Let's Design and Build a (mostly) Digital Theremin!

(Literal) Beta Testing

The mouser cancellation fiasco (cancel culture! - I keed) forced me to purchase 100pcs of generic 104 caps and 100pcs of generic 2N3904 from Amazon.  Some of the Amazon reviews said that most of the ones they got were "bad" - disturbing because you never want to be in the position of having to laboriously check and bin these types of parts yourself.  I'd much rather get pedigreed parts from a reputable distributor and not have to think about it.

Anyway, I received the generic 2N3904 parts and decided to spot check them for beta (current gain).  Found this circuit on the web:

The base current is more or less constant across devices, so all we have to do is measure the base resistor voltage once, then measure the various collector resistor voltages, divide the two (Vc/Vb) and multiply by 1000 to get a rough reading of beta.

Testing 10 of the 100 generics I read very healthy betas between 350 and 380.

So I decided to test some genuine MCC 2N3904 I got from mouser a while back and have been using in the D-Lev kits.  Testing 10 of those I read betas in the 180 to 260 range.  Hmm.

I don't know, but I strongly suspect that the bad reviews on amazon are due to the non-standard pinout of the 2N3904 / 2N3906, where E & C are swapped.  This caught me off guard with the first wave of kits where I used "super" transistors which had the more dominant pinout, I had no clue the 2N3904 I'd been using for decades was the odd man out.  And I'm and MSEE and everything and I was totally unaware that swapping E & C would actually give a functioning circuit, albeit with very reduced beta.  E.g. swapping the generic 2N3904 E & C pins in the above test circuit yields a beta of 8 - enough to maybe sorta work in many circuits, but poorly.

Over time I've come to very much appreciate the 2N3904 / 2N3906 pair, they handle the small to medium signal range really well, and make good switchers too.

Eric, are you soldering all D-Lev boards yourself?

"Eric, are you soldering all D-Lev boards yourself?"  - Buggins

Yep!  Each D-Lev is lovingly hand crafted by a skilled artisan - me! ;-)

It's all through-hole (no tweezing fleas) and not too many components, so it's not a huge deal.  The inter-board and IDC cable connectors take a lot of soldering.  The most grueling part is the encoder board cable soldering, 16 of those little ribbon cable wires per 4 encoders (hence my strong desire to reduce the interconnectivity here).

Yep!  Each D-Lev is lovingly hand crafted by a skilled artisan - me! ;-)

It's all through-hole (no tweezing fleas) and not too many components, so it's not a huge deal.  The inter-board and IDC cable connectors take a lot of soldering.  The most grueling part is the encoder board cable soldering, 16 of those little ribbon cable wires per 4 encoders (hence my strong desire to reduce the interconnectivity here).

Wow! It's really great achievement to get such a big project to production, and beat all the competitors in product quality.
As for me, biggest achievement is to not abandon it in the middle of development. I cannot stay focused on single project.

"Wow! It's really great achievement to get such a big project to production, and beat all the competitors in product quality."  - Buggins

Thanks!  It may be technically pretty on the inside, but it the cardboard box and antennas have zero curb appeal, and books get judged by their covers.

"As for me, biggest achievement is to not abandon it in the middle of development. I cannot stay focused on single project."

Ah, me too.  But this project has so many aspects there's always something trivial to focus on when the bigger things are crying out for attention (a procrastinator's dream).

The heart of D-Lev Central just got some new flair!:

SW & Librarian'n

I incorporated formant phase control in the D-Lev software.  Also doubled the noise level pmod, which seemed too weak.  No glaringly obvious changes to the preset timbres, so I think I'll leave the presets alone rather than edit them en masse.  Decided to remove all read-only negative preset and profile slots, presets are now [0:240] and profiles are now [0:5], so absolutely everything is editable via the front panel.  This gives tons more room to store new presets as almost half of the space is now blank.

Incorporated these changes in the librarian, and while I was in there I added some array bounds checking, more flags for consistency, and made the serial port setting logic more robust.  Also added a "split" command to split container files into sub containers, a "diff" command to compare two preset files, and a "reset" command to reset the D-Lev processor.  When viewing preset slot or file contents, the profile knobs are now blanked out, and vice-versa, which hopefully makes it a little clearer which are which:

Above is the trixie preset - notice that all preset knobs start with a lower case letter and have values; profile knobs start with an upper case letter and don't have values.

I wish I'd learned the Go language a long time ago, it's a breeze to program with and doesn't fight me nearly as much as C++.  And it cross-compiles like a total champ.

Happy Holidays TW!

Nice Happy Holiday mod

Could this be known as the D-Lev Holiday mod?

Anyhew, Happy Seasons greetings to all!

SW & Librarian Update

In addition to the changes described above, the NOISE:vmod is now a linear gain modulation, much like the knee, but is applied above kloc rather than below it.  Interaction with the noise level knob is much reduced so it is easier to adjust, and I think it sounds a bit more natural now.

The RESON:mode sign has been reverted back to what it was ~2.5 years ago.  That change has nagged me ever since due to inconsistency with the negative / non-negative ranges.

All of the preset defaults have been massaged to make it easier to enable stereo and pitch correction.

As a result of these changes the presets, preset list, manual, quick reference, etc. have all been updated: https://d-lev.com/support.html. 

There have been some librarian changes as well (now v5):
- New "split" command to parse container files down to their sub-components.
- New "diff" command to compare a preset / profile file with knobs / slot / 2nd file.
- New "reset" command to reset the D-Lev processor.
- Reset now issued after eeprom or spi pump.
- "split", "pump", "dump" commands now use the file extension (flags removed).
- "stob", "pump", "dump", "split" commands: -f  flag added for consistency.
- "slots" command: -d  flag added for clarity.
- "ports" command: -p  flag added for safety / clarity.
- "stob" command: -hdr flag to enable bank file header comments every 10 slots.

There are two new librarian videos demonstrating new features and how to do an update:
https://youtu.be/PpFN7D4ixZs - what's new in v5
https://youtu.be/VgPS320gQyc - updating SW & presets

I must say that these sorts of semi-major updates - where all of the presets, documentation, etc. need updating - are getting somewhat harder to perform.  Automating the preset updates can take days working the changes backwards and forwards, there are many users to inform of them, and I worry about obliterating all of the useful edits they may have made to them.  So I like to give these updates sufficient gestational time to explore all possible changes and methods, which often requires multiple deep dives into the code (and, as the saying goes: "When code is written, only God and the programmer understand it.  Six months later, only God").

"Could this be known as the D-Lev Holiday mod?"  - RoyP

Ha ha - yes!  Happy New Year everyone!

Cow Not Included

Surely on every Thereminist's bread & butter, must have list, a cow voice: https://d-lev.com/audio/2023-01-03_moo_cow.mp3

Entire barnyards are crying out to be synthesized.

And a baritone voice that came out pretty good IMO: https://d-lev.com/audio/2022-12-17_i_heard_the_baritone.mp3