Skip to content

SonKu P9.6 DRAFT — preserve powerful stage bass & treble at 2–10m - #727

Closed
masarray wants to merge 2 commits into
feat/sonku-p95-immersive-cinema-ux-20261116from
feat/sonku-p95-p96-distance-reinforcement-20261116
Closed

masarray wants to merge 2 commits into
feat/sonku-p95-immersive-cinema-ux-20261116from
feat/sonku-p95-p96-distance-reinforcement-20261116

Conversation

@masarray

Copy link
Copy Markdown
Owner

SonKu P9.6 — keep far-stage distance without sacrificing bass and treble (PRIVATE DRAFT)

Base: P9.5 PR #725 exact head b895d92c1706ab0e098bfe49447fa7b83058f865. Ahead by exactly one atomic commit; P9.1/P9.2 owner checkpoint and production main untouched. Do NOT mix in parallel/competing P9.5 PR #724.

Operator-confirmed defect

Above 2m, Large Studio / Cinema / Venue sound progressively midrange-dominant, bass punch and treble musical clarity collapse. Source audit shows mode-9 whole-band direct-distance gain 1/(1+.27*(d-1.6)), i.e. at 10m only 0.306x direct ear pressure; much of the wet room stays strong. P9.5 RMS makeup cannot correct frequency-selective losses.

Implemented, gated, zero-ABI

  • Single existing native PCM graph and measured world-positioned 2×2 KEMAR stage speakers; P9-only, ROOM ON and distance above 2m. Split direct ears only, using causal complementary 170Hz LF / 4.2kHz HF filter with smooth 35ms distance-depth retargeting.
  • Preserve midrange distance / DRR structure and independent directional early reflections + eight-line late reverb; avoid headphone-locked dry bass or a separate new processor. At 10m target direct band pressures LF/mid/HF ~0.83/0.50/0.77 before crossover; actual sinus probes at 70/800/7500Hz ~0.79/0.50/0.65, compared to P9.5 ~0.306 broadband.
  • No changes to gain/mix/state schema v11, P95 makeup or 0.98 linked output peak guard, bypass animation, selector, or preset transactions. Explicit native readback p96DistanceReinforcement makes GUI stage label truthful.
  • Disabled by default via ARMIX_SONKU_EXPERIMENTAL_P96_DISTANCE_REINFORCEMENT. Requires P9.5 and P9.0 research engine. Older builds remain source-identical. Windows research FAT turns it on only for this specific branch.

Evidence / limitations

  • Isolated C++ 44.1/48k spectral probes, ROOM-OFF and <=2m direct invariance, stereo anti-phase, finite output tested locally.
  • Exact-head Linux full P90/P91/P92/P94/P95 and P96-on source gates; Windows actual JUCE, ACK, CTest and verified artifact remain CI gates.
  • Headphone FAT pending: compare same music and trim P9.5 vs P9.6 at 2m / 5m / 7.5m / 10m, frequency-band RMS, crest/LUFS, clipping/peak guard, and perceptual stage depth. If low/high reinforcement internalizes the image, changes timbre harshly, masks vocals, or triggers limiter pumping, reject candidate. Not a claim of independent measured Cinema/Venue BRIR or dearVR parity.

See docs/SONKU_P96_DISTANCE_REINFORCEMENT_FAT_20261116.md. No merge into production or release without operator headphone approval.

@cloudflare-workers-and-pages

cloudflare-workers-and-pages Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

Deploying arsonmix with  Cloudflare Pages  Cloudflare Pages

Latest commit: 720cec6
Status: ✅  Deploy successful!
Preview URL: https://62dab5e1.arsonmix.pages.dev
Branch Preview URL: https://checkpoint-sonku-p96-operato.arsonmix.pages.dev

View logs

Copy link
Copy Markdown
Owner Author

P9.6 exact-head source evidence — 720cec6

GitHub Actions run 43816786604, SonKu P9 exact-source C++ PCM preflight: SUCCESS; fast native contracts: SUCCESS.

  • SONKU_P96_INTEGRATED_SPECTRAL_AND_EXACT_LEGACY_PCM_PASSED: byte/bit-identical full PCM hash between P9.5 and P9.6 for ROOM OFF at 10m and ROOM ON at 2m (44.1k/48k, front and rear angles). This is strict full-process sample parity, not only similar RMS.
  • Actual full native KEMAR mode-9 core with ROOM geometry ON, wet=0, source distance 10m: 70Hz 2.240–2.320×, 800Hz 1.419–1.469×, 7.5kHz 1.848–1.924× candidate vs P9.5 reference.
  • Isolated direct-band tests: 44.1k 0.79124/0.50120/0.65213 and 48k 0.79119/0.50097/0.65610 bass/mid/treble amplitude retention at 10m. Anti-phase symmetry and finite output assert.
  • P9.5 source PCM at 0/90/180/270, two sample rates, 64/256 blocks also passed with P9.6 enabled. Measured ratios are electrical spectral probe results, not measured BRIR/real headphone perceptions.

Still pending: actual Windows JUCE GUI/CTest portable FAT and owner level-matched headphone FAT; keep DRAFT and do not merge/release.

Copy link
Copy Markdown
Owner Author

Milestone A — ARCHIVED / CLOSED UNMERGED (no code merge or branch deletion). This P9.6 prototype is an earlier step of the frozen SonKu research stack; selected P9.9 was promoted via #733 without replaying the ancestry.

Immutable original head: 720cec619b41aa601414d4b5ae664cb418f04ca1. Original source branch: feat/sonku-p95-p96-distance-reinforcement-20261116; original PR target: feat/sonku-p95-immersive-cinema-ux-20261116. Full 47-PR source-head inventory: docs/governance/MILESTONE_A_PR_BRANCH_RECONCILIATION_20261117.md on chore/milestone-a-pr-history-20261117 (planned for main). Closing this review surface is not proof of importing every experimental commit and is not permission to retune released SonKu or ImmerSIB. All branch refs, checkpoints and original code are preserved. Reopening requires a new, precisely scoped owner-approved reason.

@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