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

"Just been playing with your clapp_2LC_buf_CC_2014-03-21

It really is beautiful! Rock stable, self starting, extremely low harmonics at antenna, low current, and yes - I think the CC Fet has great potential for temperature frequency correction."  - FredM

It's quite nice, just wish I had invented it originally and not partially re-invented it after the fact!

I think the CC FET itself can be dispensed with as the top differential amplifier FET will likely respond the same to variation in R1 (it does in simulation; it's in a sort of CC arrangement already).  I really wanted the CC FET design to go, but I think it maybe introduces a bit of noise (subjective) and I worry that it may introduce temperature tracking issues between the FETs? 

So the non-CC one is the way I'll likely go.  It took me forever to come over to the single transistor camp (after repeatedly bad mouthing it) - wish I'd tried this stuff long ago.  I particularly like the low current aspect which seems to eliminate any "warm-up", and the allowable higher voltages at the device terminals which somewhat relaxes ESD and tank swing sense limits compared to CMOS.

=========

Just quickly (not definitively) tried livio's Colpitts arrangement by rearranging a few things on the breadboard.  It behaves very similarly in all ways as the adapted Clapp.  But even with 10pF in series with the antenna (livio's schematic shows 18pF here) it stalls out much easier, and with this cap it is noticeably less sensitive.  The 10pF between the tank and FET necessary for larger voltage swings is largely to blame for the poorer showing IMO as it limits drive energy.  I should add that for this quick test I used 220pF caps for C1 and C2, livio specifies 33pF and 82pF respectively which would help a lot with stalling, but would dramatically lower the antenna swing. (Who knows how much swing is "enough"?  And I was going for more of an apples / apples comparison.)

Picked up one of these yesterday for ~$10 at the local Sears store (#965119):

Dimensions: 562mm x 182mm x 98mm.  With some work it looks like a Theremin could be built into it. 

It's blow-molded, but luckily the bottom is largely solid, with feet on the bottom surface giving some recess relief for screw heads and such protruding from bottom should one want to mount things in it.  Some standoffs going from the bottom up to a custom made inner control / display panel should do the trick.  The top is also solid in the recessed handle area, which would give another nice internal control / display panel mounting area.

Could use UHF chassis connectors mounted on the lower panel for the antennas.  The pitch antenna would screw right in, the volume side would require 90 degree elbow(s) to make perpendicular with the case.

I really wish the top had latches on both sides and didn't hinge, but maybe the pitch display could go in there on the left or right side behind a smoked or red plexiglas panel.  Maybe the FPGA could in the top as well on the other side, leaving the lower area for oscillators only?  This would locate much of the heat and noise away from the oscillators.  Or maybe the hinge pin acceptor on one side could be slotted out to allow easy removal of the top, which could then be used for antenna storage and such.

The microphone stand mount I have is actually threaded through, so I could bolt it to a non-metallic internal bottom stiffener plate and just provide an access hole to it through the bottom - it wouldn't have to stick out or anything so the box would still be able to sit flat (unlike the EWS which needs big feet to accomplish this).  The inner bottom area is flat, so the stiffener plate could cover the entire inner bottom (no jokes please) and be attached via the same standoffs used for the lower control panel.

Still kicking around the possibilities, but it looks as promising as anything else I've seen.  And it's durable, inexpensive, and I don't have to pay outrageous postage from Turkey.  But I have to fabricate panels and stiffeners and stuff.

[EDIT] Starving artist's conception:

I love the box ;-) - space to fit more knobs between the antennas as well ;-)

Dewster - A while back you were mentioning spread spectrum?

Have a look at this :

http://humancond.org/wiki/user/ram/electro/capsense/0main

Fred.

(the 30cm breath pattern is astoundingly impressive IMO) - REALLY wierd finding this! I was searching for philosophical matters at the time - Free will, self desception, that sort of thing .. and there in the index I find capacitive sensing... ?!

"...the 30cm breath pattern is astoundingly impressive IMO"   - FredM

Very interesting!  I need to start using the term "attoFarads"!

Still digesting it, the reasoning seems sound.  It's not clear to me how he has the transmitter and receivers arranged in the testing space.  Small capacitance in the receiving opamp feedback loop, largeish voltages on the transmitting antenna, lots of parallels to Theremin circuitry.

Output bandwidth is 1.5Hz!  Sticking a 1.5Hz LPF on the demodulated output of a generic Theremin LC oscillator would give some pretty astounding results.  When I look at my current oscillator delayed by 50ms (I think this is sort of like 20Hz BW sensitivity without the filtering SNR advantage - very analogous to heterodyning anyway) I easily and routinely see my respiration at 30cm or so even with a small rod antenna - have to back way off even with no antenna so my body doesn't overly influence things like temperature testing.  It's obviously not spread spectrum though.

His statement: "This is the receive frequency response, including the inherent single zero high-pass effect of the capacitive coupling" refers I think to the capacitive coupling of all environmental signals to the antenna, which makes added series capacitance moot at best.

" I need to start using the term "attoFarads"!" - Dewster

LOL ;-) - Not a good idea if you dont want "uninitiated" engineers to regard you as anything other than a masochistic nutter! - Ive got a "You mean pF not fF dont you?" when ive been in discussion with clients regarding cap sensors, and when one says "no - I mean sensing resolution of better than 1fF" their doubt / scepticism becomes palpable, and they start talking about inductive or optical sensors.. I suspect that if you start using aF things might be even worse ... ;-)

" I easily and routinely see my respiration at 30cm or so even with a small rod antenna"

It was the resolution that kind of impressed me (visually anyway - it may be meaningless as there may be interpolation or whatever going on)..  To be honest, I really didnt study the stuff in any depth, so it could be worthless... Just thought you should see it as it was an idea of yours a while back.

Fred.

"Just thought you should see it as it was an idea of yours a while back."  - FredM

Yeah, but I only took it as far as looking at the drive / sense differential of a resistor feeding an antenna.  This guy takes it a whole lot farther.

Some unexpected things: (1) He's using a LFSR sequence, but XORs the output with the clock, so there are just 1/2 and 1 clock period levels coming out, which constrains the harmonic bandwidth over that of a straight LFSR sequence and gives more transitions.  (2) He keeps the LFSR total sequence period rather short (10Hz) so that multiple sequences are used per reading (1.5Hz).  The sequence can be seen as an elaborate AC signal with no harmonic energy below the repeat period - no LF content to confuse things. (3) The XORed LFSR is cranked up in voltage (20V) and then low pass filtered with a simple LC.  (4) The "ground plane" copper layer on the back of the sensing plates is actually the transmitting antenna!  This is electrically connected to the aluminum pole on which all of the receivers are mounted, so it's all radiating.

(4) Gives it some directionality, and makes it largely immune to objects behind the antennas, much like your shielding ideas Fred.  He has three receiving antennas all picking up the same transmit signal, so in some sense it's also vaguely like your upside down ideas.

I imagine it could be adapted to higher frequencies so that the integration period ("spreading gain") could be more in the range of Theremin playability (~1ms).  One problem for the FPGA approach is the synchronous receivers, the output of which is an analog voltage read via A/D.

I wonder if at some point the emissions boys will force all Theremins into spread spectrum territory?

After mulling over the article a bit more:

(1) isn't unusual, the LFSR sequence is usually multiplied with a sine wave to restrict total bandwidth.

(4) is really odd.  The antennas appear to be indirectly driven capacitively via the active "ground plane".  The drive is likely somewhat strong (i.e. largish capacitance) lowering the sensitivity.

For all the talk of "spreading gain" and such, I believe one can think about these things to a first order as a moderate voltage square wave (~20V p-p) capacitively connected to an antenna, with an input +1/-1 amplifier in phase with the driving wave (detection/rectification) going to an RC LPF.  External capacitive loading of the antenna (a hand, etc.) makes the resulting DC signal smaller. 

The use of a more random signal just distributes the radiated energy over a wider band and makes the receiver statistically fairly immune to strong narrow band interferers.

Thank you, Dewster - That makes a lot of sense.. and gives me an idea..

The analogue "driven shield" stuff could probably be made to work with a similar topology without spread spectrum - its an idea I havent explored, but its certainly interesting.. Drive the shield from a low Z output from the oscillator via an inductor - shield is coupled to the "antenna" by capacitive proximity, this antenna is itself connected to ground via an inductor forming an independant parallel resonant, and it forms the frequency / tuning input to the oscillator...

But I agree that the spresd spectrum approach does look like having some big advantages IF one can decrease the latency.

Fred.

Something like this? Or perhaps even omit the drive side inductor.. Quite easy to get a big enough capacitive coupling between shield and antenna! ;-) .. This certainly simplifies shielding hugely! - with powerful enough drive one could have a ground shield over the driven shield to reduce emission.

"Something like this?"  FredM

Oh interesting!  I didn't think of integrating the coupled shielding signal back into the oscillator itself!

From what I can tell, the SS capacitance sensor works like this:

The problem is in hugely amplifying the minute difference between the antenna and the shield with both flailing around.

The SS capacitance sensor solves this rather cleverly by referencing just about everything, including chassis ground, power supplies, and A/D, to the shield signal.  The amplified difference signal is detected (rectified) with an analog switch, with the final RC trading BW for SNR.

"Oh interesting!  I didn't think of integrating the coupled shielding signal back into the oscillator itself!" - Dewster

I have run some basic simulations using a modified CMOS oscillator, and it looks promising! - Having a ground shield over the driven shield / osc drive, and giving this a capacitance of 1nF in simulation, and having a transistor push-pull output driving the shield via low R (100) .. then having shield coupling to antenna by 100pF..

A practical problem will be this shield -> antenna coupling.. This is likely to suffer huge thermal variation and also microphonics .. Active thermal correction will be essential.. Microphonics could be an advantage.

The Sensing field would AFAICS be highly directional if one had a narrow 'window' in the antenna construction, or this window could be widened as requred - This should get rid of most hum / interference / interaction issues and possibly improve linearity?

Im certainly going to play about with this!

Fred

Here is an update with included ground shield:

I get good simulation with an output inductor of 47uH (output drive = 5V from 6 parallel 74HC04s each via 2k2 giving 366 ohms) , input inductor of 680uH, Shield to ground coupling I have arbitrarily set at 1nF, Shield to antenna coupling at 100pF, total antenna - ground / player coupling at 10pF. Antenna P-P = 70V, sensitivity = ~3kHz/pF oscillator running @ ~700kHz. Drive current is sine +7ma to -7ma P-P.

The whole output stage (and osc) could be simplified - I just have a standard setup that my simulator likes! ;-)

 Flaws: Antenna voltage does vary a bit with antenna capacitance (marginal) but the real flaw in this design is that frequency will vary HUGELY with any change in shield -> antenna coupling.. Ok if this coupling was COG/NPO, but it most certainly wont be! ;-) 

One big possible advantage I am looking at is that it seems easier to put a band-pass filter on the input - Unlike normal arrangements, where anything one places in the antenna circuit deals with both "transmitted" and "recieved" signals, it seems I can have a high-Z input amplifier followed by a BPF which rejects everything below 500kHz and everything above 1MHz, and I can probably incorporate this right at the input if I want...