Data quality studies for burst analysis of Virgo

“Data quality studies for burst analysis of Virgo data acquired during Weekly Science Runs “ E. Cuoco, EGO on behalf of Virgo Collaboration GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 1

WSR goals To ease the transition between commissioning to data taking 2. 5 days during week-end, roughly one per month Provide data sets taken in stationary condition for : Commissioning studies Search analysis (online, offline and noise studies) WSR 1 08/09 -11/09 2006 : duty cycle 87. 7% WSR 2 22/09 -25/09 2006 : duty cycle 71. 2% WSR 3 06/10 -09/10 2006 (failed) WSR 4 13/10 -16/10 2006 (failed) WSR 5 10/11 -13/11 2006 : duty cycle 64. 2% WSR 6 01/12 -04/12 2006 : duty cycle 80. 5% best sensitivity reached (4 Mpc NS-NS horizon for an optimally oriented source, SNR=8) GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 2

Goal of this study Study the “quality” of the Virgo data Detect all kinds of glitches in the data which may spoil a burst search (loudest events study) Understand the origin of these glitches, when possible … Study the stationarity of the burst trigger rate Study possible vetoes strategy to suppress loud triggers Set up “tools” and provide information for the burst searches (Data. Quality flags and vetoes) GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 3

Data. Quality & Vetoes • DQ flags: list of periods during which the ITF is badly functioning (photodiodes saturation, no calibration, noisy Second Stage Frequency Stabilization loop (SSFS), DAQ problem…). • These periods can last from 1 second up to several hours. • They are applied a priori or a posteriori on the triggers list produced by a burst pipeline. • For the moment they concern only “obvious” problems but may concern in the future environmental condition (weather, …) if useful for the burst search • Vetoes: list of short periods during which a fake burst signal could be visible in the Dark Fringe channel due to any cause except a genuine GW. • They are applied on triggers’ lists. GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 4

Time domain Mean Filter (MF) on wh da ite ta ne d Virgo burst pipelines Excess of Mean Estimation on different window size Time-Frequency map Time -Frequency with different window Wo rk Power Filter (PF) Wavelet Detection Filter (WDF) Wavelet Transform Coefficient threshold Energy estimation Correlator Peak Correlator (PC) Exponential Gaussian Correlator (EGC) Wiener filtering with gaussian peak templates Wiener filtering with exponential gaussian templates GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 5

Data Quality flags study Few problems identified so far: Dark fringe outport photodiodes saturation NE/WE/BS/PR mirror coil driver saturation Second Stage Frequency Stabilization (SSFS) correction saturation SSFS electronics problem Timing problems (DAQ) (WSR 5) Photodiode shutter opening (WSR 2) Data quality segments given by the h reconstruction processing White noise injection segments Bad quality when the lines are not high enough It has been checked that all these problems create loud events in the dark fringe Definition of Data Quality segments to flag these periods Most of these flags can be applied a posteriori on trigger lists GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 6

DQ segments application on MF triggers: WSR 1 photodiode saturation SSFS saturation GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 7

DQ segments application on MF triggers: WSR 5 SSFS saturation timing GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 8

DQ segments application on MF triggers: WSR 6 Coil drivers saturation The loudest events are suppressed by using these DQ flags GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 9

Loudest glitches study We found only a few categories in WSR data: Oscillations in control loops (longitudinal and angular dof): “z” events (low frequency events <100 Hz) Noise increase on the full bandwidth due to laser frequency noise coupling temporary increase (duration <few seconds): that generates some “Burst of Burst” events (“BOB” events) In C 6/C 7 data set, the origin of the laser frequency noise coupling increase has been identified to be due to the residual angular motions of the mirrors which were too loose SSFS electronics saturation events Study of environmental channels (seismometers, magnetometers, acoustic probes …) We did not find so far any loud glitches in the dark fringe due to environmental noise … GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 10

Events in the control signals GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 11

Example of a BOB event • long duration events (up to few seconds), • large frequency band content • due to a coupling of the laser frequency noise and angular motion of the mirror GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 12

Example of a problem with the SSFS Dark fringe 2 seconds Exclude 1 second before and 1 second after the event in the SSFS channel GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 13

Veto for the SSFS saturation events WDF SNR of the triggers obtained on the channel which “monitors” the SSFS GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 14

SSFS channel as Veto SNR>85, Dt=0. 1 GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration More details in M. Delprete poster 15

Looking at the signal used to control the Power Recycling cavity length z-events Snr>30 , Dt=0. 1 GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 16

WSR 5: Dark fringe WDF events distribution GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 17

WSR 5: Cleaning the Dark fringe For SSFS SNR >85 Dt=0. 1 For z-events, SNR>30 Dt=0. 1 GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 18

Enviromental channels: Magnetometers in North End tower Glitches every 3 -4 secs GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 19

Glitches in WE-NE magnetometers GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 20

Lightning? Dark fringe GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 21

Coalescing binaries horizon and seismic noise WSR 1 WSR 6 Horizon fits quite well an empirical formula: 2. 7 (4. 3) Mpc – (wind+sea). “wind” is the North End tower top stage motion in the region 30 -100 m. Hz, “sea” is the motion in the region 100 m. Hz-1 Hz GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 22

Burst trigger rate and weather condition ? Study of the burst trigger rate as function of time (rate averaged on 10 minutes) Horizon seems to follow the low frequency (<1 Hz) seismic activity seen by the top of the suspensions Comparison with the Common Mode Rejection Ratio which gives the coupling between the laser frequency noise and the dark fringe GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 23

Burst trigger rate and weather condition: WSR 1 • The trigger rate follows the seismic noise <1 Hz • The trigger rate follows the evolution of the CMRR GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 24

Burst trigger rate and weather condition: WSR 5 GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 25

Burst trigger rate and weather condition: WSR 6 • The trigger rate follows the seismic noise (<1 Hz) • The trigger rate does not follows the evolution of the CMRR GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 26

Conclusions WSR data sets have been analyzed offline using glitches finders algorithms to provide useful information for the burst searches Identification of all problems generating huge glitches Data. Quality flags defined to suppress a posteriori these periods Identification of the loudest remaining glitches. A few categories have been identified Setup of veto strategy for the identified sources of glitches We found correlation between the transient trigger rate and the weather condition GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 27

Spare slides GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 28

Horizon vs weather condition Seismic noise: quite quiet during the week-end! GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 29

Dark fringe (whiten) A bob event Frequency (Hz) Time (s) GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 30

BOB mechanism Proposed Mechanism • • at high frequency, frequency noise dominates Pr_B 1_ACp = a * dn a related to Common Mode Rejection Ratio a varies with time by a significant amount (up to a factor 10) a variations are mainly driven by angles GWDAW 11, Postdam 18 th-21 th December 2006 E. Cuoco, on behalf of Virgo collaboration 31

Dark fringe noise increase as function as the North End mirror θy angle x 10 -8 -2 -1 0 1 2 1. 6 C 7 Pr_B 1_ACp RMS 1. 5 1. 4 1. 3 1. 2 1. 1 1. 0. 9 0. 8 0. 7 GWDAW 11, Postdam 18 th-21 th December 2006 NE_ty E. Cuoco, on behalf of Virgo collaboration 32