Skip to content

SonKu P8.2: six-wall material spectra and shared Eyring LF/MID/HF room decay - #695

Closed
masarray wants to merge 9 commits into
feat/sonku-p80-transparent-bass-acoustic-evidence-20261114from
feat/sonku-p82-sixsurface-multiband-room-20261114
Closed

masarray wants to merge 9 commits into
feat/sonku-p80-transparent-bass-acoustic-evidence-20261114from
feat/sonku-p82-sixsurface-multiband-room-20261114

Conversation

@masarray

Copy link
Copy Markdown
Owner

Scope

Owner's physical headphone feedback is P8.1 much better. Preserve P8.1 tonal/bass as a stable benchmark and focus solely on ROOM realism. All changes remain on opt-in research branch; public main and P8.1 candidate unchanged.

  • New original 6-surface material table with intensity-to-pressure coefficients sqrt(1-alpha) and LF/MID/HF 250/4800 Hz filters after each directional KEMAR reflection. Materials differ for side/front/back/floor/ceiling and three room sizes.
  • Same materials determine independently precomputed Eyring three-band reverberation times across eight stereo FDN delay lines. Example model predictions: Studio 0.717/0.392/0.231s, Chamber 0.933/0.802/0.483s, Hall 1.716/1.219/0.650s. These are parametric estimates, not measured IR.
  • Production-off CMake flag ARMIX_SONKU_EXPERIMENTAL_P82_ROOM=ON requiring P8.1; identical native JUCE + source-test DSP; no extra audio authority/worker/heap/RT coefficient design.
  • Objective tests for image-path timing and absorption, separated LF/HF FDN tails, hall vs studio decay, 44.1/48k PCM motion/chunk parity, unchanged P8.1 tonal and phase integrity. Windows full exe, ACK and v1/v2/v3 restore, ZIP/hash notices and physical FAT.
  • Native snapshot capability and GUI label only when actual accepted FX instance has P8.2 enabled.

Qualifying veto: Tests and portable build do not prove a dearVR-equivalent room or good physical ASIO. Must A/B with owner-approved P8.1 at equal volume, ROOM OFF first, then 7.2/14.4/28.8m ROOM on. Keep draft until headphone quality, p95/p99 CPU, and objective room IR are approved.

Physical FAT guide. Track 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: 751ff03
Status:⚡️  Build in progress...

View logs

Copy link
Copy Markdown
Owner Author

P8.2 source checkpoint — real P8.1 headphone improvement preserved

Owner says P8.1 music now sounds much better than previous P7.4/P7.5, and explicitly requests further P8.2 work focused on room. The improved P8.1 Windows candidate remains available unchanged (artifact 14117224898). P8.2 is isolated on draft PR #695, no public main change.

P8.2 source commit now passes the source preflight Actions run #43607515719 at 44.1/48 kHz. New real DSP components:

  • separate 6-wall LF/MID/HF intensity absorption (pressure sqrt(1-alpha)) applied AFTER each measured binaural KEMAR reflected path. The cutoff split is 250 Hz / 4.8 kHz.
  • same original materials used for frequency-specific Eyring RT60 and 8-line FDN feedback coefficients, all computed at prepare. Studio theoretical [0.717,.392,.231]s, Chamber [.933,.802,.483]s, Hall [1.716,1.219,.650]s. These are NOT measured RIR values.
  • Source PCM source tests PASS: P8.1 48k ROOM-OFF tonal transfer unchanged (63Hz .6702, 2k .8739, 4k 1.0599, 8k .1873), six-wall ER timing and room-dependence, LF/HF and studio/hall FDN tail energy, 44.1/48k chunk parity, dry bypass/anti-phase.
  • No added worker, heap, threads, FFT, secondary audio authority; new CMake flag explicit opt-in only.
  • Full Windows JUCE build and 6 scoped native CTests underway as of writing; physical ASIO, true ear DRR/IACC/T20/T30, CPU at 1/2/4 inserts and matched dearVR A/B STILL REQUIRED.

P8.2 model is room-specific parametric fidelity research; do not claim perceptual dearVR parity from numerical tests. Keep Issue #683 OPEN, PR #695 DRAFT.

Copy link
Copy Markdown
Owner Author

P8.2 exact-source Windows candidate — 6/6 native CTests PASS, owner headphone A/B still pending

Baseline: The owner confirms P8.1 sounds much better; it stays preserved in its independent artifact 14117224898. P8.2 is ONLY a new acoustic room model in this Draft, not a replacement for approved tonal/bass rendering.

Exact tested commit 751ff03fa27fbf1cf623eda002af50978c7888e5. Windows P8.2 CI run 43607515719 completed SUCCESS:

  • Linux 44.1/48k source preflight + approved GUI JS tests SUCCESS.
  • Actual Windows x64 JUCE 8.0.15 ARMixConsole full Release executable COMPILED (no source-only proxy).
  • Scoped MSVC/JUCE native CTests 6/6 PASS, including original v3 audio-applied XYZ ACK/legacy v1-v2 restore, P7.5 stereo, P8.1 tonal/dry/phase evidence compiled with P8.2 enabled, and new P8.2 real six-wall/three-band FDN time/energy/PCM parity.
  • Exact embedded approved /preview/ GUI assets SHA, PE provenance, portable ZIP verifier, BUILD_PROVENANCE.json, MIT KEMAR third-party notice, and P8.2 field-FAT guide hashes verified.
  • P8.1 ROOM-OFF 48k transfer unchanged under P8.2 source build: 63Hz 0.6702, 125Hz 0.5192, 2kHz 0.8739, 4kHz 1.0599, 8kHz 0.1873. Studio ER/late and hall ER/late differ measurably. Illustrative Eyring LF/MID/HF model RT60 Studio [0.717,0.392,0.231]s; Chamber [.933,.802,.483]s; Hall [1.716,1.219,.650]s. NOT measured physical RT60.

Portable test artifact: ArSonMix-SonKu-P82-Multiband-Room-Windows-x64-43607515719, GitHub Actions artifact ID 14118867259, 15,065,808 bytes, expires 2026-12-05. Standalone tester only, no public installer.

Release veto: Still requires owner headphone A/B versus P8.1 with ROOM OFF (must match) and ROOM ON 7.2/14.4/28.8; no physical ASIO XRUN/CPU p95/p99 qualification, no actual room RIR DRR/IACC/T20/T30, no proven dearVR listening parity. This is a material-based parametric P8.2, not dearVR IR/Clarity. Never merge/release this branch on CI alone; PR stays DRAFT, Issue #683 OPEN.

Copy link
Copy Markdown
Owner Author

P8.2 owner listening report: musical quality GOOD, spatial speaker externalization FAILS (internal/overhead). Source diagnostic P8.3 PR #697 reproduces exactly direction-invariant anti-phase SIDE with ROOM OFF at 44.1/48k while lateral mono KEMAR ear sign remains correct. Keep P8.2 artifact as approved tonal baseline; no production release or parity claim pending P8.3.

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 751ff03fa27fbf1cf623eda002af50978c7888e5, source branch feat/sonku-p82-sixsurface-multiband-room-20261114, original base feat/sonku-p80-transparent-bass-acoustic-evidence-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