The physical design of a Theremin requires equal care and consideration as the circuit design.  By design, the pitch and volume oscillators are heavily affected by any stray capacitance.  Additionally, the rather large oscillator signal amplitude makes electromagnetic isolation through physical design critical.  Proper grounding techniques must also be employed to prevent unwanted signal propagation via ground loops.  It’s not an exaggeration to state that the physical assembly of a Theremin (in particular, the GT Theremin) can make the difference between a working instrument and an expensive glowing paperweight.  Credit for the physical design of the GT Theremin goes primarily to James Lewis, who did a stellar job making detailed measurements, creating CAD layout drawings, and putting a lot of care and time into fabrication.  I’m amazed we were able to fit everything together so tightly, and this is all thanks to James’ meticulous work and just a touch of luck.  Note that the order in which the final construction is presented doesn’t really correspond with any design sequence.  The final construction came together rapidly and all at once, so the photos shown here don’t necessarily indicate the chronological progression.

Prototyping

Of course, we didn’t assemble a complete working Theremin within a couple of weeks.  Before we could even begin the physical design, I had to figure out all the circuit details so we could nail down component dimensions for the final assembly.  This required an interim form of construction since vacuum tubes do not fit into a standard IC bread board.  What we came up with was about as makeshift as these things go, but it worked reasonably well.

Hearkening back to the origins of the term “bread board”, we cut some holes in a thin piece of plywood through which we mounted several vacuum tube sockets (7, 8, and 9-pin).  To the top of our prototype “bread board” were also mounted some bus strips and several IC bread boards on which the through-hole components were assembled.  I soldered wires to each socket with a fairly consistent color code, which I highly recommend for helping keep the jumble of wires manageable.  I used green for heater pins and most of the heater wiring, yellow for heater center-taps, blue for cathodes, red for grids, orange for plates (anodes), black for grounds, and white for high voltage power.  With a few deviations I used this color code and it served its purpose well.  I used stranded wire for flexibility and tinned the ends with solder so they could be inserted into the IC bread boards.  We also mounted switches, transformers, and variable capacitors to the prototype board at various times as needed.  Antennas were separate from the main prototype board, and were secured with state-of-the-art two-by-four and C-clamp technology.  Power was provided by a 500 V Sigma vacuum tube regulated power supply that is not only my senior, but possibly my father’s as well (it worked beautifully, though).

The prototype board went through several incarnations, but these two photos taken right before I disassembled the prototype for final construction are indicative of our full test setup.  In the first you can see our test bench complete with the massive Sigma power supply on the right, a test pitch antenna (copper pipe) on the left, and a little metal box used to house and shield the big air gap tuning capacitors.  The second photo shows the actual prototype board in more detail.  The long tube wire lengths used in prototyping were somewhat troublesome in terms of interconnect capacitance, but this was unavoidable since I did not want to cut them short and then be forced to redo them for the final design (wiring and heat shrink tubing all the sockets took about 7 hours).  Moving from this prototype to the final construction with its metal case and shorter wire lengths required a little re-tuning, but thanks to careful calculation I only had to de-solder and replace one capacitor once we had the final assembly built.

Circuit assembly

Several alternatives were considered for assembling the circuitry in its final form, including point-to-point wiring using intermediate insulated solder lugs and an ancient form of mounting using terminal strips.  We initially decided to have some custom PCBs fabricated, but this proved to be impractical due to several minor last-minute circuit revisions and some unexpected project delays.  In the end, we settled on using some proto board to assemble the circuitry (PCB with pre-drilled holes and rows of copper interconnect).  This proved to be an adequate solution, though custom PCB would have still been nice.

With the exception of the power supply, all of the Theremin circuitry was assembled on four pieces of proto board.  These were arranged in two stacks to save space; each containing a pair of circuit boards separated by a sheet of grounded copper to help screen electric field and reduce unwanted coupling in the signal path.  James translated each circuit schematic into a physical layout manually, with me verifying the layout and making some minor revisions.  These layout diagrams were then used as a guide to solder each of the boards together.  Surprisingly we only made one error in the process of translating the Theremin circuitry into physical circuit boards.  Connections between the board and tubes were facilitated by header pins to allow easy disassembly (which was important given the circuit revisions we needed to make).  The photos below detail a board stack being test fit before soldering, a “top” view of the completed pitch stack, and a “bottom” view of the volume stack.

All signal wires were cut to minimum length, carefully routed, and sometimes twisted into pairs to keep parasitic interconnect capacitances to a minimum.  The layout of the header pins on the circuit boards wasn’t always optimal with respect to the location of the corresponding vacuum tube, but given how quickly the layout and construction came together, I think our results were fairly good.  The photos below show various stages of the signal circuit wiring.

Enclosure

We decided early on to use an all-metal enclosure in order to isolate the sensitive oscillators from outside influence except through the antennas.  Locating a suitable metal box for our purposes turned out to be surprisingly challenging, but we managed to cannibalize a 12×8×3 inch 14-gauge (about 0.064 inches thick) aluminum box from the parts room.  Looking at a big empty box like this one, you’d initially think that fitting everything inside would be a cinch.  Of course this turned out not to be the case, and the final construction fit together with exceedingly scant spare room.  This was only made possible by James’ careful measurement and CAD work which was of such high quality that only one aspect of the physical layout had to be re-considered during construction (and that was an unnecessary afterthought anyway). All of the mechanical CAD drawings produced by James can be found here in PDF format. I also have the drawings in DXF format; email me if you’re interested in them.

Most of the parts had to be fit around the transformers and antennas, which are bulky and require some physical separation from each other.  Out of consideration for sci-fi aesthetic appeal, we wanted to place all the vacuum tubes and at least one transformer on the top of the case for display.  Once we found a suitable high voltage transformer (no easy feat in itself!), we started sketching out ideas for the all-important top lid, on which would be mounted a majority of the circuit components.  After tossing around a few ideas, James came up with a very clever layout which managed to use our available space fairly effectively and still look extremely cool.  Below is shown James’ CAD drawing for the lid, which was printed out and used as a template to fabricate the top out of 14 gauge aluminum sheet.  Below the drawing is a photo of the lid after being drilled and punched and an image of the lid with the vacuum tubes and high voltage transformer mounted.

The antennas were made from 1/2 inch aluminum bar stock, cut and bent to form.  I have seen some discussion among Theremin builders on the merits of various types of antenna (length, solid versus hollow, etc.), but I did not have the time to explore the effects of using different antennas in detail.  At the low frequencies in question, I tend to believe such effects to be minimal, but perhaps there might be something to be said for slight changes in self-inductance.  In any case, hand-to-antenna capacitance is the salient parameter insofar as the GT Theremin is concerned, and it is probably apparent that a preferable antenna construction would utilize flat plates.  However, in deference to Termen’s original design, we opted to use the traditional “rod and loop” construction, which affords adequate performance.

We wanted the antennas to be removable, so we had to devise some mounts to facilitate this.  After a lot of brainstorming and testing, we came up with some wooden mounts and springy sliding contacts made from copper sheets which maintain good physical and electrical contact, respectively.  The antennas are isolated from the grounded case with some thick rubber grommets that we drilled out to an appropriate diameter.  All of the antenna mounting hardware is attached inside the lower case, though the pitch antenna passes through the lid.

The lower part of the case also houses the tuning capacitors, heater transformer, the high voltage power supply filter capacitors and resistors, fuses, a line power connection, the audio output jack, and the front panel switches.  James produced several CAD drawings of the case, so it shouldn’t be too difficult to figure out how we assembled it.  The photos below show the initial test-fit of the inner case components as well as its final assembled form.  Also shown in detail is one of the volume antenna mounts.  The lid is hinged on the front and retained by a length of wire.  Note the liberal usage of heat shrink tubing; this serves both to prevent wires from shorting out accidentally and as a safety precaution.  You would actually have to make a fairly concerted attempt to electrocute yourself inside this box; e.g. by touching the exposed power resistor leads directly above the written “high voltage” warnings (not recommended).

Physical circuit considerations

As was briefly touched on before, the layout and construction of the Theremin greatly affects its performance.  Careless construction can literally render the device inoperable, so some thought must be put into assembly.

One major parasitic which must be minimized is capacitive coupling between interconnects and other components.  Due to the tightly-packed nature of the GT Theremin and the relatively long lengths of parallel wires, capacitance sufficient to facilitate significant parasitic signal coupling can arise.  The most effective way to minimize this is physical isolation of critical components and minimum length interconnect wires.  The grounded copper sheets in between each pair of circuit boards screens electric field which could otherwise cause some coupling.  In addition, wires were cut to minimum lengths and routed away from each other as much as possible.  Another less obvious trick to avoid unwanted signal coupling involves twisting pairs of out-of-phase wires (like the heater circuits) together to lower their radiative ability.

Unwanted signal coupling can also occur along the high voltage DC power rails if care isn’t taken to decouple them close to signal circuitry.  The power supply filter capacitors are not able to prevent this coupling, so large electrolytic capacitors are placed across the supply rails near each critical signal circuit component.  This is especially necessary near the oscillators, otherwise the two pitch oscillators may lock on to each other or at least see degradation in harmonic performance.  To prevent ground loops from forming, a star grounding scheme was used, wherein each critical signal circuit is provided a separate path to the central signal ground on the top panel.  The decoupling capacitors also have dedicated ground connections to prevent them from building up a voltage on the low side.

It is especially important to separate the antenna connections and air gap variable capacitors from the rest of the circuitry, since these tuning components have major effects on the signal circuitry.  Therefore, the tuning capacitors were placed in the far corners of the GT Theremin enclosure, and the antennas were located as far away from the signal circuitry as possible.  There’s still some coupling, especially between the pitch antenna and its associated oscillator tube, but this didn’t turn out to be a great detriment.

Safety is another primary concern with the high voltages involved in the vacuum tube supply.  All indicated grounding points in the power supply schematic absolutely must be observed. The center taps of both transformer secondaries must be grounded to ensure that the heater filament-to-cathode voltage for the tubes is never exceeded.  I used separate points for the signal ground and power grounds, however the AC and DC impedance between these points insignificant (the path through the case is sufficiently conductive at the frequencies in use).  The case must also be connected to earth by an appropriate three-prong power cord.  This prevents the possibility of the performer’s body ever becoming the shortest path to ground in the event of an electrical fault.  The heater and high voltage supplies are separately fused with appropriate low-tolerance current ratings, and these should not be substituted for different values unless you really understand what you are doing.  In addition, to avoid inconvenient explosions, all the electrolytic capacitors should be rated for 150 V or higher.  Ideally all the capacitors should be rated for the full supply voltage, but under normal conditions only the electrolytic capacitors should ever be subjected to it.

« Previous: Circuit details part 2 | Next: Results »