PRIMARY DSP REFERENCES REVIEWED — 27 OCTOBER 2026

Mutable Instruments, Plaits feature documentation:
https://pichenettes.github.io/mutable-instruments-documentation/modules/plaits/
Relevant approach: band-limited synthesis at the base sample rate, smoothed
controls, and the relationship between amplitude and low-pass-gate brightness.

Mutable Instruments, Plaits manual:
https://pichenettes.github.io/mutable-instruments-documentation/modules/plaits/manual/
Relevant approach: variable pulse/saw shapes, a gentle middle feedback region
for FM, touch-driven dynamics and envelopes. These support curation rather
than treating every bright metallic sound as a hardware fault.

Electro-Smith / DaisySP oscillator interface:
https://github.com/daisyaudio/DaisySP/blob/master/Source/Synthesis/oscillator.h
Relevant approach: PolyBLEP options for band-limited classic waveforms on an
embedded audio platform. Evaluated as a reference, not copied or added here:
the existing Plaits VA path already provides band-limited synthesis.

Nigel Redmon, EarLevel Engineering, A wavetable oscillator — Part 3:
https://www.earlevel.com/main/2012/05/09/a-wavetable-oscillator-part-3/
Relevant approach: pitch-dependent harmonic tables and the risk of clicks at
abrupt table transitions. Our existing Satin/Spectral path already crossfades
bands. This pass also fades finite additive partials continuously near Nyquist.

APPLICATION TO THIS CHECKPOINT
The sound engine's existing licensed Mutable Instruments sources were edited
locally with their notices retained. No third-party presets, audio recordings,
product graphics or code from other firmware binaries were transplanted.
The new control slew and finite-partial fade are small original changes.
The Rings loop reorganization preserves its audio and reduces state traffic;
equivalence is tested independently against the scalar reference.

Further possibilities after this build's listening review: more original
spectral banks, constrained additive formant colours, and restrained pairwise
voice layering using the existing six tracks. A new large physical model or
multisample instrument must first justify both RAM and worst-case callback cost.
Flash space alone does not prove audio processing capacity.
