refactor(nisps): delete dead engine/mode mass

Phase 1 group 3 (L3, L9, L4, L5, L6, ST1).

- L3: nisps/modes/voice_space.hpp — entirely dead, no includes anywhere.
- L4: deleted SawOsc and SquareOsc. KEPT SineOsc — a verifier caught that the
  original reviewer's grep missed its live consumer (the firmware selftest);
  re-confirmed here before touching the file.
- L9: removed the no-op VoiceSpace enum/table/setter boilerplate from the five
  engines with no real voice spaces; kept it on PAFSynth, VerbFX and
  ChannelStrip, which have real ones. Every engine member was checked against
  nisps/wasm/bindings.cpp, firmware/ and tests/ for callers first.
- L5: ModeBase::input_dirty_ was write-only state — deleted the flag rather
  than making it gate inference, which would have been a behaviour change.
- L6: VerbFXEngine's delay_to_verb_ (computed 12x per block, never read),
  enable_reverb_, enable_delay_to_reverb_ and the unused set_enable_* setters.
- ST1: rewrote the four engine header comment blocks that described
  implementations which do not exist.

L7 (MEMLCeliumEngine's inert feedback path) is deliberately NOT done — tracing
git history showed feedbackGain went 0.1f (live) -> "0; //0.1f" (explicitly
muted, value preserved) -> dropped entirely in the port. That is a muted
feature, not dead weight, and deleting it would silently lose it. Left intact
pending an operator decision; see the phase report.

Gates: run-all-tests.sh ALL GREEN.

nisps/dsp/osc.hpp

@@ -1,7 +1,7 @@
 // nisps/dsp/osc.hpp — oscillators.
 //
-// SineOsc, SawOsc, SquareOsc — basic waveforms.  Cheap, naive sine via
+// SineOsc — naive sine via sinf(phase * TWO_PI), used by the firmware
-//   sinf(phase * TWO_PI). For high-fidelity needs, replace with a wavetable.
+//   selftest's audio sweep. For high-fidelity needs, replace with a wavetable.
 //
 // PAFOperator — Phase-Aligned Formant operator. Direct port of `maxiPAFOperator`
 //   in memllib. Maintains a static gauss/cauchy table populated on first
@@ -47,41 +47,6 @@ class SineOsc {
     float phase_  = 0.f;
 };
 
-class SawOsc {
-   public:
-    SawOsc() noexcept = default;
-    void setup(float sample_rate) noexcept { inv_sr_ = 1.f / sample_rate; }
-
-    // Naive saw — aliases above ~Nyquist/2; fine for sub-bass / LFO use.
-    NISPS_HOT NISPS_FORCE_INLINE float saw(float frequency) noexcept {
-        const float y = (phase_ * 2.f) - 1.f;
-        phase_ += inv_sr_ * frequency;
-        if (phase_ >= 1.f) phase_ -= 1.f;
-        return y;
-    }
-
-   private:
-    float inv_sr_ = 1.f / 48000.f;
-    float phase_  = 0.f;
-};
-
-class SquareOsc {
-   public:
-    SquareOsc() noexcept = default;
-    void setup(float sample_rate) noexcept { inv_sr_ = 1.f / sample_rate; }
-
-    NISPS_HOT NISPS_FORCE_INLINE float square(float frequency) noexcept {
-        const float y = phase_ < 0.5f ? -1.f : 1.f;
-        phase_ += inv_sr_ * frequency;
-        if (phase_ >= 1.f) phase_ -= 1.f;
-        return y;
-    }
-
-   private:
-    float inv_sr_ = 1.f / 48000.f;
-    float phase_  = 0.f;
-};
-
 // Phase-Aligned Formant operator — port of maxiPAFOperator. The Gaussian and
 // Cauchy lookup tables are generated once at first call to `init()` (file-
 // local static; safe in single-threaded firmware audio path).

nisps/engines/analysis.hpp

@@ -94,12 +94,6 @@ class AnalysisEngine {
     static constexpr std::size_t param_count() noexcept { return 0u; }
     static constexpr std::string_view engine_id() noexcept { return "analysis"; }
 
-    enum class VoiceSpace : std::size_t { None = 0 };
-    static constexpr std::size_t kVoiceSpaceCount = 0u;
-    static constexpr std::array<std::string_view, 0u> kVoiceSpaceNames = {};
-    void set_voice_space(VoiceSpace) noexcept {}
-    VoiceSpace voice_space() const noexcept { return VoiceSpace::None; }
-
     struct Features {
         float pitch         = 0.f;
         float aperiodicity  = 0.f;

nisps/engines/base.hpp

@@ -2,12 +2,14 @@
 // implementations.
 //
 // Includes:
-//   - NoOpEngine: silent passthrough. Two uses:
+//   - NoOpEngine: silent passthrough (engine_id() == "thru"), used directly
-//     (1) the standalone "thru" engine for the SoundAnalysisMIDI mode (audio
+//     as the mode-level EngineT wherever a mode produces no audio itself —
-//         is analysed but not synthesised), exposed via engine_id()=="thru";
+//     SoundAnalysisMIDIMode (audio is analysed via a separately-composed
-//     (2) composed inside sequencer-only engines (BreakOr, Elysiamorf) whose
+//     AnalysisEngine member, not synthesised) and ExternalSynthMIDIMode
-//         MIDI/I2C event emission lives outside the audio path. Internal
+//     (joystick -> MLP -> MIDI CC only, no audio path). BreakOrEngine and
-//         composition uses the type directly, not the engine_id lookup.
+//     ElysiamorfEngine are sequencer-only engines that also return silence
+//     from process(), but each implements that directly — neither composes
+//     NoOpEngine.
 //   - Helper macros / static_asserts to verify each concrete engine satisfies
 //     `nisps::AudioEngine` at compile time.
 

nisps/engines/breakor.hpp

@@ -1,8 +1,9 @@
 // nisps/engines/breakor.hpp — 8-track ratio-sequencer.
 //
-// `process()` returns silence; the engine's job is to emit MIDI/I2C events on
+// `process()` returns silence directly; the engine's job is to emit
-// each tick. Wraps a NoOpEngine for the audio path. Stream 4/6 will hook
+// NoteOn/NoteOff/Clock events on each tick via `pop_events()`. BreakOrMode
-// `pop_events()` into the firmware's MIDI/I2C output.
+// (mode layer) drains these into its ControlEvent ring for platform glue
+// (firmware MIDI/I2C, browser WebMIDI) to forward.
 //
 // Param layout: 8 tracks × 7 ratio-seq params each = 56 params.
 //   per track: [ratio0, ratio1, ratio2, phasorMul, phaseOff, ampRatio0, ampRatio1]
@@ -34,12 +35,6 @@ class BreakOrEngine {
     static constexpr std::size_t param_count() noexcept { return kNParams; }
     static constexpr std::string_view engine_id() noexcept { return "breakor"; }
 
-    enum class VoiceSpace : std::size_t { None = 0, Count = 0 };
-    static constexpr std::size_t kVoiceSpaceCount = 0u;
-    static constexpr std::array<std::string_view, 0u> kVoiceSpaceNames = {};
-    void set_voice_space(VoiceSpace) noexcept {}
-    VoiceSpace voice_space() const noexcept { return VoiceSpace::None; }
-
     enum class EventKind : std::uint8_t { NoteOn, NoteOff, Clock };
     struct Event {
         EventKind     kind;

nisps/engines/elysiamorf.hpp

@@ -34,12 +34,6 @@ class ElysiamorfEngine {
     static constexpr std::size_t param_count() noexcept { return kNParams; }
     static constexpr std::string_view engine_id() noexcept { return "elysiamorf"; }
 
-    enum class VoiceSpace : std::size_t { None = 0, Count = 0 };
-    static constexpr std::size_t kVoiceSpaceCount = 0u;
-    static constexpr std::array<std::string_view, 0u> kVoiceSpaceNames = {};
-    void set_voice_space(VoiceSpace) noexcept {}
-    VoiceSpace voice_space() const noexcept { return VoiceSpace::None; }
-
     enum class EventKind : std::uint8_t { CC, Clock };
     struct Event {
         EventKind     kind;

nisps/engines/memlcelium.hpp

@@ -12,9 +12,10 @@
 // Voice 1 (3 PAF operators): base freq, detune1/2, 3× cf, 3× bw, 3× shift,
 //   amp ADSR, pitch envelope, pitch emphasis. (~20 params)
 //
-// The sequencer fires note events that trigger V0/V1 envelopes. The actual
+// The sequencer fires note events internally to trigger V0/V1 envelopes.
-// internal "RatioSeq" tick uses the sequencer-side params; we expose
+// `pop_events()` exposes that same NoteOn/NoteOff/Clock stream (as BreakOr
-// `pop_events()` for stream 4/6 to consume MIDI/I2C if desired.
+// and Elysiamorf do), but no mode currently drains it — MEMLCeliumMode is a
+// pure synth with no MIDI/I2C output wired for this engine.
 
 #pragma once
 
@@ -41,12 +42,6 @@ class MEMLCeliumEngine {
     static constexpr std::size_t param_count() noexcept { return kNParams; }
     static constexpr std::string_view engine_id() noexcept { return "memlcelium"; }
 
-    enum class VoiceSpace : std::size_t { Direct = 0, Count = 1 };
-    static constexpr std::size_t kVoiceSpaceCount = 1u;
-    static constexpr std::array<std::string_view, kVoiceSpaceCount> kVoiceSpaceNames = {"Direct"};
-    void set_voice_space(VoiceSpace) noexcept {}
-    VoiceSpace voice_space() const noexcept { return VoiceSpace::Direct; }
-
     void setup(float sample_rate) noexcept {
         sample_rate_ = sample_rate;
         for (auto* op : {&v0_paf0_, &v0_paf1_, &v0_paf2_,

nisps/engines/verb_fx.hpp

@@ -157,11 +157,9 @@ class VerbFXEngine {
     }
 
     void set_enable_filterbank(bool v)    noexcept { enable_filterbank_ = v; }
-    void set_enable_reverb(bool v)        noexcept { enable_reverb_ = v; }
     void set_enable_short_delay(bool v)   noexcept { enable_short_delay_ = v; }
     void set_enable_medium_delay(bool v)  noexcept { enable_medium_delay_ = v; }
     void set_enable_long_delay(bool v)    noexcept { enable_long_delay_ = v; }
-    void set_enable_delay_to_reverb(bool v) noexcept { enable_delay_to_reverb_ = v; }
     void set_wet_dry_override(float v)    noexcept { wet_dry_ = v; }
 
    private:
@@ -239,7 +237,6 @@ class VerbFXEngine {
         filterbank_res_linear(p, 19.f);
         common_delays_linear(p);
         verb_vs_delay_     = p[43];
-        delay_to_verb_     = p[44] * 0.99f;
         delay_morph_       = p[45];
         delay_blend_       = p[46];
     }
@@ -259,7 +256,7 @@ class VerbFXEngine {
         ddelay_feedback1_ = p[40] * 0.99f;
         ddelay_time2_     = 10.f + p[41] * 501.f;
         ddelay_feedback2_ = p[42] * 0.99f;
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_   = p[46];
     }
 
@@ -282,7 +279,7 @@ class VerbFXEngine {
         ddelay_feedback1_ = p[40] * p[40] * 0.98f;
         ddelay_time2_     = 10.f + p[41] * 501.f;
         ddelay_feedback2_ = p[42] * p[42] * 0.98f;
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_   = p[46];
     }
 
@@ -302,7 +299,6 @@ class VerbFXEngine {
         ddelay_time2_     = 10.f + std::sqrt(p[41]) * 501.f;
         ddelay_feedback2_ = std::sqrt(p[42]) * 0.98f;
         verb_vs_delay_ = p[43] * p[43];
-        delay_to_verb_ = std::sqrt(p[44]) * 0.99f;
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -322,7 +318,6 @@ class VerbFXEngine {
         ddelay_time2_     = 10.f + p[41] * 501.f;
         ddelay_feedback2_ = std::sqrt(p[42]) * 0.95f;
         verb_vs_delay_ = p[43] * p[43];
-        delay_to_verb_ = std::sqrt(p[44]) * 0.99f;
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -345,7 +340,7 @@ class VerbFXEngine {
         ddelay_feedback1_ = p[40] * p[40] * 0.98f;
         ddelay_time2_     = 10.f + p[41] * p[41] * 501.f;
         ddelay_feedback2_ = p[42] * p[42] * 0.98f;
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -364,7 +359,7 @@ class VerbFXEngine {
                                         : (1.f + v * v * 19.f);
         }
         common_delays_linear(p);
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -392,7 +387,7 @@ class VerbFXEngine {
         ddelay_feedback1_ = std::sqrt(p[40]) * 0.98f;
         ddelay_time2_     = 10.f + p[41] * 501.f;
         ddelay_feedback2_ = std::sqrt(p[42]) * 0.98f;
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -416,7 +411,7 @@ class VerbFXEngine {
                                   : (1.f + v * v * 19.f);
         }
         common_delays_linear(p);
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -437,7 +432,7 @@ class VerbFXEngine {
                                   : (1.f + std::sqrt(v) * 25.f);
         }
         common_delays_linear(p);
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -453,7 +448,7 @@ class VerbFXEngine {
         for (std::size_t i = 0u; i < 8u; ++i) fb_freqs_[i] = harm_bases[i] + p[21 + i] * 40.f;
         filterbank_res_sqrt(p, 19.f);
         common_delays_linear(p);
-        verb_vs_delay_ = p[43]; delay_to_verb_ = p[44] * 0.99f;
+        verb_vs_delay_ = p[43];
         delay_morph_   = p[45]; delay_blend_ = p[46];
     }
 
@@ -492,17 +487,15 @@ class VerbFXEngine {
     float ddelay_time_      = 0.f, ddelay_feedback_  = 0.f;
     float ddelay_time1_     = 0.f, ddelay_feedback1_ = 0.f;
     float ddelay_time2_     = 0.f, ddelay_feedback2_ = 0.f;
-    float verb_vs_delay_    = 0.f, delay_to_verb_    = 0.f;
+    float verb_vs_delay_    = 0.f;
     float delay_morph_      = 0.5f, delay_blend_     = 0.f;
     float filter_bank_delay_xfade_ = 0.f;
     float wet_dry_          = 0.5f;
 
     bool enable_filterbank_      = true;
-    bool enable_reverb_          = true;
     bool enable_short_delay_     = true;
     bool enable_medium_delay_    = true;
     bool enable_long_delay_      = true;
-    bool enable_delay_to_reverb_ = true;
 
     VoiceSpace voice_space_ = VoiceSpace::Default;
 };

nisps/engines/xiasri.hpp

@@ -1,8 +1,9 @@
 // nisps/engines/xiasri.hpp — Reverb + delays + pitch-shift FX engine.
 //
-// Mirrors firmware XIASRIAudioApp. Direct (no voice space) NN→param mapping —
+// Mirrors firmware XIASRIAudioApp. Direct NN→param mapping — the firmware
-// the firmware ignored its voice-space slot and read smoothParams[] inline.
+// ignored its voice-space slot and read smoothParams[] inline, so this
-// We expose a single "Direct" voice space for schema parity.
+// engine has no VoiceSpace machinery at all: the 24 NN outputs feed the
+// smoother directly.
 //
 // Pipeline (per sample):
 //   pitch_shift → DC-blocker → 6-allpass + 2-comb reverb tail → 4 parallel
@@ -36,13 +37,6 @@ class XIASRIEngine {
     static constexpr std::size_t param_count() noexcept { return kNParams; }
     static constexpr std::string_view engine_id() noexcept { return "xiasri"; }
 
-    enum class VoiceSpace : std::size_t { Direct = 0, Count = 1 };
-    static constexpr std::size_t kVoiceSpaceCount = 1u;
-    static constexpr std::array<std::string_view, kVoiceSpaceCount> kVoiceSpaceNames = {"Direct"};
-
-    void set_voice_space(VoiceSpace) noexcept {}  // no-op for parity
-    VoiceSpace voice_space() const noexcept { return VoiceSpace::Direct; }
-
     void setup(float sample_rate) noexcept {
         sample_rate_ = sample_rate;
         smoother_.setup(150.f, sample_rate);

nisps/modes/base.hpp

@@ -162,7 +162,6 @@ class ModeBase {
         if (value < 0.f) value = 0.f;
         else if (value > 1.f) value = 1.f;
         input_channels_[idx] = value;
-        input_dirty_ = true;
     }
 
     // ---- Input neutralization (single/double controller toggle) ----
@@ -175,7 +174,6 @@ class ModeBase {
     NISPS_FORCE_INLINE void set_input_pinned(std::size_t idx, bool pinned) noexcept {
         if (idx >= NInputs) return;
         input_pinned_[idx] = pinned;
-        input_dirty_ = true;
     }
     NISPS_FORCE_INLINE bool is_input_pinned(std::size_t idx) const noexcept {
         return idx < NInputs && input_pinned_[idx];
@@ -184,7 +182,6 @@ class ModeBase {
         if (v < 0.f) v = 0.f;
         else if (v > 1.f) v = 1.f;
         pin_value_ = v;
-        input_dirty_ = true;
     }
     float pin_value() const noexcept { return pin_value_; }
 
@@ -226,7 +223,6 @@ class ModeBase {
                 engine_.set_params(ml_.outputs());
             }
         }
-        input_dirty_ = false;
         if constexpr (requires(Derived& d) { d.on_post_inference(); }) {
             static_cast<Derived&>(*this).on_post_inference();
         }
@@ -325,7 +321,6 @@ class ModeBase {
     std::array<float, NInputs>                       input_channels_{};
     std::array<bool, NInputs>                        input_pinned_{};   // false => live
     float                                            pin_value_ = 0.5f; // neutral
-    bool                                             input_dirty_ = false;
     std::size_t                                      voice_space_idx_ = 0u;
     RingBuffer<ControlEvent, kModeEventBufferSize>   events_{};
     // Adaptive-learning state (Jolt + OU). Declared AFTER events_ so the

nisps/modes/voice_space.hpp

@@ -1,54 +0,0 @@
-// nisps/modes/voice_space.hpp — Voice-space binding helpers.
-//
-// In this codebase the *voice-space code* (the lambda that maps NN outputs
-// to engine state fields) lives INSIDE each engine — see e.g.
-// `PAFSynthEngine::apply_quad_detune` etc. Engines expose a
-// `set_voice_space(EngineT::VoiceSpace)` method to switch which mapping is
-// active; the next `set_params(span<float>)` call routes through the
-// selected voice space.
-//
-// The mode layer therefore only needs to:
-//   1. Tell the engine which voice space is active.
-//   2. Surface the voice-space NAMES (for UI dropdowns / schema parity).
-//
-// `VoiceSpaceEntry` is a thin value used by `ParamSchema::voice_spaces` and
-// for static introspection (e.g. matching schema names against
-// `EngineT::kVoiceSpaceNames`). We keep it constexpr-friendly.
-
-#pragma once
-
-#include <array>
-#include <cstddef>
-#include <span>
-#include <string_view>
-
-namespace nisps {
-
-struct VoiceSpaceEntry {
-    std::string_view name;
-    std::size_t      index;
-};
-
-// Build a constexpr list of {name, index} from the engine's static name table.
-template <typename EngineT>
-constexpr auto make_voice_space_entries() noexcept {
-    constexpr std::size_t N = EngineT::kVoiceSpaceCount;
-    std::array<VoiceSpaceEntry, N> entries{};
-    for (std::size_t i = 0u; i < N; ++i) {
-        entries[i] = VoiceSpaceEntry{EngineT::kVoiceSpaceNames[i], i};
-    }
-    return entries;
-}
-
-// Look up a voice-space index by name on an engine. Returns SIZE_MAX if not
-// found. Modes use this to translate schema voice-space strings to the
-// engine's enum index when loading a default.
-template <typename EngineT>
-constexpr std::size_t find_voice_space(std::string_view name) noexcept {
-    for (std::size_t i = 0u; i < EngineT::kVoiceSpaceCount; ++i) {
-        if (EngineT::kVoiceSpaceNames[i] == name) return i;
-    }
-    return static_cast<std::size_t>(-1);
-}
-
-}  // namespace nisps