TherAsynth - New product under development - What do YOU want?

Hi folks - I 'let the cat out of the bag' in "Roll Call 2008" so will elaborate a bit here on the product I am developing.
I will give a brief overview of my project, and would REALLY like feedback about ANYTHING YOU would want to see implemented, or anything you think is a bad idea..

PART 1:

TherAsynth has:

1.) Two ACTIVE "antennas" - these are housed in soft-sided plastic boxes which are attached to brass antenna poles. The poles, and antenna boxes (henceforth called sensors) can be swiveled and tilted, and have glands which can be hand tightened to lock their positions.
The conventional (Vertical pitch, horizontal amplitude) and other antenna positioning can be easily set - for example, both antennas can be placed horizontal or vertically , and the unit played while seated at a table, or the unit can be hung around ones neck, and the sensors placed just above the knee (when player is standing) so that one can perform whilst providing dramatics (Lead Guitarists beware!! :)
The sensors are directional, and can be swiveled to any angle, so effects of stray / background capacitances are reduced - Also, the sensors are continuously self-calibrating, so that the full selected capacitive sensing range is maintained.

QUESTION: A switch to swap sensors (left<-right)? or does pitch on the right / level on the left suit everyone ? (I will offer a left-handed version if anyone wants this - but is an extra switch a good idea, or would it be a bother?)

2)Sensor update speed: The sensors have been the most complex part of this design - and the only part which uses any digital circuitry.. The position -> capacitance relationship is non linear, but one wants a linear position -> musical interval relationship.. Theremins (even the best) only achieve reasonably close to this
over a few octaves (from what I have seen, the Moog Pro is the best at this relationship, which is one of the things which makes it such a superb instrument).
Each sensor has a PSoC (Programmable System On Chip) Processor which is working flat-out to compute the correct output voltage for the capacitance being detected - The maths is quite horrendous (Floating point with extremely wide range of input values) - it must accurately convert values ranging from <255 to >65000 and output the result to a 16 bit DAC (Output voltage is from 0 to 10V, with resolution of 400uV [0.0004V] to give pitch tracking to 0.5 cent.
The above takes time - I have it running with an update rate of 10ms at present, and am hoping to achieve 5ms - Steps are integrated (smoothed) but this is the only aspect which worries me about my design..

Anyone have any ideas about whether a 10ms (100 times a second) update might cause a problem?

TherAsynth Part 2:

3: Sound Generation:
The inspiration for TherAsynth came when I was playing the lead to Emerson Lake + Palmers "Lucky Man" (The superb Moog lead) on a crude Theremin, and wanted to have a 'fat' oscillator and control of a VCF - yep - it all started there..
So my primary focus was to make a Mini Moog driven by an antenna system.
Since then, having played with and built several more sophisticated Theremins than I had used before, I realised that I also needed to keep the unique Theremin hetrodyning sound in addition to analogue synthesis..
I have noticed that on some Theremin designs, the HARMONICS of the waveform changes slightly as the pitch changes - I suspect that this is to do with the reference and pitch oscillators interacting with each other (crudely 'breadboarded' theremins seem to exhibit this 'distortion' more profoundly)

I believe that I could synthesise the same waveforms using voltage controlled modules - but I have also designed a high frequency sine VCO, and a HF sine reference oscillator, and together (through a 4Q modulator)these perform EXACTLY like the hetrodyning theremin sound source (including the ability to put the pitch null-point at any position from the antenna - which might be quite interesting if run simultaneously with a VCO directly tracking the absolute position WRT the antenna)

However.. The above hetrodyning oscillator/s are expensive, and I am not sure if they are needed - I will wait until I have my prototype fully running and then do comparison between them.

Do YOU have any opinion on the above? Would YOU be more inclined to buy the TherAsynth if it had a real hetrodyning sound source than an analogue synthesised Theremin output?

The Synthesis engine is totally analogue, and consists of:

Dual tracking VCO's with common tuning, and detuning of VCO2 WRT VCO1. Output from the pitch antenna drives the VCO's, and this is adjustable in range and polarity. Waveforms and VCO mixing is fully independantly adjustable between square and triangle, allowing a differentiated square+ramp waveform in the centre position. A noise generator is also provided.

Dual VCF's - These are an entirely new design using diodes as the VR elements, and are 12db/8ve and have adjustable 'emphasis' (Q) from 1 to self-oscillation. Switch options allow selection of a 24db/8ve single VCF, or 2 seperate 12db/8ve VCF's acting on different sources. These VCFs sound like a mixture between the Moog ladder VCF and the SSM VCF.

The VCF frequency can be controlled by either or both antennas, with variable +Ve or -Ve signal from either/both.. This allows harmonic content to track pitch and be varied by amplitude antennas

VCA - this is also an original design, and uses diodes as the VR element, has control from 0 to -88db.

Sound sources can be mixed with externally input audio, or sound sources can be turned off, allowing TherAsynrh to act as an effects unit.

One of the things (there are many features not listed above) I forgot to mention was that VCO2 can modulate VCO1 giving complex FM waveforms

A while back, a few of us on TW talked about finding a way to make theremins wearable.

The idea met with some resistance on the grounds that a wearable theremin would be too unstable to play accurately while moving about.

My question regarding this device is: How many octaves will it allow?

Is it truely portamento?

Would it have midi, and or CV outs?

Would it be intended more for ambient effects than a melodic instrument?

I wouldn't mind giving this instrument a test drive.

A while back, a few of us on TW talked about finding a way to make theremins wearable.

The idea met with some resistance on the grounds that a wearable theremin would be too unstable to play accurately while moving about.

>> Each sensor is a rounded soft-sided plastic box 20mm thick, 70mm wide, 115mm long, with a capacitive plate 90mm x 50mm placed inside on one face, and the rest of the inside faces of the box metalised and grounded - This results in a conical sensing field widening from 90mmx50mm outwards, and is therefore directional -
I believe that if the sensitivity is turned down to allow (say) a 3 octave playing range, one should be able to perform with fair accuracy.

My question regarding this device is: How many octaves will it allow?

>> The sensors output a full scale of 10V - If this was taken directly to the synthesis section without attenuation this (on a 1V/octave system) would give 10 octaves linearly over the sensing range.

As it is a voltage controlled system, the 'sensitivity' controls (this issue is more complex - as these are in fact independant controls for the VC elements) adjust the number of octaves spanned over the sensing area.. widening the distance between any given musical interval.

Let us say* that the sensing area extends to 300mm, and one had sensitivity at max, each octave would be seperated by 30mm, giving 10 octaves. If one reduced the sensitivity to span 3 octaves, each octave would be seperated by 100mm.
Tuning is independant of the above - say one spanned 3 octaves, these could be set to any frequency in the audio range.

Is it truely portamento?

>> Yes. The voltage from the sensors is directly proportional to the position of the player, and changes in steps of 0.5 cent (1/200th of a semitone) and is updated every 10ms

Would it have midi, and or CV outs?

>> The sensors outputs are voltages linearly proportional to distance. These voltages (and other signals) will be taken to an expander socket at the back of the instrument.

A MIDI version is in the pipeline - but here I am cautious - MIDI is not well suited (in my opinion) to this application - I could simply take the CV's and convert these to MIDI - but intend instead to examine outputting serial data from the sensors and having another processor deal with message forming/transmission and reciept of incoming MIDI and conversion of this to relevant CV's to drive the synthesis engine.

Would it be intended more for ambient effects than a melodic instrument?

I feel that the market is utterly cluttered with "Theremins" which make noises, but are grim (or impossible) to play musically - Unless I can produce the most musically playable Theremin related instrument ever seen, I will be displeased with myself -
Having said that - There is a full analogue synthesiser in this instrument - so it will certainly be possible to produce "effects" ;)

I wouldn't mind giving this instrument a test drive.

>> Nor would I ! - I hope to have the first complete prototype running in the next few months (a lot depends on what other work comes in - My finances are dire, so on one hand I need work - but on the other, I dont want to slow down on this project.... ):

[i]QUESTION: A switch to swap sensors (left<-right)? or does pitch on the right / level on the left suit everyone ? (I will offer a left-handed version if anyone wants this - but is an extra switch a good idea, or would it be a bother?)[/i]

As a southpaw, I would definitely be interested in a left-handed version. I like the idea of a switch -- It would allow anyone to play my instrument while I owned it, and it would improve the re-sale value should I ever need to sell it. ;)

Add me to the list of those willing to beta-test this baby out.

Thanks for the questions and comments Thomas + Jon.. You are both on my list of potential testers.. For obvious reasons people in the U.K. will get the first opertunity to play with these..
Looks like I will probably need an agent (or a few! LOL) in the States... And probably need to find a bank where the manager plays a Theremin..

I have another question for you all - I intend to house this unit in a high quality case.. But I also want to keep costs down -
The difference in price between a unit in a strong ABS box versus one in a wooden cabinet is about £60 ($100).. (in small quantities).. This will translate to about 8% of the cost.
I would like to hear your ideas about the importance of the aesthetics to you -

Also - I am also thinking about supplying the unit as a kit - this would consist of 2 fully assembled (and encapsulated to stop anyone coppying them) sensors, and 3 assembled and tested circuit boards each having 7 potentiometers mounted on them, and a few wired switches and sockets etc, and a drilled aluminium panel - The kit builder would provide their own cabinet which will need to be completely screened internally with aluminium or copper foil or tape. Power supply (for both the assembled and kit versions) will be an external box I buy in - sorry, I hate these too - but the costs involved in certifying the safety of an in-built PSU would add about £50 to the price of the first 100 units.

I welcome any comments on any of the above.. It may be a few days before I get to check this site again.. Need to spend some time with the children.. and need to get back to the real job of making this unit as quickly as possible..

Thank you all for your interest - "see" you soon - probably about Tuesday.

I'm also left-handed, so ether a dedicated left handed model,I tend to hang on to stuff, or a switchable one would be cool too.
I am a complete idiot on electronics so pretty much everything you said was gobbably goop (or ever you spell it) But I made out it would have a ten octave and be playable, so that's enough for me.
So yes, I am interested.

Andy

Gosh. This does sound interesting.

I have a few comments.

PSoC doesn't appear to have floating point on board from a cursory glance at the documentation. It's been a long while since I dabbled with programming embedded processors but if it doesn't and you're looking for a faster method I implemented a basic rational number suite, complete with mediant rounding - as per Knuth's Semi-Numerical Algorithms - and dropped it into the public domain some years ago. It's in Forth so the code is structured yet inscrutable, but it is heavily documented and in a non-fp environment it might be just the ticket.

I bet some of the top classical players could detect a 10ms latency. They're [i]sharp[/i].

There are some guys working on theremin waveform synthesis in csound at the moment. I found it here (http://www.thumbuki.com/20080116/big-briar-etherwave-waveforms.html) - they might have some thoughts on the subject. My comments there are not particularly relevant to your product, but there might be something by following the link to the csound forum.

I would probably be interested in the more economical version. (Unless the fancy box version totally blows me away - I'm a bit of a sucker for sleek elegant minimalism.)

vonbuck :

"But I made out it would have a ten octave and be playable, so that's enough for me."

>> I doubt whether even the most skilled player could play accurately with sensitivity set for 10 octave coverage..
I want to be clear about this so there are no false expectations..

1.) [skip this if you don’t want in-depth analysis!] At this time I am not sure what the best and worst case sensing distances are.. What the sensor does is this.. it continuously looks at the minimum capacitance it has sensed, and at the maximum capacitance it has sensed (after turning the instument on, one would move ones hands over the entire range from closest to furthest from the antenna, to set its initial calibration)
There is a minimum value of capacitance change (at present about a femto farad I think - but this is quite difficult to measure - cannot be measured directly but can be computed from change in oscillator frequency) which is usable and can be operated on (mathematically) to give a linear output .. Also, at the most distant from the sensor, accuracy is reduced due to other field effects being large compared to the capacitance change.
I have yet to determine the ‘cut-off’ point beyond which reliable and usable capacitance change is not available – I THINK that with my present 90mmx50mm sensor plate this is 30cm (12”) – This could be increased by increasing the size of the sensor (in fact, just gluing a 50cm (2”) x 30cm (12”) piece of aluminium to the sensing face [no electrical connection required – capacitive coupling] doubles the sensing range – but it also almost entirely eliminates the directional nature of the sensor, which is (in my opinion) one of the best features of my sensor.
---------------------
QUESTION: What distance (sensing) range do people want? - Remember, this is a linear playing range, so the entire area can be played with linear control..

If the sensing range is 30cm, to play 10 octaves (which is a bit excessive!) each octave would be 3cm, each semitone 2.5mm.. Reduce the sensitivity to give 5 octaves range, and each octave is separated by 6cm, and each semitone by 5mm ..


GordonC:

“PSoC doesn't appear to have floating point on board from a cursory glance at the documentation.”

You are correct – the PSoC is not ideally suited to complex maths – however it does have a hardware MAC which helps a bit, and it has UM’s (both analogue and digital) – one can do some clever things by using the analogue modules to form a hybrid analogue + digital computer – for example, (I don’t want to give too much away..) one can use the transfer curve of a transistor in combination with analogue modules which provide thermal error compensation, to realize non-linear functions.
In fact, I was not giving the whole story when I blamed the maths for taking 10ms – The ‘problem’ is more related to averaging the capacitive sensor .. The PSoC is handling everything related to the sensor – it is configured to provide the reference oscillator, the sensing oscillator, the equivalent of a 4q modulator, and (with some trickery and external bits) a 16 bit DAC.

Speeding up the maths would be nice (speeding anything up would be nice – there are a few bottlenecks I am still having problems with) but even doing this may not solve the latency ‘problem’ – I may manage 5ms, but this particular problem is not due to maths computation time.