bcox/mini-800
1
0
Fork 0
Code Issues 36 Pull requests Projects Releases Packages Wiki Activity Actions

Analog quirks: shape every instrument toward its hardware #4

New issue
Open
opened 2026-10-11 23:39:08 +00:00 by bcox · 1 comment
bcox commented 2026-10-11 23:39:08 +00:00
Owner
Copy link

Every instrument here is built "clean": ideal waveforms, linear filters but for the four-pole filter's limited feedback, straight-line envelopes, and placeholder curves. CLAUDE.md says this is deliberate until the hardware has been measured, and that each quirk of an original gets a switch of its own, not one switch for the whole instrument.

This ticket stands for all of that work, on all five instruments. It is deliberately not a list. As a quirk is understood (measured from hardware, read from a service manual, or heard on a recording), split it out into a ticket of its own and do it there.

Where the known gaps are written down today:

  • each instrument's "Deliberately not done" paragraph in CLAUDE.md;
  • the opening comment of each header in engine/ and instruments/*/, which says what is a placeholder and what is a plain choice;
  • "Open decisions" in CLAUDE.md.

Rules that apply to every split-out ticket: provenance rule 3 (curves come from published manuals or from measuring a real instrument's audio, never from a ROM or firmware), and no shaping parameters invented ahead of the measurements.

Every instrument here is built "clean": ideal waveforms, linear filters but for the four-pole filter's limited feedback, straight-line envelopes, and placeholder curves. CLAUDE.md says this is deliberate until the hardware has been measured, and that each quirk of an original gets a switch of its own, not one switch for the whole instrument. This ticket stands for all of that work, on all five instruments. It is deliberately not a list. As a quirk is understood (measured from hardware, read from a service manual, or heard on a recording), split it out into a ticket of its own and do it there. Where the known gaps are written down today: - each instrument's "Deliberately not done" paragraph in CLAUDE.md; - the opening comment of each header in `engine/` and `instruments/*/`, which says what is a placeholder and what is a plain choice; - "Open decisions" in CLAUDE.md. Rules that apply to every split-out ticket: provenance rule 3 (curves come from published manuals or from measuring a real instrument's audio, never from a ROM or firmware), and no shaping parameters invented ahead of the measurements.
bcox commented 2026-10-18 05:14:24 +00:00
Author
Owner
Copy link

Parking #4 here on 2026-10-18, left open. Where it stands, so it can be picked up again.

Done (on main, through 1cd6096; CLAUDE.md has the detail of each):

  • Filter saturation, from documents and not measured: OTA cells (kOta) and the transistor ladder (kLadder) in engine/four_pole_filter.h; also in engine/sallen_key_filter.h, engine/state_variable_filter.h and engine/multimode_ladder.h. Each instrument has filter.saturation, at nothing as it comes, and renders byte for byte as before at nothing. --config=no_analog compiles it out. mini-Bit's filter.chip chooses between the CEM3328 and the SSM2044, and the SSM2044 is there for the Mono/Poly and Polysix to reuse.
  • Mixer drive (engine/mixer_drive.h) on every instrument with a mixer ahead of an analog filter, and the 2600's filter input overdriving.
  • Oscillator shaping: 3340 symmetry (engine/vco_symmetry.h) on the Prophets, the Jupiters and the SH-101; the JX ramp spread; the Juno ramp tracking and bend; the 2600's VCO-2 fold and sine shape; the Prodigy's triangle fold.
  • Voice drift (engine/drift.h) on fifteen instruments, with the variance.* drift settings. Retune comes from the page button, MIDI Tune Request (F6), --retune-cc and the automatable variance.retune. The Juno has the full variance.* set.
  • Bench: mini-XP's saturation on the Pi costs about a fifth more a voice, and up to about 24 voices fit.

Decided not to model (recorded in CLAUDE.md):

  • The Prodigy's sawtooth reset time, which at about 1 µs is a twentieth of a sample at 48 kHz.
  • The CEM 3340's top-octave flattening, which its HF track trimmer corrects. The datasheet gives no figure for what is left.
  • The CEM 3340's slow pulse fall, about a microsecond, with no figure in the datasheet.

Still open on #4, not started:

  • The ARP filters as ladders, which is a reading of the circuit still to be done.
  • Whole-instrument drift for the Omni and mini-RS, whose divide-down tones would want a master-oscillator drift.
  • The 3340's reversing hard sync, deferred until an instrument here is wired that way.
  • mini-CS's and mini-RS's filter chips, which no document read so far names.
  • The Poly-800 and DW-8000 filter, to be measured last on Ben's Poly-800.
Parking #4 here on 2026-10-18, left open. Where it stands, so it can be picked up again. **Done** (on main, through 1cd6096; CLAUDE.md has the detail of each): - Filter saturation, from documents and not measured: OTA cells (`kOta`) and the transistor ladder (`kLadder`) in `engine/four_pole_filter.h`; also in `engine/sallen_key_filter.h`, `engine/state_variable_filter.h` and `engine/multimode_ladder.h`. Each instrument has `filter.saturation`, at nothing as it comes, and renders byte for byte as before at nothing. `--config=no_analog` compiles it out. mini-Bit's `filter.chip` chooses between the CEM3328 and the SSM2044, and the SSM2044 is there for the Mono/Poly and Polysix to reuse. - Mixer drive (`engine/mixer_drive.h`) on every instrument with a mixer ahead of an analog filter, and the 2600's filter input overdriving. - Oscillator shaping: 3340 symmetry (`engine/vco_symmetry.h`) on the Prophets, the Jupiters and the SH-101; the JX ramp spread; the Juno ramp tracking and bend; the 2600's VCO-2 fold and sine shape; the Prodigy's triangle fold. - Voice drift (`engine/drift.h`) on fifteen instruments, with the `variance.*` drift settings. Retune comes from the page button, MIDI Tune Request (F6), `--retune-cc` and the automatable `variance.retune`. The Juno has the full `variance.*` set. - Bench: mini-XP's saturation on the Pi costs about a fifth more a voice, and up to about 24 voices fit. **Decided not to model** (recorded in CLAUDE.md): - The Prodigy's sawtooth reset time, which at about 1 µs is a twentieth of a sample at 48 kHz. - The CEM 3340's top-octave flattening, which its HF track trimmer corrects. The datasheet gives no figure for what is left. - The CEM 3340's slow pulse fall, about a microsecond, with no figure in the datasheet. **Still open on #4, not started:** - The ARP filters as ladders, which is a reading of the circuit still to be done. - Whole-instrument drift for the Omni and mini-RS, whose divide-down tones would want a master-oscillator drift. - The 3340's reversing hard sync, deferred until an instrument here is wired that way. - mini-CS's and mini-RS's filter chips, which no document read so far names. - The Poly-800 and DW-8000 filter, to be measured last on Ben's Poly-800.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
No results found.
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
bcox/mini-800#4
No description provided.