Repository navigation
Conversation
… late field control
…ching saved scene
…ity and direct music fidelity
… and P8.6 rejection gates
Deploying arsonmix with
|
| Latest commit: |
884e144
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://867eb4d7.arsonmix.pages.dev |
| Branch Preview URL: | https://research-sonku-p87-musical-a.arsonmix.pages.dev |
P8.6 native DSP preflight evidence (2026-11-15)Source-backed test run: https://github.com/masarray/arsonmix/actions/runs/43665939272, exact PR SHA PASS: actual C++ six-wall measured-KEMAR IR path, 44.1/48 kHz. Four horizontal wall path energy ratios P8.6/P8.5 = These are limited reproducible source tests (room preset 1, distance 1.6m). They are not perceptual externalization evidence and do not validate the owner's LARGE 28.8m / distance 6.1m / reflections 100% headphone scene. Full Windows JUCE executable / CTest / embedded resources / artifact qualification is checked by the separate Windows job; do not mark headphone FAT passed. Direct signal/music levels and HRTF are preserved, but wet reductions can decrease measured overall level; A/B must level-match after FX, without adding hidden automatic makeup. P8.5 remains golden; no main merge or public release. |
|
MILESTONE A: CLOSED UNMERGED as historical P8.x SonKu research. This opt-in/private experiment predates the approved SonKu P9.9 source selectively integrated in #733; no permission exists to bulk-replay the earlier research ancestry into production. Original source branch |
Isolated headphone-FAT candidate. DO NOT MERGE / NO PUBLIC RELEASE.
Owner physical feedback: P8.5 has slightly improved externalization, but sound is overly room-filling and comes from floor/ceiling. Target: focused ear-height virtual loudspeakers rotating coherently horizontally, music body preserved.
Hypothesis from actual source: the 6-surface KEMAR room (floor & ceiling) plus strong separate ER and diffuse 8-line FDN presence gains can dominate direct speaker cues at LARGE / reflections 100% / reverb 47%. Directional HRIR is already head-yaw-coherent in P8.5. This change intentionally does not invent new crossfeed, suppress stereo SIDE, alter 220 Hz bass or 750 Hz P8.4 fusion, touch user scenes, or claim proprietary dearVR equivalence.
P8.6 mode=5, behind private
ARMIX_SONKU_EXPERIMENTAL_P86_NATIVE_RTOFF by default: maintain all 4 horizontal early image paths at 100%, floor 24%, ceiling 15%, ER presence 76%, FDN tail presence 38%, all applied only on mode5. All six physical paths remain nonzero and one authoritative JUCE PCM graph remains. Native v7 continues 59-byte state v1–v6 compatible; modes 0/3/4 preserved, P8.3 1/2 forbidden. Real native bridge identity-bound GUI A/B adds P8.6 beside P8.2/P8.4/P8.5.Tests: native source per-wall energy ratio [1,1,1,1,0.0576,0.0225]; P8.5 vs P8.6 dry PCM bit-exact 44.1/48k at azimuth/yaw combinations, anti-phase and 64/128 blocks; room-on PCM contrast. Actual JUCE v7 queued audio ACK, schema, v6 restore and rejected-mode veto. Scoped Windows portable workflow builds real GUI and verifies embedded resources/provenance before ZIP upload.
This PR is stacked deliberately on P8.5 research PR #710, not main. No positive headphone FAT yet; coefficients are experimental. See
docs/SONKU_P86_HORIZONTAL_SPEAKER_FOCUS_20261115.md. Clean integration needed after user FAT. No release/merge.