Moog Theremini!

Peter wrote:

"I can usually adjust my playing after a couple of minutes to accommodate the characteristics of an instrument no matter how bizarre it may be. There is one thing, however, about the THEREMINI that I cannot compensate for and cannot adjust to - its lack of response."

Indeed, with well-developed hand/ear coordination, one can adjust to different theremins.  However, it takes very little latency to throw everything off.  Only a few milliseconds is all it takes.

I suspect that the theremini may utilize a table-lookup to accomplish its calibration -- similar to how the Little Phatty works.  In the case of the LP, the calibration procedure populates a table of control voltages that, in turn, drive a VCO that may or may-not be running at precisely 1v/8ve.

In the case of the Theremini, they would have to match MIDI notes to voltages derived from the pitch circuitry.  Of course, this approach is doomed from the standpoint of latency.   Now, I don't know if this is how the Theremini works -- but the calibration procedure suggests something along those lines.

I had considered something like this sometime ago: namely that the frequencies from the variable oscillator could be mapped (converted) to output control-voltages such that a linear response (1v/8ve) would result.  However, for me, this was a mental exercise in that I've never done any DIY microcontroller stuff and, well, the theremin that I have (an Epro) works fine as is.  I suppose with a fast enough microcontroller that, in turn, feeds ANALOG (no noticeable latency) circuitry one might accomplish the deed.

So, can it be done?  Can one create a playable theremin that can quantize pitch while allowing vibrato that has no (noticeable) latency?  Well, yes -- it is possible.  However, the tone would come from a VCO and NOT from the theremin's detector circuit.  However, in my experiments, I have found that the theremin's response and shading define the "theremin-sound" much more than the actual waveform.  

Alas, my equipment is packed up at the moment however when I finally am set up again, I will do a video demo of this (most likely a winter project, though).

Hah, got one better! Make sure a Tesla coil is not in operation on the floor above during your demos at a Maker's Faire!!! (Orlando 2013).

I'd be playing and the sucker would suddenly go from sharp to flat.

FEELEEP

"Indeed, with well-developed hand/ear coordination, one can adjust to different theremins.  However, it takes very little latency to throw everything off.  Only a few milliseconds is all it takes."  - kkissinger

If my recent BW measurements are to be believed, Theremini pitch latency is on the order of 100ms, which is amazingly bad for a continuous instrument that relies so heavily on the human in the feedback loop.

Dewster has done some real comprehensive measurements of the response time, and pasted these in all their glory!  - they can be seen here.

From these I extract the following (my highlighting):

" 0.15sec full settling time, or around 0.1s to hit 63%.  This RC time constant corresponds to a cutoff frequency of 1/2*pi*RC = 1.6Hz.  Wow." - Dewster

-----------------------------------------------------------------------------

"However, in my experiments, I have found that the theremin's response and shading define the "theremin-sound" much more than the actual waveform.  " - kkissinger

I think that this is the missing salient fact has probably cost many theremin developers a load of time - and am coming to believe that the single most important aspect of the sound is the way that the actual waveform can be subtly modulated by the thereminist.

Fred.

For me, the theremini has highlighted the dilemma facing a developer.. The true theremin does not have any instant appeal to non musical people, in order to appeal to these you probably need to make a  toy.. They dont care about 100ms latency, they dont care bout nuffin except what it looks like, that it makes synthy sounds, and that they can actually get 'musical notes' out of it even when they have no idea about what they are doing..

And the theremin, on the other hand, requires more skill and co-ordination, and better pitch awareness, than most musical instruments.

We have been in this illusory state - talking in terms of "globalization" and "theremins for the masses" - subliminally, we probably bought into the RCA LIE that "if you can hum a tune, you can play a theremin" - Its NOT TRUE! - it never has been true, and it never will be true!

An instrument that everyone can play, regardless of their musical ability or sense of pitch - such an instrument will not ever be a theremin, IMO!

100 ms of latency!?  If it is true then the Theremini absolutely can not be called a theremin.  I had strong doubts that Wilco's estimation of 15ms was accurate. 15 ms of latency is generally acceptable for most people that play keyboard software instruments.  

I've plugged my Epro into software effects in ableton Live and even with 20ms of total input/output latency, I was able to play and hear myself okay. 

I still haven't had a chance to sit down with the Theremini, but when I finally do, I will test the latency using my own method.

If anyone from Moog Music is reading this... Hello! This is absolutely unacceptable! There are many things that I would let slide about the Theremini, but this kind of latency... This is a TOTAL embarassment.  If it is not fixed, the Theremini project should be completely scrapped.  At the very least the marketing copy should reflect exactly what the customer is buying... A toy. It is not a theremin, not with 100+ms of latency!

I apologize, I should qualify what I've inexactly referred to as latency as the settling time response to a step input.  I should know better as this issue is analogous to transport delay vs. inertial delay in electronics, and I imagine this is more of an inertial issue (i.e. limited bandwidth, whereas something like MIDI latency is a transport delay and rather unrelated to bandwidth).  What you experience playing it will likely be less, as playing steps are generally small and lower bandwidth (e.g. ramps).  That said, the bandwidth itself is likely much too low for those used to instant response from a heterodyning analog Theremin, and you might notice things like fast vibrato getting somewhat damped out.  I understand why digital Theremin designers end up utilizing very low bandwidth as it is something of a cure-all, but it tends to throw out the baby with the bathwater.

There may be people who play the Theremin in very low bandwidth ways that might find the Theremini response time not too onerous?  I probably find smallish size of the pitch field, its non-linearity, the noise pickup issues, and the fiddly calibration more troublesome, but then again I don't really play.  But I don't consider <10Hz pitch side gesture bandwidth to be acceptable in a "serious" instrument, and I think modern players should be aware of the potential (but not inherent) limitations of more modern non-heterodyning approaches, particularly if not designed well.

Thanks for the clarification, Dewster.  I hope to provide some kind of feedback for players regarding the latency as soon as I can get time with the instrument.    I'm not entirely convinced it's a complete lost cause (not until I have a go at it for myself). 

In a parallel universe Moog Music engineers are working right along side proficient theremin players to create the next professional Moog Theremin. The theremini does not exist in that reality. 

"I apologize, I should qualify what I've inexactly referred to as latency as the settling time response to a step input." - Dewster

I dont think in the context of a musical instrument, there is really anything "inexact" about the term "latency" being used in this way..

Its like having a permanent 100ms portamento on a keyboard - a portamento which you cannot overcome with pitch or mod wheels - set the modulation to any rate, and it will be slurred by this portamento, bend the pitch with the wheel and it will be slurred...

Even a keyboard with this latency would be unplayable! - Add to that pitch quantization, and on a keyboard you would have consistent timing errors of 100ms!

There is no way to water this down - there's no way that this response is acceptable for any modern musical instrument, and there is particularly no way that the most free and expressive (in terms of control) instrument ever developed could excuse this kind of mutilation!

Whoever developed or caused this monstrosity to get to market should join the ranks of the unemployed or be demoted to some function like cleaning the toilets or packing boxes IMO.. But they wont be - they will probably be promoted because the public is buying this rubbish!

If anyone is responsible for making a big deal of the technical IDIOCY of this toy, its me - if anyone needs to apologize, its me - But I dont believe there is any justification in any apology from you or I .. We have merely disclosed a SCAM!

Look at it - Not only unforgivable errors, but also DELIBERATE deception! - "same as an etherwave" "using heterodyning". BAH! --- And those in any parallel world better wake up and realize that whatever they develop will be tarnished by the shit their company has dumped on the market!

Fred.

"Its like having a permanent 100ms portamento on a keyboard"  - FredM

That's a very good analogy!  What I believe would be worse is something like a 100ms MIDI delay, which might make even slow gestures difficult to control due to the delay in the feedback loop. I wanted people to be aware of the difference.  (A reference I like to keep in my head regarding pure delay is the speed of sound in air, which is ~1' per millisecond.  So when you are playing something like an acoustic piano, the sound is hitting your ears with a ~2ms delay.)

The Theremini acts like a low pass filter and probably not so much as a delay.  An ideal low pass filter passes frequencies (rates of change) below cutoff pretty much intact and without delay, and above cutoff attenuates them to the degree that they are above the cutoff frequency.  So if you play a slow to moderately fast frequency sweep on Theremini you probably won't notice anything wrong, but if you try to change notes quickly the ramp response to your step input may make it difficult to hit the note if you are using audio feedback as your guide.

OK, one more audio file:

http://www.mediafire.com/listen/0d36y9lrtm3zhat/Theremini_touch_response_2014-09-23.mp3

For the above, instead of using the FM interference method with square wave modulation, I calibrated the Theremini close pitch position with my finger touching the pitch antenna, and the far position just 1" away.  After calibration the Theremini seemed to "know" when I was touching the antenna and muted the volume when I was doing so.  To circumvent this I held a piece of paper next to the antenna, which prevented the muting. 

I tried to make my finger act like a square wave, bringing it as quickly as I could in contact with the antenna (& paper), holding for 1/2 sec, then away as quickly as I could, holding for 1/2 sec, to give a roughly 1Hz rate.  If (and only if) the rate of change of the finger movement is significantly above the filtering cutoff frequency, then this method can give a sense of the LP filtering of the response going on inside the Theremini.  Results are essentially the same as I got with the FM method (which is a relief). 

Not recorded here, when I brought my finger to the paper and away again as fast as I could in one quick motion, the Theremini barely registered it (barely changed pitch).  In contrast, a heterodyning analog Theremin will squeal like a pig no matter how quickly one touches and withdraws from the pitch antenna.

These types of extreme changes don't necessarily happen during play, but sluggish response could likely aggravate things even if the player isn't doing quick things like this that boldly display the very low pass nature of the Theremini pitch response.  As they say in the south: "Caveat emptor y'all".

[EDIT] I suppose there is the chance that a small amount of portamento is being "musically" applied on the Animoog side of things?  Without access to the synth engine I can't check this, though I quickly tried some of the other voice presets (with pitch correction and delay dialed off) and they subjectively seem to have the same sluggish step response.

Regarding latency -- the latency is the sum of the theremin's latency, digital processing latency, and the delay from the monitor to the thereminist.  An overall latency more than 8-10 ms is enough to cause problems and at 12-15ms a theremin is utterly and completely unplayable.

The problem is due to 'overshoot'.

Consider an example where I am performing an ascending jump to a note (say 'C' above middle-C).  When I reach the correct pitch, I will stop moving my hand closer to the rod.  However, with latency, I will hear the correct pitch, stop my hand, and the pitch will continue to move up!  In fact, with vibrato the output may be the reverse of what one plays in that the high part of the vibrato may be heard when the hand is at the low part.   How in the world can the performer adjust to pitch in such a topsy-turvy world?

A freeform/avant garde performer will experience the same problems -- namely that what is heard will not match the player's actions.

Pitch-correction cannot help this, either.  A pitch will come out but it may differ from the performer's target pitch.

Besides, pitch, rhythm suffers.  When looping and attempting to overlay parts, the latency will accumulate and it will be impossible to loop-record anything.

Tone quality can be adjusted with filters, wave-shapers, and EQ.  Minor linearity problems can be dealt with by the performer. Latency is a show-stopper.  Latency is the one issue that cannot be overcome.