Build Project: Dewster's D-Lev Digital Theremin

Fascinating Roger!

As an experiment I ungrounded mine and tried various capacitors to ground.  0.01uF (103) gave pretty similar results to hard grounding, smaller values gave a ton more hum.  I believe it would take an enormous plate to get 0.01uF intrinsic (a 180m diameter sphere by my calculations).

I think the large plate reduces the noise because it's flapping around in the same 60Hz environment the Theremin is in.  Too bad it isn't a more physically practical solution.

I don't understand what the source of the 4.6Hz could be, nor why the 120Hz spur series always seems to be larger than the 60Hz series.

Also, thanks for that third analyzer view!  I was expecting no significant energy at even harmonics, but some at odd harmonics due to the square wave drive of the LC tank, which acts like a low pass filter.  This suggests that a 4:1 coil ratio, with the volume axis frequency placed below the pitch axis frequency by a 1:2 ratio would probably work without interference.

New AFE2 PCBs and Inductors

Here are some pictures of the new second-generation Analog Front-End (AFE2) boards that I laid out from Dewster's updated design. On the outside they are not very different from the AFE1 boards except for the use of a more compact IDE ribbon connector in place of the previous RJ45 connector.  Although an 8-pin cable connector is used this board requires only four lines, one less connection than the previous design, with the rest of the pins either unused or allocated to ground lines adjacent to signal lines. Where the previous board sent both zero and quadrature outputs back to the FPGA, this one sends only the XOR'd result of these signals with the intention of eliminating any threshold sensitivities that may have been caused by voltage variations within the FPGA itself. So for inputs the new board board requires ground, +5v (locally regulated to 3.3v), the same square wave oscillator drive provided by the FPGA board, and one RF sense signal taken directly from the antenna via the green terminal block connector.  There are two outputs: the inductor drive (also at the green connector) and the XOR output sent back to the FPGA via the ribbon cable.  

PDF schematic of the AFE2.

This is probably the last DIY version of this that I will make since I want to send out for a test batch of both through-hole and SMT boards:

Although not directly related to the AFE2 design change, Dewster is also experimenting with decreased inductance values to push both the volume and pitch oscillators to higher frequencies.

Here is my new 0.5mH pitch inductor made with 26AWG wire, replacing a 2mH inductor.  The pitch oscillator is now running at about 1.89MHz.

And the new 1.0mH volume inductor (previously 4mH).  The volume oscillator runs at roughly 1.23MHz.

I'm still using the AFE1 boards to test these new inductors, but with Dewster's new pitch-tracking hum filter and other changes, the stability at low frequencies is uncanny.  As a test the pitch field can be calibrated to "zero beat" way out past the volume antenna and you can go down to a 1Hz metronome tick with the tuner still showing a steady display.  If you have gorilla hands and you like to play with 12" octave spacings, you can probably set the D-Lev up to do just that, with adjustable linearity as well. 

If this isn't the closest thing to the Holy Grail of pitch-field configurability, you'll have to show me what is.

For some reason just discovered this building thread. Damn nice case, with the wooden side panels!
What effect does placing the coils close to the antennae have?

You mentioned winding them either on the lathe or by hand.
For what it's worth, for the rare occasions that I wind coils, I use the simple wooden contraption I built, to fix my Li-Ion battery powered drilling/screwdriving thingy (whatever it's called in English) onto, with "analog" speed control on the grip, which is nice for not ripping things apart.
The rod that's inside there instead of a drill, to which a coil base gets fixed, also has a painted-black cardboard disc around it, with one index hole in it. There is an LED on each side, one just shines, the other of same type is abused as receiver, amplified by 1 transistor or so, connected to the counter input of my el cheapo Chinese signal generator - et voila, my short attention span is not a hindrance anymore to make a coil with a certain number of turns
(no, the thing does not produce moon shine)

I saw you etched the PCBs yourself. Have you tried services like JLCPCB? People are very happy with the quality, they're fast & cheap, and they make the vias for you

"For some reason just discovered this building thread. Damn nice case, with the wooden side panels!
What effect does placing the coils close to the antennae have?" -tinkeringdude

The presence of the coils within the pitch or volume fields on the D-Lev does not negatively affect playability after the effects are calibrated out.  That is not to say that they don't have a large effect on the oscillator frequencies; they do, but this is the case with just about any theremin that uses compact inductors in series with the antenna capacitance.  The impedance over the length of the inductor itself transitions from low at the drive end to very high at the antenna end, and so you have to make a decision where you place the inductor.  If you put it near the circuit board and run a long wire to the antenna, that wire becomes part of the whole antenna structure and can distort the pitch field because of the horizontal distance before you reach the vertical antenna.  The better choice, although still a compromise, is to place the inductor at the base of the antenna.  In this way the antenna itself is the dominant high impedance element that is influenced by the player and has a predictable pitch field determined by the antenna shape.  The downside is that the low impedance drive wire feeding the inductor is now a fixed capacitance element in the pitch field that lowers the oscillator's frequency and to some extent can desensitize the influence of distant hand movements.

I look at the Etherwave (or generally any compact theremin), where the inductors are on the circuit board and absolutely everything is sitting right in the pitch field (even the volume loop, unfortunately), and I'm amazed that it all works.

I have found the D-Lev to be extraordinarily tolerant to significant changes in the antennas or the theremin's proximity to conductors. You can make pretty major changes to antenna lengths (or change from rod to plate antennas) or set up the theremin near a metal bench, and once you turn it on and perform an "acal" auto calibration, it just works.

"You mentioned winding them either on the lathe or by hand.
For what it's worth, for the rare occasions that I wind coils, I use the simple wooden contraption I built, to fix my Li-Ion battery powered drilling/screwdriving thingy (whatever it's called in English) onto, with "analog" speed control on the grip, which is nice for not ripping things apart.
The rod that's inside there instead of a drill, to which a coil base gets fixed, also has a painted-black cardboard disc around it, with one index hole in it. There is an LED on each side, one just shines, the other of same type is abused as receiver, amplified by 1 transistor or so, connected to the counter input of my el cheapo Chinese signal generator - et voila, my short attention span is not a hindrance anymore to make a coil with a certain number of turns "

A lathe is really overkill, and your setup, or even the hand-cranked fixture is perfectly good for winding these simple coils.  When I was a kid back in the 1960s I hand-wound a high voltage spark coil that required a full pound of #36 magnet wire for the secondary (don't know how many feet that is), and this was entirely wound in single tight layers with paper separators between layers.  This was all done with a simple wooden-crank fixture.  So these theremin coils seem easy!

"I saw you etched the PCBs yourself. Have you tried services like JLCPCB? People are very happy with the quality, they're fast & cheap, and they make the vias for you "

That's coming.  I only do the boards at home because I can have them finished in a couple hours to test the circuits (versus 3 weeks or so).  But the DIY boards need wider traces, larger pads, and have the via problem, all of which prevent making proper layouts with compact routing and ground planes to help prevent noise issues.  The next boards will be commercial prototypes, and I've been working on new layouts over the last couple of weeks.  I have a couple services in mind, but I'll look at the one that you suggested too.  Thanks.

It's been a while since I've posted any progress on the D-Lev "Pro", which is my cross between a mostly digital D-Lev innards and an Etherwave Pro-style cabinet that I started to build a long time ago.  I've been getting tired of working on this copycat cabinet that was started as more of a personal woodworking challenge than it was a good idea, but I do need to finish it. There are other more interesting and original cabinets to make once this thing is done.

Cabinet

After trial fitting and disassembling the cabinet a bunch of times I thought it would be safe to start finishing it, so I removed all of the decorative brass parts, and polished and then nickle plated them.  One batch of parts that I forgot to electro-clean started showing adhesion problems, so I had to strip and re-plate them, which is something I try to avoid because of the effort involved.  Here are all of the custom machined parts after plating:

The Peruvian walnut cabinet and flame-maple front panel were finished in semi-gloss lacquer and the front sanded and rubbed to a high gloss.  The walnut came out darker than I would have liked, but lighter black walnut was in short supply around here when I made this.  I also should have applied the graphics and lettering prior to applying any finish, but I had originally planned to screen-print black ink on a lacquer base coat and then overcoat it with more clear.  Instead at the last minute I decided to do toner transfer graphics, which required application of many more finish coats than I had planned, but the results were worth it.

Electronics

Besides taking care of some springtime household duties and trying to get some work done on my other Melodia theremin project, the push has been to get some pc boards made for the D-Lev. I was new to Diptrace software for schematic capture/pwb layout and took my sweet time getting things together so that all of the boards could be ordered at once to save on shipping.  The idea here was to test out a more optimized line and signal routing for the main board with an eye on EMI/EMC considerations, at least to the extent possible under the constraint of having only two board layers.  A modular, interconnected board approach was deemed the safest route for this second pass, since potential mistakes would be isolated.  My cabinet needs to have the layout flexibility that a single large board would not offer, and while any cabling to connect individual boards adds to the cost and labor, it was planned so that easy-to-assemble ribbon cables could be used in most cases.

I was having a problem getting answers out of the board vendor that I had intended to use for fabrication, and just by chance I happened to watch an EEVBlog video where Dave Jones was talking some boards that he received from JLCPCB, which is where I ended up going.  But as I was preparing to post this today I noticed that tinkeringdude had suggested the same vendor a while back and I had completely forgotten about it.  Anyway, even though my memory failed, luck prevailed and it all worked out.  I couldn't be happier with the results from JLCPCB.  They had the boards processed within 24 hours from the time I sent the gerbers (with online updates of every process milestone throughout fabrication), and I received everything six days after ordering!

This shipment consisted of 5 each main FPGA, encoder, and tuner boards, and 10 each discrete AFE3 (Analog Front End) and AFE3 smd boards.  The encoder boards are small and were panelized 12 per 4" x 4" board for a total of 60 individual boards.  With $17 for 3-4 day DHL shipping the grand total for 90 usable boards was about $53.  And they look very good under inspection as well.  I don't think I be etching my own pcbs anymore.  Here are the boards, minus a set that was populated:

This is the back of the D-Lev Pro front panel with the Main FPGA board mounted directly to the back of the LCD.  The encoders are mounted on individual interconnect boards to fit the curved panel, and the ribbon cables are soldered directly to the boards (they could use 4-pin connectors as well):

The volume AFE board that will be located at the base of the antenna connects to the Main FPGA board through an 8-conductor ribbon cable.  This cable will end up being quite short with a longer one on the pitch side to travel the length of the extension arm.  You can see one of my boo-boos on the main board - the alternate LCD connector footprint falls under the overhang of the FPGA demo board (I never made a new component for the demo board that would show the correct outline and I forgot about the overhang):

In this shot you can see the interconnect between the Main FPGA board and the LCD.  I have a pin extender in between to allow the main board to stand higher to allow for tuner and optical cable clearance:

This is an individual AFE3 board.  One is located at the base of or near each antenna:

And this is the SMD version of the same board, necessary only for my Pro cabinet design:

The remote tuner, front and back.  I used green and red because I wanted to trim leads only on LEDs that I don't intend to use later on the real thing.  I am a big fan of the remote tuner placed in the field of view while gazing past the pitch antenna, but it will probably be offered as part of the main board in later designs.

And giving credit where credit is due to the inventor, Dewster:

The really good news is that everything is working so far.  All the boards have been tested, and the reassignment of all the FPGA pins to fit the new optimized layout went off without a hitch (except for one error that I have made before and will make again).  No cuts or blue-wire jumpers, although the main board has some clearance issues that will get fixed on the next pass.  We'll see if it all works when put together... 

Wow Roger, lookin' real good!  You constantly amaze me with your fabrication skills!  That flame maple sure has a nice depth to it.  Love the logo too!

As I adhere to a rigid goofing-off schedule, I've been able thus far to dodge PCB layout - I'm sure the whole experience is pretty much no fun.

Caution, extreme nit picking: For the LP2950-33LPRE3 3.3V regulator I'd recommend tantalum caps at the I/O pins.  Oddly, the one at the output is critical for feedback (and oddly requires a certain range of ESR to work), so I'd mount it reasonably close to the regulator.  And tantalums don't have the short lifespan issues that electrolytics tend to experience.

I have a limited number of tantalums on hand, and since these boards are destined to be scrapped I took quite a few liberties on component types and values throughout the build, and that included electrolytics everywhere except on the SMD board.  My Mouser order didn't come until last Friday and I had these built by the previous Wednesday night using whatever was on hand.

As it turns out I have a space conflict between the main board and a metal fitting inside my case, so the compact stacked installation of the main board behind the LCD is no more and I have to remake the encoder wiring harnesses.  But on a happier note the surface-mount AFE is all packed inside the phenolic extension tube and wired up and seems to be working. It was sort of like building a ship in a bottle.  How well it's working I don't know because the audio isn't yet connected. But the tuner looks stable. 

Ah, OK, just wanted to make sure you were aware of that minutia.  (Note to self: investigate whether one could design an analog Theremin around oscillating 3-terminal voltage regulators...)

"It was sort of like building a ship in a bottle."  Ha, ha, great analogy!  I can imagine that it's a tight fit in there.  Surface mount sure saves on PCB real-estate, it make DIP stuff look like land of the giants.

DIY SMD Reflow Oven

This is a little off-topic, but it is directly related to my D-Lev build.  I have a habit of going off on tangential projects while in the middle of doing something else, so building a reflow oven for doing small-scale surface-mount prototype work just sort of happened while I was between D-Lev board runs.  Until now I have been doing surface-mount work with a PID-controlled hot plate supplemented by a handheld hot air rework tool, and I've been quite happy with the results.  The hot plate heats the board to just short of the solder paste's reflow temperature, and the hot air is then worked around to bring the component pads fully up to the reflow point.  With this method it is easy to ensure complete reflow on all sections of the board without overheating, but it does not allow the board temperature to follow the profiles recommended for the components and the solder paste.  It was time to acquire an oven.

At first I wanted to simply buy one of the many reflow ovens like this available on eBay:

But after reading about them it sounded like they had a number of shortcomings both in uniformity of heating and in the controller software.  While they appear to be well-made mechanically, and some users claim that they work fine out of the box, others feel that they need quite a bit of work to function as they should.  And they aren't particularly cheap;  this one is around $400.

On the other hand, toaster ovens are cheap, and even after modifications to turn them into reflow ovens they are still 1/3 to 1/2 the cost of the above unit (which still needs work).  Of course there is a ton of information out there on how to do this, so I read up on the subject and got started on my own.

The toaster oven I chose to use had a generous volume capacity and most importantly had four quartz heating elements (total 1500W) which can heat more quickly than the metal/ceramic elements that are found in most toaster and full size ovens. It has also been suggested that convection ovens with internal fans that circulate the heat provide more uniform board temperatures. This is the Black+Decker T0325XSB oven found on Amazon
for $62 that met all of the above criteria:

Conversion of a toaster oven to an effective reflow oven generally involves adding insulation to the air spaces around the walls and gutting the timer controls and replacing them with some type of temperature controller and relays to control the heating elements.

This B+D oven uses double wall construction for the top and ends with an air space to provide some minimal insulation to keep the outer shell from getting dangerously hot.  The rear "pizza bump" and the bottom are single-walled, and of course the glass door does not provide much insulation.

After removing the cover (this was not easy; it's designed to never come off) the first step was to gut all of the controls and the front panel.  The top and bottom quartz heating elements were then swapped to put the higher wattage elements on the bottom.  Next, about 1" of high-temperature ceramic wool insulation was added to the top and ends, and a sheet of woven fiberglass cut from a welding curtain was added around that to keep the insulation contained and the dust level down.

Insulating the back and the bottom presented more of a problem because they were each just a single wall of formed steel.  I decided that the only way to do it was to make new outer walls by fabricating sheet metal boxes to contain the insulation and attach them to each of the single walls using standoffs.  The bottom box is about 1" deep, just clearing the table when the oven's feet are attached. This is the rear box that fits over the pizza bump with a minimum of 1" insulation depth at the height of the bump:

I should add that these are not step-by-step photos.  They were taken when the oven was nearly complete and just before the cover when back on, and the above photo has all of the finished electronics in place.

Next up: Oven electronics and other details.

DIY SMD Reflow Oven (cont.)

Before I had even started on this project I had decided to use the open source Reflow Master temperature profile controller designed by Seon Rozenblum.  If you look for available controllers you will find a few that have come and gone in relatively short production runs. The above unit is (or was) still available and has a nice color display that shows both the target profile temperature curve and the current temperature progress against it.  It has four built-in temperature profiles for some common leaded and lead-free pastes, and more can be added if you get into the code.  It also has provision for an exhaust fan that can be switched on as a cooling aid after the reflow peak, helpful because the oven cool time is much longer than the temperature profile demands.  I chose not to use this, however, because an exhaust fan does nothing if there is no air intake, and if I have to be present to open the door for that intake, I don't really need the fan.

One of the other available temperature controllers out there can trigger a standard RC servo to prop the door open at the appropriate time, but it's still best to monitor things in person for the whole five minutes or so that it takes to reflow a board.

The space on the control end of the oven was limited, so I knew better than to try to fit this controller inside the oven body.  The solid state relays (SSRs), their heat sink, and a 5v power supply would be located inside the oven and the controller and display would be housed in a separate box (I haven't started on this yet).  A multi-pin connector on the back of the oven provides the connection port for the controller.

Even though this particular controller does not have separate outputs for the top and bottom heating elements, I decided to use separate SSRs for the top and bottom with separate toggle switches on the front panel so that I could manually preheat the oven or override the controller if the temperature for any reason lagged the desired profile.  The SSRs require heatsinking, so I machined a plate from 3/8" thick aluminum that would fit the space and also provide a mounting surface for terminal blocks and the power supply.  This plate had to fit around some obstacles (note the cutouts for the convection motor and the bell).

The only original control that I used was the timer switch.  A new front panel was made from stainless steel and a small fan was put in the bottom to push air over the heat sink and out the bottom and rear vents.

The choice of solid state relays wasn't trivial.  The most commonly available brand on eBay is Fotek, which is a real manufacturer of quality relays, but as far as I can tell nearly all on eBay are Chinese clones that have a habit of failing, often in the shorted state.  As it turns out, post-mortem analysis by a few inquisitive users has shown that some of the clones use severely underrated triacs - in one case someone found the clone SSR-40 amp relay board populated with a 6 amp triac.  Fortunately I found these Panasonic relays from Mouser for about $14 each.

I was now able to fire up the oven for the first time to get some idea of the heating and cooling rates and to see if it could keep up with the controller profile.  With all of the elements turned on, it could generally reach the peak temperature within the required number of seconds, but keeping up with those sections demanding a higher-slope temperature rise would be difficult.  Plus there was a lot of radiant heat escaping through the front glass, so I applied a sheet of polished aluminum with a viewing window to the inside of the glass to help reflect most of the energy before it even reached the glass. Here is the inside of the oven showing the clip-on thermocouple and the reflective panel on the door:

This was a huge help for increasing the rate of the temperature rise, but it presented a new problem.  The front glass was much cooler, except in the area of the window, where it was still very hot.  This could have exploded the glass in my face, but fortunately it held together and I was able to make a couple of sliding shutters (also polished aluminum) that can be left closed until the crucial moment where you want to view the reflow action, and then closed again after that.

As a final test before buttoning up the oven I timed the heating rates and found that I could go from room temperature to 185 Celsius in 145 seconds, which should be able to keep up with the profiles.  Some people have lined the inside of their ovens with a very expensive adhesive gold film (gold is highly reflective at infrared wavelengths) which improves both the heating and cooling rates.  And it looks impressive as well.  But I thought I would try going without at first because it does add probably over $100 just for the material.

Now I have to finish an enclosure for the Reflow Master controller.  I'm thinking it would be a good idea to put a standard PID temperature controller in the same box so I could use the oven for other purposes as well (powder coat baking, drying).  I also really need to go back and put some type of hardware temperature limit switch inside the oven too.  With the manual override switches on the front panel I don't need to have a China syndrome in my basement if I should forget and walk away with the heating elements left on...