Jordi Burguet-Castell for the Calibration Team

Slides:



Advertisements
Similar presentations
11/12/2003LIGO Document G Z1 Data reduction for S3 I Leonor (UOregon), P Charlton (CIT), S Anderson (CIT), K Bayer (MIT), M Foster (PSU), S Grunewald.
Advertisements

Calibration of the gravitational wave signal in the LIGO detectors Gabriela Gonzalez (LSU), Mike Landry (LIGO-LHO), Patrick Sutton (PSU) with the calibration.
GWDAW 16/12/2004 Inspiral analysis of the Virgo commissioning run 4 Leone B. Bosi VIRGO coalescing binaries group on behalf of the VIRGO collaboration.
LIGO-G D LIGO calibration during the S3 science run Michael Landry LIGO Hanford Observatory Justin Garofoli, Luca Matone, Hugh Radkins (LHO),
Calibration of data in the time domain (or how to generate 1800s long SFTs from time domain data) XS, Bruce Allen, Mike Landry, Soumya Mohanty, Malik Rachmanov,
9/23/08G Z Status of the S5 calibration Michael Landry For the Calibration Committee LSC September 2008 Michael Landry For the Calibration Committee.
1 Time-Domain Calibration Update Xavier Siemens, Mike Landry, Gaby Gonzalez, Brian O’Reilly, Martin Hewitson, Bruce Allen, Jolien Creighton.
LIGO-G0200XX-00-M DMT Monitors: Beyond the FOM John Zweizig LIGO/Caltech LLO August 18, 2006.
Calibration of TAMA300 in Time Domain Souichi TELADA, Daisuke TATSUMI, Tomomi AKUTSU, Masaki ANDO, Nobuyuki KANDA and the TAMA collaboration.
LIGO-G E ITR 2003 DMT Sub-Project John G. Zweizig LIGO/Caltech Argonne, May 10, 2004.
LSC Segment Database Duncan Brown Caltech LIGO-G Z.
GDS Software Status John Zweizig LIGO/Caltech LSC/Virgo DetChar Session Cambridge, July 25, 2007 LIGO-G Z.
S4/S5 Calibration The Calibration team G Z.
Stefan Hild Ilias WG1 meeting, Hannover, September 2004 DATA TOOLS ‚how can data exchange work?‘ Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut)
LIGO- XXX Bruce Allen, LSC Talk LIGO Scientific Collaboration - U. Wisconsin - Milwaukee 1 Effects of Timing Errors and Timing Offsets in Pulsar.
Calibration of LIGO data in the time domain X. Siemens, B. Allen, M. Hewitson, M. Landry.
3 May 2011 VESF DA schoolD. Verkindt 1 Didier Verkindt Virgo-LAPP CNRS - Université de Savoie VESF Data Analysis School Data Quality and vetos.
Data Characterization in Gravitational Waves Soma Mukherjee Max Planck Institut fuer Gravitationsphysik Golm, Germany. Talk at University of Texas, Brownsville.
LIGO- G Z 03/23/2005LSC Meeting March BlockNormal Near-Online S4 Burst Analysis Keith Thorne Penn State University Relativity Group (Shantanu.
3/19/08G D S5 Calibration Status Brian O’Reilly For the Calibration Committee LSC March 2008 Brian O’Reilly For the Calibration Committee LSC.
LIGO-G I S5 calibration status Michael Landry LIGO Hanford Observatory for the LSC Calibration Committee LSC/Virgo Meeting May 22, 2007 Cascina,
LIGO-G9900XX-00-M ITR 2003 DMT Sub-Project John G. Zweizig LIGO/Caltech.
Preparing GEO600 for Gravitational Wave astronomy (A status report) Martin Hewitson, AEI Hannover for GEO600.
Jan 12, 2009LIGO-G Z1 DMT and NDS2 John Zweizig LIGO/Caltech Ligo PAC, Caltech, Jan 12, 2009.
Martin Hewitson and the GEO team Measuring gravitational waves with GEO600.
LIGO-G Z S2 and S3 calibration status Michael Landry LIGO Hanford Observatory for the Calibration team Justin Garofoli, Luca Matone, Hugh Radkins.
May 29, 2006 GWADW, Elba, May 27 - June 21 LIGO-G0200XX-00-M Data Quality Monitoring at LIGO John Zweizig LIGO / Caltech.
S.Klimenko, LSC, August 2004, G Z BurstMon S.Klimenko, A.Sazonov University of Florida l motivation & documentation l description & results l.
LIGO-G Z Data Quality Segments Repository K. Riles - University of Michigan 1 Data Quality Segments Repository Keith Riles (University.
S5 Data Quality John Zweizig LIGO/Caltech. Contents ● Segment Database ● Automatically generated segments ● Other segments ● Segment summary.
SFT data preparation using GEO E7 data m.alessandra papa max planck institut f. gravitationasphysik.
ALLEGRO GWDAW-9, Annecy 16 December, Generating time domain strain data (h(t)) for the ALLEGRO resonant detector or calibration of ALLEGRO data.
LSC meeting – Mar05 Livingston, LA A. Di Credico Syracuse University Glitch investigations with kleineWelle Reporting on work done by several people: L.
LIGO-G050197LSC Collab. Mtg, LHO, August 16,20051 S4 Data Quality & S5 Preview John Zweizig LIGO/Caltech.
LIGO-G9900XX-00-M DMT Monitor Verification with Simulated Data John Zweizig LIGO/Caltech.
Online DQ Segments and Triggers John Zweizig LIGO/Caltech.
Igor Yakushin, December 2004, GWDAW-9 LIGO-G Z Status of the untriggered burst search in S3 LIGO data Igor Yakushin (LIGO Livingston Observatory)
S.Klimenko, LSC meeting, March 2002 LineMonitor Sergey Klimenko University of Florida Other contributors: E.Daw (LSU), A.Sazonov(UF), J.Zweizig (Caltech)
3/21/2007 LIGO-G Z S5 Data Quality Investigation Status John Zweizig LIGO/Caltech LSC/Virgo DetChar Session MIT, July 25, 2007.
Martin Hewitson and the GEO team Measuring gravitational waves with GEO600.
LIGO-G E S2/S3 Data Quality Flagging John G. Zweizig LIGO/Caltech.
S5 Data Quality John Zweizig LIGO/Caltech Hanover, October 25, 2007.
LIGO-G W The knownpulsardemod DSO August 2002 LSC Update PULG Session 08/23/02 Gregory Mendell LIGO Hanford Observatory.
Chapter 3 Data Representation
Development of KAGRA Burst Pipeline
P.Astone, S.D’Antonio, S.Frasca, C.Palomba
Introduction to Computer Science
The Stream Model Sliding Windows Counting 1’s
Photon Calibrator Investigations During S6
LIGO Scientific Collaboration meeting
Overview of E10 / S3 Hardware Injections
Calibration of data in the time domain (or how to generate 1800s long SFTs from time domain data) XS, Bruce Allen, Mike Landry, Soumya Mohanty, Malik.
Design of Digital Filter Bank and General Purpose Digital Shaper
WaveMon and Burst FOMs WaveMon WaveMon FOMs Summary & plans
Measuring Gravitational Waves with GEO600
Calibration of TAMA300 in Time Domain
Targeted Searches using Q Pipeline
Lab 5 Part II Instructions
CSCI1600: Embedded and Real Time Software
CS 201 Fundamental Structures of Computer Science
First look at Injection of Burst Waveforms prior to S1
h(t): are you experienced?
M Ashley - LSC Meeting – LIGO Hanford, 12 Nov 2003
John Zweizig LIGO/Caltech
Optimal on-line time-domain calibration of the
Preparing GEO600 for Gravitational Wave astronomy (A status report)
CSE 1020:Software Development
CSCI1600: Embedded and Real Time Software
ENEE222 Elements of Discrete Signal Analysis Lab 9 1.
Hardware Injection Status Update
Presentation transcript:

Jordi Burguet-Castell for the Calibration Team LIGO-G0900893-v1 LIGO h(t) Calibration LSC-Virgo Meeting Budapest, 23 September 2009 Jordi Burguet-Castell for the Calibration Team

Overview Challenges and novelties with h(t) generation in S6 Actuation Residuals New Sensing Filter Production and distribution of h(t) Standard LDAS Low-Latency DMT

Introduction: The Calibration Loop

Calibration Loop C D A h ~ 1/C (1+CAD) derr Sensing h + - derr Servo Explain how we reconstruct h from derr Actuation h ~ 1/C (1+CAD) derr Need filters for each element

New Challenge: Actuation Residuals

Actuation Residuals FIR Filters IIR Filters Model Use zeros&poles, combine elements Output of a d Life used to be easy...

Actuation Residuals Model + Actuation FIR Filters IIR Filters Output of a d Use zeros&poles, combine elements Imre Bartos Ilya Beloposki Max Factourovich Jamie Rollins But we now know better

+ Actuation Residuals FIR Filters IIR Filters Model Use zeros&poles, combine elements + Output of a d and combine We have frequency-domain only actiation residuals, directly measured and without model. We take them into account by our ifft method. Actuation Residuals New method IFFT-based

Actuation Residuals Relative error if Form of the residuals and Mention methods to control de instability, taking less points and playing around. Form of the residuals and response of IFFT filter Relative error if ignoring the residuals or correcting for them

New Sensing Filter

New Sensing Filter Using the IFFT method we have created the inverse sensing filter. The new filter doesn't need to be upsampled. Better precision and more speed (lalapps_ComputeStrainDriver > 2x faster than before!) Now, total h(t) errors < 5% between 30 Hz and 7 kHz

Production and Distribution of h(t)

h(t) Production Two approaches: LDAS produced calling lalapps – Greg Mendell and Xavi Siemens. This is the official h(t). 5-10 min latency. DMT produced, using lal: hoftMon – JB, Shourov Chatterji, Xavi Siemens, John Zweizig. This is the online h(t). 30s-1min latency.

Online h(t) generation Produced with low latency in special-purpose computers at the sites, and distributed immediately to analysis centers. Generated with a DMT monitor that calls exactly the same (LAL) function as LDAS-produced h(t). Also produces some simple data quality information in the form of a channel, from which segments are automatically inserted into the segment database. Creates frame files of 16s duration, with a predictable file name and path (starting gps times are multiples of 16). Example: /online/frames/hoft/L1/L-L1_DMT_C00_L2-9129/ L-L1_DMT_C00_L2-912993136-16.gwf Runs continuously, regardless of detector state. Calibrated h(t) produced with low latency in special-purpose computers at the sites, and distributed immediately to analysis centers. Generated with a DMT monitor that calls exactly the same (LAL) function as LDAS. Produces data quality information too in the form of a channel. From it, segments are generated and automatically inserted into the segment database. Creates frame files of 16s duration, with a predictable file name and path (starting gps times are multiples of 16). Example: /online/frames/hoft/L1/L-L1_DMT_C00_L2-9129/ L-L1_DMT_C00_L2-912993136-16.gwf It runs continuously, regardless of detector state.

Online h(t) Aggregation Low-latency h(t) aggregated to: CIT, UWM, Hannover, Syracuse A set of rsync-based scripts (Igor Yakushin) gather the frame files from the sites to CIT, and distribute them from CIT to TIER-2 centers. Latencies < 1m Find the data: ${ONLINEHOFT}/<IFO>/latest

More Information For more details about the S6 generation of calibrated data, see: https://www.lsc-group.phys.uwm.edu/daswg/wiki/ S6OnlineGroup/CalibratedData For data aggregation, see: S6OnlineGroup/DataAggregation For more details about generation of calibrated data, see: https://www.lsc-group.phys.uwm.edu/daswg/wiki/ S6OnlineGroup/CalibratedData For data aggregation, see: S6OnlineGroup/DataAggregation

Summary A new method to generate filters from frequency response allows us to account for the actuation residuals and have a better sensing filter. h(t) calibration is 2x faster, and also more precise. Online analysis possible thanks to a DMT monitor that produces h(t), with latencies < 1m from data taking to presence at clusters.

Backup Slides

Actuation Residuals: Relevant Code %%% From DARMmodel.m %Detect if AI residuals (model/meast) are defined, if not, set them to 1 %since these are model/meast, and we want the meast, % thus we need to *divide* our actuation model by these residuals: if isfield(Xn,'aif') aixres = interp1(Xn.aif,Xn.aiexres,ff'); aiyres = interp1(Xn.aif,Xn.aieyres,ff'); else aixres = 1; aiyres = 1; end % [...] %new actuation functions: actx = squeeze(freqresp(pendx*drumhead*ai, 2*pi*ff)) ... .*tdelay .* hdx .* d2a./dwxres./aixres; acty = squeeze(freqresp(pendy*drumhead*ai, 2*pi*ff)) ... .*tdelay .* hdy .* d2a./dwyres./aiyres; actuation = Xn.darm2etmx*actx - Xn.darm2etmy*acty; %%% From actuation_fir_<EPOCH>.m % Now the new part: residuals resx = load('S6H1FilterResidualsX_<EPOCH>.txt'); yX=filter(resx,[1;zeros(length(resx)-1,1)],yX); resy = load('S6H1FilterResidualsY_<EPOCH>.txt'); yY=filter(resy,[1;zeros(length(resy)-1,1)],yY); % Net result is the difference y=yX-yY; %%% From fit_residuals.m %% Points used as reference (to later compute the FIR filter). N=8; ref1_f = f(1:2^N:end); ref1_rf = rf(1:2^N:end); %% Make it so the ifft is real. ref1_f = [ref1_f; ref1_f(end)+ref1_f(2:end)]; ref1_rf = [ref1_rf; conj(ref1_rf(end:-1:2))]; %% Create the FIR filter. x1 = ifft(ref1_rf); %, 2^nextpow2(length(ref1_rf))); x1 = real(x1); % it should be basically real, even before this %SR = (ref1_f(2) - ref1_f(1) ) * length(ref1_f); %x1 = [x1(floor(end/2)+1:end); x1(1:floor(end/2))]; % Window N=size(x1,1); wind=tukeywin(N,0.75); wind=hann(N); for ii = floor(N/2):N x1(ii)=x1(ii)*wind(ii); end

h(t) Frame Files Three channels: IFO:DMT-STRAIN: calibrated h(t) Double precision (8 bytes). 16384 Hz. IFO:DMT-STATE_VECTOR: state vector as constructed by the Control and Data System Float (4 bytes, need to cast to Int for use). 16 Hz. IFO:DMT-DATA_QUALITY_VECTOR: data quality vector, stores several data quality flags Integer (4 bytes). 1 Hz. LDAS generates Level 2 frames only (Level 1 disappeared). They also contain calibration factors. Three channels: IFO:DMT-STRAIN: calibrated h(t) Double precision (8 bytes). 16384 Hz. IFO:DMT-STATE_VECTOR: state vector as constructed by the Control and Data System Float (4 bytes, need to cast to Int for use). 16 Hz. IFO:DMT-DATA_QUALITY_VECTOR: data quality vector, stores several data quality flags Integer (4 bytes). 1 Hz.

Data Quality Vector Definition Bitmask definition (DQ = S 2b) Bit Name Meaning 0 SCIENCE StateVector_SCIENCE & LIGHT 1 INJECTION 0=no injection, 1=injection 2 UP StateVector_UP & LIGHT 3 CALIBRATED UP & (not TRANSIENT) 4 BADGAMMA Calibration is bad (not 0.8<gamma<1.2) 5 LIGHT Light in the arms is ok 6 MISSING Data was dropped in DMT For any state to be true, the condition must be true for all samples within the second. Bit Name Meaning 0 SCIENCE StateVector_SCIENCE & LIGHT 1 INJECTION 0=no injection, 1=injection 2 UP StateVector_UP & LIGHT 3 CALIBRATED UP & (not TRANSIENT) 4 BADGAMMA Calibration is bad (not 0.8<gamma<1.2) 5 LIGHT Light in the arms is ok 6 MISSING Data was dropped in DMT

Segment Information The h(t) monitor turns Data Quality flags SCIENCE, CALIBRATED_ONLINE, and INJECTION into segments of the same name. The segments will have version 0 so they are flagged as online segments.

State Vector Definition Bits 0-4 define the states. Bits 5-15 are always set to 1, such that the word 0xffff means science mode: Bit Name Meaning 0 SCI Operator set to go to science mode 1 CON Conlog unsets this bit in non-harmless epic changes 2 UP Set by locking scripts 3 !INJ Injections unset this bit 4 EXC Unauthorized excitations cause this bit to be unset Bit Name Meaning 0 SCI Operator set to go to science mode 1 CON Conlog unsets this bit in non-harmless epic changes 2 UP Set by locking scripts 3 !INJ Injections unset this bit 4 EXC Unauthorized excitations cause this bit to be unset

New Sensing Function Sensing he sensing function is trickier. We use our ifft method and get it right, without having to follow the model so closely, and getting a shorter filter that doesn't need upsampling.

Total Response