Finding that pleasing illusive sweet note

Christopher: The students get a direct link to the page when they are enrolled.  Only guests have to traverse the long arduous route to the grail. ;-) I think the noise has to do with the speakers I hooked the theremin line out up to.  I should have fed the audio directly into my recorder.  I don't use switching regulators. The unusual waveshape at 40Hz comes naturally from the FET source follower after the oscillator circuit. the 460 Hz wave is high enough in frequency to avoid the non-linearity of the input capacitance of the FET used in the follower. If a sine wave is what one wants then I have already tested a compound NPN bipolar transistor buffer instead of the FET based source follower. It results in very low distortion sine waves to below 5Hz. I'm not sure pure sinusoidal is the right path unless I can apply the right analog signal processing to the sine. Yes I looked at the link as well as some others you have posted and I am presently thinking that an asymmetrical acting soft limiter will yield signals that look a lot like the signals in your linked audio signals.  I will be trying this soon. I already tried asymmetric without the soft limit and I like the sound much better. The soft limiter is soon to follow. And finally indeed the humor is essential.

Dominic: I looked at simpler mixers for some time last year.  I agree completely that a lot can be done there. One problem with the simple mixers is that they are based on FET transistors which have a horrible consistency of specs. I noticed a few bipolar mixer designs that could be more suited to successful repeating. I really like the LT2940 if for no other reason than it was never intended to be a mixer as such. You can also overdrive it to generate some of the same non-linearities that simpler mixers create when over driven. It also consistantly overloads. I was trying to generate a very pure sine wave to smash up later with "pitch processing" when I chose the LT2940. I might reconsider creating more non-linearities using simpler mixers but not just yet.  I realize the rev 1 design is more involved and more complicated than it has to be. Perhaps I will abandon the Theremin lab. Naaaaahhhhh. I am not ready to do that yet I just started. It isn't really a lab on Theremin design it is a circuits lab that uses a Therenmin as the engine to introduce "neat" interesting analog processing. I have even triedusing a ring modulator which I'm guessing is Sacrilege when constructing/playing a Theremin.

Mark, because I don't build analog Theremins, please take my comments with a grain of salt.

First, I think the Theremin is an excellent way to introduce budding EEs to the real world of inductance and capacitance.  Inductance in particular was quite mysterious to me until I started down this road.

I simulated an NPN version of your pitch side variable oscillator in LTSpice and am seeing an antenna voltage swing of ~10Vp-p @ ~400kHz, is this approximately what the circuit does in real life?

Is there any reason you picked a PNP version of the oscillator?  I don't think it matters much, but NPN silicon transistors are generally superior, and the tank can be referenced to any low impedance source, including VCC of the circuit.

Did you pick this oscillator because it is similar to the one in the Moog Etherwave?  I don't think it is necessarily the best choice if you don't use a linearizing inductor (you don't, correct?).  I think the oscillator is probably padded with too much tank capacitance, but I'd have to do a study to know for sure.  How large and linear does the pitch field feel to you?

It's a matter of taste I suppose, but with modern supply voltages down in the deep single digit volt range it might be more interesting to use circuitry that can function off of lower supply voltages.  Much of the fun for me as a designer is finding and testing oscillators that use a few volts, a few milliamps at most, and produce many tens of volts at the antenna.  A real candidate for battery power.

Why a FET as the oscillator output buffer?  A BJT would likely work as well here, particularly if the antenna voltage is high.  I'm glad you are buffering here, coupling can give interesting timbre variation but it is troublesome.

In terms of mixing, I would try a double stacked BJT.  There is an interesting looking clipped sine wave across the common emitter resistor that might give you a decent start at vocal timbres.  Beyond that I'd probably try asymmetrically clipping the sine wave to get a fixed PWM, and alter the duty cycle to taste.  I would keep keep it all a simple as possible.

Also, if I were me I'd get the students to build it on a breadboard with through hole parts so that they can monkey around with things and see first hand the parasitics at work on the breadboard.  The breadboard would need to be bare and not have a metal plate on the back though.

And, as I mentioned over on the PWB thread, I'd aim for a more rectangular PWB, with the volume and pitch circuits more physically separate from each other at the ends, and the processing circuits in the middle (ala the Etherwave)

Best of luck, and thanks for bringing this to our attention!

Dewster: Thanks for taking the time to respond.

I do get almost 12 Vpp from the tank circuit at a little less than 400KHz. In simulating, prototyping and on the PCB. I have since changed the bias voltage at the base  of the tank driver to give me 10Vp headroom for the tank circuit. I lowered the resistor in series with the inductor and see almost 20Vpp at the tank. I also went to 100pF for about 450KHz oscillator frequency. Also as you probably know the series resistor changes the tank voltage greatly. I am of the opinion that a small series resistor gives the tank more ability to deliver current to the next stage. This is why I raised the bias point on the transistor.

I choose the PNP version because at the start of all this I was looking at Art Harrison's 145 circa 2000 design notes he published. I also noticed that Bob Moog had the NPN version in a '96 article I saw. I use Vcc as an AC ground on the mixer. I decided to use GND for the AC ground because I believe it is a better ground if one doesn't use a plane for Vcc. I didn't on the board.  I do use a plane for GND though. I doubt that it makes much difference given the other things that are going on. So I guess the real answer is a made the deliberate decision to use GND for the AC ground. I agree with you that it could easily be Vcc.

I did not use linearizing inductance as in the Etherwave this last time.  I have purchased some "high" frequency self resonance 10mH inductors (1.3MHz) to try them. I agree this oscillator is not the best without the linearizing inductors. I have yet to sit down with proper elbow body position to see how the pitch changes with distance. I like the way it responds which isn't saying much since I can't play music on a theremin nor have I tried an Etherwave or any other. I am interested in an oscillator that can work well without the linearizing inductors. I am spending way to much time going over as many posts as I can on this and other subjects directed towards analog design. I am also interested in digital signal processing as I have much industrial experience in this area. Analog signal processing and electro-acoustics also not that this has anything to do with the discussion.

I used an FET because that is what I saw in the Harrison 145 design. I used an NCH MOSFET instead of the JFET because I like source/emitter followers for buffering. I commented on this just 1 post above. I have simulated and prototyped the compound NPN/NPN transistor for the buffer. It is far more linear than the FET with the FETs quadratic nature of V/I.

dewster: In terms of mixing, I would try a double stacked BJT.  There is an interesting looking clipped sine wave across the common emitter resistor that might give you a decent start at vocal timbres.  Beyond that I'd probably try asymmetrically clipping the sine wave to get a fixed PWM, and alter the duty cycle to taste.  I would keep it all a simple as possible.

I saw your example in a much earlier post. See my response in my previous post as to why I picked an LT IC for this. I think you have it correct that I have choosen a mixer that does not allow me to easily smash up the signal to generate descent vocal timbres in the present design. I have since attempted to overdrive the mixer with promising results. I made circuit boards for the students to evaluate the mixer by itself since it cannot be purchased in a DIP package. This allows me to run real time tests on the part.  LTSpice has trouble with their own LT2940 though it can be coaxed into running. I tell the students how in the lab manuals. I have tried asymmetrical clipping (see previous post) clipping and will soon add soft limiting to the mix. I am waiting for the sample THAT parts to come in later this week. I haven't figured out what you really mean by fixed PWM.  I am very familiar with PWM when used say in a type D amplifier or in switching power supply design. Through what process are you putting the asymmetrical sine wave to get a variable duty cycle? Yes simple is divine. I probably caused my own trouble by making the oscillator mixer chain as linear as possible. Fortunately this is alterable.

The lab tries to have the students simulate with LTSpice, prototype on breadboards with example layouts and ultimately build the circuit board. The breadboards we use are OK but they could be better. The metal plates are at least gone. The circuit board uses surface mount parts because through hole parts are getting harder to come by and IMO learning how to build SMC prototypes is valuable.

Rev 2 board has a much more rectangular appearance as opposed to the square board from last semester.

I am now seeking usable analog processing techniques before I crank out Rev 2 of the PCB. I have tried creating a square, triangle, ramp in phase with the sine and adding or subtracting them in ways that more closely emulate the waveforms posted that "sound good". I also have tried the asymmetrical clipping followed soon by soft limiting all in an effort to create that "illusive sweet note" through analog signal processing. I do realize that includes doing things to the oscillators and mixers. I probably won't have enough time this summer to try "everything" before I make rev 2. I am making a quick stab at it now though.

I know there are many things that I do that go against best practice. I.E. ferrite surface mount inductors. The Q's are high but temperature stability is an issue. I have a place for both silver mica through hole and ceramic surface mount caps on the board. As far my quick tests show that the CGO/NPO ceramic caps work as well as the silver mica. Of course they haven't been tested over large temperature variations. I'm not yet concerned about these issues as far as putting people off the Theremin. This the lab is supposed to generate interest in the circuitry and the actual Theremin. It certainly has the first semester of the lab.

Christopher. I interpreted what dewster meant was that it was mysterious until he looked into it.  Although I could be wrong about this. It is not my day to be omniscient. I think that occurs in about 2051 sometime. All of the students have the same feeling as dewster. That is why some of us particularly use inductors in labs. There doesn't have to be as much mystery as all that. It is a good complimentary device to the capacitor with a few more real world caveats.

Mark

dewster: I forgot to mention that there is a difference in the Harrison PNP oscillator and the Moog NPN. I found that Harrison's starts up very quick compared to Moog's. The Harrison seems more tolerant to changing the inductance R. This certainly comes from the collector to base coupling differences used in these oscillators.

Mark

Mark, I will respond to each of your points in time, but I only have a little left this evening (I adhere to a very rigid goofing off schedule ;-).

Now that you mention it, I do recognize the Art Harrison oscillator design.  I eschew oscillators which supply operating current to the drive transistor via the tank inductor.  Not so much because there is DC going through the inductor (though this could be an issue if it is large enough to cause heating or saturation) but mainly because it limits the antenna voltage swing due to clipping via reverse conductance of the drive collector / drain.  In the Etherwave this isn't an issue because the tank isn't the primary LC resonance (the series "linearizing" coil and the antenna intrinsic + hand C are the primary LC resonance) but I believe it may be an issue e.g. in the Thierrymin.

I thought the 100 ohm series resistance in your tank was parasitic, but you say it is real.  You are using it to keep the voltage swing from clipping?  If so this is hurting Q and thus hurting your SNR.  IMO the aim for high quality Theremin oscillator design should be to do whatever to keep Q high and phase error low, and thus maximizing antenna voltage swing and minimizing oscillator current draw.  Lower voltage swing, low Q Theremin oscillators may seem to work OK, but they don't reject environmental interferers so well.  Oscillators that employ transistors in the inverting configuration (sans cascode) tend to cause phase trouble.

Designs that use a linearizing series coil require both the fixed and variable oscillators be tunable so as to best accommodate the third and primary LC resonance.

"I forgot to mention that there is a difference in the Harrison PNP oscillator and the Moog NPN. I found that Harrison's starts up very quick compared to Moog's. The Harrison seems more tolerant to changing the inductance R. This certainly comes from the collector to base coupling differences used in these oscillators."  -- Mark

Yes, I believe I've seen this in simulation, where the Moog oscillator is right on the edge of not oscillating due to too much feedback.

Mark, I will respond to each of your points in time, but I only have a little left this evening (I adhere to a very rigid goofing off schedule ;-).

Yes I am lucky this is pertinent to my actual job.

Now that you mention it, I do recognize the Art Harrison oscillator design.  I eschew oscillators which supply operating current to the drive transistor via the tank inductor.  Not so much because there is DC going through the inductor (though this could be an issue if it is large enough to cause heating or saturation) but mainly because it limits the antenna voltage swing due to clipping via reverse conductance of the drive collector / drain. 

Yes this is why I changed the bias point for the tank drive transistor in the next version. By going to 10 volts instead of 6 I can reduce the R to 10R0 and achieve 20Vpp at the tank. I am also considering letting the tank voltage be limited by the becoming too large to add "pleasing" hopefully harmonics.

In the Etherwave this isn't an issue because the tank isn't the primary LC resonance (the series "linearizing" coil and the antenna intrinsic + hand C are the primary LC resonance) but I believe it may be an issue e.g. in the Thierrymin.

I will have to consider the Moog oscillator. I like the Harrison design for its ability to oscillate so easily.

I thought the 100 ohm series resistance in your tank was parasitic, but you say it is real. 

The inductor has 3R9 ohms max I think. The rest is added by a physical resistor. The inductor is high Q. I remember 78 at 122KHz.

You are using it to keep the voltage swing from clipping?  If so this is hurting Q and thus hurting your SNR.  IMO the aim for high quality Theremin oscillator design should be to do whatever to keep Q high and phase error low, and thus maximizing antenna voltage swing and minimizing oscillator current draw. 

I agree which is why I changed the design to allow a higher unconstrained tank voltage. I still put a 10R0 ohm resistor in series. As usual there are compromises one has to make. I will seriously consider an alternate oscillator. As I mentioned above I like the lower Q designs. I'm not sure I care about the small phase change by inserting resistance. After all this is a sine wave entering a mixer that doesn't care what the phase is when generating the difference frequency. Battery powered is a great reason to have the largest Q you can get. I don't have that constraint for the lab.

Lower voltage swing, low Q Theremin oscillators may seem to work OK, but they don't reject environmental interferers so well. 

Yeah I haven't turned on the Mig welder to test interference yet. ;-)

 Oscillators that employ transistors in the inverting configuration (sans cascode) tend to cause phase trouble. Designs that use a linearizing series coil require both the fixed and variable oscillators be tunable so as to best accommodate the third and primary LC resonance.

OK I'll definitely consider at a different oscillator.

Hi Mark,

- In terms of tank capacitor formulation, I believe C0G / NP0 are probably the best.

- Changing the oscillator bias point to get more tank swing room makes good sense. 

Placing a resistor in series with the inductor is one way to control tank swing, but it isn't the best IMO as it is analogous to adding rust.  When I talk about phase issues I'm not referring this resistor, but to the phase of the drive.  If in the lab you drive a parallel LC tank with a square wave via a small capacitance (~1pF) and observe it through a similarly small capacitance, and along with the drive on a second channel, if the coil is high Q then the peak swing of the tank will happen over a very narrow ~quadrature phase range.  Ten or so degrees either way and the swing really starts to suffer.  The same thing is often happening in oscillators if the drive timing is off due to delays in the feedback.  Inverting transistor configurations tend to be the slowest due to Miller effect.

Anyway, these are the reasons I tend to avoid the Moog differential pair oscillator: reverse conduction of the collector, drive delay, and full HV tank voltage across the collector.

For jollies I tried isolating the collector with a diode, which should allow swings that would normally cause reverse conduction.  This is simulation only!  I'm able to wangle 70Vp-p out of it, and I'm pretty sure 50V at the collector is getting too much for a 2N3904.  FredM (RIP) reported some tens of volts p-p swing is enough to swamp environmental interferers, but I say the more the merrier!

There is a nice fixed duty cycle rectangular waveform across the emitter resistor that could be of use in producing a more buzzy voice.

Here is the Spice file I've been playing with: http://www.mediafire.com/download/hdye9v8pevfop3w/ece370_osc_2016-05-26.asc

In terms of timbre, I think you are perhaps persuing things more from an analog synthesis angle than a Theremin angle.  I believe the EWPro uses analog synthesis, and this is likely why I find it's voice drab and sterile.  We've all heard analog synths way too much IMO, I suppose I'm getting to the point where I never want to hear another one.

A good rule of thumb is to double the DCR of the coil as the parasitic R in simulation.  I believe RF losses put a practical cap on absolute Q, so looking for sky-high Q is very much diminishing returns after some point.  Relying on Q too much for oscillator operation (very low current drive) seems to make for oscillators that poop out easily when the hand is near or touching, and that are slow to restart.

Mark: what is the range of the pitch field for this Theremin?  E.g. can you set the null point out at ~1/2 meter or so and have roughly linear pitch response?  One test is to open and close your hand, listening for the interval, then move your hand closer and do the same thing and listening for the same interval.  It's normal for the pitch to cramp up both near and far from the antenna, with a roughly linear central region if things are well designed & adjusted.

Dewster, Perhaps this is something you covered long ago. Is your icon a picture of the cat that thinks you are his slave? ;-) We have several cats in the house with an occasional visiting cat.

I saw your recent posts in Simple Analog ... I played around a little with the NPN/PNP mixer with pretty decent results using a 12V supply. I added an emitter resistor to the pair. I think that was all I changed.

In this same thread you presented some oscillator circuits. Are these what you like to use or are they only what you use with 3.3V?

I only used the term Timbre because that was the term used in the PAIA schematic for the analog processing circuit were the sine and square were added. I am really referring to any kind of analog processing I might use to "smash up" the signal. Asymmetrical clipping, limiting adding portions of sine, square, ramp or triangle. I never thought of this as synthesizing. Instead of trying to achieve this in the oscillator/mixer stage I am trying after the low frequency is extracted from the mixer. Not as efficient and perhaps not as good. It is just what I am trying at the moment.

I tried to posy an actual image in Chinese pcb thread. I was never able to get it to show up. Obviously you know how to do it. Please tell me the secret. I think I used a Microsoft browser during my attempt.

Also is it OK to post wav files or do they get to big and annoy the forum.

Mark, the cat is an animated GIF I snagged from somewhere - the cat is playing slide guitar.

If you're looking in the analog thread then I think you can find everything significant I've looked into so far. The NPN Colpitts with PNP buffer is one of the best oscillators I've run across.  It might work a bit better with a J113 in place of the NPN (and with the gate biased to ground via a 470k resistor rather than to VCC) but the output levels are less predictable as they rely heavily on VT.

I haven't done much work at all at higher VCC and don't intend to at this point.  Though there are intriguing things that can be done around 20 to 30V as the antenna wave can be created directly, and shielding via buffering is then a real possibility.

And I've done precious little on the bench with mixing, so I'm no authority there, though that totem pole mixer seems to work well at least in simulation.  Mixing is I believe related to convolution.  FredM was really into using various analog and digital topologies in his mixers to generate the usual suspect waves (ramp, sine, triangle, etc.) but I'm not sure how many he actually built.  I know he did some vocal synthesis work as well because I pumped him for that.

A couple of weeks ago I researched metal detectors a bit, the very early types were very similar to Theremins but the field has advanced much more since.  The schematics look like FM radios.  I'm glad Theremins can be built so simply.

To paste an image from the photo album: go to the photo album and make sure the image you want is showing, not just a thumbnail.  Right click on it and "Copy Image Location" to get the URL.  In the forum there is a picture button (mountain and a moon).  Here you past the URL into the "Source" field.  I usually click in the "Dimensions" to see what size it will be, and limit this to 600 or 800 wide, or whatever seems best.  You can edit it later, but I don't think there is a way to reset the size back to the native without redoing things.