Let's Design and Build a (mostly) Digital Theremin!

"for the Yamaha 4-op and 6-op synths to get to voices which would respond in interesting ways to breath control." - Explorer

Yeah - This is an area I need to look at closely.. At this stage, I am thinking in terms of breath just as a replacement for the theremin volume control, perhaps with a simple VCF added to change the harmonics somewhat as a function of breath, or as a morph control between two additive waveforms on the more advanced / expensive instruments..

Hopefully I will get advice from the woodwind crowd - But I am a little fearful of going too far in R+D at this stage - a breath controlled theremin is managable and could be put together quite cheaply and quickly - And I need to start making money quickly to pay the rent on my new home, as I still pay for the one I was forced to leave, and the family..

Fortunately I have other paying work for the next 6 months - but when this ends I need to be shifting some products if more work doesnt come in..

I wonder if breath pressure rather than volume of air past a constriction might be a better input?  By that I mean make the breath controller part a breath dead-end rather than a through path.  There wouldn't be any wind sounds you could mix in, and it would be abnormal feeling for a true wind player, but the player could isolate their mouth from their lungs and still breathe normally through their nose while playing (and there would be less spit to deal with).

Fred, are you using a dual antenna in a split / tapered configuration?  That is, are there two antennas fed from opposite ends of the strip, with the feed end wide and tapering to nothing as it gets to the other end?  I've seen this method used for 2D (3D if you include height from the sensor) capacitive inputs as well.

I didn't say so explicitly, but I was thinking about continuous breath control and volume control interchangeably.

I don't know if one needs a filter. It was trivial to map the amplitude of one or more operators to the mod wheel in order to do pseudo filter sweeps. As for morphing, I imagine that one could have an absolute volume control, and then run one operator stack in inverse volume to another, so that as one looses volume the other increases.

I'm definitely a fan of FM/iPD, especially of how inexpensive it is in terms of resources compared to resulting complexity....

"Fred, are you using a dual antenna in a split / tapered configuration?" - Dewster

No - That was my first scheme, it worked, but not well (getting the taper right was extremely difficult, and could only linearize for a fixed 'span'. This was the topology I used for my "Epsilon" - I had several "opposing" tapered strips, from which I derived X,Y and Z..

Cannot say right now how I am doing this new "Antenna" - I have filed a UK patent application, but unlike almost every other patent I have filed, I have handed this idea to a patent agent, to check my application and advise me, so I could yet file another updated application if errors or ommisions are found in my first..

So, at this time, I cannot disclose anything! (I am not paying for the patent stuff - I have effectively sold any rights to this sensor to my client for his application - I am free to use the idea for my application.. My client requires patents..)

Fred.

"I wonder if breath pressure rather than volume of air past a constriction might be a better input?" - Dewster

I dont know - Its one of the things I hope to get some feedback on.. I started by using pressure sensors where there was no air flow, but it didnt feel right to me.. and also, the sensors were damn expensive and I worry that moisture etc will screw them up quite quickly.

What I have now is cheap and simple - I have a tiny electret capsule mounted on (what used to be) the exit hole of a nicorette inhalator, the way these inhalators are constructed is as two parts, the mouthpiece, and the exit shell - both ends have pointed inserts inside, so one places the nicotine capule inside the assembly, and when its closed, the pointed tubes pierce the film on either side of the nicotine capsule, and one can suck air through it -

I have arranged the electret so that the mouthpiece air is focussed on it - the exit hole is blocked and the wire to the electret goes through that.. Then I have drilled a hole on the bottom section of the mouthpiece - this is the air escape hole..

The electret is sensitive up to about 20kHz, but has a good sensitivity at about 16kHz (I originally looked at ultrasonic transducers, but could not find one at the right size - sure I could get these if I really needed them) - I amplify the noise, which (more by luck than judgement) has quite a lot of HF in it.. (my original idea was to construct a "dog whistle" type resonator, but I couldnt be bothered with all the work just to test the idea) .. So I HP filter frequencies above about 14kHz, rectify this, and get my CV..

I wanted as high an audio frequency produced as possible (within reason) because I didnt want external audio signals to be picked up and produce a CV.. But it seems that I have been perhaps a bit pedantic on this.. By biasing the output CV -Ve a bit, I can completely cut off any spurious output even with my crude cheap assembly..

I am quite pleased with the construction - the inhalators can be easily opened and the sensor cleaned (it sits behind a thin plastic film which I have roughed up to give more agitation when hit by the air flow, but its watertight and isolated) and the shape of the mouthpiece suits me well - I tried several different mouthpiece assembly ideas (bought a load of melodia and other spares off ebay) before I found an old nicorette inhalator from years ago when I was trying to give up smoking (I have now given up, and am addicted to E-Cigs ;-)

But as to what the players will think or want, I have no idea.. I can adjust the size of the exit hole by twisting the mouthpiece, and think this "feature" will probably be a good thing to keep.. but personally I like the feel of air flow far more than just using my mouth to control pressure in the mouthpiece - and I like a lot of air flow, as this seems to give better dynamic control - one can go really soft, and need to blow really hard to get maximum output.

"I'm definitely a fan of FM/iPD, especially of how inexpensive it is in terms of resources compared to resulting complexity...." - Explorer

Yeah, I had a lot of fun with FM/iPD back in the '80s - even had a Yamaha MSX "PC" with its FM sound engine.. Then got a some small Yamaha FM synths and a Casio CZ101 (in fact I still have that - keyboard removed and heavily modified as a MIDI module with knobs.. ;-)

But I find analogue far simpler - and for my new stuff, its all mainly analogue (particularly at the controller end) and interfacing the signals to MIDI or  digital sound engines is not "my thing" - not at this time at least..... When I have it all running as I want with voltages I can measure (rather than watching decoded MIDI data on my analyser)  and pitches I can hear, then I MIGHT start looking at MIDI or OSC..

However.... Analogue FM using topology similar to the Yamaha "operator" scheme is well within reach using HF VCOs, multipliers, samplers  etc, and this is an area I have been exploring quite a lot..

But I cannot afford to be a full-time explorer anymore :-(

Fred.

Fred, thanks very much for the detailed responses.  It's clear that you've put a fair amount of development time, effort, and thought into the mouthpiece.

And even if you are contractually prohibited from revealing the details of your antenna, it's interesting to know that the dual taper didn't work very well for you.

The Yamaha FM algorithms are a model of efficiency (doing things in the log domain so multiplications are simple adds, etc.) though they do have some aliasing going on in the implementations.  I owned an FS1r for a while and it was pretty incredible what one could get out of it (though the UI was most horrid).  I often wonder how difficult this stuff would be to do in an FPGA.

Sine wave generation is super easy to do digitally with two multiplications, one add, one subtract, and one delay.  The multiplications set the frequency, the amplitude is set by the initial value in one of the registers, and sine and cosine are both available simultaneously.  In an FPGA one could time multiplex one or two multipliers connected to one or more BRAMs and generate huge gobs of them at 44.1kHz.  Summing them would require only a simple running accumulator.

I believe there was a DAW audio board made a while back that used an FPGA to do the mixing.  They claimed it beat the pants off of a DSP processor.

Hi Dewster,

I will email you soon with details - I am not actually contractually bound per-se, but its unwise to publish anything publicly at this stage.. I have a patent specification filed, but claims are being drafted - and sometimes the specification needs re-wording in order to align with the claims... I have never progressed a patent myself beyond the specification stage, and understand that disclosure of anything could prevent changing something if I needed to.

"And even if you are contractually prohibited from revealing the details of your antenna, it's interesting to know that the dual taper didn't work very well for you."

Please understand that it may work fine if implemented correctly - I had lots of other issues and ended up abandoning Epsilon before I had really explored everything fully.. My reason for dumping the project was not primarily due to the antenna, it was due to the fact that I had tried to cram too much into the concept.. I was trying for single handed control over pitch (X axis), dynamics (Z axis) and harmonics (Y axis) - The antennas may have been ok if I had left out the Z axis (volume) and used the other hand or some other mechanism for volume control - but I decided that combining volume and pitch made the instrument impossible to play (I think that if you take the difficulty of each on a theremin, combining the two on a single control doesnt add them, it probably multiplies them).. And was so disheartened that I never really looked moving volume to a seperate controller - I had an idea for this instrument in my head, and the instrument was unplayable - I chose to dump it and possibly threw the baby out with the bath water.. thats a habit of mine! ;-)

I have had the theremin EWI in my mind for a long time - right back when I started with theremins.. My first experiments used hand proximity to the tube, and I still think this is usable ..

But in the process of trying to solve a clients capacitive sensing problem, I came up with this incredibly simple way to change (and shape to taste) a linear capacitive "antenna" - Its so simple I was sure it must have been done before - But so far the closest found to the idea is the tapered antenna - and thats miles off what I am doing. The beauty of this "antenna" is that it can be used single or double (or in fact, multiple) ended - its entirely passive, and in single ended mode could be connected to an unmodified theremin (it would need to be built to suit that theremins operating frequency and response)..

But this antenna scheme is just perfect for my TEWI and for a monophonic continuum type theremin, or controller for monophonic synths or whatever - it gives me pitch control on the X axis, with some (hopefully deliberate ;-) pitch control (vibrato or trills etc) extending into the Z axis.. And I could provide an aditional control signal from movement on the Y axis..

Unlike Epsilon, this controller is not intended for theremin "no touch" type playing - I envisage the pitch hand usually being at least lightly in contact with the "antenna" - I have no way of compensating for proximity non linearities, so the whole linearity of the antenna depends on proximity (coupled hand capacitance) being quite constant.. As such, one cannot mark the antenna with fixed semitone positions, one still has a theremin, one still needs pitch preview, and it still wont be as easy to play as a keyboard.

"It's clear that you've put a fair amount of development time, effort, and thought into the mouthpiece."

LOL ;-) .. Not really!  I have no idea if its any good - My first moutpiece I did some 5 years ago, and I was blowing directly onto a piezo transducer - My first serios attempt I bought a pressure transducer for, but didnt like it at all - Its been one of those "junk box" projects, with almost no design except the minor circuit for amplifier and rectifier.. If, in the last 5 years, I have spent more than a couple of days on it, I would be surprised..

So I am extremely happy to get any ideas or feedback from anyone - I would really like to hear from anyone here who plays wind instruments.

Fred.

"Sine wave generation is super easy to do digitally with two multiplications, one add, one subtract, and one delay.  The multiplications set the frequency, the amplitude is set by the initial value in one of the registers, and sine and cosine are both available simultaneously.  In an FPGA one could time multiplex one or two multipliers connected to one or more BRAMs and generate huge gobs of them at 44.1kHz.  Summing them would require only a simple running accumulator." - Dewster

This sounds really interesting!

If I could get a bunch of say 24 sines in a simple harmonic series at fixed frequency, with the fundamental at any frequency above say 80kHz, and have digital control over the amplitude of each, all done in an FPGA, then with the "analogue sampling" (or as some here have claimed it to be, and with which I STRONGLY disagree "simple heterodyning") scheme I presented here a while ago, a multi-purpose additive engine could easily be constructed.

This additive engine could be used on monophonic instruments (the simplest being direct control of the sampling frequency from a theremin or theremin type instrument) but by using phase locking techniques, HF VCO's can be driven, and the resulting output waveform from the sampler would be a free-phase waveform frequency of which is determined by the VCO's CV, and this VCO would have an exponential 1V/Octave law - so one could put a polyphonic additive synthesiser together..

My FPGA experiments are in their infancy - I have managed to produce a square wave harmonic series.. But to use these there is a LOAD of analogue required.. Each of these harmonics needs a good filter and a means to control its amplitude.. I have simplified it a bit by driving an analogue switch and using this to switch the level CV - so I am effectively controlling the amplitude of the square wave, rather than having a VCA per harmonic.. The filter incorporates this switch, so I get a sine out with DC controllable level - Its quite simple (a lot simpler than having a HF VCA per harmonic) but still involves 8 analogue switch IC's for 24 harmonics,and an inductor and capacitors and resistors per harmonic, and then the summing amplifier and other stuff... All I have done is breadboarded the stuff for 3 harmonics just to test the concept... The above is just for production of one waveform - if one wants to morph waveforms, all the analogue stuff would need to be duplicated..

If it was possible to have the FPGA do the majority of the above, even if one needed a FPGA per waveform, that would bring the total complexity and cost down hugely. Alas, I am so way out of my depth when it comes to anything like you are describing, that I think I am probably beyond any hope of learning it and doing it in my lifetime... I think one gets into a niche and can get really good - get a real "feel" for ones area of expertise - and it becomes easy and one can see things and innovate in ways others can only dream of..... But this takes years.. At some point (and I am beyond it) its too late to become an expert in another discipline - one may be able to "engage" with that discipline, but thats it...

And digital synthesis, DSP and all that stuff is out of my reach - I need to leave that to those like you who have a real expertise and "feel" for it... Oh, I know what you will say ;-) .. The same thing I say to those who dont have a "feel" for analogue.. "Its easy when you get into it.. " etc .... I find it hard to understand people having difficulty with what I find so easy - Its easy to forget the decades it took me to get where I am.

Fred.

dewster: "Sine wave generation is super easy to do digitally with two multiplications, one add, one subtract, and one delay".

moreover,  one of multiplications can be excluded if the amplitude symmetry between sine and сos is not required.