Theremin robotics

Posted: 10/16/2011 1:35:40 PM

From: Denmark

Joined: 8/1/2010

After I had to take my Theremin back because of issues described elsewhere, I decided to give it another go because I really would like to play the Theremin.

I find it very difficult to stand still enough to get a good pitch, and even though I have come far, I still have a very long way to go.

It is so much easier to just pick an instrument I can play freely on. Now I mostly use the Theremin as ear training which it does way better than the piano.

But today, as I was playing with reprogramming an old Sony Aibo I thought "Why not build a Theremin-playing robot?"

I have found two videos of toy robots playing Theremin, and it sounds terrible. Has anyone ever tried building a dedicated Theremin-playing robot, midi-controlled or autonomous, with volume-control?



Posted: 10/17/2011 7:20:35 AM

From: Canada

Joined: 8/1/2008

Jason, the founder of Theremin World, built a theremin-playing robot back in the mid 1990's. One of many challenges that builders of these kinds of devices face is that the EM field of a theremin is not stable. It is constantly shifting and changing depending on everything from temperature and humidity to how long the instrument has been turned on. Human players adapt to this naturally and can adjust their instruments (and themselves) to accommodate these shifts.

In order to be successful at playing the theremin, a robot would have to be equipped with pitch sensitivity and an ability to adapt to the unpredictable and constantly changing configuration of the instrument, adjusting the intervals as it plays, just as a human player would. If the robot's actions were mechanically fixed, as they would be for a piano, it wouldn't work.

I have seen a number of theremin-playing robots but they have all been cheap plastic gizmos that looked like they were bought on sale in the toy department at Walmart. What we need is Mr. Data's "positronic" brain! And even Dr. Soong was "often wrong".

Posted: 10/17/2011 7:41:55 AM

From: Denmark

Joined: 8/1/2010

I have some experience with industrial robotics, and in that field feedback mechanisms are quite common and works well. I would expect the robot to establish the pitch exactly like I would do myself. Even to use pitch preview if available.

Or, is it possible to not use moveable parts at all, and just use fields to interfere with the Theremin fields? Then two antennas, pointing towards volume and pitch antennas, could create fields of varying properties.

Could be fun to try, and much cheaper to build as no moving parts would be required.


Posted: 10/17/2011 7:54:46 AM

From: Colmar, France

Joined: 12/31/2007

Coalport is right. Perhaps not about Mr. Data and his positronic brain but a theremin robot needs to have at least a "classical" cybernetic pitch feedback circuit:

a) Initialization: The pitch arm goes through the whole pitch field while a frequency detector is connected to the pitch preview output of the theremin in order to get a first virtual map of the pitch field.

b) Choosing a tone: The control logic looks up in the map and puts the pitch arm rapidly into the computed position.

c) Optimizing pitch: The control logic checks the audio frequency, corrects the arm's position and eventually the virtual map in order to optimize further positioning.

d) Repeat c) while the desired tone should be kept, else go to b)

And then nice effects would be possible, for example lowering the volume by 8 to 10dB while pitch is outside the desired tone +/-50cent, and only by 3 to 4 dB while between the +/- 20cent and the +/-50cent corridors, thus simulating the dip with the left hand while moving from one tone to another if desired.

Oh, if I only had the time and the resources...
Posted: 10/17/2011 12:34:45 PM
RS Theremin

From: 60 mi. N of San Diego CA

Joined: 2/15/2005

Vespandy said:

[i]Or, is it possible to not use moveable parts at all, and just use fields to interfere with the Theremin fields?[/i]

You mean the appearance of a physical robot playing a theremin for amusement or the internal generation of robotic theremin style music with the precision of a Nickelodeon Player-Piano?

I use to experiment with automatic theremin pitch tuning but found a lot more than that with my research. I would use calibrated emitter currents through the transistor, not manipulate the magnetic pitch field directly, of the heterodyne pitch oscillator. This is similar to the way potentiometer pitch tuning is often done. On my perfectly linear theremin designs this allowed for the control of selecting the correct notes and created a soulless theremin that attempted to communicate with me!@#%^

If you open this sample below in Audacity you would see it is a single continuous analog theremin tone, though it may sound otherwise due to the speed robotics can control note jumps... buuuut this approach I found unnatural, more like gimmickry for circuit benders!

Robotic Precision ( .mp3 137k

This sound byte does not have the spirit of a true theremin, though it is RF heterodyned, and does not have the proper sound of "her voice". To some it may be pleasant or just a hypnotic flute/whistle!

Between each series of notes you hear the unmuted theremin hunt or center on G4 - 392Hz which almost sounds like a slow vibrato. Once this marker is locked on, it no longer hunts, all of the note jumps in a linear theremin pitch field (PWM Matrix) are relative to this reference note. This avoids any latency with musical notes selected in a millisecond, leaving plenty of time for auto volume, analog note sliding or the perfect "illusion" of a synchronized physical robotic movement.

This note centering is similar to you setting a specific musical note 16" from your pitch field antenna during setup, then checking it again between music pieces.

As a non-musician I find the relationship of musical notes to be an interesting phenomenon between human nature and our perception of what is musical.

I think this is where the real problem rests with robotic theremin playing, especially for me... evolving as a purest; a true theremin voice needs the intimacy of a skilled human to give her voice life or it misses the point, playing the theremin is a symbiotic relationship as my theremin designs are alive!

Oh-oh let's not go there again. ( ‘

Edit: I abandoned auto pitch tuning because simple tube/valve pitch oscillators properly designed are virtually "drift free" after warmup and have that organic sound! The characteristic in a transistor that makes it unstable and allows for the broad range of robotic theremin "note picking" does not exist in it’s cousin the steadfast vacuum tube/valve theremin!


Posted: 10/18/2011 11:44:38 AM

From: Hillsborough, NC (USA)

Joined: 2/13/2005

Here's a short article from the archives about the robot we built at the Intelligent Robotics Laboratory at Vanderbilt back in the day.

We used an FFT algorithm to listen to the audio coming in from the theremin and adjusted real-time. You could either sing to it and it would match your pitch on the theremin, or you could play it with a MIDI keyboard and it would track your notes. The particular robotic arm we used was constructed of rubber actuators which had a natural "bounce" to them, so adding vibrato was also pretty easy.

The nice thing about a real-time feedback system was that if someone entered the control field of the theremin and changed the response characteristics, the robot could adjust in real-time and stay in tune.

That was a LONG time ago... feels like a lifetime.

You must be logged in to post a reply. Please log in or register for a new account.