Listening trial: blinded A/B thruster noise sweep (static page) #7
Loading…
Reference in a new issue
No description provided.
Delete branch "aqua/listening-trial"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Static listening-trial page. Live at https://minds.gamecult.org/aqua/trial-52f9012/ (200 on every asset; MIME types checked).
Stacked on
aqua/playground(PR #6); retarget after #6 merges.tools/aqua-trial/web: consent checkbox gates Start; four 4 s renders compiled and rendered in the browser (faustwasm, same runtime as the playground); six comparisons (every pair once, order and A/B position randomized from a random seed); per sound a 1-5 "does it land?" rating and one word. Variant identity is never shown. Nothing is sent or stored; Export downloads JSON (patch, params, peak, rms, rating, word, time, UA, sample rate).thruster-trial.dsp:patches/thruster/thruster.dsp+noisecontrol scaling rasp and rattle (rumble untouched). rpm 600; noise 100/50/20/0. The "tone-forward mix" reading is mine, flagged in the README. Not loudness-matched.analyze.mjs: dependency-free; per-variant n/mean/sd/min/max/peak/rms/words, and Spearman rho + permutation p for peak and rms vs rating (predicts = |rho|>=0.5 and p<0.05).node test.mjs: 15 checks (design, validation, export, analyzer maths on synthetic data, CLI, and the patch through faustwasm incl. noise=100 sample-identical to thruster.dsp).Finding before listening: rms barely moves across the sweep (48 kHz: 0.0734/0.0712/0.0706/0.0705; peak 0.189/0.161/0.148/0.140). The variants differ mostly in texture, not level.
Never run in a browser. Nothing has been heard. The DOM glue, AudioContext playback and the download are untested; the analyzer has only seen synthetic exports.
🤖 Generated with Claude Code