Korg MS-20 synthesizer + theremin

Posted: 5/28/2013 4:57:40 AM
Joe Max

From: Oakland, California

Joined: 1/2/2009

I just got a vintage Korg MS-20 modular synthesizer, a piece I've been wanting for a long time, because I've long suspected it would be the perfect synth to use with a theremin. It even looks like the kind of machine a thereminst should have:


I used to sell these beasts back in the early 1980s, when I worked at a pro keyboard dealer (Don Wehr's Music City in San Francisco). Once upon a time you could find these for a fairly cheap, but now they're sought-after collector's items - very popular with house and dub-step artists. It's monophonic, like a theremin. But I didn't play theremin back then, and since I took up the instrument, I've never had access to an MS-20 until now.

The reason I suspected it would be perfect with a theremin is that row of six knobs along the bottom right side of the control panel. That's the "External Signal Processor" (ESP) module, a pitch-to-voltage converter and envelope follower. What it does is convert the incoming pitch of a sound source, and the volume of the source, into control voltages that can control the pitch, volume and/or timbre of the synth.

Of course, the top of the line Etherwave and the Theremax have direct CV outputs built in for this purpose, to directly drive the control voltage inputs of a modular synth. I've not tried it with the Etherwave, but I spent an afternoon playing with a Theremax and a Roland SH-09 synthesizer and I could not get the theremin signal and the synth signal to stay in tune with each other over the whole playable range.

So I had high hopes that the MS-20's tracking would work better. For one thing, the old Korg synthesizers, like Yamaha CS series, use a linear frequency-to-voltage scaling system: to raise the pitch an octave, double the control voltage (just like an octave is a doubling of the frequency.) Moog, ARP, Roland and Sequential all use the logarithmic 1 volt-per-octave scale (to raise the pitch an octave, add 1 volt.) The Etherwave and Theremax both also use this same logarithmic scaling system. The point is that frequency-to-voltage scaling is far more stable and the circuits are much simpler and more reliable.

So today I finally got to test my suspicion, and the results are fantastic! I set up the "basic" patch from the owner's manual and it just... worked. Almost no twiddling around with the scaling adjustments that drove me crazy with the Theremax-Roland combination. Just one cable from the theremin, two patch cords in the panel, and the whole palette of the MS-20's sounds was mine.

The MS-20 oscillators track the pitch of my Burns B3 Deluxe theremin over 5 octaves with only a minor deviation in pitch (less than 5 cents at the top of the scale.) The volume tracking is somewhat less exact, there's a bit of a lag compared to the theremin's own output, but it's not hard to deal with.

Now, the MS-20 I have is an original, but Korg has just released a "retro" version of the MS-20 called the "Mini", which is the same original circuitry, but it's 86% smaller in size than the old machine. But most importantly it's got the original ESP module in it. (And as a bonus, the new version has MIDI and USB built in too.) If you've ever had a desire to play a synthesizer with a theremin, this is the machine for you!



Posted: 5/28/2013 11:28:53 AM

From: Eastleigh, Hampshire, U.K. ................................... Fred Mundell. ................................... Electronics Engineer. (Primarily Analogue) .. CV Synths 1974-1980 .. Theremin developer 2007 to present .. soon to be Developing / Trading as WaveCrafter.com . ...................................

Joined: 12/7/2007

Hi Joe,

Interesting! - One question I have is about the MS-20 pitch-to-voltage converter - You say this tracks 5 octaves with only a 5 cent error - thats really good! ... What I wonder is how low the audio input frequency can go and tracking still be maintained ?

Also, at the lower audio frequencies (input) is there any noticable latency (time lag) between the theremins audio and the synths audio ?

Voltage control is something I spent a lot of time on, and is the direction I have taken with my more recent "hopefully destined for production" theremins - I have opted for a front-end which outputs a linear voltage-distance, and voltage controlled heterodyning theremin sound modules.. All 1V/octave..

So your playing with a V/Hz synth and getting good results is extremely interesting to me.


Posted: 5/29/2013 3:29:28 AM
Joe Max

From: Oakland, California

Joined: 1/2/2009

Hi Fred,

I haven't been posting for a while and I'm glad to see you're still here and carrying on with your project!

The MS-20 tracks the Burns signal steadily on the high pitch side until the tone really gets into "dog whistle" range (about an inch from the antenna on the Burns) and then it just stops. I think it's reached the tuning limit of the oscillator's voltage range, so I can't do the "bird tweets" and "laser shots". At the very low end I can hear a bit of fluctuation, like faint "motorboating" of the sound, but the straight B3 sound does that too. The tracking is surprisingly solid through most of the musical range of the B3.

I suspect there are two things that lend the F-to-V converter to work so well: first, the B3 produces a pretty clean triangle wave with few overtones, which makes it easy for the converter to find the fundamental pitch. Second, the continuous nature of the theremin sound lends itself to easier tracking. On most instruments, the note commences suddenly and the F-to-V has to "find" the note, and there's always some lag time. For guitar players (the ones who probably used the MS-20 the most for this in the past) the picking of the note is very well-defined (a sudden start) and the guitarist can tell there's a tiny bit of lag before the corresponding note comes out of the synth. By nature, the notes commence on the theremin as a "fade in", no matter how fast you pull away from the volume antenna to play staccato, there's still a bit of fade-in time. So the tiny bit of flutter as it searches for the pitch is hardly noticeable, as the sound is still "fading in".

What I was really concerned about was, could the P-to-V circuit track a continuously portamento-ing pitch without lagging, and the answer is yes.

The old analog Korgs have a reputation for excellent tuning stability, and it's really due to the V/Hz scaling. As it was explained to me, ALL oscillator circuits start out as V/Hz, and then the logarithmic converters are designed into it to scale it to 1v/Oct. It's those converter circuits that are so sensitive to temperature changes and "drift". I leave the MS-20 on overnight and the oscillators are still right on pitch in the morning. Or turn it off for hours and when it's turned on, it's still in tune. Try that with a 1v/Oct analog synth!

This MS-20 came to me from a friend-of-a-friend who had it gathering dust in a closet for well over ten years. Once I confirmed that it worked - powered up, all the controls were functional, - the first thing I did is download the service manual and recalibrate it. I've repaired and calibrated lots of 1v/Oct synths (I once owned a Mini-Moog and a Sequential Pro-One) and it is a bitch of a process, lots of going back and forth to tweak the high and low ends of the scaling to get the intervals to be right. And you've got to do it on a regular basis to keep the synth scaled and tuned.

With the MS-20, I hooked up the tuner and played the low note. Ten years in a closet and the pitch was right on the money. I checked the high note, and it was about 7 cents flat. I tweaked the "high pitch" set screw until it was right, and expected to go back and tweak the low again, then the high again, over and over until it was correct, like I had to do with my Mini-Moog. But no! The low end was still right on pitch. It couldn't have been any easier. Then I realized that unlike the Moog, there is no "scaling" adjustment. It doesn't need one! It's straight-up V/Hz.

The biggest tweak I had to do was setting the self-oscillation pitch of the filters. With no CV into it, the filters are supposed to resonate at 500Hz. They both were high, a bit over 600Hz. So I tweaked them to within -/+ 2 Hz (it was harder to get it perfect with the set-screw adjustment).

Besides that, I opened up the case and sucked out the dust bunnies, and gave all the pots a bath in deoxidizer spray. MS-20, ready to go!

So far as I know, the only true V/Hz scaled synths on the market today are the new Korg MS-20 Mini, and the PAIA Fatman. Korg used to make a V/Hz-to-1v/Oct converter (the MS-02) to use their machines with the rest of the world, but those are extremely rare and almost impossible to find nowadays.

However, both the MS-20 and the Fatman instruments are in current production and easily available. You can buy a Fatman in kit form, with a nice desktop case, for less than US$300. The MS-20 Mini is a quite a bit more, around US$800 retail, but I'll tell you, I'm having a blast with mine. It's got me really excited about practicing my theremin more often. The tracking is so good that I haven't had to adjust my aerial fingerings much at all. (Mostly with the volume control, it's less responsive than the Burns output, but it's not that bad.)

I'll see if I can record a video demonstrating the combination and what can be done with it (give me some time to practice with them, but soon.) If you've got any questions or want me to try something, let me know!

Posted: 5/29/2013 4:50:12 AM
Joe Max

From: Oakland, California

Joined: 1/2/2009

I'm noticing now that there is a bit of lag as the pitch glides up and down, but it's very minor and only obvious when playing the B3's sound and the synth sound together. In fact, it's kind of musically nice - it puts a bit of "chorus" on the combined sound and makes it fatter. When I stop on a pitch it's fine.

And I've found a way to play a wicked sounding plucked bass with it!

Posted: 5/29/2013 9:45:25 PM

From: Eastleigh, Hampshire, U.K. ................................... Fred Mundell. ................................... Electronics Engineer. (Primarily Analogue) .. CV Synths 1974-1980 .. Theremin developer 2007 to present .. soon to be Developing / Trading as WaveCrafter.com . ...................................

Joined: 12/7/2007

[Warning to those wanting a light non-technical read.. This post aint that! ;-) ]

Hi again Joe,

Yeah - V/Hz is intrinsically less prone to drift, but at a high technical cost IMO.. Oscillator / filter tracking / scaling being based on multiplication (which is a complex function to achieve) rather than much simpler addition.. All fine if one has only one primary CV and dont want anything too fancy in terms of modulation - but try mixing the output from an analogue sequencer with the keyboard CV and driving the synth, and the problem becomes obvious!

Back when I was about 12 (A LONG time ago ;-) my first synth I built was (although I would not have known the terminology) V/Hz, and my attempts to get other additional 'modules' to sync with it drove me nuts.. Then I got the details of Bob's V/Octave system and it all clicked..

There is no reason why exponential or log converters should drift.. You just need to keep them at a constant temperature (I really dont get on with matched transistor + Tempco resistor designs - they are never good enough IMO ).. Just use the matched transistors in a transistor array IC, and use the other transistors as temperature sensor and heater - then run the whole thing in a precise closed loop with the substrate regulating at 70C, wrap the assembly in polyurithane foam encapsulation for thermal insulation (to reduce the power consumed maintaining the temperature) and all drift problems are gone.. The humble CA3083 is good enough for about 6 octaves, and the THAT 300 series arrays are good for 10 octaves or more, but 6* the price.

I do agree with your comments regarding tuning vintage V/Octave synths.. an absolutely frustrating pain of a job.. The ideal exponential converter requires just an 18mV/Octave into the base of the conversion transistor, and will double its collector current for each 18mV.. If one has good transistors with large dynamic range, and if the temperature is fixed so your base voltage does not need to be corrected using a specific and expensive PTC thermally coupled to the transistor/s, then trimming is just down to dropping the 1V/octave to the required 18mV and trimming this for precise conversion.. Then there may be one extra trimmer for HF compensation on VCO's.

So, back to the MS20.. I have a confession - I have never understood how the external signal processor behaves in any reasonable way.. Looking at the circuit fills me with horror! .. But, it does seem to work, and work well.. And yes, what you say about theremin signals (particularly if taken from before the theremins VCA) being easier to track than many other signals is true.

"I'm glad to see you're still here and carrying on with your project!" - Thanks ;-) LOL .. Not really sure any more what my "project" is though..

My present "plans" are to put my prototypes into production - or at least into demonstrable form so that someone else might put it into production... Your CV based discussion is particularly relevant to my "direction"..

I spent quite a lot of time, and had high hopes for the "Lev theremin clone" stuff I got into (via the "TW Theremin" thread) - But the result of this (and some dissapointments with not being able to get it to achieve what I had hoped) led me to look at the whole theremin topology again..

I have been flip-flopping between voltage control and 'direct from antenna' tone generation - my basic idea being to have a fairly standard heterodyning front-end with standard mixer producing classic theremin audio output (call this voice 1)..

In addition to the "direct analogue heterodyne audio" the Reference and Variable oscillators from "voice 1" were squared and multiplied using PLL's, so the frequency of each was 16x their original frequency - these were then "mixed" (using a D F-F) to produce a clean logic level difference frequency 16* the audio frequency from voice 1.

This multiplied frequency was used to produce a V/Hz output voltage which could track voice 1's frequency rapidly - Voice 1 @ 16Hz produced 256Hz, and the CV is updated every 2 cycles so even at 16Hz, there is only 7.8ms latency (as opposed to 62ms if one determined the CV from a single cycle of the audio from voice 1, or 125ms if you required 2 cycles). The V/Hz output was converted to V/Octave, but both are available.

I then developed voltage controlled theremins - Full theremins except that there are no antennas - antennas are replaced with a means of remotely changing the oscillator frequency.. and a closed analogue loop so that the resultant audio output matched the incoming CV.

My plan was to have a really good linear theremin at the front, one where the linearity and "span" could be adjusted, and where register switching could be implemented... And my hope was that the Lev designs would provide such a "front end".

It didnt.

And I then realized that I was missing the damn obvious..

The "problem" with theremin design, the thing which makes is so tricky, is that one is integrating the controller and the sound source as a single entity.. Any "trim" on the controller will (or can) affect the sound generator, and trim or change on the sound generator, and you can affect the controller... Typical example of this is oscillator synchronisation - it changes the sound, but also changes the linearity... Its a bit like the Melotron - a sample player entirely dependant on the keyboard mechanics for its operation.

I have now opted for a 1V/Octave system - The controller (antenna's / front-end) can be clearly defined and seperateld designed / implemented.. You need a voltage from this front-end / controller which is linearly proportional to distance - NOT musically proportional.. At the antenna you want say 10V, 10cm from it you want 9V, and every further 10cm reduces the voltage by 1V until you get to 0V..

Now, you can attenuate the above output by whatever factor you want - at "full gain" you get 6 octaves over 60cm.. attenuate to 50% and you get 1 octave for every 20cm, giving 3 octaves span over 60cm.. so making attenuation from 50% to 100% gives adjustement of span from 3 octaves to 6 octaves..

And you can add or subtract whatever voltage you want to the above - add one volt, you increase the frequency of whatever is connected to the CV by one octave, subtract 1V and you drop 1 octave..

Unlike a conventional theremin or V/Hz CV system, you do not need to mess about to obtain "musical linearity" - Musical linearity (an exponential control law) has real problems because if you change anything by any means other than multiplication or division, the linearity will change.. With V/Octave, multiplication or division affects the 'span' and one can add any tuning voltage one likes - so if you have an actual linear voltage relationship between distance and voltage, you can maintain a musical relationship with any exponential (V/Octave) modules driven from it.

Having a front-end like this, one is free to do whatever you like WRT the "voices" - You can use standard CV syth modules - but I have designed an interface circuit which accepts a 1V/Octave CV and drives any conventional theremin from this..

Its not a simple circuit! ;-) .. Each of the HF oscillators is devided and processed with an analogue circuit that "looks ahead" and determines the difference frequency far faster than by measuring the audio output - This frequency is converted to a voltage which is differentialy compared (after exponentiation) to the incoming CV, and the error signal is used to adjust the theremins variable oscillator frequency - its a sort of PLL - extremely fast, but not actually phase-locking (there is no "reference" frequency to lock against, only a CV to match)..

The massive and surprising audio benefit of the above is with regard to oscillator "pulling" - Conventional theremins go into 'lock' as the oscillators approach the same frequency, and in order to allow bass to be produced, the oscillators cannot be coupled together too much.. With the closed loop topology of my CV controlled theremins, one can tightly couple the oscllators, and get the frequency dependent harmonic influences this imparts without having to take the extreme care usually required to balance sonic advantages against the problems tight coupling can bring. (I suspect that the Lev Theremins somehow acomplish better syncing and that this is one of the things that makes them sound great)

And one can implement register switching simply by adding or subtracting a volt per octave.. A true analogue theremin with register switching and as many voices as one can afford.. each detunable against the other.. ;-)

Easy to do V/Hz though if anyone wants to drive a Korg.. Just tap the V/Hz output from one of the voices before its exponentiated..

Anyway .. thats where I am with "carrying on with your project" right now - Laying out multi-layer PCB's at the moment - theres so much in the damn prototype which is redundant due to multiple changes, and its such a mess, that I need to get it all onto PCB.. This particular design probably needs to go to SMD to make it economically viable.

This is, however, my last attempt at getting somewhere with the theremin - This one covers all my targets - It gives me a linear capacitive long-range sensor for use in other applications, and it gives voltage controlled theremin modules for the synth market, and provides the means to produce a combined composite pro theremin for this community..


 (and yes - I do want to go out and buy a MS20!  ;-) ... But I will spend any money I can scrape together on PCB's for my monster instead.. I am kind of hoping to have it in time for HO-2013 in a couple of months, but im not going to give myself another heart attack over it - going to take it slow.

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