Miniature Harrison 145

"My current setup is testing with Sensepeek probes, hopefully that'll be a little more stable, and I'm actually attaching antennas to the thing now."  - ekahn

Are those 10x probes?

"One thing that's been on my mind lately: there's a lot of literature out there about compensating inductors for temperature - both analog and digital - could potentially be useful for the D-Lev? I can think of a lot of ways it could be implemented, depending on the specific needs of a given circuit, but that's probably a topic for another thread."

With air core solenoids I believe any first order temperature effects are due mostly to factors of the air surrounding the antennas, which one would have a difficult time compensating for.   So probably not worth the trouble, but a possible avenue to explore when ferrite is in the picture.  FredM experimented with DC saturation to vary the inductance, and thus tune the coil (a position I hope to never find myself in).

> Are those 10x probes?

Yes!

> With air core solenoids I believe any first order temperature effects are due mostly to factors of the air surrounding the antennas, which one would have a difficult time compensating for.

Yep - well, I'm experimenting with these surface-mount ferrites - for future volume-produced theremins made by anyone but Moog, it's probably safe to assume the pi-wound stuff isn't coming back.

A couple ways to approach it - all in my head, of course:

For a 'full analog' implementation: Use a shielded SMT inductor with a lot of thermal mass and put it on a ground plane with one low-ohm, high-watt 1206 resistor on either side and a thermistor on the bottom, build an op-amp circuit that turns on the heaters when the temperature is below whatever point thermal drift becomes a problem, just enough to gently bring it into a 'safe' operating range.

Or: something like this - no idea how hard it would be to implement.

(Though this is still probably easier to do with a jellybean microcontroller like ATTiny85, if a too-fast analog control loop introduces noise, or if you don't want to worry about starting a fire.)

For a digital theremin it's much easier: don't bother with a heater, just thermal-ground a surface-mount inductor with a thermistor, then actively monitor the temperature with the main processor and compensate continuously if the T/L/DCR drift equations are known.

Sounds!

Congrats Evan!

Updates!

Pictures and video
Beauty shot.

- The video in my previous post was recorded before I tuned the capacitance on C14 and C21, which are the parallel capacitors to both antennas. I am using 0603 ceramics instead of mica as recommended in the build instructions - and maybe that's causing problems - but I do need to solder and desolder them as part of the build process, and once I have my final antenna geometries dialed in I'll swap them out for the (100x more expensive) mica ones.
Here's what it sounds like post-capacitor tuning.

Antenna designs
- Iterating on a final design for the housing and bottom PCB antenna. I expect it to look something like this:

- My puny antenna (4.5 diameter x 300mm length) isn't really cutting it for me, you can see how narrow and nonlinear the pitch field is from the Instagram video. Switching it out for an Etherwave antenna, or a large plate, then re-tuning, makes the playing experience much better. I'm hoping my next project will involve a 'posable' dual plate antenna system, but I'm not confident enough in my mechanical design skills to go for that at the moment, so I'm sort of stuck with a rod for now. That said, I found a series of adapters that will let me go up to an 8mm antenna, and since this system is intended to be played primarily in the near field anyways, that may be enough.

Coil drift

In a previous post I mentioned that Vpp at the oscillator outputs (TP1 and TP3) is about half what it should be, with losses compounding through the circuit. Best I can tell, this is part of why the sound is rather thin and reedy, and I imagine it also contributes to the subpar pitch field sensitivity / linearity. At least from reading through dewster's posts, there's only so low your oscillator voltages can go before you start bumping up against the noise floor; the other Coilcraft inductor with a higher rated DCR didn't work at all.
I now have the entire system simulating in ngspice, and realized that I could improve the accuracy of my inductor model by incorporating the frequency-dependent series and parallel resistive equations that Coilcraft provides. Doing this got me even closer to the actual behavior in simulation. I also built the oscillator on a breadboard and swapped in different components to try and figure out how to bump up the oscillator voltages to the ones that the build instructions expect (step 10): 4Vpp centered at 3.5V. In simulation and on the breadboard, I was able to achieve that by increasing the value of R1 (and R24, and R36, which bias the other oscillators) - 6K on the breadboard and 10K on the PCB let me hit the correct peak voltages exactly.
Here are some of my scope captures from those experiments if anyone is curious.
However. After a long honeymoon period I believe I've finally run into the infamous 'ferrite vs phenolic' brick wall. Before swapping out those resistors, the pitch field was disappointingly narrow and non-linear, but quite stable. Changing those resistors did indeed give the final output a much smoother and richer tone, and the sensitivity and linearity improve dramatically even with my crappy antenna.. but now the pitch drifts terribly.
That said, I suppose there's always the chance I just nudged the wrong thing with my iron and blew out a BJT by accident. Who knows!
Either way, 'more investigation is warranted' - my current plan is:

- Just for fun, repeat the process for my backup board which uses the higher-DCR inductor and would require more biasing; see if I run into the same problem
- Keep simulating, see if there are components at other points in the circuit I can swap out to up the peak voltage and do that instead
- Build the single oscillator as a modular unit on perfboard; use that as a testbed for inductors; figure out how to reproduce the drift in a reliable way
- Maybe pick up a cheap VNA to see what's going on with the antennas
- Do an 'inductor shootout' with all the coils in roughly the correct form factor

In any case. This has all been a lot of fun!

Thanks for sharing your work Evan!  Theremin oscillators are fun to research.

As another data point you might take the pitch coil out of your D-Lev kit, it's 1mH @ 10ohms DCR.  Or I could make another one for you.

It's been my experience that almost all oscillators have significant phase error, which keeps them off resonance, and this lowers the Q voltage boost.  When transistors get used in the inverting arrangement (base to collector) delay gets introduced, which I suppose will be temperature dependent.  Operating at or near resonance likely helps with stability.  The hottest oscillator I've seen so far (for digital use) is Buggins' which is based on a fast hex inverter - it's also very stable, even with small L and C (in series).

Also, the Harrison uses plates, so the oscillators are expecting somewhat more intrinsic and mutual C than usual.

Thanks for sharing your work Evan!  Theremin oscillators are a fun to research.

Couldn't have done it without you, obviously.

As another data point you might take the pitch coil out of your D-Lev kit, it's 1mH @ 10ohms DCR.  Or I could make another one for you.

Already done That was how I verified that it was the inductor causing the attenuated swing on the oscillator and not something like ESD or a misplaced resistor. Image 5 in the 'scope captures' album is the oscillator with the D-Lev coil, compare it to image 6.
By the way, my D-Lev now has a Coilcraft in place of your air coil (sorry - i'll swap it back soon!) - and spoiler alert, it works fine, though I believe you include Q as a firmware parameter so I'm sure it could work better. But per your experiments I'm sure it's subject to the same issue as the Harrison so I'm on the hunt for the 'best possible' inductor in the right form factor - willing to compromise and experiment re: size. Nonetheless, still more viable given that the D-Lev takes 5 seconds to tune rather than 5 minutes.

BTW, I was able to hit what I felt was a comfortable balance between drift and oscillator voltage by going to 4.7K on R1/R24/R36. The instrument doesn't stay in tune over long periods, but I don't hear moment-to-moment pitch drift like in the video. I think the next step is probably to set up a long-term scope capture to measure frequency over time, if only to figure out which oscillator is drifting and by how much.

Just empirically, I'll probably also just order all the 1812 inductors I can find and swap them in and out until one of them works better than what I've got. I think if I looked a bit harder I could find something a little closer to ideal. At least get a comparison between iron and ferrite.

Also, the Harrison uses plates, so the oscillators are expecting somewhat more intrinsic and mutual C than usual.

Yep, although the design lets you compensate for the antenna geometry by changing the value of C14, so frequency-matching the oscillators is trivial. Though I expect the distance-to-pitch relationship has been tuned for the 8x5" plate design. If you have any suggestions for components to adjust to reshape that curve, let me know, I'm getting pretty good at swapping out 0603 resistors with 2-3 mm of iron clearance.

I'm beginning to bump up against the limits of ngspice as a purely command-line interface, I like working with plain text but it takes me hours to figure out how to actually probe the parts of the circuit that I want to look at. I probably will need to switch to something with a usable GUI. (And LTSpice doesn't fall into that category IMO, haha. Qucs-S looks promising.)

I think this design works really nicely with a small antenna in the near-field, as it can be played with primarily finger-based movements rather than wrist-based movements, sitting at a desk. Not so good for grand articulations, but I'm not sure any other theremin can make that claim.

I'm also wondering if 12-layer PCB inductors might be a cheap and more-widely-available substitute for the designs that use pi-wound inductors; they're coreless if not as compact, and should be cheaply manufacturable for the foreseeable future, regardless of the whims of the market. Carl Bugeja has an open-source design that could presumably be expanded to 1mH without too much trouble. Though I'm aware you'll run into the expected issues with mutual C, and that everything is a compromise compared to an air coil.

If I do another analog theremin design I assume I could make it compatible with my existing board shape and form factor - I may pull in some of the elements from the analog theremin master thread. But I'd like to get this as good as it can be first.

"Already done That was how I verified that it was the inductor causing the attenuated swing on the oscillator and not something like ESD or a misplaced resistor. Image 5 in the 'scope captures' album is the oscillator with the D-Lev coil, compare it to image 6." - ekahn

Ha!  I didn't catch that.

"By the way, my D-Lev now has a Coilcraft in place of your air coil (sorry - i'll swap it back soon!) - and spoiler alert, it works fine, though I believe you include Q as a firmware parameter so I'm sure it could work better."

The AFE C divider ratio also depends on a certain voltage swing (Q).  If the voltage is too low there could be trouble.  This is one adjustment in the design I dearly wish I could eliminate.  It seems a shame to knock the voltage down just so logic can sample it and gain it back up again.  I've thought about it forever with nothing to show for it.

"But per your experiments I'm sure it's subject to the same issue as the Harrison so I'm on the hunt for the 'best possible' inductor in the right form factor - willing to compromise and experiment re: size."

Are there any commercially available options that aren't super compact?

"Yep, although the design lets you compensate for the antenna geometry by changing the value of C14, so frequency-matching the oscillators is trivial. Though I expect the distance-to-pitch relationship has been tuned for the 8x5" plate design. If you have any suggestions for components to adjust to reshape that curve, let me know, I'm getting pretty good at swapping out 0603 resistors with 2-3 mm of iron clearance."

Sorry, no suggestions.  Though for a parallel resonant tank one would think just about anything would work, as the parasitic capacitance shouldn't be a factor, so it must be magnetization killing Q?  What are your thoughts here?

"(And LTSpice doesn't fall into that category IMO, haha. Qucs-S looks promising.)"

The interface is kinda clunky.  But it works (after a fashion, as all Spice it seems).

"I'm also wondering if 12-layer PCB inductors might be a cheap and more-widely-available substitute for the designs that use pi-wound inductors; they're coreless if not as compact, and should be cheaply manufacturable for the foreseeable future, regardless of the whims of the market. Carl Bugeja has an open-source design that could presumably be expanded to 1mH without too much trouble. Though I'm aware you'll run into the expected issues with mutual C, and that everything is a compromise compared to an air coil."

IMO the pot of gold at the end of the rainbow is a single layer air core solenoid. Thank goodness they're fairly easy to wind!

Sounds!

Hello everyone!

Congrats on the 145 progress, Evan! Sorry that the parts are so hard to get.

I have a design similar to this in mothballs (now for about 5 years!) which is about the same size. It is essentially a reduced-size 302 design which uses very small (0402, 0201) SMT technology. There wasn't enough demand to produce them though.

If and when market conditions improve, and there's a minimum order of 1000 units (not likely!) then I may go into production, with a wholesale price goal of about $150.00 per unit, perhaps $100.00 for the populated PCB.

It is easier to sell harpsichords than theremins.

"It is easier to sell harpsichords than theremins"

Hi Art, glad to have you here. The answer to the question of what puts the theremin on a wider basis is driving most of us a bit.
Everyone has their approach to solving it. For 100 years. What might be its appeal to young musicians? I don't think classical, pop, jazz or electronic music per se. But what then?

It is as a monophonic sound tool, and technically almost impossible to compete with the versatile sound spectrum and dynamics of classical instruments.
But it can evoke strong feelings, produce extremely rich basses and climb singing heights, completely break away from the listening habits of the western scale. The player is the composer and also the arranger of all these things, like a singer song writer. Theremin to Midi, CV out enhance the possibilities, but the original charakter breaks then down.

Having to practice 10,000 times to play limited parts of pieces well may not be the way to go.