Build Project: Dewster's D-Lev Digital Theremin

Posted: 1/1/2019 5:35:26 PM
pitts8rh

From: Minnesota USA

Joined: 11/27/2015

"I really like what I'm hearing (nice playing by the way).  The vocal formant sounds excellent, much better than my experience with the Electro-Harmonix box.  There's some real genius behind this design that I can truly appreciate now that I can hear it played well (and without accompaniment).

If you or anyone else decides to produce these theremins I would be very interested in purchasing one as this is way beyond my skill level to attempt.

Thanks for sharing." -johnthom

Thanks for your comments.

The particular vocal sound in the video is just the tip of the iceberg.  I think you could spend weeks/months just creating new vocal sounds with different male and female characteristics.  The EHX talking machine has one or two theremin-useful vocals that appear to use 4 formant filters, and as near as I can tell the formants are fixed and unable to be modified with louder or higher pitch notes.  On the other hand, the D-Lev has 12 dedicated formant filters that can be set to any frequency and relative amplitude, and the resonances are adjustable for each of the 4 groups of 3 filters.  In addition, the first bank of 3 formants can be modulated up or down in frequency by pitch or volume or both, which gives a powerful  means of opening up the vocal sound realistically.  All settings can be saved on a per-preset basis.

On top of the the dedicated formants, there are oscillator and noise filters, an inharmonic resonator, and now the variable even/odd harmonic mix that in itself can add thousands of new variations.  This is getting into serious synthesizer territory. And the vocals are just one small part of the overall sound capabilities. 

I can't speak for Eric, but as I said earlier I'm pretty sure that he will offer some path for getting one of these in hand in the very near future.  Eric is furiously adding new features as we work through some issues on the forum or via email, and I think it's more a matter of getting the basic architecture stable so that a library of representative presets can be supplied to users. I have no stake in this, but I can't hide that I am pretty enthusiastic about the whole thing.  I think it will be worth your wait.

Roger

Posted: 1/13/2019 8:13:04 PM
pitts8rh

From: Minnesota USA

Joined: 11/27/2015

D-Lev Tuner
I finally made an enclosure for the D-Lev tuner out of heat-formed acrylic with a steel base.  The clear acrylic was masked and painted from behind in the same manner as the main enclosure, except here the only masked feature is the clear window for the 7-segment octave display.  A strip of sheet steel with the ends bent up 90 degrees forms the main chassis over which the acrylic front panel is mounted.  I used steel so that magnets could hold the tuner enclosure to a tilting mount placed on the top of a mic stand.  Early on I had decided that the only way that I would use a tuner would be if it could be placed out past the pitch antenna where I usually gaze while playing, and I'm happy with the way it turned out.

This support is made from a block of Delrin with two countersunk magnets.  It attaches to a tilting mount on the top of a microphone stand.

The back of the tuner enclosure is steel so that it can be quickly popped into place on the magnetic mount.  The adjustable tilt mount came from small LCD display used for backup cameras. The dark end panels of the enclosure are pieces of black acrylic glued to the formed ends of the steel base.

Pictures of the boards inside the tuner display were posted here earlier.

I had envisioned changing the layout of the tuner's LEDs from Eric's original layout but on his urging I decided to try it out without any modification, and I agree that it is a good layout the way it is.  If anything, the circle of LEDs might need to grow a little because I'm finding that I tend to view it only peripherally without focusing or fixating on the individual LEDs (more on this later), and a little larger display might help that.

The response of this tuner is unmatched for speed.  If you have ever tried using any guitar or instrument tuners that are inserted in the audio chain (usually on a signal that is not volume modulated), you will know how slow and unpredictable the response can be.  Low pitches are particularly laggy, due in part because it takes longer to measure the period of a few cycles in order to establish the frequency.  The D-Lev tuner operates in the digital domain and can display pitches down to fractional Hertz as easily as it does higher pitches.  The response is virtually instantaneous, and certainly far faster than any hand or arm movements can be made.

I didn't really understand Eric's LED layout at first, but it makes a little more sense now that I have used it.  It consists of two concentric circles of 6 LEDs, with one circle rotated 30 degrees from the other.  A center LED is illuminated at all times to act as a reference.  Each of the surrounding LEDs represents one of the 12 scale notes, and pulse width modulation of the LED drive provides a smooth brightness transition from one LED to the next.  Each semitone step in pitch alternately illuminates an outer- and inner-circle LED, so you get a clear indication of pitch changes because of the extra alternating in/out movement.

But judging peak brightness of an LED in itself is no way to determine if you are truly on pitch, and this is one of the reasons I was skeptical of the usefulness of a tuner with a granularity of only 12 LEDs.  But this is where the design gets really smart.  As it turns out, while the LED for a particular pitch is approaching peak brightness, the adjacent LED is also illuminated but decreasing in brightness.  When you are exactly on-pitch, the two adjacent LEDs are just going dark, and if you play with a little vibrato, they will flash like railroad warning lights on either side of the bright LED that represents center pitch.  This graded intensity overlap makes the seemingly-coarse 12 light display capable of a much higher degree of pitch accuracy than you would expect, and certainly more than adequate for theremin use. And as Eric points out, after a while you begin to see recognizable patterns while you are playing.

The 7-segment display also shows a smooth transition between octaves, with one number blending into the next instead of abruptly changing.  The volume display consisting of the column of four blue LEDs on the left is useful primarily for setting up volume curves.

I can't play up to speed using a visual aid such as this tuner, but I still think it is extremely cool and I see the value in this even for my own theremins.  I find that my brain seems to respond more quickly to audible feedback with pitch preview than it does to visual cues.  Like looking at keyboard keys while typing, I find that fixating and playing to a visual aid is not helpful, at least the way my brain is wired. Nevertheless, I do like having the display out past the pitch antenna where I can see the flashing of the two notes adjacent to the target pitch.  It's a great aid for ear training and for for observing your long-term pitch stability without having any musical accompaniment for reference. 

In summary, I like the design and I will keep the visual layout nearly the same when I make the next one.  I also like the convenience of the magnetic mount and the positive (even if a little bulky) RJ45 connectors and cat5 cables.  Some board changes need to be made, however.  Since the separate tuner module was originally envisioned to be mounted on or close to the main enclosure, no care was taken to pay attention to the assignment of the RJ45 connector pins other than to make the correct interconnects. It has become apparent that for reliable behavior over longer cable lengths I will need to use a differential transmitter and receiver to transfer data over the twisted pairs of the Cat5 cables, or at least make a half-hearted attempt to terminated the ends properly if tests show that I am able to stick with a single-ended drive.


Other Experiments
After spending even more time playing the D-Lev without finding any major issues (not only that, it really is quite amazing), I have decided to start thinking about building a more interesting enclosure.  I think that have come up with one that I like (it's a woodworking challenge nightmare), but before I do that I've decided to try to shoe horn the D-Lev into an EW Pro-like case that I had built a little over a year ago.  This was originally intended to house a substantially embellished analog theremin that I had been working on for some time that was based on Etherwave oscillators, but that project had grown into such an abomination of knobs that I was afraid to start drilling holes in my curved flame-maple front panel.  With the D-Lev I now have everything that I had implemented in analog, and much more, and I have no qualms about using it for this enclosure. 

One of the first tasks to make this work is to find some more compact antenna and volume inductors, so I have been testing a bunch of commercial and hand-wound coils.  Surprisingly there are quite a few that seem to work just fine, although the best that I have found are the Hammond/Bourns/Miller type of pi-wound inductors used in the Etherwave (inductors no longer manufactured as of Q2 2017) and some simple hand-wound coils made on grooved forms.

Shown below is a DIY ferrite-core inductor used to replace the air-core pitch coil used on the D-Lev prototype. It was necessary for this coil to fit inside a 5/8" i.d. phenolic tube used for the pitch arm of the EW Pro clone cabinet. The spool is made of 9/16" wood grooved on a lathe.  The core is 8mm x 50mm Mn-Zn ferrite.  As usual I wound this to fill up the grooves and then unwound to the required inductance, so the exact turn count is not known for now.  If this all works out I will make the final inductors on acetal plastic bobbins instead of wood.

Here is the inductor being tested on the pitch side.  This one measures about 2.6mH, which works out well with the particular pitch rod length that I am using. 

And here is a different inductor being tested on the volume side.  Although this one has only 4 grooves, the slightly large diameter and finer gauge wire give it a higher inductance of about 4.6mH.

Ultimately the pitch inductor and the board for the capacitive voltage divider must fit inside a 7/8" o.d.,5/8" i.d. phenolic tube used as the core for the extension arm of my EW Pro clone.  With the feedback wires it ended up being a tight fit.

The end of the pitch extension arm that plugs into the cabinet only needed one electrical connection when the plan was to use it with an analog theremin.  There is a brass cam (not shown) that was intended to serve as both the electrical connection and the mechanical lock for the arm.  But the D-Lev requires more connections to the pitch antenna.  Mounting the D-Lev analog front end board at the antenna base (using a miniaturized SMT version of the board) would have required too many electrical connections, but having the analog board inside the cabinet would require only 3 connections, so I decided to go that route.  Three grooves were machined in the phenolic tube and solid silver rings were inset into the grooves to be flush with the phenolic surface.  Three spring-loaded plunger contacts were placed inside the pitch arm receptacle to make contact with the slip rings when the arm is inserted and locked in position.

It all works, but it was a bear to make and assemble.  Definitely a one-off only.

Here is the merger of the prototype enclosure with the pitch arm of the new cabinet.  The volume loop has also taken the place of the aluminum plate.  It's all working, but I'm pretty sure that I need to optimize things.  I'm sort of working blind without access to or knowledge of the phase lock loop requirements, so I have to make changes and look for instabilities to really know where I am.

 


Posted: 1/13/2019 9:35:01 PM
JPascal

From: Berlin Germany

Joined: 4/27/2016

I am very impressed by everything that has been done here. The sound examples give a vague idea of what is still possible.

Posted: 1/14/2019 10:53:08 AM
pitts8rh

From: Minnesota USA

Joined: 11/27/2015

Although I have a different and completely unique design in mind for a later D-Lev cabinet, for now this is the EW Pro "clone" cabinet built over a year ago that I will be attempting to use for the next D-Lev build.  Instead of the flat back panel of the real EW Pro I chose to make something a little more stylish and appropriate to go with the curved front panel. The back and sides are made of Peruvian walnut and the front is of course flamed maple.  The LCD display will be roughly centered in the space above the knobs and two columns of four encoder knobs will be placed below the wood knobs.  If the large wood knobs weren't already previously fixed in place I might have arranged things differently, but I am looking into some means of making the encoder knobs visually correlate to the respective LCD display elements in lieu of having the encoders placed directly adjacent to the display (there is not enough room to do this).  The large knobs will probably be used as analog controls for Master Volume and Pitch Preview Volume. 

The removable pitch arm is unique as well.  There are no threads used for the antenna or the extension arm that plugs into the housing.  The phenolic tube shown earlier is free to slide a little inside the outer maple shell, and this is used as a clamping mechanism for the pitch antenna.  To assemble the pitch extension arm you begin by simply slipping the antenna into the hole in the brass cone. Then while holding the antenna horizontally and facing toward the player you insert the extension arm fully into the main housing and rotate the antenna to the vertical position where it clicks into position.  Cam-lock tension on the internal phenolic tube holds the arm tight and also clamps the pitch antenna tightly in the brass cone.

The next two pictures show the cam-lock fitting at the housing end of the rod and the antenna clamping mechanism at the antenna end (note that this picture was taken before the three silver slip rings were added to accommodate the extra D-Lev connections) :

The step that you see inside the hole is actually a separate movable cylinder that grips the antenna rod by offsetting slightly as shown.

The combination of a fully digital theremin and my pretty blatant ripoff of a Moog cabinet design should give something to offend the sensibilities of just about any theremin purist .  No offense intended, but I really do like the EW Pro cabinet style and I'm never going to buy a real one now that I can have something that is (in my opinion) far better.

Posted: 1/14/2019 3:00:54 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

Roger, you're outdoing yourself!  

Your tuner box and mount are really pro looking!  Having the LEDs protrude from the case is interesting, and shows their position quite well, unlike the case of sticking it behind a dark bezel.  I've contemplated lightly smoked Plexiglas here covering the whole thing to accommodate various colors and some positional indication.  The LED serial / parallel drivers can be cascaded via data in / out, something I avoided because of trouble with setup / hold race with very different parts on a very different product, but I'd be open to trying it here if you like.  That would give four active drive signals in the CAT5, which you could ground / +5V the other side of the pairs to provide power and signal integrity.  The board wiring change for the ICs would be minimal, and the SV FPGA code probably wouldn't take too much effort.  Let me know if you want to try that route.

Your wound ferrites look quite nice!  You evidently didn't have to scramble wind them to keep the self-C low?  Scramble winding is a high cost to entry for DIYing these parts, so if it could be avoided that would be great.  I also wonder how much C is added by having the inner turns close to the ferrite?

Interconnect on antennas is always an issue, and at one point long ago I had an analog oscillator that differentially sensed the quadrature current through a drive resistor instead of via a capacitive divider, which conveniently removed the need for ground and sense going to the antenna.  There are differential inputs on the FPGA that might be able to do this (? - I've never used them) but I think the scheme is likely inferior to the current C divider approach, and series R is a Q killer.

Since the perversity of the universe tends towards a maximum, I half expect by this point next year you'll be using the LED tuner exclusively for pitch and I'll have abandoned it for pitch preview! ;-)

Posted: 1/14/2019 3:41:38 PM
pitts8rh

From: Minnesota USA

Joined: 11/27/2015

 "Let me know if you want to try that route" -Dewster

If it would help avoid having to add drivers/receivers, then yes.  I seem to get odd behaviors with different cables, and some but not all can be attributed to noise pickup near the pitch field.  Anything that would improve common mode rejection would help.

"You evidently didn't have to scramble wind them to keep the self-C low?"

I tried, but there is only so much scrambling that you can do in a .100 wide channel.  One was wound on the lathe and the other by hand with no particular care except to prevent the wire from winding into smooth layers.

The ferrite itself has a high permittivity as well as permeability, and does promote coupling between pi sections, but apparently not enough to cause problems.  I looked for the SRF of one of the coils and it appears to be in the 1.3-1.5MHz range, similar to the Hammond coils (or whatever I have) for the same value of inductance.  I was looking at it with a tracking generator on a spectrum analyzer though, and while the absolute reading is questionable, it was similar to that of the Hammond measured the same way.

"Interconnect on antennas is always an issue, and at one point long ago I had an analog oscillator that differentially sensed the quadrature current through a drive resistor instead of via a capacitive divider, which conveniently removed the need for ground and sense going to the antenna.  There are differential inputs on the FPGA that might be able to do this (? - I've never used them) but I think the scheme is likely inferior to the current C divider approach, and series R is a Q killer."

It took some effort, but in the end the signal integrity of the sense signal is as good with this compacted form-factor as with the original spacious air-core inductor arrangement. I had to change the 100pf shunt cap to account for the capacitance of the twisted pair.  I now have a trimmer at the antenna, the distributed twisted-pair capacitance down the arm, and a final capacitor on the AFE board, all in parallel adding up to roughly 100pf, tunable with the trimmer to set the voltage amplitude.

I still have a number of things to look at before I call this good. It seems to play just fine, but I want to know more about it.

Posted: 1/14/2019 4:28:54 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"If it would help avoid having to add drivers/receivers, then yes.  I seem to get odd behaviors with different cables, and some but not all can be attributed to noise pickup near the pitch field.  Anything that would improve common mode rejection would help."  -- pitts8rh

Oh I see, it's more a matter of interference with the pitch field.  Do they make CAT5 cables with a shield around the whole thing?

"I still have a number of things to look at before I call this good. It seems to play just fine, but I want to know more about it."

I have a diagnostic SW load I'll send you that spits out high passed (~10Hz, 4th order) volume & pitch operating points out to SPDIF audio.  If you don't see sticking points or other weirdness with it, and if you don't exceed the maximum operating frequency of the LC DPLLs, then you're probably fine.  Try doing a smooth pitch sweep with your arm rigidly sticking straight out of your body and rotating your torso so your hand goes from far, to very near, to far from the pitch antenna, and examine the results with the vocal pitch app.

Posted: 1/14/2019 5:44:38 PM
pitts8rh

From: Minnesota USA

Joined: 11/27/2015

 "If you don't see sticking points or other weirdness with it, and if you don't exceed the maximum operating frequency of the LC DPLLs, then you're probably fine."

In the course of trying different inductors I have run into the sticking points that you describe, so I know what to listen for.  I have been doing the smooth sweeps by setting up a metal tripod near the pitch antenna to act as a coarse tuning element while I walk around some distance away to do the fine tuning.  Any diagnostics that you have would definitely be welcome at this point.

I did have a question come up yesterday though.  I have been noticing that my tuner is showing what I can barely hear, and that is a little pitch jitter that I don't know if I have had previously.  It seems to be related to environmental RF because it drops when I turn my fluorescent lights off, but it is still there to some degree even in quiet environments.  Thinking it might be something related to my antenna coil experiments, I reverted back to the air-coil/plate arrangement and it is still there.

I was wondering if this might have cropped up when you fixed the volume response-I seem to remember that you may have eliminated some smoothing on the pitch side as well?  Anyway, when you have a note centered on your tuner do you see a completely smooth fade out of the adjacent LEDs, or do you see a little pitch-noise jitter causing the adjacent LEDs to flicker randomly?  I am tempted to revert back to the SW version prior to the volume response fix just to check this, but I thought I would ask first.

Posted: 1/14/2019 7:54:01 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"In the course of trying different inductors I have run into the sticking points that you describe, so I know what to listen for.  I have been doing the smooth sweeps by setting up a metal tripod near the pitch antenna to act as a coarse tuning element while I walk around some distance away to do the fine tuning.  Any diagnostics that you have would definitely be welcome at this point."  - pitts8rh

I've got the dither cut to bare minimum, which seems to work OK with plates and air core inductors.  Any sticking points are most likely caused by low coil Q or low coil self-resonance frequency (i.e. high coil parasitic C).  Coil Q is also a gain factor in the LC DPLL feedback loop, so it can directly affect DPLL BW (bandwidth).

I'll send you a diagnostic load today.

"I did have a question come up yesterday though.  I have been noticing that my tuner is showing what I can barely hear, and that is a little pitch jitter that I don't know if I have had previously.  It seems to be related to environmental RF because it drops when I turn my fluorescent lights off, but it is still there to some degree even in quiet environments.  Thinking it might be something related to my antenna coil experiments, I reverted back to the air-coil/plate arrangement and it is still there.

I was wondering if this might have cropped up when you fixed the volume response-I seem to remember that you may have eliminated some smoothing on the pitch side as well?  Anyway, when you have a note centered on your tuner do you see a completely smooth fade out of the adjacent LEDs, or do you see a little pitch-noise jitter causing the adjacent LEDs to flicker randomly?  I am tempted to revert back to the SW version prior to the volume response fix just to check this, but I thought I would ask first."

I see some jitter when my hand is maybe 18" or more from the pitch plate, but I don't hear it.  That the jitter increases with larger hand distances is largely masked by the pitch dropping.  You're likely using much higher sensitivity so you're probably seeing more.

I did increase the pitch side BW, both in HW and SW, when I increased the volume side BW.  I think the volume side BW issue was likely more caused by the envelope generator, which I believe has been fixed.  Really early SW loads won't work anymore as the HW/SW interface has changed substantially.  I'll give you a SW load that reduces the pitch side SW BW some for you to test.

Posted: 1/15/2019 3:26:31 PM
dewster

From: Northern NJ, USA

Joined: 2/17/2012

"The ferrite itself has a high permittivity as well as permeability, and does promote coupling between pi sections, but apparently not enough to cause problems.  I looked for the SRF of one of the coils and it appears to be in the 1.3-1.5MHz range, similar to the Hammond coils (or whatever I have) for the same value of inductance.  I was looking at it with a tracking generator on a spectrum analyzer though, and while the absolute reading is questionable, it was similar to that of the Hammond measured the same way."  - pitts8rh

I don't have any numbers for guidance, but I might worry about that SRF being too low?  Any idea of the Q at the operating point?

I'm curious, have you tried measuring the SRF of your pitch side air core?

I believe ferrite allows you to have longer, skinnier coils that are still fairly well coupled?  Not trying to pushback or poop on your parade, and much easier to say than to do, but if I were going that route I might try a single layer solenoid wound with fine wire and placed some distance radially from the ferrite - rather like what you've done with the wooden bobbin but ungrooved and wound on top of it. Then play with the length & wire gauge until I got the target value.  The single layer and the radial distance should help minimize self-C.  If I were scramble-winding I might consider a double coated wire just to space out the windings a bit more, and/or somehow put spacers between the layers in the donuts.

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