jheaps/JoshHeaps.Net

Echo acoustic ranging #5

Open

jheaps wants to merge 5 commits from echo-acoustic-ranging into master

Author	SHA1	Message	Date
Josh-Heaps	7f664e1b38	add deploy yml for gitea .NET / build (pull_request) Successful in 53s Details Playwright Tests / playwright-tests (pull_request) Canceled after 16s Details	2026-08-26 15:13:45 -06:00
Josh-HeapsandClaude Opus 5	5501c84e9b	Test the ranging pipeline against a simulated room The DSP tests drive the same files the page loads, so there is no second implementation to drift. The room tests put two real browsers through a real room with a synthetic microphone, which covers everything except the acoustics: slot rotation, leave and rejoin, and the stalled-device case where a chirp must be withheld rather than sent late. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>	2026-07-30 14:42:16 -06:00
Josh-HeapsandClaude Opus 5	d78251c928	Add the /echo page, capture and playback Worklet capture into a ring buffer indexed by AudioContext frame, with the chirp scheduled on the audio clock. Timers are avoided throughout: a hidden tab has its timers throttled to roughly once a second, which is long enough to put its chirp in another device's slot and misattribute every arrival in the round. A device that cannot reach its slot in time sits the round out rather than chirping late. Echo cancellation, noise suppression and auto gain are all requested off and warned about if the device refuses, since echo cancellation exists to delete exactly the sound being measured. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>	2026-07-30 14:42:05 -06:00
Josh-HeapsandClaude Opus 5	94054c0f1b	Add echo rooms, round scheduling and peak relay In-memory rooms with a background service that opens a round per room, rotates which device chirps first, and broadcasts the raw peak table once everyone has reported or the deadline passes. Rounds are announced ahead of time with a server timestamp so devices schedule against a shared reference rather than against message arrival. The server only ever holds sample indices; the geometry is solved on the clients and no audio is uploaded. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>	2026-07-30 14:41:52 -06:00
Josh-HeapsandClaude Opus 5	183a53234f	Add acoustic ranging maths and a room simulator Chirp generation, FFT matched filter, first-arrival detection with sub-sample peak fitting, the BeepBeep pair solve, triangle-inequality outlier rejection and classical MDS, plus a virtual room that synthesizes what each device would have recorded so the pipeline can be exercised without microphones. Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>	2026-07-30 14:41:39 -06:00