TEXTES SUR LES MODELES 3
PSEUDO-CODE DE LA REDUCTION DES MODELES 4
-1- Absolute criteria 4
-2- Masking effect for one given time 4
parameters: 4
principle: 4
Algorithm: 4
-3- Masking effect for several different times 5
-4- Masking effect for impulse response 5
-5- Possible improvements 5
PSEUDO-CODE DE L'ANALYSE 6
-1- fft 6
-2- peakextraction 6
-3- analysis 6
-3A- Elementary analysis : peak matching and calculation of parameters 6
-3B- merge new elementary model and previous merged model 7
-4- Possible improvement 8
NOTES ON C VERSION 10
POSSIBLE CODE IMPROVEMENTS 10
PEAK EXTRACTION 10
ELEMENTARY ANALYSIS 10
MERGED ANALYSIS 10
SUMMARY OF OPTIONS FOR CODE VARIANTS: 11
PARAMETRES UTILISES POUR LES MODELES OBTENUS A L'IRCAM: 12
CLOCHE A VACHE 12
CLOCHES TUBULAIRES 12
CONTREBASSE 14
CROTALES 15
CYMBALUM 15
GLOCKENSPIEL 16
MARIMBA 17
PIANO 17
TAMTAM 20
TIMBALE 20
TUYAUX BEAUBOURG 20
VIBRAPHONE 21
DOCUMENTS
SUR L'ANALYSE POUR LES MODELES
DE RESONANCES
version C
doc programmeur
TEXTES SUR LES MODELES
(1) P-F. Baisnée, J-B. Barrière, O. Koechlin, M. Puckette & R. Rowe (1986) Real-time interaction between musicians and computer: performance utilizations of the 4X, Proceedings of 1986 International Computer Music Conference, La Haye, Berkeley: Computer Music Association, pp.237-240.
(2) J-B.Barrière, Y. Potard & P-F. Baisnée (1985) Models of Continuity between Synthesis and Processing for the Elaboration and Control of Timbre Structures, Proceedings of 1985 International Computer Music Conference,Vancouver, Berkeley: Computer Music Association, pp.193-198.
(3) Y. Potard, P-F. Baisnée & J-B.Barrière (1986) Experimenting with Models of Resonance Produced by a New Technique for the Analysis of Impulsive Sounds, Proceedings of 1986 International Computer Music Conference, La Haye, Berkeley: Computer Music Association, pp.269-274.
(4) Yves Potard, P-F. Baisnée, J-B. Barrière (1986) Interaction matériau organisation - version préliminaire d'un rapport de recherche, document interne Ircam.
(5) Yves Potard, P-F. Baisnée, J-B. Barrière (1988) Méthodologie de synthèse du timbre: L'exemple des modèles de résonance, A paraître dans le livre rendant compte du séminaire sur le timbre (Ircam 1985) chez Christian Bourgois.
(6) Baisnée P-F. Mode d'emploi des outils développés à l'Ircam
(7) Baisnée P-F. Mode d'emploi de la version Csound
(8) Baisnée, Barrière, Baudot, Freed ....(1989) Texte ICMC 89
PSEUDO-CODE DE LA REDUCTION DES MODELES
-1- Absolute criteria
Discard resonances for which
freq < min_freq
freq > max_freq
ampl < min_ampl
band < min_band
band > max_band
-2- Masking effect for one given time
parameters:
model
time
offset
slope
principle:
The mask for a given resonance is linear in log(freq) log(linear ampl) coordinates.
It is characterized by an offset(dB) and slope(dB/octave).
The mask produced by a given resonance (freq ampl band) at a given frequency f
is calculated as follows:
(1) mask = (ampl - offset) + (slope / log(2.)) * (log f - log freq)
Slope is positive or negative depending on whether the mask is computed
to the left or right side of the resonance, in a log-freq db-ampl representation).
Algorithm:
-models must be sorted by increasing frequency order
-calculate amplitudes of resonances at given time (the initial amplitude
stored in the model corresponds to time 0, amplitude at time t is given
by A(t) = A(0) exp (- ½ band t)
-lower resonances mask higher resonances:
slope = - (abs (slope));
for each resonance (from first to last)
if ampl > mask
keep resonance
if (ampl-offset > mask) /* current resonance produces a bigger mask */
current resonance is used for computing the mask for next resonance
(its ampl and freq are used in formula (1))
else discard resonance
-higher resonances mask lower resonances:
slope = (abs (slope));
for each resonance (from last to first)
if ampl > mask
keep resonance
if (ampl-offset > mask) /* current resonance produces a * bigger mask */
current resonance is used for computing the mask for next resonance
(its ampl and freq are used in formula (1))
else discard resonance
-3- Masking effect for several different times
Discard every resonance that is discarded for each of the given times.
-4- Masking effect for impulse response
Discard every resonance that is masked during the whole impulse response (until it reaches mask_floor dB).
-5- Possible improvements
new absolute criterium: min bandwidth = f(freq) <-> min filter coefficient
PSEUDO-CODE DE L'ANALYSE
-1- fft
input = [normalized] mono sound file
parameters = fftsize
window size
window type
In Lisp version:
window-size <= fftsize(2**n >= window) = [2,128000] bins
window = blackman-harris 4 for every window size, window values are normalized (* 1/sum of window values)
output = power-spectrum as db floatsams, not normalized
-2- peakextraction
input = db power spectrum as floatsams. Peaks are interpolated (4th order pylynomial interpolation)
parameters= min deltaDB between maxpoint and surrounding points, number of points for peak def (on each side of maximum), min dB ampl for a peak;
different values for windows 1 and 2:
mdb1 (-80) and mdb2 (-96)
output = pairs (frequency, dB power)
-3- analysis
N iterations (usually 3 to 8), with increasing window size
STEP A : ELEMENTARY ANALYSIS
STEP B : MERGE
input for each iteration: two set of peaks (same window & FFT sizes, t1 and t2)
(freq, dBpower)
previous model
parameters :
window size (plus FFT size if different) [from FFT]
t1 begin time for first window [from FFT]
t2 begin time for second window [from FFT]
freq threshold for peak matching
freq threshold for resonance matching (same as previous in Lisp version)
check option
double option
max b2 value for doubling (12Hz)
[min delta b2-b1 for doubling (new)]
-3A- Elementary analysis : peak matching and calculation of parameters
input = 2 sets of sorted peaks (freq, dB) for window1 and window2
parameters: freq-threshold
t1
t2
window-size in seconds
parse peaks1 and peaks2 in parallel
if abs(peak1-peak2) < theshold
band = (log(10)/20½) * (a1-a2/t2-t1)
if band >0
new resonance
band
freq=freq(peak2)
[sound is supposed to be less chaotic]
ampl = db_in_C_version (
(½*band*lin(a1)*window) /
(exp(-½*band*t1)*(1-exp(½*band*window)) )
else throw resonance away
else skip peak1 and peak2
output = elementary model:(freq, ampl(db in C, lin in Lisp), band) triplets
-3B- merge new elementary model and previous merged model
input = elementary model (sorted) (freq ampl(lin in Lisp, dB in C) band)
previous merged model (sorted except for the doubled resonances in Lisp version) (freq ampl(lin) band)
parameters: freq-threshold (same as threshold used for elementary analysis in
Lisp version)
check option
double option
max b2 value for doubling (12Hz)
[min delta b2-b1 for doubling (new)]
t2: time of right window used for check mode(t2= tw2+window/2)
CALCULATION OF PARAMETERS
nb: 1 = previous merged model 2 = current elementary model
if match f1 f2 (cf below) /* resonance in previous model is present */
tune-resonance:
if doubling-conditions f(b1 b2 max-doubling-b2 [min-%delta-b1-b2])
/* with doubling conditions =
analyse0 ((and double-mode (< b2 b1) (< b2 12.) (eqn double 0)) ;1-> 20.
analyse01 ((and double-mode (< b2 b1) (< b2 12.) (eqn double 0)) ; 1-> 20.
analyse ((and double-mode (> (/ (- b1 b2) b1) .05) (< b2 2.) (eqn double 0))
*/
keep resonance1
add resonance2
record doubling
record history
else if already doubled
keep resonance1
add resonance2 (take f a b from model2)
/* analyse : for 2 take amplitude from model 1 */
record history
else (if not doubling condition and not already doubled)
keep a1
take b2
take f2
record history
else if f1 has no match /* resonance has disappeared in current analysis */
keep resonance1 (including possible double resonance)
if check : check band (cf below)
checking applies to double for doubled resonance
record in resonance history
else if (or (> f1 f2) (not f1) /* a new resonance has been found
* in current analysis */
add resonance 2
record history
RESONANCE MATCHING
abs(f1-f2) < threshold OR
abs(f1(double)-f2) < threshold
CHECK OPTION
band = max (theoritical min band, b1)
amplitude should be corrected too but wasn't
theoritical min band (a1, t2, ml2)
= log(a1 / ml2) / ½*t2
avec ml2 = lin(mdb2(from peak extraction))
t2 = tw2 + window/2
DOUBLE OPTION
FORMULAS FOR PARAMETER ESTIMATION IN ELEMENTARY ANALYSIS
band = (log(10)/20*½) * ((a1-a2)/(t2-t1))
lin initial amplitude = (½*band*lin(a1)*window)
/ ( exp(-½ band t1) * (1 - exp(-½ band window)) )
-4- Possible improvement
SLIGHT CHANGES
- [WINDOWING] use energy plot and 3D spectrum for setting t1 and t2, for small window sizes
- [FFT] different window types for different sizes
- [PEAK] try different peak estimation techniques (from interpolation to analysis
techniques)
for interpolation: use FFTsize > window size for improving interpolation
accuracy
for peak extraction: use inflexion points ??
- [ELE] iterate for amplitude estimation
- [ELE] take FFT window into account for initial amplitude estimation
- [MER] when doubling, decrease amplitude of first resonance (implies keepig track of
peaks amplitudes ??)
- [MER] when checking, check amplitude
- [MER] experiment with doubling (eg check if doubled resonance match with new peak...)
HEAVY CHANGES
- [PEAK-MATCH] [RES-MATCH] interactivity and graphics for peak matching
- [PEAK] [PEAK-MATCH] [RES-MATCH] 1 full size FFT could give tips
for peak estimation (non FFT techniques)
for peak matching
- [ITERATION] reverse process: start with large windows (1 full size), end with small ones
-> peak matching and peak estimation improvement
-> improvement of initial amplitude estimation (substract already existing energy for new resonances)
NOTES ON C VERSION
-an elementary model results from the matching of two sets of peaks.
-a merged model results from the matching of one model (elementary of merged)
and an elementary model obtained from bigger FFTs.
-in double resonance, doubled is the original one (more damped), doublon is the
other one (less damped).
POSSIBLE CODE IMPROVEMENTS
- don't use MODEL_SIZE for allocation of new models: allocate max possible size
given left peakset and right peakset, or left model and right model
- M_ana.c could be changed to work with two merged models
PEAK EXTRACTION
new controls in C version:
peak extraction mode can be one of:
-DAN (originally used in Lisp version): each of the width (on each side)
neighbour of maximum, but the two immediate ones, must be threshdB smaller than maximum.
-ONE: one neighbour must be thresh dB smaller than max
-MEAN: MEAN difference between maximum and mean amplitude of neighbours must be at least equal to thresh.
ELEMENTARY ANALYSIS
#define IRCAMBUG (usually not defined)
in file E_ana.c
restores Lisp bug in elementary analysis
mode 'first match' works as in Lisp version, except for IRCAMBUG
mode 'closest match' is new, and should improve results on short windows
-in firts match, frequency of new resonance is that of right peak
-in closest match, frequency of new resonance is mean frequency of
left and right peaks if MEANFREQ is #defined, or frequency of right
peak otherwise.
MERGED ANALYSIS
new controls:
- double mode 1: doubling occurs if new band is < max_double_band
- double mode 2: idem and additional condition: minimum percentage of
change in bandwidth.
mode first match works as in Lisp version except for new controls
mode closest match is new. Matching conditions (distance between two resonances is lower than a given threshold) may differ from first match:
- In first match, distance between one double res and one simple res was the distance from simple res to closest element of double res. This is used if MINDISTANCE is #defined
- if MEANDISTANCE is #defined mean freq of double is used for computing distance.
- if DBLONDISTANCE, distance is measured from doublon resonance to simple resonance (this should work better, because doubles are usually first obtained with samll FFTs, and with a low frequency resolution, and the goal of further matches is to tune doublon resonance, regardless of doubled resonance).
SUMMARY OF OPTIONS FOR CODE VARIANTS:
#define IRCAMBUG E_ana.c not defined
restores Lisp bug in elementary analysis
#define MEANFREQ E_ana.c not defined
controls the way the frequency of a new resonance is set
(meanfreq of left and right peaks if defined, freq of right
peak otherwise).
{
#define MEANDISTANCE M_ana.c not defined
#define MINDISTANCE M_ana.c not defined (equivalent to first match)
#define DBLONDISTANCE M_ana.c defined
}
PARAMETRES UTILISES POUR LES MODELES OBTENUS A L'IRCAM:
generalement:
- deux fenetres adjacentes, la premiere etant calee sur le debut du son, ou dans
quelques cas, sur le maximum d'amplitude du son.
- taille des fenetres *= 2 a chaque etape
-lorsque le son est court et que la resolution des FFT n'est pas suffisante:
utiliser des fenetres superposees, utiliser plusieurs fois la meme taille de fenetre
avec des ecartements differents]
- extraction de pic = extraction des maxima
- seuil dB pour extraction de pics: toujours (sauf marimba tres bruitee)
-80 a gauche, -96 a droite.
- match threshold pris bien plus petit que la resolution des FFT dans les premieres
etapes, a cause du bug de la version ircam, et a cause de la methode "first match".
- match threshold etait identique pour analyse elementaire et analyse merge de meme
niveau.
- 3 a 9 etapes -> 6 a 700 resonances
OPTIONS DE L'ANALYSE:
- check mode: generalement utilisee, si double mode ne l'etait pas
- double mode: utilisee sans check-mode pour piano, contrebasse.
- fftsize > wsize: utilisee pour le piano
CLOCHE A VACHE
AS2
quatrieme et derniere etape =
t1=0. t2=1. window=.5 l=2 thresh=3 seuil=6 niveau=3
F2
cinquieme et derniere etape =
t1=0. t2=3. window=.5 l=2 thresh=4 seuil=6 niveau=4
CLOCHES TUBULAIRES
do2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 6)
(2 0. .05 .05 1 3 6 t () 80 96 29)
(3 0. .1 .1 1 3 6 t () 80 96 58)
(4 0. .2 .2 1 3 6 t () 80 96 114)
(5 0. .5 .5 1 3 6 t () 80 96 236)
do2bis
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 6)
(2 0. .05 .05 1 3 6 t () 80 96 29)
(3 0. .1 .1 1 3 6 t () 80 96 58)
(4 0. .2 .2 1 3 6 t () 80 96 114)
(5 0. .5 .5 1 3 6 t () 80 96 236)
(6 0. 1. 1. 1 3 5 t () 80 96 297)
(7 0. 2. 1. 1 3 4 t () 80 96 303)
(8 0. 4. 1. 1 3 3 t () 80 96 305)
fa
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 10)
(2 0. .05 .05 1 3 6 t () 80 96 35)
(3 0. .1 .1 1 3 6 t () 80 96 73)
(4 0. .2 .2 1 3 6 t () 80 96 142)
(5 0. .5 .5 1 3 6 t () 80 96 237)
(6 0. 1. 1. 1 3 5 t () 80 96 273)
(7 0. 2. 1. 1 3 4 t () 80 96 289)
(8 0. 4. 1. 1 3 3 t () 80 96 295)
fas2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 10)
(2 0. .05 .05 1 3 6 t () 80 96 37)
(3 0. .1 .1 1 3 6 t () 80 96 99)
(4 0. .2 .2 1 3 6 t () 80 96 187)
(5 0. .5 .5 1 3 6 t () 80 96 319)
(6 0. 1. 1. 1 3 5 t () 80 96 358)
(7 0. 2. 1. 1 3 4 t () 80 96 360)
(8 0. 4. 1. 1 3 3 t () 80 96 363)
(9 0. 6. 1. 1 3 3 t () 80 96 363)
las
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 7)
(2 0. .05 .05 1 3 6 t () 80 96 29)
(3 0. .1 .1 1 3 6 t () 80 96 85)
(4 0. .2 .2 1 3 6 t () 80 96 141)
(5 0. .5 .5 1 3 6 t () 80 96 237)
(6 0. 1. 1. 1 3 5 t () 80 96 299)
(7 0. 2. 1. 1 3 4 t () 80 96 301)
(8 0. 4. 1. 1 3 3 t () 80 96 306)
mi
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 7)
(2 0. .05 .05 1 3 6 t () 80 96 31)
(3 0. .1 .1 1 3 6 t () 80 96 99)
(4 0. .2 .2 1 3 6 t () 80 96 183)
(5 0. .5 .5 1 3 6 t () 80 96 285)
(6 0. 1. 1. 1 3 5 t () 80 96 357)
(7 0. 2. 1. 1 3 4 t () 80 96 365)
(8 0. 4. 1. 1 3 3 t () 80 96 367)
re2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 9)
(2 0. .05 .05 1 3 6 t () 80 96 33)
(3 0. .1 .1 1 3 6 t () 80 96 71)
(4 0. .2 .2 1 3 6 t () 80 96 133)
(5 0. .5 .5 1 3 6 t () 80 96 253)
(6 0. 1. 1. 1 3 5 t () 80 96 328)
(7 0. 2. 1. 1 3 4 t () 80 96 336)
CONTREBASSE
E0
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .05 .05 1 3 6 () t 80 96 3)
(2 0. .1 .1 1 3 6 () t 80 96 18)
(3 0. .2 .2 1 3 6 () t 80 96 31)
(4 0. .4 .4 1 3 6 () t 80 96 47)
(5 0. 2. 4. 1 3 6 () t 80 96 71)
BF0
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .05 .05 1 3 6 () () 80 96 5)
(2 0. .1 .1 1 3 6 () t 80 96 19)
(3 0. .2 .2 1 3 6 () t 80 96 41)
(4 0. .4 .4 1 3 6 () t 80 96 67)
(5 0. .8 .8 1 3 6 () t 80 96 94)
(6 0. 2. 4. 1 3 6 () t 80 96 104)
(7 0. 4. 4. 1 3 6 () t 80 96 108)
E1
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .05 .05 1 3 6 () () 80 96 10)
(2 0. .1 .1 1 3 6 () t 80 96 21)
(3 0. .25 .25 1 3 6 () t 80 96 38)
(4 0. .5 .5 1 3 6 () t 80 96 62)
(5 0. 2. 4. 1 3 6 () t 80 96 66)
BF1
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .05 .05 1 3 6 () () 80 96 5)
(2 0. .2 .2 1 3 6 () t 80 96 30)
(3 0. .5 .5 1 3 6 () t 80 96 59)
(4 0. 1. 1. 1 3 6 () t 80 96 75)
(5 0. 2. 2. 1 3 6 () t 80 96 84)
E2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .05 .05 1 3 6 () () 80 96 8)
(2 0. .1 .1 1 3 6 () t 80 96 15)
(3 0. .2 .2 1 3 6 () t 80 96 31)
(4 0. .4 .4 1 3 6 () t 80 96 59)
(5 0. .8 .8 1 3 6 () t 80 96 81)
(6 0. 2. 3. 1 3 6 () t 80 96 85)
BF2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 3)
(2 0. .05 .05 1 3 6 () () 80 96 10)
(3 0. .1 .1 1 3 6 () t 80 96 20)
(4 0. .2 .2 1 3 6 () t 80 96 35)
(5 0. .4 .4 1 3 6 () t 80 96 53)
(6 0. .4 1. 1 3 6 () t 80 96 68)
E3
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 2)
(2 0. .05 .05 1 3 6 () () 80 96 10)
(3 0. .1 .1 1 3 6 () () 80 96 22)
(4 0. .15 .15 1 3 6 () () 80 96 31)
(5 0. .2 .2 1 3 6 () () 80 96 38)
E0 cdb4, cinquieme et derniere etape
son original: /snd/pf/pE0.norm
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0. t2=2. window=.5 l=1 thresh=3 seuil(Hz.)=6 double-mode=t niveau=4
CROTALES
C5
troisieme et derniere etape:
t1=0. t2=.5 window=.5 l=2 thresh=4 seuil=6 niveau=2
50 partiels retenus |#
C6
troisieme et derniere etape:
t1=0. t2=.5 window=.5 l=2 thresh=4 seuil=6 niveau=2
49 partiels retenus |#
C7
troisieme et derniere etape:
t1=0. t2=.5 window=.5 l=2 thresh=4 seuil=6 niveau=2
28 partiels retenus |#
CYMBALUM
C1
t1=0. t2=1. window=.5 l=2 thresh=4 seuil=6 niveau=3
161 partiels retenus |#
C2
t1=0. t2=2. window=.5 l=2 thresh=4 seuil=6 niveau=4
167 partiels retenus |#
C3
son original: /snd/pf/balum-do2.norm
t1=.0 t2=2. window=.5 l=2 thresh=4 seuil=6 niveau=4
235 partiels retenus |#
C4
sixieme et derniere etape
son original: /snd/pf/balum-do3.norm
t1=0. t2=4. window=.5 l=2 thresh=4 seuil=6 niveau=5
208 partiels retenus |#
C5
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 3)
(2 0. .05 .05 1 3 6 t () 80 96 22)
(3 0. .1 .1 1 3 6 t () 80 96 63)
(4 0. .3 .3 1 3 6 t () 80 96 250)
(5 0. .6 .6 1 3 6 t () 80 96 409)
(6 0. 1. 1. 1 3 4 t () 80 96 478)
(7 0. 2. 2. 1 3 3 t () 80 96 568)
(8 0. 6. 2. 1 3 2 t () 80 96 578)
(9 0. 7. 2. 1 3 2 t () 80 96 583)
GLOCKENSPIEL
G3
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 7)
(2 0. .05 .05 1 3 6 t () 80 96 35)
(3 0. .1 .1 1 3 6 t () 80 96 104)
(4 0. .2 .2 1 3 6 t () 80 96 118)
(5 0. .5 .5 1 3 6 t () 80 96 119)
CS4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 11)
(2 0. .05 .05 1 3 6 t () 80 96 39)
(3 0. .1 .1 1 3 6 t () 80 96 89)
(4 0. .2 .2 1 3 6 t () 80 96 103)
(5 0. .5 .5 1 3 6 t () 80 96 105)
(6 0. .7 1. 1 3 5 t () 80 96 105)
G4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 4)
(2 0. .05 .05 1 3 6 t () 80 96 18)
(3 0. .1 .1 1 3 6 t () 80 96 88)
(4 0. .2 .2 1 3 6 t () 80 96 103)
CS5
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 5)
(2 0. .05 .05 1 3 6 t () 80 96 28)
(3 0. .1 .1 1 3 6 t () 80 96 85)
(4 0. .2 .2 1 3 6 t () 80 96 140)
G5
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 9)
(2 0. .05 .05 1 3 6 t () 80 96 29)
(3 0. .22 .22 1 3 6 t () 80 96 53)
MARIMBA
C2
derniere etape
t1=0. t2=1.1 window=.5 l=1 thresh=4 seuil=6 niveau=5
6 partiels retenus |#
C3
t1=.083 t2=.8 window=.5 l=2 thresh=4 seuil=6 niveau=4
48 partiels retenus |#
PIANO
A0
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 .5 1 3 2. () () 80 96 1)
(2 0. .06 .06 .5 1 3 2. t () 80 96 11)
(3 0. .12 .12 .5 1 3 2. t () 80 96 71)
(4 0. .25 .25 .5 1 3 2. t () 80 96 167)
(5 0. .5 .5 .5 1 3 2. t () 80 96 244)
(6 0. 2. 4. 4. 1 3 .25 t () 80 96 328)
(7 0. 5. 4. 4. 1 3 .25 t () 80 96 330)
(8 0. 8. 4. 4. 1 3 .25 t () 80 96 330)
variante A0
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 .5 1 3 2. () () 80 96 1)
(2 0. .06 .06 .5 1 3 2. t () 80 96 11)
(3 0. .12 .12 .5 1 3 2. () t 80 96 72)
(4 0. .25 .25 .5 1 3 2. () t 80 96 195)
(5 0. .5 .5 .5 1 3 2. () t 80 96 321)
(6 0. 2. 4. 4. 1 3 .25 () t 80 96 414)
(7 0. 5. 4. 4. 1 3 .25 () t 80 96 443)
(8 0. 8. 4. 4. 1 3 .25 t () 80 96 443)
A1
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .06 .06 .5 1 3 2. t () 80 96 27)
(2 0. .12 .12 .5 1 3 2. t () 80 96 97)
(3 0. .25 .25 .5 1 3 2. () t 80 96 168)
(4 0. .5 .5 .5 1 3 2. () t 80 96 388)
(5 0. 1. 1. 1. 1 3 1. () t 80 96 513)
(6 0. 2. 2. 2. 1 3 .5 () t 80 96 634)
(7 0. 4. 4. 4. 1 3 .25 t () 80 96 694)
(8 0. 15. 4. 4. 1 3 .25 t () 80 96 697)
A2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .06 .06 .5 3 3 6. t () 80 96 0)
(2 0. .12 .12 .5 3 3 6. t () 80 96 0)
(3 0. .25 .25 .5 3 3 6. () t 80 96 3)
(4 0. .5 .5 .5 1 3 2. () t 80 96 151)
(5 0. 1. 1. 1. 1 3 1. () t 80 96 242)
(6 0. 4. 1. 1. 1 3 1. t () 80 96 246)
(7 0. 10. 1. 1. 1 3 1. t () 80 96 249)
(8 0. 10. 2. 2. 1 3 .5 t () 80 96 259)
A3
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .03 .03 .5 1 3 2. () () 80. 96. 9)
(2 0. .06 .06 .5 1 3 2. () t 80. 96. 31)
(3 0. .12 .12 .5 1 3 2. () t 80. 96. 85)
(4 0. .25 .25 .5 1 3 2. () t 80. 96. 131)
(5 0. .5 .5 .5 1 3 2. () t 80. 96. 198)
(6 0. 1. 1. 1. 1 3 1. () t 80. 96. 235)
(7 0. 2. 2. 2. 1 3 .5 () t 80. 96. 251)
(8 0. 4. 4. 4. 1 3 .25 () t 80. 96. 258)
(9 0. 8. 4. 4. 1 3 .25 () t 80. 96. 258)
A4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .03 .03 .5 1 3 2. () () 80. 96. 3)
(2 0. .06 .06 .5 1 3 2. () () 80. 96. 13)
(3 0. .12 .12 .5 1 3 2. () t 80. 96. 49)
(4 0. .25 .25 .5 1 3 2. () t 80. 96. 106)
(5 0. .5 .5 .5 1 3 2. () t 80. 96. 156)
(6 0. 1. 1. 1. 1 3 1. () t 80. 96. 173)
(7 0. 2. 2. 2. 1 3 .5 () t 80. 96. 181)
(8 0. 4. 4. 4. 1 3 .25 t () 80. 96. 184)
(9 0. 6. 4. 4. 1 3 .25 t () 80. 96. 185)
variante A4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .03 .03 .5 1 3 2. () () 80. 96. 3)
(2 0. .06 .06 .5 1 3 2. () () 80. 96. 13)
(3 0. .12 .12 .5 1 3 2. () t 80. 96. 49)
(4 0. .25 .25 .5 1 3 2. () t 80. 96. 106)
(5 0. .5 .5 .5 1 3 2. () t 80. 96. 156)
(6 0. 1. 1. 1. 1 3 1. () t 80. 96. 173)
(7 0. 2. 2. 2. 1 3 .5 () t 80. 96. 181)
variante A4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .03 .03 .5 1 3 2. () () 80. 96. 3)
(2 0. .06 .06 .5 1 3 2. () () 80. 96. 13)
(3 0. .12 .12 .5 1 3 2. () t 80. 96. 49)
(4 0. .25 .25 .5 1 3 2. () t 80. 96. 106)
(5 0. .5 .5 .5 1 3 2. () t 80. 96. 156)
(6 0. 1. 1. 1. 1 3 1. () t 80. 96. 173)
(7 0. 2. 2. 2. 1 3 .5 () t 80. 96. 181)
(8 0. 4. 4. 4. 1 3 .25 t () 80. 96. 184)
A5
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .03 .03 .5 1 3 6. () () 80 96 15)
(2 0. .06 .06 .5 1 3 6. t () 80 96 44)
(3 0. .12 .12 .5 1 3 6. () t 80 96 102)
(4 0. .25 .25 .5 1 3 2. () t 80 96 133)
(5 0. .5 .5 .5 1 3 2. () t 80 96 167)
(6 0. 1. 1. 1. 1 3 1. () t 80 96 167)
(7 0. 2. 2. 2. 1 3 .5 () t 80 96 172)
(8 0. 4. 4. 4. 1 3 .25 () t 80 96 173)
A6
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .03 .03 .5 1 3 6. () () 80 96 13)
(2 0. .06 .06 .5 1 3 6. () t 80 96 34)
(3 0. .12 .12 .5 1 3 6. () t 80 96 94)
(4 0. .25 .25 .5 1 3 2. () t 80 96 154)
(5 0. .5 .5 .5 1 3 2. () t 80 96 188)
(6 0. 1. 1. 1. 1 3 1. () t 80 96 196)
(7 0. 2. 2. 2. 1 3 .5 () t 80 96 199)
A7
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0.00828 .01128 .003 .5 1 3 50. () () 80 96 0)
(2 0.00828 .01828 .01 .5 1 3 40. () t 80 96 6)
(3 0.00828 .03828 .03 .5 1 3 40. () t 80 96 53)
(4 0.00828 .06828 .06 .5 1 3 30. () t 80 96 142)
(5 0.00828 .2 .2 .5 1 3 10. () t 80 96 339)
(6 0.00828 .4 .4 1. 1 3 5. () t 80 96 424)
(7 0.00828 .6 .6 1. 1 3 2. () t 80 96 444)
(8 0.00828 .5 1. 1. 1 3 1. () t 80 96 451)
variante A7a
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0.00828 .01128 .003 .5 1 3 50. () () 80 96 0)
(2 0.00828 .01828 .01 .5 1 3 40. () t 80 96 6)
(3 0.00828 .03828 .03 .5 1 3 40. () t 80 96 53)
(4 0.00828 .06828 .06 .5 1 3 30. () t 80 96 142)
(5 0.00828 .2 .2 .5 1 3 10. () t 80 96 339)
(6 0.00828 .4 .4 1. 1 3 5. () t 80 96 424)
(7 0.00828 6. 6. 1. 1 3 2. () t 80 96 424)
(8 0.00828 .5 1. 1. 1 3 1. t t 80 96 433)
variante A7b
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0.00828 .01128 .003 .5 1 3 50. () () 80 96 0)
(2 0.00828 .01828 .01 .5 1 3 40. () t 80 96 6)
(3 0.00828 .03828 .03 .5 1 3 40. () t 80 96 53)
(4 0.00828 .06828 .06 .5 1 3 30. () t 80 96 142)
(5 0.00828 .2 .2 .5 1 3 10. () t 80 96 339)
(6 0.00828 .4 .4 1. 1 3 5. () t 80 96 424)
(7 0.00828 6. 6. 1. 1 3 2. () t 80 96 424)
(8 0.00828 .5 1. 1. 1 3 1. () t 80 96 433)
variante A7c
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0.00828 .01128 .003 .5 1 3 50. () () 80 96 0)
(2 0.00828 .01828 .01 .5 1 3 40. () t 80 96 6)
(3 0.00828 .03828 .03 .5 1 3 40. () t 80 96 53)
(4 0.00828 .06828 .06 .5 1 3 30. () t 80 96 142)
(5 0.00828 .2 .2 .5 1 3 10. () t 80 96 339)
(6 0.00828 .4 .4 1. 1 3 5. () t 80 96 424)
(7 0.00828 .6 .6 1. 1 3 2. () t 80 96 444)
(8 0.00828 .5 1. 1. 1 3 1. t t 80 96 451)
TAMTAM
derniere etape
t1=0. t2=2.5 window=.5 l=2 thresh=3 seuil=6 niveau=8
183 partiels retenus |#
TIMBALE
derniere etape
t1=0. t2=2. window=.5 l=2 thresh=3 seuil=6 niveau=6
223 partiels retenus |#
TUYAUX BEAUBOURG
derniere etape
pipe 2
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0 t2=.3 window=.8 l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=3
219 partiels retenus |#
pipe 3
son original: /snd/benjamin/pipes/P3a
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0. t2=1. window=1. l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=3
221 partiels retenus |#
pipe 3b
son original: /snd/benjamin/pipes/P3b
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0 t2=.3 window=1 l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=3
285 partiels retenus |#
pipe 4
son original: /snd/benjamin/pipes/P4
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0. t2=.5 window=1. l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=2
118 partiels retenus |#
pipe5-1
son original: /snd/benjamin/pipes/P5
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0 t2=.75 window=.75 l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=4
229 partiels retenus |#
pipe5-2
son original: /snd/benjamin/pipes/P5
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0 t2=.75 window=.75 l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=4
229 partiels retenus |#
pipe 5
son original: /snd/benjamin/pipes/P5
seuil-dB-gauche=80 seuil-dB-droit=96
t1=0 t2=.22 window=.22 l=2 thresh=4 seuil(Hz.)=6 double-mode=() niveau=3
116 partiels retenus |#
VIBRAPHONE
A2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 8)
(2 0. .05 .05 1 3 6 t () 80 96 23)
(3 0. .1 .1 1 3 6 t () 80 96 70)
(4 0. .2 .2 1 3 6 t () 80 96 142)
(5 0. .4 .4 1 3 3 t () 80 96 161)
(6 0. .6 .6 1 3 2 t () 80 96 174)
(7 0. 1. 1. 1 3 1 t () 80 96 183)
(8 0. 4. 4. 1 3 1 t () 80 96 190)
G2
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .05 .05 1 3 6 () () 80 96 12)
(2 0. .1 .1 1 3 6 t () 80 96 55)
(3 0. .2 .2 1 3 6 t () 80 96 102)
(4 0. .4 .4 1 3 3 t () 80 96 117)
(5 0. .6 .6 1 3 2 t () 80 96 132)
(6 0. 1. 1. 1 3 2 t () 80 96 136)
(7 0. 2. 2. 1 3 2 t () 80 96 146)
D3
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 5)
(2 0. .05 .05 1 3 6 t () 80 96 26)
(3 0. .1 .1 1 3 6 t () 80 96 62)
(4 0. .2 .2 1 3 6 t () 80 96 91)
(5 0. .4 .4 1 3 3 t () 80 96 97)
(6 0. .6 .6 1 3 2 t () 80 96 109)
(7 0. 1. 1. 1 3 1 t () 80 96 115)
(8 0. 4. 4. 1 3 1 t () 80 96 121)
G3
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 7)
(2 0. .05 .05 1 3 6 t () 80 96 24)
(3 0. .1 .1 1 3 6 t () 80 96 56)
(4 0. .2 .2 1 3 6 t () 80 96 70)
(5 0. .4 .4 1 3 3 t () 80 96 80)
(6 0. .6 .6 1 3 2 t () 80 96 88)
(7 0. 1. 1. 1 3 1 t () 80 96 96)
(8 0. 2. 2. 1 3 2 t () 80 96 110)
DS4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 5)
(2 0. .05 .05 1 3 6 t () 80 96 26)
(3 0. .1 .1 1 3 6 t () 80 96 64)
(4 0. .2 .2 1 3 6 t () 80 96 94)
(5 0. .4 .4 1 3 3 t () 80 96 106)
(6 0. .6 .6 1 3 2 t () 80 96 110)
(7 0. 1. 1. 1 3 1 t () 80 96 128)
(8 0. 4. 4. 1 3 1 t () 80 96 161)
G4
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 6)
(2 0. .05 .05 1 3 6 t () 80 96 19)
(3 0. .1 .1 1 3 6 t () 80 96 44)
(4 0. .2 .2 1 3 6 t () 80 96 90)
(5 0. .4 .4 1 3 3 t () 80 96 114)
(6 0. .6 .6 1 3 2 t () 80 96 131)
(7 0. 1. 1. 1 3 1 t () 80 96 145)
(8 0. 2. 2. 1 3 2 t () 80 96 170)
E5
(niveau t1 t2 window peak-def peak-thresh seuil check-mode
double-mode seuil-dB-gauche seuil-dB-droit nb-de-partiels)
(1 0. .02 .02 1 3 6 () () 80 96 4)
(2 0. .05 .05 1 3 6 t () 80 96 23)
(3 0. .1 .1 1 3 6 t () 80 96 53)
(4 0. .2 .2 1 3 6 t () 80 96 107)
(5 0. .4 .4 1 3 3 t () 80 96 136)
(6 0. .6 .6 1 3 2 t () 80 96 150)
(7 0. 1. 1. 1 3 1 t () 80 96 166)
(8 0. 4. 4. 1 3 1 t () 80 96 215)