Build Project: Dewster's D-Lev Digital Theremin

Posted: 1/16/2019 12:30:29 AM

From: Minnesota USA

Joined: 11/27/2015

Edited 1/16/2019:
Although I have re-measured all of the inductors pictured below and verified that the numbers are correct (at least the way I'm testing them), I think that the Q measurements of all the coils taken at 100kHz on a DE-5000 LCR meter are misleading for this application.  After checking some the Q of several coils around resonance the clear winner is the original air-core coil on the left, even though the 100kHz Q readings would suggest otherwise.  The next best is the long single layer coil and the wood spool coil is in third place.  This can also be seen in the relative E-field strength around the pitch antenna with each coil in the circuit; the left coil clearly has the largest voltage swing and consequently the highest Q.   Even though all of the coils except the rightmost one seem to work without any issues in the theremin, I think I may revisit the coil that I am using...

"Not trying to pushback or poop on your parade..." -Dewster

Here are a few of the many coils that I tried out for the pitch inductor.  All of these except the Genco on the far right seemed to work without any sticking or other anomalies, and I gave each of those that worked a fairly thorough exercise.  There were many more that did not make the cut for various reasons, but most of those were clearly wrong for the job (some were power supply inductors).  I just happen to have 3 full cabinets of Coilcraft non-RF inductors, so I thought I would give a few a try.

All Q measurements were taken at 100kHz. I was not set up to measure Q at the operating points.

A commercial 3 pi-wound inductor that I also tested (not shown) is a 2.5mH 1535B (I don't remember which manufacturer that would be) and the Q=103 and SRF=1.518MHz.  This part is no longer produced, but may be available for a price from one of the surplus vendors. The radial-leaded 1.0mH parts were purchased recently from one of the surplus sources, but I'd have to check my purchase history to determine what they are.

The smaller single-layer coil shown below (2nd from left, which was wound for 4mH) only needed about 3cm of a 10mm diameter ferrite inserted to achieve the 2.053mH inductance value shown.  Replacing the 10mm ferrite with an 8mm x 50mm fully inserted and centered in the core to be equidistant from the windings actually reduced the SRF to a little under 1.3MHz, presumably because the ferrite provided more end-to-end capacitive coupling.

What I did see when testing the compact single layer coil on the theremin itself was more undesirable perturbation of the pitch field down the pitch arm because of the slightly extended length of high impedance.  I do this test by inserting a metal ball on the end of a long insulating stick at a random point close to the upper half of antenna while observing the pitch.  I then try to place the ball in the same region in the lower half (the new position being symmetric to the original about a horizontal plane passing through the antenna mid-point), while observing the typically higher pitch.  This helps to point out the distortion of the pitch field around the coil and interconnect.  Since I play seated, I'm always fighting between keeping the volume loop high enough to stay away from my leg but at the same time trying to keep the pitch arm low enough so that when your hand naturally lowers at higher pitches you don't start entering this distortion zone.  If the pitch extension arm becomes electrically part of the pitch antenna, I can generally feel it, not because I have any great sensitivity but because of the way I play.

I think for the 2mH pitch side inductor I might have been able to make a ferrite-cored single layer coil that would be about the same size as the 5-section that I made (at the expense of a lower Q), but is there a benefit?

I'm not advocating one type of coil over the other, but since size mattered in this case I don't see a problem with the 5-section coil that I used for my EW Pro style arm. The heavier wire (#28 I think) and fewer turns required because of the high diameter/length aspect ratio of each of the 5 sections results in higher Q.  The obsolete commercial pi-wound coil described above is somewhat better for SRF, and I could have used it instead, but where's the fun in that...

Posted: 1/16/2019 5:14:04 AM

From: Northern NJ, USA

Joined: 2/17/2012

Hi Roger,

I just removed the pitch axis 1.999mH inductor from the prototype and tried to measure the SRF.  I did this by setting it on a tall plastic box, driving the lower end with a function generator set to sine, and measuring the field near it with a test lead clipped to a scope probe placed a few inches away from the coil but otherwise unconnected.  I scrunched the top unconnected lead of the inductor down to make it physically small and less antenna-like. Ground for the setup was provided via the scope.  This is a rough-and-ready setup somewhat similar to one described in a paper I read by Knight, though he used antennas on both ends (as stimulus and sense) with no direct connection.

My function generator doesn't go above 2MHz, but resonance was clearly somewhat above this as I had to bring my hand ~1" or so near the top end of the coil to make it fully resonate at 2MHz.  So I can't really measure anything beyond L via my LC meter and DCR via my DMM for the pitch axis coil.

I took a couple of similarly constructed coils on my bench with larger inductance values and so was able to measure the SRF and Q at resonance:

L=3.813mH, DCR=92.8, SRF=1.488MHz, Q=148 (squat aspect ratio)
L=6.186mH, DCR=74.2, SRF=1.164MHz, Q=166 (tall aspect ratio)

I measured Q by finding the -3dB (0.707) voltage point on either side of resonance, taking the frequency difference, and dividing the SRF by the difference.

Could it be that I'm doing something wrong in the measurement department?  If not, then Q might actually be higher than this because the function generator output impedance is 50 ohms, which is damping things somewhat.  And SRF would be a bit higher if the free top lead were clipped shorter.

Because Q is a direct multiplier at resonance, Q's a bit greater than 100 are consistent with the AFE drive voltage (+/-1V square) and the voltage seen at the AFE capacitive divider (1:100 sine).  I don't know what the Q of the antenna intrinsic capacitance is, but I believe it isn't excessively damping things when no hands or such are nearby.

IIRC I've had less luck measuring the SRF and Q of Bourns 6300 series chokes with this method, I don't know if it's the ferrite containing the magnetic field more or what, but they don't seem to generate as high voltages at the scope probe at resonance, and their resonances are broader.  Just seeing several of them in series inside the EW tells me their SRF and Q probably aren't all that hot.

[EDIT] Just measured a Bourns 6310 I had laying around:

L=51.04mH, DCR=111.3, SRF=479.4kHz, Q=60

Not too shabby.  There may be something wrong with my methodology, but it's testing like this that drove me to single layer air cores.

"I just happen to have 3 full cabinets of Coilcraft non-RF inductors..."  - pitts8rh

Your place sounds like an EE heaven!  :-)

"If the pitch extension arm becomes electrically part of the pitch antenna, I can generally feel it, not because I have any great sensitivity but because of the way I play."

Yes, I've often thought that the geometry of the pitch coil placement might be somewhat obvious to some players.  This is one reason I place the coil smack dab in the center behind the plate, and likely why Theremin placed the coil below and on the same axis as the antenna.  His "Enron" case further ensconced this arrangement artistically.

[EDIT2] Interestingly, if you work the LC resonance equation backwards to calculate the parasitic C of the coils I measured, my two air cores work out to ~3pF, and the Bourns 6310 works out to 2.15pF.  I know that "interwinding" C is likely an incorrect way of imagining / calculating parasitic C, but I believe the larger C of the air cores is due to the larger "plate area" of the high-Z business end windings acting like one side of an intrinsic capacitor with the environment.  It certainly behaves that way, and isn't necessarily a bad thing if properly integrated into the physical design of the Theremin.  And I assume the Bourns Q being ~1/3 that of an air core is due mainly to magnetization losses in the ferrite, though I always wonder to what degree the ferrite "looks" rather like a shorted winding to the coil?

I must say, the exquisitely sensitive and highly interactive nature of the Theremin has given me intuition (however incorrect) into capacitance and inductance that I likely never would have obtained elsewhere.

Posted: 1/16/2019 9:08:39 PM

From: Minnesota USA

Joined: 11/27/2015


I revisited the coil comparison that I made previously and added some qualification notes rather than just delete the post.  The numbers as stated have been rechecked and all seem to be valid, but the results are misleading, at least to me. The voltage swing from your coil is clearly the highest, so I think I need to ignore the Q results from my LCR meter readings and just compare calculated Q at resonance.  I wanted to clarify that my conclusions were wrong, and even though the more compact coils seem to function in the D-Lev, I'm concerned about the reduced field strength/voltage swing.

However I still get a low SRF for your coil no matter how I measure it except in the case where I connect the generator directly to it as you describe.  In all other cases I isolate the generator and the probes through 10meg resistors to keep the stimulation and the voltage observation as non-intrusive as possible. And even though I see a considerably higher SRF when connected directly to the generator, insertion of a 10dB pad after the generator again reduces the SRF, although not to the 1.5 MHz numbers that I am quoting above.  This tells me 1) that my generator's "50 ohm" output is reactive to some degree, and 2) that a 50 ohm generator source (through the pad) gives a different SRF result than that with a generator fed through a 10meg isolation resistor.

To be honest I still don't really know what I'm doing at low frequencies when I have to rely on a scope and a generator.  I keep wishing I could just check the coils on one of my previous company's vector network analyzers.  SRF, Q, and s-parameters de-embedded right up to the fixture plane are all right there at the push of a button. 

Posted: 1/16/2019 9:40:20 PM

From: Northern NJ, USA

Joined: 2/17/2012

"The voltage swing from your coil is clearly the highest, so I think I need to ignore the Q results from my LCR meter readings and just compare calculated Q at resonance."  - pitts8rh

Yeah, the more I looked into getting a fancier LCR meter, the more they didn't seem all that useful for Theremin work.  The ones I was lusting after had rigid fixed frequencies for Q testing, and couldn't resolve deep sub pF.  I got a toy LC meter for cheap that works pretty good for this stuff because the measurement method is resonance-based rather than calculated, though C measurements are temperature dependent (likely due to a poor quality internal L reference) and there is no Q reporting.  It can even sorta directly measure antenna intrinsic C, which is pretty amazing.

SRF, which gives you an indication of operating point headroom, is a really tender measurement.  I think one basically wants to know Q at the operating point, i.e. with a 10pF or so dummy antenna load, and not so much at SRF (though high Q at SRF is nothing to sneeze at).

Posted: 1/29/2019 12:31:40 PM

From: Minnesota USA

Joined: 11/27/2015

The prototype D-Lev is still working absolutely great, and I'm continuing work on a more permanent board and enclosure design for my own use. Progress has been slow on the new build because I'm trying to adapt the D-Lev into a previously-built EW Pro-style enclosure (I'm just going to cleverly call it the D-Lev Pro from now on).  The earlier prototype cabinet with the PVC arms was a quick build because I didn't need to be very careful.  But for this enclosure I had only made one curved flame-maple front (and it was difficult to make), so I'm being very careful in fixturing and machining holes.  Fortunately the hard work is finished.

Since I had originally made this cabinet to house an analog theremin, I had to work around the big wooden knobs that were already in place.  For lack of anything better they will now be used for analog volume controls (for master volume and pitch preview) instead of the traditional theremin volume and pitch oscillator controls.  The eight encoders for the LCD display have been placed below these knobs, which granted is a less than ideal placement for visualizing which of the eight display LCD functions that they control. To compensate for this placement (and assuming everything works out) an array of four LEDs will be placed adjacent to the LCD panel on both sides, with each LED aligned with the corresponding row of pixels on the 20x4 display.  The encoder knobs will be touch sensitive, and the LED for a display cell will illuminate only when its corresponding knob is touched.  This setup works on the bench, and the only possible show stopper might be if the touch-sensitive circuit oscillator causes any interference.  I don't know if this scheme will be useful or even necessary, but we'll see.

Here is the front panel as of today:

The metal display bezel and a back-painted acrylic window will hide the eight surface-mount LEDs (not shown) that will surround the LCD display:

Here are the custom-machined brass knobs and their separate bezels (they are just slipped together in this picture).  Three days of manually machining these parts had me wishing that I had a CNC lathe.  The knobs themselves are DC isolated from the bezels but capacitively coupled to allow the touch sensitive circuits to work. All of the brass parts - knobs, bezels, and antennas will be bright nickle plated instead of chrome (I'm not set up for chrome).

This back view of the LCD display shows how nicely it is buried in the 3/4" front panel.

I need to walk away from the build stuff for a while and get caught up on Eric's latest design changes.  I'm not good at multi-tasking, and it's usually one task or the other, but not both.

Posted: 1/30/2019 1:57:23 PM

From: Northern NJ, USA

Joined: 2/17/2012

Wow Roger!  I'm running out of "wow" type words for your design and construction skills!  The bezel and knobs are things of beauty - and you're going to nickel plate them!

What circuit are you using for touch sensing of the knobs, RC and latch?

Posted: 1/30/2019 3:13:32 PM

From: Minnesota USA

Joined: 11/27/2015

"What circuit are you using for touch sensing of the knobs, RC and latch?" -Dewster

Never spend a few days designing what you can buy ready-made for $2.05, although it remains to be seen if the internal oscillator on this board creates any problems.  It even has eight inputs, as if designed for the D-Lev:

The 2x5 pad array on the upper left allows for sensitivity jumpers and there is some programmability that I won't be using.

Here is the LED board that will stick on to the front of the LCD panel bezel.  As an afterthought I cut the traces to put the current limit resistors right on the board so that I can run directly from the 5v outputs of the sensor board.  The LED board was made on a 10mil substrate (some very! expensive teflon microwave substrate that I had lying around) to keep the acrylic window masking as close to the panel as possible.  Hope it all works...

Posted: 1/31/2019 12:30:44 PM

From: Minnesota USA

Joined: 11/27/2015

And here are two photos showing the finished acrylic window and mask with the eight LEDs that surround the LCD display. 

This shows all eight LEDs powered on to check for the mask alignment:

but in actual use the LED(s) will only light when the corresponding knob or knobs below are touched:

I suspect that this whole extra effort is probably not really necessary and that the knob positions will be easy enough to learn and correlate to the display without any LED indicators.  If something doesn't work out with the touch-sensitive detection I'll just abort this little side project and get on with other business. But at least for now it looks pretty cool.

Posted: 1/31/2019 1:38:31 PM

From: Minnesota USA

Joined: 11/27/2015

Here's a tip if you want to make the LCD display described here for the D-Lev more readable from above.  As supplied, the polarizing filter on the panel favors a range of viewing angles ranging from slightly above straight-on to significantly below perpendicular to the face.  This range is more suitable for something like a top mounted display (like a calculator would have) than it is for a front panel where you tend to view downward.  LCD displays can often be purchased with any viewing angle that you like, but for $10 displays like this one you have to take what you get.

Fortunately the LCD panel in this display is fully symmetric and can be easily flipped end-for-end to favor top-down viewing angles.  Before you get started, note that if you are likely to get interrupted or if you confuse easily it may be wise to make a mark on the glass display to indicate the original right and left sides.  The glass panel itself has no markings to indicate orientation if you should lose track. 

On the back side of the display board you'll find six metal tabs that are part of the bezel.  These protrude through slots in the board and are bent to lock the bezel to the board:

Carefully straighten all six of these tab so that they are free to pass through the board slots.  Try to do this right the first time; these tabs can only be bent a few times before they fatigue and break off:

Now pull the metal bezel away from the board while keeping the LCD panel held in place on the board.  When the bezel is removed you will see the glass panel supported on both long edges by rubbery conductive strips.  Note the placement of the rubber strips in relation to the silk screened outlines.  You will want to be able to replace the rubber conductive strips in the same relative position later.  Take something soft and pointed to coax the rubber strips away from the board at one end, and then lift the glass panel away with the conductive strips still attached to it:

Now simply rotate the glass panel end-for-end and replace it on board with the pads in the same relative position on the silk screening.  Pay particular attention to making sure that the board slots are all clear and that the conductive pads don't block them.

Now insert the bezel tabs through the slots and push it down tight to the board.  Go around to each tab and bend it back to a 45 degree angle while pushing the bezel tightly to the board.  If there is a gap between the bend and the board and it doesn't pull the bezel tight, the whole tab can be bent over slightly in the direction of the bent leg to increase the pressure.  The connections between the board and the glass panel rely on the bezel being pulled tight, so if you find after reassembly that you have some missing pixels on one side or corner of the display, it is probably caused by insufficient pressure at one or more tabs.

If you were successful you will now have a display with greatly improved contrast from a top viewing angle.  Prior to the panel reversal the screen at the viewing angle shown below would have been solid white with no text:

Posted: 1/31/2019 11:37:15 PM

From: Northern NJ, USA

Joined: 2/17/2012

Roger, I bet no piece of electronics is safe around your house!  :-)

You must be logged in to post a reply. Please log in or register for a new account.