Now that you have a reasonable idea of how the GT Theremin works and how its functional signal path flows, I’ll give some specifics of the circuit design and try to explain where some of my choices came from.  I’ll also try to point out any areas I think might be improved or modified to suit different preferences.  Once again, here’s a copy of the GT Theremin full schematic (PDF) in case you want a printable version.

Pitch circuit

The pitch circuit contains two more-or-less identical Colpitts oscillators, formed by 12AU7A (datasheet) tubes U1 and U2 and surrounding periphery.  The “A” section of each tube is connected to a resonant feedback network comprised of an inductor and a few capacitors (basically an LC tank).  The values of the inductors and capacitors, along with parasitic capacitances in the tube and its connecting socket and wires, determine the frequency of oscillation according to the following formula:

Where is the inductor value, is the total capacitance in shunt with the inductor, and and are the values of the two “series” capacitors in parallel with the inductor.  The nominal pitch oscillator frequency for the GT Theremin is 500 kHz.  Note that both the variable air-gap tuning capacitor C14 and the pitch antenna are connected in shunt with an oscillator’s feedback network, effectively modifying the value of above.  The “series” capacitances and (C1,C4 and C11,C15) form a sort of AC voltage divider which determines the amount of feedback in the oscillator and primarily determines whether it can start up and sustain oscillation.  R1 and R8 are fairly large values to keep the plate current low and the oscillator output as sinusoidal as possible.   Adjusting these bias resistors can increase the oscillator amplitude, but will also significantly affect the harmonic content in the output.  The “B” section of each pitch oscillator tube is configured as a cathode follower (or current amplifier), which helps reduce unwanted coupling between the two oscillators by providing a low-impedance oscillator output.  The voltage divider formed by R3, R4, R5, and R11 further isolates the two pitch oscillators.

I chose to use this particular oscillator topology after playing with the performance of several similar designs.  Colpitts oscillators require only one inductor (as opposed to a Hartley) and can provide fairly sinusoidal output despite their simplicity.  Of the numerous possible arrangements for a triode-based Colpitts, I found this “common-plate” configuration to be the easiest to stabilize and control harmonic content.  An alternative “common-grid” arrangement was considered, but it required the use of an additional vacuum diode to consistently stabilize, which I felt broke symmetry and would have been an unnecessary waste of a tube.

Here are screenshots of simulations in EWB Multisim which compare some of the oscillator topologies I tried.  The first shows the common-grid arrangement, the second shows the high harmonic content resultant from a common-plate arrangement using too much plate current, and the third shows the lower harmonic content of the topology I settled on.  I used Norman Koren’s phenomenological triode models modified for Multisim, which do an excellent job of describing the tubes’ behavior.  All of these oscillators behaved almost identically to simulation when they were built, which is impressive considering tube variations and the chaotic nature of oscillators.

The radio-frequency pitch oscillator signals are combined and heterodyned in a triode mixer formed by the “A” section of U3.  This tube is biased in a highly non-linear (that is, cut-off) region so that it produces the audible difference frequencies we want.  Designing the mixer was accomplished mostly by trial-and-error, and I think a better topology which yields a nicer-sounding low (audible) frequency response may be possible.  Keep in mind that modifying the mixer circuit will significantly alter the sound of the Theremin, and may cause it to cease working altogether.  You can change out the tube in the mixer position as long as it is a twin triode and is also suitable for use a cathode follower (used in another section of the circuitry).  I’ve successfully used a 12AU7, 12AT7, and 12AX7 in this position, which all yield a slightly different timbre to the sound.  I recommend using the 12AX7 (datasheet) mostly because its high amplification factor provides the highest signal amplitude at the Theremin’s output.

Components C12, R7, and C13 form a two-pole RC lowpass filter which removes the inaudible RF signal.  Care must be taken such that the filter’s input impedance is high enough not to load down the very weakly-driven triode, which has a high effective output impedance.  I tweaked this filter innumerable times, and I don’t suggest changing the component values unless you are trying a different mixer topology.  The audible pitch signal exists at the output of the filter, but the output impedance is high so care must be taken not to heavily load it.

Volume circuit

The volume oscillator uses nearly the same topology as a pitch oscillator, with feedback component values selected for a frequency of 455 kHz.  The oscillator triode is biased a little stronger to provide a larger signal, at the expense of higher harmonic output.  The reason for this trade off will be explained momentarily.

Frequency detection is accomplished with a specialized little component which wasn’t exactly designed for this purpose.  It’s a Murata 455 kHz ceramic ladder filter intended for use as an IF filter in AM radios (part number CFULB455KH1A-B0, datasheet).  The same function could be accomplished with an RLC filter, but the ceramic filter affords a rapid fall-off which can’t be touched by the Q-factor of typical RLC tanks.  This is very important because the quicker the filter fall-off, the wider and more sensitive the Theremin’s volume response.  The ceramic filter must be impedance matched at its input and output in order to provide the performance listed on its data sheet.  However, I did not want to add the additional complexity of matching networks for a number of reasons.

Without matching, the filter must be isolated from the oscillator with a fairly large resistance, R13.  This is to prevent the low-input-impedance filter from pulling down the high-impedance oscillator output, causing problems with the oscillator and maybe even preventing it from starting.  Finding an appropriate value for this resistor took some trial-and-error, and changing it will greatly affect the volume circuit performance. This method causes high signal loss at the filter input due to the AC voltage divider formed by R13 and the ceramic filter’s 1.5 kΩ input impedance.  It’s for this reason that the volume oscillator is biased to produce a higher signal amplitude; in order to compensate for the large insertion loss in the filter.  The extra harmonic content generated by the oscillator thus configured is not much of a concern since it is almost fully removed by the filter (plus this signal is rectified into DC anyway).  The volume response could probably be improved by utilizing actual matching networks to keep the ceramic filter happy, but at these low RF frequencies that would probably require bulky transformers (I think… maybe there’s a simpler way).

The amplitude-varying signal at the filter’s output is passed through a collector-follower buffer to prevent the rest of the volume circuitry from loading down the filter.  The signal is then clamped below ground by C25 and section A of U4, a 6AL5 twin diode (datasheet).  The B section of U4 along with C24 and R16 are a half-wave rectifier that turns the volume signal into a DC value which corresponds to the player’s hand distance from the volume antenna.  This DC voltage is used to alter the gain of the voltage-controlled amplifier section.  A diode-connected twin triode (grid connected to plate) like the 12AU7 may be used in lieu of the 6AL5 if fewer tube types are preferred.

« Previous: Design overview | Next: Circuit details part 2 »