From c3ebda07cc4c02d2ef60bc8afbd1ccceaeecd848 Mon Sep 17 00:00:00 2001 From: Aqua Date: Thu, 12 Nov 2026 17:27:27 +0000 Subject: [PATCH] Research doc: label pre-fix (c0e6b98) and post-fix (af0e44f) burst figures Co-Authored-By: Claude Sonnet 5.5 --- docs/research/articulatory-vocal-tract-sources.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/docs/research/articulatory-vocal-tract-sources.md b/docs/research/articulatory-vocal-tract-sources.md index b45fdc8..deba354 100644 --- a/docs/research/articulatory-vocal-tract-sources.md +++ b/docs/research/articulatory-vocal-tract-sources.md @@ -125,20 +125,20 @@ changes the level by -1.3 to +1.8 dB by vowel, so a single vowel is a noisy meas Plosive: word-initial /b/ closure is the tract wall's voice bar, -29.2 dB re the following vowel, /p/ -40.1 dB (`PlosiveClosureCarriesTheVoiceBarOnlyWhenVoiced`); /bɪn/ closure peak 0.0036 (was 0.0000, below 16-bit resolution), release 60-95 ms peak 0.0695, vowel 0.129. The wall radiation reads the first-cell pressure of the ladder (R0 + L0), band-limits it to 90-300 Hz and adds it at gain 0.7 (a DESIGN value for a voice-bar level of -30 dB; the primary level was not retrieved, and the 2.4 manual's skin radiation, Birkholz & Zhang 2025, is the reference to read). In a vowel the term moves the f0 line by under 1 dB; the 200 and 300 Hz harmonics move up to +1.5 dB on one open vowel (wall-only measurement, Soul pass). The band is two second-order sections, not a brick wall: with the mouth and the velum shut the bar sits about 25 dB down above 300 Hz. The burst -itself (difference of renders with and without `release_flow`) peaks at about 0.006-0.008, -24 dB re the vowel peak; the release peak in the table above is the voicing breaking through as the contact opens. Burst gain 0.06 is a DESIGN VALUE. +itself (difference of renders with and without `release_flow`), before the #47 fix (c0e6b98; see the re-measure below for the post-fix figures), peaked at about 0.006-0.008, -24 dB re the vowel peak; the release peak in the table above is the voicing breaking through as the contact opens. Burst gain 0.06 is a DESIGN VALUE. Weak spots found in the renders, 2026-11-08 (not re-measured on 2026-11-12 unless named in the next paragraph): (1) the vowel level is not steady, it rises for about 100 ms after onset (ɑ: rms 0.028 at 80 ms to 0.079 at 160 ms) so any window early in a vowel reads low; the cause is not isolated (surface smoothing and the pressure ramp are the suspects). (3) The voice bar's level is a design value and its band (90-300 Hz) assumes f0 of 90 Hz and up. (4) Only f0 and level of the first vowel are checked per cue; the cues' perceptual strength is unknown. Everything here is a measurement, nothing was heard. -Re-measure 2026-11-12 on main c0e6b98 (`tools/formant-check/weak_spots.py` over renders from `tools/aqua-listen/kits/tract-weak-spots/render.sh`; the renders reproduce the table above: /ɪ/ 375/2024 Hz, /ʊ/ 486/894 Hz, murmurs -11.81 and -9.64 dB, /i/ in /nibu/ +0.21 dB). Unheard. +Re-measure 2026-11-12. The stress, nasal and lax-vowel figures were taken on main c0e6b98, before the #47 burst fix (which does not touch them); the plosive-release bullet gives the c0e6b98 figures as "before" and the main af0e44f figures, after the #47 fix, as "after" (`tools/formant-check/weak_spots.py` over renders from `tools/aqua-listen/kits/tract-weak-spots/render.sh`; the renders reproduce the table above: /ɪ/ 375/2024 Hz, /ʊ/ 486/894 Hz, murmurs -11.81 and -9.64 dB, /i/ in /nibu/ +0.21 dB). Unheard. - Stress level (100-170 ms of each vowel, stressed against unstressed, f0 rise split from the rest by a render of the unstressed vowel at 148.4 Hz): total i +1.22, ɑ +1.67, u +2.07, e +2.99, o +1.69 dB (mean 1.93, sd 0.60); the f0 rise alone -0.80, +0.03, +0.17, +1.23, -0.07 dB; the rest (the amplitude factor 1.30 = 2.28 dB nominal, with the longer vowel) +2.02, +1.64, +1.90, +1.76, +1.76 dB (sd 0.14). The scatter between vowels is the f0 rise; the amplitude cue itself is even. No level the f0 rise cannot move was found cheap, so the five-vowel-mean basis stands (`aqua-body:question:stress-amplitude-calibration-basis`); the rest sits 0.46 dB under the nominal 2.28 dB, cause not isolated. - Plosive release: a defect was found and fixed. `release_flow` was held at the burst gain from the plosive's gesture on (a surface holds its first value before its first point and its last after its last), so every later contact state made noise: the difference of the renders with and without `release_flow` peaked at 0.0302 in the final /n/ and tail of /bɪn/ (above its 0.0226 at the release, 81 ms), 0.0126 in /bʊn/; in /nibu/ (flat) the leading /n/ carried it backward: peak 0.0688, rms 0.0067 against the /n/'s own 0.0058. - After the fix (`BurstGainAroundGesture`: zero before and after the plosive, the gain through it) the same difference is 0.0 from 120 ms after the release and the burst itself is unchanged (0.0226 at 81 ms in /bɪn/). Burst alone, after the fix: /bɪn/ 0.0226 (-15.1 dB re its vowel peak 0.129), /ba/ 0.0142 (-24.0 dB re 0.225), /bʊn/ 0.0130 (-24.3 dB re 0.213), - each falling to a tenth in 15-17 ms; the earlier "0.006-0.008, -24 dB" figure above does not hold for /bɪn/. The release peak of the full render (voicing breaking through) is -5.4, -6.3, -5.9 dB re the vowel peak. Test `PlosiveBurstExistsOnlyAroundItsOwnRelease`. + After the fix (`BurstGainAroundGesture`: zero before and after the plosive, the gain through it; figures from here to the end of the bullet are post-fix, main af0e44f) the same difference is 0.0 from 120 ms after the release and the burst itself is unchanged (0.0226 at 81 ms in /bɪn/). Burst alone, after the fix: /bɪn/ 0.0226 (-15.1 dB re its vowel peak 0.129), /ba/ 0.0142 (-24.0 dB re 0.225), /bʊn/ 0.0130 (-24.3 dB re 0.213), + each falling to a tenth in 15-17 ms; the pre-fix "0.006-0.008, -24 dB" figure above does not hold for /bɪn/. The release peak of the full render (voicing breaking through) is -5.4, -6.3, -5.9 dB re the vowel peak. Test `PlosiveBurstExistsOnlyAroundItsOwnRelease`. - Nasals: /m/ -11.81 dB, /n/ -9.64 dB re the flanking vowels, Burg N1 295 and 332 Hz. The 2 dB is not harmonic sampling: /n/ minus /m/ is 2.01, 2.17, 2.05, 1.61 dB at f0 110, 140, 170, 200 Hz. Band levels of the murmur (0-400, 400-800, 800-1600, 1600-3200 Hz): /m/ 36.3, 23.6, 21.3, 2.7; /n/ 38.4, 28.2, 16.8, 0.9 dB. The cause (the shorter oral side branch, 7.1 cm against 9.6 cm from the velum junction) is a candidate, not shown. Open. - Lax vowels: /ɪ/ and /ʊ/ are in the DSL table (`StoryVowelAreas`) and render in `WordRender` (/bɪn/, /bʊn/, isolated). Burg F1/F2: /ɪ/ 375/2024 against the tube chain 346/1987 (+8.5/+1.8 %), Hillenbrand 429/2034 (-12.5/-0.5 %), Peterson & Barney 392/1993 (-4.3/+1.5 %);