Skip to content

SonKu P8.0: stereo LF transparency, actual L/R room paths, objective acoustic evidence and physical Windows tester - #693

Closed
masarray wants to merge 15 commits into
fix/sonku-p75-phase-safe-music-20261114from
feat/sonku-p80-transparent-bass-acoustic-evidence-20261114
Closed

masarray wants to merge 15 commits into
fix/sonku-p75-phase-safe-music-20261114from
feat/sonku-p80-transparent-bass-acoustic-evidence-20261114

Conversation

@masarray

Copy link
Copy Markdown
Owner

Why

Physical listening demonstrated that P7.4 cancels stereo and P7.5 can be shrill/thin with an insufficient spatial room. Neither is dearVR-equivalent. Follow architecture RFC docs/SONKU_P8_DEARVR_BENCHMARK_ACOUSTIC_ARCHITECTURE_20261114.md.

P8.0 production-safe research-only implementation

  • New opt-in ARMIX_SONKU_EXPERIMENTAL_P8_TRANSPARENT_BASS=ON requiring native XYZ + P7.5. Public LIGHT, old binary, sessions and user GUI unchanged by default.
  • Original LF headphone L/R survives in a prepared 220 Hz one-pole crossover, while only above-crossover direct signals use measured 3D KEMAR HRIR. High-band SIDE preserved without double-counting LF. No gain hack.
  • True L/R into six-surface early reflection image renderer and stereo diffuse FDN, replacing P7.5 duplicated MID/MID room feeds; maintains existing realtime cost model and JUCE single authoritative callback.
  • Exact 0%-MIX bypass invariant and anti-phase (60° WIDTH) exact waveform even for low bass + high frequencies.
  • New per-frequency 63–8k objective native PCM log, hard-left vs mono, 44.1/48k, ROOM-tail impulse response, motion/block size reproducibility and native Windows CTest.
  • Exact SHA portable Windows research artifact with native v3/ACK and explicit p8TransparentBass GUI readback (no false production capability). MIT KEMAR attribution retained.

Non-goals and release veto: This does not complete true spatial stereo source objects, multi-band calibrated decay, head-related tonal transparency, human dearVR AB or ASIO physical acceptance. Keep Draft and no public release/merging. P8.1/P8.2 remain required; report if objective test fails rather than loosening unknown spectral thresholds.

Field guide, issue #683.

@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Nov 14, 2026 •

Copy link
Copy Markdown

Deploying arsonmix with  Cloudflare Pages  Cloudflare Pages

Latest commit: d2d8fb0
Status: ✅  Deploy successful!
Preview URL: https://63143bfd.arsonmix.pages.dev
Branch Preview URL: https://feat-sonku-p82-sixsurface-mu.arsonmix.pages.dev

View logs

Copy link
Copy Markdown
Owner Author

P8.0/P8.1 implementation and reproducible 48kHz objective spectrum (2026-11-14)

Draft PR #693 implements research-only original stereo LF pressure below a prepared 220Hz crossover, highpass KEMAR spatial direct, highpass-only stereo SIDE, and true stereo sources for 6-wall ER + late FDN. The public LIGHT/default, old V1/V2/V3 FX schema, WIRE graph and installer remain unchanged.

A real FFT-free HRIR-bank calibration SonKuHrtfTonalCalibration.h computes common-ear corrective biquad coefficients in prepare from the 37-direction x two-ear MIT KEMAR horizontal bank. 44.1/48k measured presence cut -4.645 dB (2.2k Q .85), high-air shelf +3.306 dB (6.2k). Same coefficients for both ears, direct HRTF only; preserves L/R directional ratios without asserting a dearVR Clarity match. No per-sample trig/FFT/memory allocation.

Actual native PCM response at 48 kHz FRONT, 100% wet, ROOM OFF, 0.75 trim, duplicated mono (RMS output/dry input, synthetic sine):

  • P8.0 before calibrated EQ: 63Hz 0.6704; 125Hz 0.5164; 1k 0.6139; 2k 1.5207; 4k 1.3034; 8k 0.1447.
  • P8.1 after bank-derived EQ: 63Hz 0.6702, 125Hz 0.5192, 1k 0.5325, 2k 0.8739, 4k 1.0599, 8k 0.1873. Dominant 2k spike is significantly attenuated without a guessed bass boost.
  • Source preflight PASS 44.1/48k: actual phase-inverted stereo waveform at width60 EXACT after settled control; hard-left LF remains mostly in left ear; MIX0 dry bit-exact after clickless smoothing; six-wall+FDN impulse tail nonzero, multi-chunk parity; unit per-frequency diagnostics print raw evidence.

Important: this is objective evidence, not subjective sound acceptance, not a fully externalized stereo object renderer nor a calibrated real-room BRIR. Owner's P7.4/P7.5 poor headphone verdict stands. Physical ASIO and level-matched dearVR A/B still pending; no production/default changes. Full P8 JUCE Windows portable CI run 43603900923 is pending while this comment is written; artifact must pass integrity checks before being shared.

Keep issue OPEN and PR DRAFT; do not label "dearVR equivalent" until human listening approves actual musical tonality and room depth.

Copy link
Copy Markdown
Owner Author

P8.0+P8.1 measured LF / KEMAR tonality / stereo room — Windows candidate DONE at exact SHA

Result: experimental full Windows x64 CI #43603900923 SUCCESS, exact source d2d8fb03c82fa1d678b850ec240e435e34bfe019.

Test artifact: P8-Bass-Room-Research-Windows-x64; GitHub artifact ID 14117224898, size 15,061,261 bytes, retained through 2026-12-05. Full native JUCE ARMixConsole Release compile PASS (not just source tests). FIVE scoped MSVC/JUCE CTests PASS (v3 PCM ACK + previous state migration + P7.5 phase integrity + new P8 per-band/room evidence), PE/embedded approved GUI hashes and portable provenance/ZIP verifier PASS, hash-checked MIT KEMAR notice and P8 physical FAT instructions included. Source output measurements 44.1/48k PASS.

Measured 48kHz FRONT mono ROOM OFF PCM (RMS output/dry input, .75 trim, synthetic tones): 2 kHz BEFORE P8 bank EQ 1.5207x, AFTER 0.8739x; 4k 1.3034→1.0599, 8k .1447→.1873, 63Hz remains .6702x. Room impulse generates actual post-direct reflection+FDN tail; low-L/R stereo and polarity-inverted width-60 waveform preserved in source test.

Physical release veto still OPEN: no verified ASIO audio or headphone owner acceptance of timbre and cubic room, no blinded dearVR A/B, no real music frequency-by-band capture, no 1/2/4 insert XRUN p95/p99, and P8.2 more physically calibrated room + P8.3 genuine stereo binaural source-object architecture remain open. P7.4/P7.5 feedback stays FAILED. This build is research-only, not public installer; PR #693 Draft. Public/main unchanged. Do not market as 'equal dearVR' based only on green CI.

Copy link
Copy Markdown
Owner Author

MILESTONE A — research archival (closed, NOT merged). This P8.x SonKu research candidate is historical/experimental and is not the owner-approved P9.9 production path. The real approved main integration is #733, by selective forward-port, not by importing all earlier research commits. Original head d2d8fb03c82fa1d678b850ec240e435e34bfe019, source branch feat/sonku-p80-transparent-bass-acoustic-evidence-20261114, original base fix/sonku-p75-phase-safe-music-20261114 are pinned in docs/governance/MILESTONE_A_PR_BRANCH_RECONCILIATION_20261117.md on chore/milestone-a-pr-history-20261117. No branch, tag, or checkpoint deletion; no binary/DSP/GUI/session change. Archive the review only, retain complete source evidence. Reopen if explicit new research is approved.

@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