Repository navigation
SonKu P7.4: reconciled native XYZ Windows physical FAT executable and verified portable artifact - #688
SonKu P7.4: reconciled native XYZ Windows physical FAT executable and verified portable artifact#688masarray wants to merge 83 commits into
Conversation
…red HRTF + room authority Add callback-owned hysteretic energy/cross-correlation recovery of input M/S difference after binaural convolution, never mono-summing that content. Active only for strongly negative correlation, gain-slewed, finite/RT-safe; zero for correlated source and spatialMix=0. Apply to ROOM-OFF, experimental distance and native SPACE branches with consistent output trim, reuse scene limiter, preserve JUCE state/ACK ABI.
…PCM chunk parity at 44.1/48k Numerical tests exercise native impulse, FRONT/RIGHT/REAR/LEFT collapse rescue, independent L/R, 128-frame chunk parity, finite peak guard, positive-correlation no-op, silence reset, device prepare and spatialMix=0 dry identity.
…evidence Render original harmonic material at matched time/level as coherent front, anti-phase front and anti-phase rear with exact native core. Do not normalize WAVs or claim dearVR externalization.
…out full room renders
…i-stage quality gate
…, not Chromium/Windows FAT Use existing native core fixture in Fast UI plus a separate original three-WAV evidence and same-runner CPU benchmark. Full UI runner still protects GUI-changing SonKu branches. No network HRTF fetch or slow native installer blocks.
…ase side Use the existing per-sample 20ms correlation evidence for complementary positive-coherence hysteresis; no extra buffering, allocation, or new thread. Expose coefficient correction target strength to native FIR scheduler.
Prevent coherent stereo comb nulls by steering common L/R content toward measured center HRIR, while algebraically preserving left-minus-right directional transfer. Hysteretic coherent evidence ramps weights; coefficient updates remain on existing 16-frame cadence and 6ms per-tap slew. Reuse same two native source convolutions, SPSC/graph and state ABI.
…comb notch Assert per-ear 48k KEMAR spectrum old 60-degree stereo sum has a deep notch; coherent native FIR now converges to center HRIR within 14%, retaining side difference and preserving partition/phase/mono guards at 44.1/48k.
…23 baseline Use compile-time pinned baseline include only in the offline tool (never native JUCE) and add 2k harmonic plus 60-degree coherent FRONT/REAR cases. Existing full listening suite remains supported.
…3 in 5 matched WAVs Sparse checkout only immutable baseline SonKu headers; compile both native cores and assess exact coherent center, directional 2kHz comb recovery, anti-phase L/R RMS and peak safety. Keep benchmark same runner and upload both sides without Windows/Chromium rebuild.
…-FIR Mid/Side correction Document exact-phase cancellation, frequency-domain KEMAR comparison, invariant side transfer, bounded coherence ramps, matched baseline native WAV evidence and explicit limits beyond a real BRIR.
…e baseline-vs-candidate CI
…itrary PCM block splits Assert native Chamber 5m reflection+FDN still preserves both ears, phase/finiteness, bounded peak and exact 16k-sample 256-frame partition parity.
…n at deep HRTF notches
…t externalization limitations
…d render twelve headphone WAVs
…–4 instance callback cost
…tap epoch branch; prove PCM LSB parity
… remove FIR warning; constrain OFF-scene XYZ
…YZ on targeted Windows MSVC
…ce before optimizing verdict
…e JUCE restore behind strict opt-in
…CK-state compatibility
…Tests behind default-OFF option
…CK gates, keep public LIGHT/v2
…ative v3 readback and bounded geometry
…e without layout inflation
…rejection, preserve legacy UI assertions
…ct in scoped native CI
…al field trial Two-parent exact reconciliation (main + P7.3). Merge three disjoint shared files without rolling back the approved GUI, Player, Guitar FX or WIRE. Experimental research remains opt-in and all public defaults unchanged.
…dows physical tester
…rVR comparison gates
…T guide to experimental ZIP
Deploying arsonmix with
|
| Latest commit: |
c0c866d
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://c9daccfc.arsonmix.pages.dev |
| Branch Preview URL: | https://fix-sonku-p75-phase-safe-mus.arsonmix.pages.dev |
…s and retain explicit XYZ evidence
P7.4 exact-SHA Windows physical tester — artifact published (14 Nov 2026)Exact PR HEAD: Independently verified on this exact commit
CRITICAL: This is only a verified hardware-test candidate, NOT evidence that real ASIO load, multi-insert XRUN, .smix hardware Save/Open, subjective externalization, or dearVR parity passed. Keep PR #688 DRAFT, Issue #683 OPEN and public CMake default OFF. Perform actual field protocol before sound-quality or product-release approval. Source: field FAT guide. |
REAL HEADPHONE FIELD FAILURE — P7.4 rejected (14 Nov 2026)Owner has now listened to the P7.4 experimental Windows build and reports severe phase cancellation / music sounding destroyed. This is a FAILED physical audio qualification, overriding previous CI-only "test candidate" framing. The green build / three CTests / artifact SHA only established software packaging, not musical correctness. Do not recommend P7.4 artifact for live monitoring, do not promote/merge/release it and do not assert dearVR parity. Initial source-level failure mechanisms: per-ear sum of two offset directional HRIRs can cancel components of real stereo music; MID center correction is gated by 20ms correlation evidence and may miss spectrally complex partial coherence; side-rescue is a separate phase-unaligned direct addition. P0–P7.4 synthetic stereo tests did not prove natural complex-music preservation. These are mechanisms to investigate, not a verified physical root-cause attribution without matched input/output capture. A quarantined P7.5 research candidate is under review in draft PR #689: measured centered MID + unconditional SIDE, dedicated music-phase PCM tests, full Windows native build and headphone qualification still REQUIRED. Original production/default LIGHT must remain unchanged. Need actual dry/wet sample captures and ASIO + independent listening before any quality statement. Issue #683 stays OPEN. |
|
MILESTONE A: closed WITHOUT MERGE — historical audio research. This P7/P8 version was a private opt-in SonKu research stage, superseded as an active product review by selected owner-approved P9.9 #733. Production #733 selectively ported the accepted latest sound, not the full chain. Preserved original source: |
P7.4 — complete experimental native Windows mixer for physical 8D-SonKu testing
Branch exact head:
e3381f88a8481d48c6c1d3db94fef69d27bc56cb(at PR creation). Two-parent reconciliation: commit9374719fe71ed310027610782515c701d71cc072preserves latestmain@baf88ddd+ native XYZ P7.3cf34921d. All public CMake defaults remain OFF; alpha.23/installer/landing unchanged.Actual engineering changes
src/Main.cpp,CMakeLists.txt,.github/workflows/ui-smoke.ymlwith preserved Player/Main additions and P7.3 bridge/GUI; no force-push and no duplicated DSP engine..github/workflows/sonku-p74-windows-field.yml: fast pure PCM/GUI preflight, real MSVC/JUCE 8.0.15 fullARMixConsoleRelease executable withARMIX_SONKU_EXPERIMENTAL_NATIVE_XYZ_RT=ON, three focused JUCE source/ACK/state tests, exact source SHA, pinned WebView2 SDK, independent Windows GUI/embedded asset PE/provenance verifier, portable field artifact (no installer).sonkuExperiment.enabled=true, v3 native state, physical ASIO and dearVR parity unproven; no implicit production admission.Why DRAFT
A compiled executable and passing CTests are not actual hardware qualification. Await GitHub Action full-build completion, then run user-owned Windows ASIO field trial and loudness-matched blind listening. Do not call it dearVR-equivalent, do not merge or release without physical evidence, CPU admission, provenance review and explicit owner approval.
Closes neither the product gate nor #683. This PR exists to deliver the testable physical Windows field candidate, not a second research algorithm.