Skip to content

SonKu P8.8 (DRAFT): ear-height HRTF calibration independent of physical XYZ + better far distance cues - #714

Closed
masarray wants to merge 14 commits into
research/sonku-p87-musical-air-body-20261115from
research/sonku-p88-perceptual-height-distance-20261115
Closed

masarray wants to merge 14 commits into
research/sonku-p87-musical-air-body-20261115from
research/sonku-p88-perceptual-height-distance-20261115

Conversation

@masarray

Copy link
Copy Markdown
Owner

Headphone FAT — owner screenshot diagnosis, NOT released

Observed: P8.7 large 28.8m source at distance 7.3m and physical elevation -7.4° is still heard above and near. At 7.3m / -7.4°, original room physics places source at ~0.51m above floor (the minimum is 0.5m!) rather than 1.45m ear height. Lowering SOURCE ELEVATION is the wrong remedy. KEMAR measured generic HRTF can be perceived as vertically displaced even when physically level. Original P8.7 direct attenuates too little at far distances and un-HRTF stereo LF return is distance-INVARIANT. The mixer render screenshot also shows strong reverb 100%; no automatic preset changes.

P8.8 mode7 private opt-in only, P8.2/P8.4/P8.5/P8.6/P8.7 unchanged, rejected P8.3 1/2 remain disabled. Separate SOURCE ELEVATION (physical source XYZ, wall paths) from new EAR HEIGHT ALIGN · HRTF (-30..+20° user calibration of true measured KEMAR 3D spectral elevation). This changes only the direct HRIR lookup; at 0° correction mode7 equals mode6 with ROOM OFF, and physical0 with -12° HRTF trim equals P8.7 physical -12° direct PCM, while room source remains at ear level. No arbitrary phase inversion, no invented direction/HRTF. No guarantee a specific user will localize any chosen trim lower: individual pinna/headphone transfer is not KEMAR.

New P8.8-only distance curves ROOM ON: direct HRTF above LF 1/(1+0.27*(r-1.6)); original stereo LF 1/(1+0.10*(r-1.6)) to preserve bass body yet remove invariant near-head LF. P8.7 musical air and FDN LF resonance remain, P8.6 floor/ceiling suppressions remain. No global reverb increase, no second FX graph. Native v9 appends one 4-byte float to prior 59-byte .smix state (63 bytes); v1-v8 restore modes intact and rejected P8.3 never revived. JS selector/compact slider capability-gated on actual v9 readback/ACK, original scene/yaw untouched.

C++ direct PCM tests 44.1/48k and 64/128, old-mode parity, physical-vs-spectral elevation equivalence, 7.3m far-distance effect, anti-phase music and 100% ER/reverb LARGE. Actual JUCE v9 audio-applied ACK, state v9/legacy restore, 18 scoped Windows native CTests and PE/provenance resource ZIP. See docs/SONKU_P88_PERCEPTUAL_HEIGHT_DISTANCE_20261115.md. Headphone listening FAT still required; no public release/merge and not a claim of dearVR parity. Stacked on P8.7 PR #712 and #711/#710 research; extract safe clean patch only after accepted FAT.

@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Nov 15, 2026 •

Copy link
Copy Markdown

Deploying arsonmix with  Cloudflare Pages  Cloudflare Pages

Latest commit: 0bd3162
Status:⚡️  Build in progress...

View logs

Copy link
Copy Markdown
Owner Author

Actual owner screenshot + P8.8 native source evidence, 2026-11-15

Owner screenshot has 7.3 m source / -7.4° SOURCE ELEVATION / 28.8m LARGE / ER 100% / reverb 100% / yaw0° / speaker ANGLE 24° / ORBIT active (~10.3s), yet still close and localized above. The 24° control is stereo speaker spread, not world azimuth; ORBIT continually changes azimuth. Correct native spherical projection at 7.3m/-7.4° gives speaker height ~0.51m, essentially floor minimum0.5m. Do not automatically lower physical elevation further: separate generic KEMAR perceptual HRIR lookup from speaker geometry and give a user-calibrated reference after STATIC A/B.

C++ source test on Actions #43680250085 passed 44.1/48k and 64/128: ROOM OFF P8.8(mode7,trim0) PCM-exact P8.7(mode6), P8.8 physical0°/trim-12° PCM-exact P8.7 physical-12°/trim0° yet different room XYZ height; stereo anti-phase and yaw matrix; far 7.3m direct-only mixed scene output relative P8.7: 44.1k correlated −3.711dB, antiphase −2.426dB; 48k correlated −3.701dB, antiphase −2.425dB. The P8.8 LF distance law is gentler than the direct HRTF distance law to avoid thin bass; both are scalar/finite. No claims of subjectively solved height or dearVR parity.

Early source CI JS contract initially failed because a helper referenced UI capability state in a standalone unit harness; repaired pure geometry function and second preflight fully PASS at SHA cdb0276be164aad235fca8cba20edca1eedde50b. Full Windows native JUCE v9/18 CTests/GUI PE/portable ZIP pending. Baselines 0/3/4/5/6 untouched. No merge/release until owner physical FAT.

Copy link
Copy Markdown
Owner Author

P8.8 Windows FAT artifact: verified upload, NO release/merge

Source SHA: 0bd316246c054a5041d7ea48fdb89045f8ed371c. Successful Actions run 43680665844. 18/18 scoped native Windows CTests PASS, including P8.8 measured ear-height and distance PCM at 44.1/48k, 64/128 parity, native JUCE v9 audio-applied ACK and v1–v8 state restore. The genuine Windows mixer GUI x64 compiled (PE machine 0x8664, 15,065,088 executable bytes), approved embedded /preview/ GUI resource byte hashes verified, hardware/ASIO/headphone FAT unproven.

Real portable artifact: ArSonMix-SonKu-P88-Height-Distance-Windows-x64-43680665844, artifact ID 14148478262, 15,122,356 zip bytes, 10 entries. Independently downloaded, ZIP CRC all 10 entries PASS, Windows PE machine x64 verified, included portable_mode.txt, BUILD_PROVENANCE.json, SONKU_P88_FIELD_FAT.md, KEMAR license notice. Provenance sourceCommit exact SHA, sonkuExperiment.p88PerceptualAlignment=true, stateSchema=9, publicRelease=false, old modes P8.2/P8.4/P8.5/P8.6/P8.7 included. Artifact ZIP SHA256 00a1abe39f9ba6da339f88b16adbabf9bc7d5dc754593008c6ef13d80afb056f, independently equals runner digest. Runtime ASIO, audio device, headphone perceptual height and dearVR quality NOT YET physically qualified.

Owner headphone acceptance: compare P8.7 vs P8.8 with ORBIT off / STATIC, physical SOURCE ELEVATION=0°, head yaw0°, spread60°, ROOM OFF then LARGE. P8.8 EAR HEIGHT ALIGN 0°, −5°, −10°, −15°, −20°. At ROOM ON compare near1.6–3m and far7.3m; test reverb47% before stress reverb100% and level-match output. Do not interpret trim as a physical source position or promise perceptual correction from generic nonindividual KEMAR. No merge, public release or automatic preset/session changes. Research stack includes dormant rejected P8.3 ancestry; clean forward-port only after approved owner FAT.

Copy link
Copy Markdown
Owner Author

MILESTONE A: historical research archived / CLOSED UNMERGED. An earlier unqualified P8.x experimental candidate in the stacked SonKu research chain; current approved P9.9 production authority is selective integration #733, not this research ancestry.

Source branch: research/sonku-p88-perceptual-height-distance-20261115; exact original head: 0bd316246c054a5041d7ea48fdb89045f8ed371c; target: research/sonku-p87-musical-air-body-20261115. Frozen in docs/governance/MILESTONE_A_PR_BRANCH_RECONCILIATION_20261117.md on chore/milestone-a-pr-history-20261117. This closure does NOT import all commits, discard any work, delete any branch/checkpoint, alter audio, or authorize a public release. Future reactivation requires separately approved, scoped evidence.

@masarray masarray closed this Nov 17, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant