Some other observations..
With an AM system there is no need for seperate oscillators - one oscillator with seperate buffers would do the job..
The coupling capacitors are at least 2 orders of magnitude too big.. 100n will couple 50Hz as if one was directly injecting it.. a few hundred pf will reduce coupled LF.
One could probably use a simpler oscillator - even a square wave.. going through a TLO72 at the frequencies involved, one would probably get a reasonable sine wave out anyway!
The problems I encountered were only discovered after I thought it was all working.. Simulation works perfectly, and the modulation effects are easily missed when doing a bench test on a breadboard.. It is really only when one drives a VCO and hears the horrible 50Hz modulation, that alarm bells start to ring.. Then I kicked myself for missing the bloody obvious, and not taking care to evaluate the design fully..
The reason for this sort of oversight probably has a lot to do with using simulation - it is easy to run a simulation using a capacitor as a 'human' model - setting the capacitance to sweep a range of values, and then gloating over the near perfect results.. after this, building an actual circuit, and watching it perform ALMOST identically to the simulation perhaps leads one to miss flaws.. Even without simulation, this process can occur if one sees almost exactly what one expects to see, and if what one expects to see has not accounted for the 'real' world.
I now have a more realistic 'human' model - this is a capacitor in series with a LF AC signal source.. I did this because at first I could not see why induced LF did not affect the thresholds of oscillators operating in the conventional 'FM' mode.. The reason turned out to be simple.. a half cycle displacement is corrected the following half cycle due to the oscillator maintaining its P-P amplitude.. In an AM system, there is no such mechanism.
The AM system does work - but has the problems detailed in my last.. The most noticable being that a lot of integration (filtering) is needed to get rid of induced LF.. One could use a higher order filter to reduce the latency down to perhaps 50ms, but this causes problems with 'ringing'.. The result of which is a damped sinusoid on the CV output every time a rapid change (movement) occurs in the sensing field.
A lesson I am continually forced to re-learn is that Theremins are unlike most other devices.. They LOOK extremely simple, they LOOK like their design is begging for improvement, they inspire INCREDULITY in engineers who look at the 'simple' principles involved, the 'high' prices they sell for, and some foolish engineers like me set off to redesign the Theremin and make a fortune in a few months... And the truly foolish engineers (like me) fall in love with the damn instrument in the process, and a year later, 10 years older, and broke, are still striving for completion of their dream instrument.. [b]because there is, and never was, anything about Theremins which is simple, straightforward logical or free from quark strangeness and charm![/b]
>> Edit 20 Dec 2008 by FredM..
Whilst I do not understand how / why, Mark has assured me that his Theremin does not suffer from any of the problems I saw.
