Embed Size (px)
DESCRIPTION
WaveBurst upgrade for S3 analysis S.Klimenko, I.Yakushin University of Florida Motivation Multi-resolution analysis Parameter space Summary. Implementation of “original” WB algorithm add analysis at multiple TF resolutions more flexible time shift analysis - PowerPoint PPT Presentation
Citation preview
S.Klimenko, August 2004, LSC, G040394-00-Z
WaveBurst upgrade for S3 analysis
S.Klimenko, I.Yakushin
University of Florida
Motivation Multi-resolution analysis Parameter space Summary
S.Klimenko, August 2004, LSC, G040394-00-Z
motivation
Implementation of “original” WB algorithm add analysis at multiple TF resolutionsmore flexible time shift analysissingle and multiple detector options
use different analysis environment: DMT+Condormake development cycle shortersimplify debugging, validation and testingreduce data processing time
S.Klimenko, August 2004, LSC, G040394-00-Z
Wavelet scalograms decomposition in basis {(t)}
d4
d3
d2
d1
d0
a
a. wavelet transform tree b. wavelet transform binary tree
d0
d1
d2
a
dyadic linear
time-scale(frequency) spectrograms
critically sampledDWT
fxt=0.5 LP HP
S.Klimenko, August 2004, LSC, G040394-00-Z
Wavelet Resolution
representation of SG 850Hzwith Symlet wavelet
resolution 64 Hz X 1/128 sec
=1 ms=100 ms
=10 ms
optimal
S.Klimenko, August 2004, LSC, G040394-00-Z
Optimal Resolution
run analysis in a range of TF resolutions
(rotation in larger space)
=1 ms=100 ms8 Hz X 1/16 sec
=10 ms64 Hz X 1/128 sec
optimal
=1 ms256 Hz X 1/512 sec
optimal optimal
S.Klimenko, August 2004, LSC, G040394-00-Z
WaveBurst multi-resolution analysis
t
f
t
f/2
cover range of resolutions by selecting nodes from wavelet tree.
select pixels with maximum excess power
wavelet tree
S.Klimenko, August 2004, LSC, G040394-00-Z
Coverage of the parameter space
Signals with large TF volume ? for example windowed white noise.
need multi-detector pipeline coherent excess power (model independent, implemented in
WB) similarity of waveforms (r-statistics, can be model dependent)
S.Klimenko, August 2004, LSC, G040394-00-Z
signal/noise parameter space
“LIGO cheese”: fc, f, t (Lazzarini, Sutton) Two-dimensional space: fc, V=f x t
A time series sample has volume rs/2 x 1/rs = 1/2 (rs sampling rate)
1 10 100V
f, H
z
LIGOband
sources
noise
S.Klimenko, August 2004, LSC, G040394-00-Z
LIGO noise parameter space
“
S2
White noise
S3pg
S.Klimenko, August 2004, LSC, G040394-00-Z
ROC
Burst signal with SNR 25:
V=2
V=20
V=5
matchedfilter
P=10%
P=31%
d
nSM
)(|)(|
21
2
S.Klimenko, August 2004, LSC, G040394-00-Z
S2 rates
S.Klimenko, August 2004, LSC, G040394-00-Z
S2 efficiency for SG
See Igor’s talk
S.Klimenko, August 2004, LSC, G040394-00-Z
S3 vs S2
WB version v3
S.Klimenko, August 2004, LSC, G040394-00-Z
S3 vs S2
WB version v5
S.Klimenko, August 2004, LSC, G040394-00-Z
Summary & Plans
WaveBurst upgrade Implemented multiple TF resolutionsSensitivity improvement 20%-100%single and multiple detector options
S3 playground studies (both v4 & v5 versions)sensitivity & ratesnoise non-stationarityresolve S3 analysis issues before next LSC
Preparations for S4
