Plates, Shielding, Spread Spectrum, etc.

Ever since I encountered that crazy spread spectrum capacitance sensor (SSCS) site (http://humancond.org/wiki/user/ram/electro/capsense/0main), the Penn & Teller spirit chair paper FredM pointed to (pubs.media.mit.edu/pubs/papers/96_04_cmj.pdf‎), and most recently with the various noodling over on the staccato pedal thread, I've been thinking a lot about plate sensors, shielding, and spread spectrum, trying to make sense of it all. 

1. I believe if the shield plate is driven directly and perfectly with a buffered version of the signal FROM the sensor plate, then the capacitance effectively disappears for the shielded side of the sensor plate. 

2. If there is any delay or differential in this drive then some influence - however residual - will remain, lowering sensitivity and exposing the temperature dependence of the capacitance between the plates.  (I believe this is an issue with the SSCS, where the shielding plate is driven with the stimulus rather than the sensor signal, which creates troublesome per-cycle and long term differentials between the plates.)

3. The closer the shield plate is physically to the sensor plate (i.e. the higher the mutual capacitance), the more accurate the shield drive needs to be in order to cancel/bootstrap the capacitance.  (I believe this also is an issue with the SSCS, where the shielding and sensor plates are the two sides of a PWB.)

4. One could have multiple sensor plates, all driven with the same stimulus, but each shield plate would need separate buffered drive from its associated sensor plate signal to be most effective.

5. These methods are usually implemented as AM, which means, unlike with conventional FM Theremins, there is the opportunity to easily squash mains hum.  (For something like the staccato pedal this is vital, because the human body is a massive hum source when in very close proximity to high impedance inputs.)

6. Stimulus can be just about anything, though band limited spread spectrum (as described on the SSCS site: the XOR of an LFSR output and its clock, with LC lowpass) probably makes the most sense as it doesn't correlate well with any environmental signals.

7. If the stimulus is spread spectrum then the receiver should be synchronous AM to reject / average out interferers.  This could be as simple as an analog switch arrangement followed by RC LPF (set to 500Hz or so).

8. Sensor drive is kind of problematic.  A healthy voltage swing at the sensor plate is necessary in order to sense external capacitance, but driving this conflicts somewhat with the the high impedance nature of the sensing itself.  ANY drive or sense loading will lower sensitivity.  Resonant drive via series inductance in a multiple sensor arrangement is also problematic.

The above precludes using the shield to capacitively drive the sensor for the most sensitive arrangements, but for a close proximity detector like the staccato pedal, stimulating the sensor via the shield would likely work.  In fact I set up a simple RC oscillator with feedback based on the sensor plate voltage and it worked pretty good on the bench (though doing so was only a rough feasibility test, there was hum all over the place, and this of course is a FM approach).

If you go the AM route and are planning on feeding the data to digital stuff downstream, you need to have a good A/D converter, possibly with a log converter in between (this is the spirit chair HW approach).  FM approaches can often be dealt with directly in the timing domain and so don't necessarily have this requirement, but interference between multiple sensors is more of an issue with FM.

I should also add that the spirit chair is a transmission approach, where the user is electrically stimulated via a capacitive plate in the seat of the chair, and the multiple sensors are unshielded receivers.

staccato pedal -> See Dewsters next post! ---->

Hi Dewster

1. I believe if the shield plate is driven directly and perfectly with a buffered version of the signal FROM the sensor plate, then the capacitance effectively disappears for the shielded side of the sensor plate. 

Yes, it does - completely

2. If there is any delay or differential in this drive then some influence - however residual - will remain, lowering sensitivity and exposing the temperature dependence of the capacitance between the plates. 

This is a big problem area - the buffer must have the sensor as its input signal, and if there is any error between its input and output, strange problems can occur because any distortion coupled from the shield to the sensor can load the sensor or produce HF ringing and one ends up with a mess..

3. The closer the shield plate is physically to the sensor plate (i.e. the higher the mutual capacitance), the more accurate the shield drive needs to be in order to cancel/bootstrap the capacitance.  (I believe this also is an issue with the SSCS, where the shielding and sensor plates are the two sides of a PWB.)

Yes - One can tolerate some delay / error between signal on the shield and signal on the sensor, but the closer they are (the greater the coupling of any error) the worse the problems - The ideal is probably to have a shielding "screen" at least 10cm from the antenna. 

4. One could have multiple sensor plates, all driven with the same stimulus, but each shield plate would need separate buffered drive from its associated sensor plate signal to be most effective.

Absolutely! Its not only about changes to the sensor signal - its about changes to ground capacitance loading the shield/s.

5. These methods are usually implemented as AM, which means, unlike with conventional FM Theremins, there is the opportunity to easily squash mains hum.  (For something like the staccato pedal this is vital, because the human body is a massive hum source when in very close proximity to high impedance inputs.)

This was one reason I went for constant frequency PLL "antennas" (My upside-down topology)

6. Stimulus can be just about anything, though band limited spread spectrum (as described on the SSCS site: the XOR of an LFSR output and its clock, with LC lowpass) probably makes the most sense as it doesn't correlate well with any environmental signals.

For digital, this makes sense - I am wondering if I could use SSCS to drive the "upside-down" topology.

7. If the stimulus is spread spectrum then the receiver should be synchronous AM to reject / average out interferers.  This could be as simple as an analog switch arrangement followed by RC LPF (set to 500Hz or so).

Yeah - I like that idea.

8. Sensor drive is kind of problematic.  A healthy voltage swing at the sensor plate is necessary in order to sense external capacitance, but driving this conflicts somewhat with the the high impedance nature of the sensing itself.  ANY drive or sense loading will lower sensitivity.  Resonant drive via series inductance in a multiple sensor arrangement is also problematic.

With conventional and upside-down analogue topologies, my biggest problem has been providing a buffer (and supply for it) capable of "copying" the high antenna voltage. For the sake of reducing EMI, the shield needs a ground shield behind it which increases the capacitance the buffer must drive (apart from EMI, driving a shield with a nearly constant capacitive loading makes design of the buffer "easier" in that one is always dealing with the "worst case" capacitance),  - in all, it gets quite hairy! I am however finding that with a shield, "required" antenna voltage seems to be a lot lower and I could get away with about 20V P-P which simplified things greatly as I could use +/- 12V supply to power the buffer.

The above precludes using the shield to capacitively drive the sensor for the most sensitive arrangements, but for a close proximity detector like the staccato pedal, stimulating the sensor via the shield would likely work. 

How do you overcome capacitance change between the drive and the sensor? - This was the issue that killed the idea for me.

In fact I set up a simple RC oscillator with feedback based on the sensor plate voltage and it worked pretty good on the bench (though doing so was only a rough feasibility test, there was hum all over the place, and this of course is a FM approach).

If you go the AM route and are planning on feeding the data to digital stuff downstream, you need to have a good A/D converter, possibly with a log converter in between (this is the spirit chair HW approach).  FM approaches can often be dealt with directly in the timing domain and so don't necessarily have this requirement, but interference between multiple sensors is more of an issue with FM.

I should also add that the spirit chair is a transmission approach, where the user is electrically stimulated via a capacitive plate in the seat of the chair, and the multiple sensors are unshielded receivers.

I think the "spirit chair" probably only "worked" because there was no critical coupling between the "transmitter" and the "recievers" - there wasnt a "transmitter / shield" plate coupling to a "sensor plate" through some fixed distance / dielectric (which would be highly subject to thermal and other capacitive variations) -  the "transmitter" plate was the player, and by moving, the proximity to the sensor plates was being deliberately changed.

Fred.

(FredM) - FredM is still broken and cannot log in...

ps - I should just say that im not exploring any of this stuff at the moment - I decided its just too complex and expensive with almost no chance of any return - The electronics and mechanics required to implement a good reliable directional theremin would, I believe, price the product way outside any commercially viable point - Better I think to concentrate on whats really wanted - sounds and playability.. But I am still interested in the ideas! ;-) .. The "low price" end is where any market is IMO - So a capacitive ribbon for those who want to play the theremin without having to play the theremin or be bothered by people walking past, and a box you can plug your EW audio into and make it sound like an RCA or whatever - those are my present theremin-related directions (work on industrial application of capacitive sensing is taking most of my rime and providing a tricle of money right now - so it has highest priority) ... IMO, the E-Field theremin interface is just too quirky - If one makes a theremin that isnt quirky those who love the theremin interface will miss the quirkyness and lose some of their excuses for bad playing, so they wont buy it!  LOL ;-)

but for a close proximity detector like the staccato pedal, stimulating the sensor via the shield would likely work

staccato pedal -> See Dewsters next post! ---->

"How do you overcome capacitance change between the drive and the sensor? - This was the issue that killed the idea for me."  - FredM

You don't! ;-)  To me, the staccato pedal is more of a crude, snappy action, short range controller - as opposed to a precision, continuous action, long range controller.  Design things so the response takes place where the foot-to-top-plate capacitance is in the same ballpark as the bottom-plate-to-top-plate capacitance (this is around 36pF for my rough-and-ready trial setup) and linearity etc. of the response doesn't even enter in to the equation. 

If the plate-to-plate capacitance varies 10% over temperature, who cares?  Any variation in response will likely be more than offset with a substantial dead zone so as not to influence things if the foot is say ~1 cm or more above the top plate.

"To me, the staccato pedal is more of a crude, snappy action, short range controller - as opposed to a precision, continuous action, long range controller.  ......  If the plate-to-plate capacitance varies 10% over temperature, who cares?  Any variation in response will likely be more than offset with a substantial dead zone ... " Dewster

Ok - I see what you are saying..!

Perhaps the relevant parameters for the pedal are more to do with attack 'velocity' and release 'velocity' - and these can be determined regardless of any slow drift in actual capacitance sensed by the pedal - in fact, even mains hum would have only minimal effect if the attack and release velocities were 'strobed'.. (one would perhaps be looking at a change in capacitance from movement of the foot of 1cm, which would be quite substantial - this 1cm could be anywhere within say a 2cm field from the plate - and this 1cm change in capacitance would swamp any changes caused by thermal or induced effects)..

And the driven-shield capacitively coupling to the sensor plate could be the best way to implement this.. (in fact, a small PSoC CapSense IC has everything needed to fully implement this - shield drive, sensor, ready-configured "User Module" and API's to derive capacitance delta, DAC's  or PWM or UART to output the data, or perhaps even MDAC so one may not even need a VCA)

Yeah - all the problems go away if you arent looking for an absolute measure, but are just looking for a "large"  change - and are effectively only using that... Or if you have the "sensing zone" "floating" above the plate and dont need to worry too much about its absolute position or say 10% variation of this position.

Fred.