed by a parent. Oscillation takes a while to build up to maximum voltage swing, it doesn't happen all at once.
7. Q1 is biased through L5, so you have to take this current into account when picking the one you will be using in your circuit. DC bias current can cause the inductor to look smaller than it is if it causes saturation. Also, this circuit can "stall" which can be troublesome.
I'd recommend you stay away from discrete transistor designs and go with something like CMOS 4000B series buffered logic for drive and inversion. And don't use the kind with schmitt trigger inputs. Look at crystal oscillator circuits for ideas, they are designed to be very stable. A CMOS gate will give you tons of gain for a snappy response at the output (can't stall) which you reduce with a resistor before driving the tank. The CMOS threshold is somewhere around VCC/2, but it doesn't really matter that much as this is just an arbitrary operating point. Perhaps use a capacitive divider in the tank to get hundreds of volts of swing.
I should add to all of the above that analog Theremins are not my strong suite - there are others here with much more real world experience who can offer you excellent hard-won advice.
Dewster has given great answers to your questions and I have only this to add..
6. There does need to be a capacitor between +V and ground - it is absolutely essential -
The top of the inductor L5 needs to be connected to ground so that the players capacitance (which is referenced to ground) is seen across the tank inductance (via the antenna inductance) - but the top of L5 is connected to +V... C4 effectively shorts +V to ground for AC (high frequency) signals. I have found that the quality of the decoupling capacitance is important - Low ESR and low inductance = better coupling.
I disagree with Dewster on the matter of discrete transistor vs CMOS.. ;-)
IMO, both technologies can be equal for constructing oscillators, and neither is intrinsically better - a poorly designed CMOS oscillator will be inferior to a well designed transistor osillator, a poorly designed transistor oscillator will be inferior to a well designed CMOS osillator, and if both designs are equal , the results will be equal.
IMO, Bob Moogs EW/EM transistor oscillator (UNMODIFIED) is well designed and stable / reliable, and difficult (not impossible) to match with a CMOS design - In particular, CMOS is quite fussy about voltage levels - the voltage on the collector / inductor junction can be higher than the limit for CMOS, and one wants the voltage (amplitude) here to be high - CMOS restricts one IMO, discreet is comparatively liberating! ;-).
Ps .. I have some LT-Spice simulations on the Element-14 Theremin Group you may be interested in.. LT-spice is excellent and free..
"I disagree with Dewster on the matter of discrete transistor vs CMOS.. ;-)" - FredM
Another negative point about the Moog circuit: I could be wrong, but I believe the tank swing voltage is restricted to the power supply voltage. And it probably works best with a dual supply, which is one more rail that you should regulate to reduce frequency drift.
Discrete transistor LC oscillators are often designed with spectral purity in mind (for use as a local oscillator in an RF transmitter or receiver). The way to do this is to only give a gentle tug on the LC tank every cycle. So they can stall if moved sufficiently off of the resonance point (e.g. with external linearizing coil and hand capacitance) and / or if the gain is reduced a little.
CMOS oscillators have gobs of gain about the input threshold point which keeps them from stalling, and gobs of push-pull output current that you have to knock down to use as tank drive. Other than possible ESD issues, I'm not sure why they aren't employed more often in Theremins. It could be more for historical reasons than anything else I suppose. But then again IANAATD (I am not an analog Theremin designer).
"In particular, CMOS is quite fussy about voltage levels - the voltage on the collector / inductor junction can be higher than the limit for CMOS, and one wants the voltage (amplitude) here to be high - CMOS restricts one IMO, discreet is comparatively liberating! ;-)." - FredM
Not really a limit if you pick a topology that allows this. My AFE has a capacitive divider that knocks the tank voltage down by a factor of 50 before being presented to the CMOS input. And that input is protected against overvoltage with the same device used to absorb ESD.
"Another negative point about the Moog circuit: I could be wrong, but I believe the tank swing voltage is restricted to the power supply voltage. And it probably works best with a dual supply, which is one more rail that you should regulate to reduce frequency drift." - Dewster
Agreed, for optimum performance one does need dual regulated supply, which is a pain.. But the rest of the EW/EM circuit also requires dual supply - if one is looking for a single supply circuit, then one isnt looking at the EW/EM circuit - one would need to redesign everything.
The waveform (voltage) across the tank exceeds the supply voltage (goes to about +24V) which is beyond CMOS limits. I have built CMOS oscillators which have been as good as, perhaps better than the EW/EM oscillator - but these have had a discreet transistor driving the tank, allowing the voltage across the inductor to exceed CMOS limits..... However, if one needs a transistor and CMOS device, I see the 2 transistor oscillator as simpler. (should just say that the tank voltage does not present a problem if one uses a tapped tank inductor, and the better CMOS oscillators I have seen often use one)
IMO, there is no reason to stay clear of CMOS oscillators, and no reason to avoid discreet oscillators - the choice comes down to application requirements and a pinch of personal (often irrational ;-) preference.
It think the advice "I'd recommend you stay away from discrete transistor designs " is biased, and (sorry ;-) a bit irrational, just as if I was to say "I'd recommend you stay away from CMOS designs " I would be being biased and a bit irrational - But I am inclined to say such irrational thing sometimes! ;-)
The only real advantage I can see for CMOS oscillators, where the inductor is in the feedback path, is that it is easier to design for zero dc current through the inductor.. In practice, this current is no problem unless it fluctuates. I have built EW oscillators where the emitter resistor (R2,R6) are replaced with a constant current sink - but had no noticable or measurable improvement.
Not to beat a dead horse (my disagreements with you Fred are mainly just quibbles) but having a lot of excess gain means the oscillator will pretty much run no matter what, which is good for newbies. And my AFE doesn't require a tapped inductor to reach hundreds of volts in the tank.
Speaking of tapped inductors, I saw a design on-line (Russian I think, the oscillator went to a divider, then to the input of a PC soundcard input for further processing) which struck me as quite interesting. It looked like the linearizing inductor was incorporated into the tank as a tap (with many more windings than the driven winding). The slug then (I believe) mainly just balanced the response between linearizing inductor / hand capacitance and tank inductor / capacitor. It used CMOS gates as actives.
"Not to beat a dead horse (my disagreements with you Fred are mainly just quibbles) but having a lot of excess gain means the oscillator will pretty much run no matter what, which is good for newbies." - Dewster
ROFLMAO!! - ;-) It think you should change your icon to a picture of a rotwieler ;-) ..
First, I wish to say that your " my disagreements with you Fred are mainly just quibbles" is the way I feel about you - except I would even remove the "mainly".
Ok, I give up! CMOS oscillators are the greatest, they always work, never have the kinds of problems one gets from discreets, and the only reason why they are rarely used in theremins is because most classical theremin designers are either old or stupid or both, or because they are afraid of having some theremaniac brand their theremin "digital" (there may be some truth in this, sadly) or because they want to hide the operation from those who dont understand transistors, or because they want to appear clever to the ignorant masses who want to build a theremin... .. Nah! I don’t give up! ;-)
Sadly, there are probably some instances where some of the above may be true, but, in the main, if there was some overwhelming advantage to CMOS, I think people like Bob Moog would have used CMOS oscillators in products like the E-Pro and EtherVox which have loads of CMOS IC's doing other functions - but no - both use transistor oscillators.
I am not saying here that because Bob didn’t use CMOS oscillators they cant be any good - What I think I can deduce is that the choices he made were usually good rational ones when it came to design, and that he was well aware of CMOS oscillators, and well aware that, from a manufacturing perspective, IC's are a lot easier (cheaper) to fit in a board than discreet components. One possible reason for his choice may have been board area - Transistors take a lot less space - But I don’t think he would have made this choice if CMOS was a clearly superior choice from a stability / purity / quality perspective.
And here I go again.. Nerdolympics! LOL ;-) [nerd rating > 10 - but this discussion probably needs a geek rating!]
1.) Gain .. Yes, there is tons of gain in CMOS - More than one needs.. there is more than enough gain in almost every small signal BJT.
2.) Transistors have 3 legs, all used! - No unused pins possible which need to be tied to a rail otherwise produce shit - Require no independent supply or supply decoupling - All the sort of mistakes "newbees" are likely to make with a CMOS oscillator which produce all kinds of difficult to find problems.
3.) Newbees are best sticking to tried and tested circuits - there is probably no more thoroughly tried and tested theremin oscillator than that in the EM/EW design
4.) It is far more difficult to do a good PCB layout for a multi-element IC than it is to use discreets - and if one is building several oscillators (VFO,REF,VOL) you really don’t want to share elements in an IC between oscillators - therefore you need 3 14-pin ICs for these 3 oscillators - Far easier to use a pair of transistors per oscillator, this way uses less board area and allows oscillators to be placed apart and optimally laid out.
5.) Competent designers will choose the best solution for any particular application, based on requirements and costs - CMOS is clearly superior for your AFE application - but ..
I do think you are wrong to assert that CMOS is the only 'good' choice for theremin oscillators, or to try to put "newbees" off the EM/EW oscillators - these oscillators IME are good reliable and easy - there are dozens of really crap oscillator circuits on the www (both CMOS and transistor) - let newbees play with something which works straight off the schematic - I have never heard any reports from anyone having a problem with an EW/EM oscillator, but many reports of other oscillators which stall or squeek or otherwise screw up.
"the oscillator will pretty much run no matter what, which is good for newbies." - Applies to the EW/EM oscillator probably more than to any other - certainly more than any other that I have seen.
(did I get the gold - or do I have to settle for the silver?) - As there are only two contestants, it must be one or the other.. ;-)
thanks a lot for all the (slightly conflicting) advise ;) . We will probably be sticking to BJT's and/or other discrete components for now (Maybe we'll use an op-amp at the final amplification stage but at least for the oscillator/mixing part we want to use basic components). That said, CMOS certainly seems interesting and maybe worth it to look into in the future. For now we mostly enjoy playing with the electronics to see what it does and building the whole Theremin has become almost secondary. We have built two types of oscillators so far abut we aren't really happy with those so it might take a while before we have our final design ;) .
In any case, I have one more question:
"Perhaps use a capacitive divider in the tank to get hundreds of volts of swing."
Why would you want hundreds of volts of swing? I suppose the signal must not be too weak but why would such high voltages be useful?
"I have never heard any reports from anyone having a problem with an EW/EM oscillator..." - FredM
IIRC (can't find the post) RS Theremin had his EWS conk out when using the "maximum voltage with a scope probe in the vicinity" tuning method.
Probably displaying my vast blind sides by overly commenting on this (I'm something of a newbie myself when it comes to oscillator design) but another personal nit I have to pick with the EWS oscillator is that the drive and sense points are one and the same connection to the tank. This kind of 0 degree / 360 degree thing strikes me as asking for stalling / non-self starting behavior. I guess I prefer designs that use 180 degree phase via a tap or the LC in series.
"Why would you want hundreds of volts of swing? I suppose the signal must not be too weak but why would such high voltages be useful?" - edanielf
High voltages will help swamp outside interfering noise sources (increase SNR). Not sure if anyone has studied this, but it makes sense. Using a linearizing inductor (like the EWS design) can jack this up even higher at the antenna.