LIGO-G060553-00-Z LSC Nov 5, 2006 1 Bruce Allen, U. Wisconsin - Milwaukee and AEI

Slides:



Advertisements
Similar presentations
S3/S4 BBH report Thomas Cokelaer LSC Meeting, Boston, 3-4 June 2006.
Advertisements

GWDAW 11 - Potsdam, 19/12/ Coincidence analysis between periodic source candidates in C6 and C7 Virgo data C.Palomba (INFN Roma) for the Virgo Collaboration.
LIGO-G Z 1 S2/E10 pulsar injection analysis Réjean Dupuis University of Glasgow 13 November 2003 LSC Meeting, LHO.
Adaptive Hough transform for the search of periodic sources P. Astone, S. Frasca, C. Palomba Universita` di Roma “La Sapienza” and INFN Roma Talk outline.
LIGO-G Z Update on the Analysis of S2 Burst Hardware Injections L. Cadonati (MIT), A. Weinstein (CIT) for the Burst group Hannover LSC meeting,
1 Hough transform Some Fourier basics: –Nyquist frequency: 1/2 , with  the difference between time samples. If signal is bandwidth limited below Nyquist.
GWDAW-8 (December 17-20, 2003, Milwaukee, Wisconsin, USA) Search for burst gravitational waves with TAMA data Masaki Ando Department of Physics, University.
Moving towards a hierarchical search. We now expand the coherent search to inspect a larger parameter space. (At the same time the incoherent stage is.
LIGO-G W Gregory Mendell and Mike Landry LIGO Hanford Observatory The StackSlide Search Summary: November 2003.
Fitting a Model to Data Reading: 15.1,
LIGO-G W Gregory Mendell, LIGO Hanford Observatory on behalf of the LIGO Scientific Collaboration StackSlide Summary LSC Meeting, August 2005.
Searching for pulsars using the Hough transform Badri Krishnan AEI, Golm (for the pulsar group) LSC meeting, Hanford November 2003 LIGO-G Z.
LIGO Meeting LIGO Scientific Collaboration - University of Wisconsin - Milwaukee LIGO-G Z SuperComputing 2003 grid-distributed wide-area.
Event 18 Dec Bruce Allen, UWM and AEI History and Status.
LIGO-G Z Coherent Coincident Analysis of LIGO Burst Candidates Laura Cadonati Massachusetts Institute of Technology LIGO Scientific Collaboration.
LIGO-G W Gregory Mendell, LIGO Hanford Observatory on behalf of the LIGO Scientific Collaboration Stackslide search for continuous gravitational.
The Demodulation Code The Demodulation Code For the 1 st and 3 rd stage of hierarchical searches and for targeted searches of known objects. LIGO-G Z.
LIGO- G Z LSC ASIS Meeting LIGO Scientific Collaboration - University of Wisconsin - Milwaukee 1 S2 Hardware Pulsar Injections Bruce.
GWDAW /12/161 Activities of the LIGO Scientific Collaboration’s Continuous Wave Group Xavier Siemens for the LSC University of Wisconsin -- Milwaukee.
Searching for gravitational radiation from Scorpius X-1: Limits from the second LIGO science run Alberto Vecchio on behalf of the LIGO Scientific Collaboration.
Optimized Search Strategies for Continuous Gravitational Waves Iraj Gholami Curt Cutler, Badri Krishnan Max-Planck-Institut für Gravitationsphysik (Albert-Einstein-Institut)
1 Spectral filtering for CW searches S. D’Antonio *, S. Frasca %&, C. Palomba & * INFN Roma2 % Universita’ di Roma “La Sapienza” & INFN Roma Abstract:
LIGO-G Z GWDAW10, December 16, S3 Final Results Bruce Allen, for the LIGO Scientific Collaboration.
10/06/2003LSC WIDE COLLOQUIUM1 LSC Pulsar Group and P Shawhan, S Marka, and S Koranda presented by B Allen, R Dupuis, N Christensen, and X Siemens S2 Hardware.
LIGO-G D Status of Stochastic Search with LIGO Vuk Mandic on behalf of LIGO Scientific Collaboration Caltech GWDAW-10, 12/15/05.
GWDAW-8 December 18, 2003LIGO Scientific Collaboration, UW - Milwaukee 1LIGO-G Z Broad-band CW searches (for isolated pulsars) in LIGO and GEO.
Yousuke Itoh GWDAW8 UW Milwaukee USA December 2003 A large value of the detection statistic indicates a candidate signal at the frequency and.
Searching for Gravitational Waves from Binary Inspirals with LIGO Duncan Brown University of Wisconsin-Milwaukee for the LIGO Scientific Collaboration.
LIGO-G E Network Analysis For Coalescing Binary (or any analysis with Matched Filtering) Benoit MOURS, Caltech & LAPP-Annecy March 2001, LSC Meeting.
S.Klimenko, G Z, December 2006, GWDAW11 Coherent detection and reconstruction of burst events in S5 data S.Klimenko, University of Florida for.
LIGO-G v2 The Search For Continuous Gravitational Waves Gregory Mendell, LIGO Hanford Observatory on behalf of the LIGO Science Collaboration The.
LIGO-G Z LSC ASIS Meeting LIGO Scientific Collaboration - University of Wisconsin - Milwaukee 1 Update on November 2003 grid-distributed.
LIGO-G Z 1 Setting upper limits on the strength of periodic GWs using the first science data from the LIGO and GEO detectors Bruce Allen, University.
Optimized hierarchical continuous-wave searches Badri Krishnan Curt Cutler, Iraj Gholami AEI, Golm G Z LSC meeting, March 2004 ASIS session.
2005 Unbinned Point Source Analysis Update Jim Braun IceCube Fall 2006 Collaboration Meeting.
LIGO-G Z GWDAW9 December 17, Search for Gravitational Wave Bursts in LIGO Science Run 2 Data John G. Zweizig LIGO / Caltech for the LIGO.
S. Frasca INFN – Virgo and “La Sapienza” Rome University Baton Rouge, March 2007.
15-18 December 2004 GWDAW-9 Annecy 1 All-Sky broad band search for continuous waves using LIGO S2 data Yousuke Itoh 1 for the LIGO Scientific Collaboration.
LIGO-G All-Sky Burst Search in the First Year of the LSC S5 Run Laura Cadonati, UMass Amherst For the LIGO Scientific Collaboration GWDAW Meeting,
Periodic Sources Session M. Alessandra Papa - Albert Einstein Institute - Golm, Germany The LSC UL group is using all (and only) of the codes that we (GEO)
LIGO- G Z AJW, Caltech, LIGO Project1 A Coherence Function Statistic to Identify Coincident Bursts Surjeet Rajendran, Caltech SURF Alan Weinstein,
1 Work of the LSC Pulsar Upper Limits Group (PULG) Graham Woan, University of Glasgow on behalf of the LIGO Scientific Collaboration GWDAW 2003.
LIGO-G Z Peter Shawhan (University of Maryland) for the LIGO Scientific Collaboration Special thanks to Michael Landry and Bruce Allen Eastern.
LIGO-G Z Status of the LIGO-TAMA Joint Bursts Search Patrick Sutton LIGO Laboratory, Caltech, for the LIGO-TAMA Joint Working Group.
CW Search Group Status Summary: August 2005 M.Alessandra Papa for the CW group G05yyxx-00-Z.
Tuning the hierarchical procedure searching for the key where there is light S.Frasca – Potsdam, December 2006.
All-sky LIGO Search for Periodic Gravitational Waves in the Fourth Science Run (S4) Keith Riles University of Michigan For the LIGO Scientific Collaboration.
LIGO Insert Name of the Meeting KAWABE K., 2006/Apr/25 APS Meeting Dallas, W DCC: LIGO-G Z Coherent searches for periodic gravitational.
SC03 failed results delayed FDS: parameter space searches
Review production of 30min calibrated SFTs well under way (see talks by V.Dergachev., X. Siemens in ASIS/DetChar). Also expect calibrated h(t) in the next.
Xavier Siemens for the LSC University of Wisconsin -- Milwaukee
A 2 veto for Continuous Wave Searches
All-Sky broad band search for continuous waves using LIGO S2 data
Overview of the CW Search Group Efforts – S2 data analyses Michael Landry LIGO Hanford Observatory on behalf of the CW Search Group LSC Meeting March.
S4 Pulsar Search Results from PowerFlux
Bounding the strength of gravitational radiation from Sco-X1
Searching for pulsars using the Hough transform
GWDAW - 8 University of Wisconsin - Milwaukee, December
S4 Pulsar Search Results from PowerFlux
Coherent wide parameter space searches for gravitational waves from neutron stars using LIGO S2 data Xavier Siemens, for the LIGO Scientific Collaboration.
(Y. Itoh, M.A.Papa,B.Krishnan-AEI, X. Siemens –UWM
Bounding the strength of gravitational radiation from Sco-X1
A.M. Sintes for the pulgroup
Status of the Hough CW search code - Plans for S2 -
Broad-band CW searches in LIGO and GEO S2 and S3 data
Broad-band CW searches in LIGO and GEO S2 and S3 data
LIGO Scientific Collaboration, UW - Milwaukee
Broad-band CW searches in LIGO and GEO S2 and S3 data
Coherent Coincident Analysis of LIGO Burst Candidates
CW Search Group Efforts – overview Michael Landry LIGO Hanford Observatory on behalf of the CW Search Group LSC Meeting June 6, 2004 Tufts University,
Presentation transcript:

LIGO-G Z LSC Nov 5, Bruce Allen, U. Wisconsin - Milwaukee and AEI

LIGO-G Z LSC Nov 5, Overview Status of -Users, credits, size -Officially funded as a project by NSF! Status of the “old style” S3 analysis -Done and reviewed! Status of the S4 analysis -Postprocessing underway since mid-summer, still ongoing -First careful estimation of search sensitivity (for practice, not for publication!) Status of the S5 analysis ( workunits x 5 CPU hours x 2) -Processing 49% complete -About 100 more days of processing to go server/project up and operational with no glitches for 143 days! Status of the upcoming hierarchial S5 analysis -First CW analysis that integrates long (> 30 min) coherent and an incoherent methods -Gives us the optimal sensitivity for our CPU power -No large data transfers to or from the host machines.

LIGO-G Z LSC Nov 5, How big is

LIGO-G Z LSC Nov 5, User/Credit History

LIGO-G Z LSC Nov 5, Users/Hosts History

LIGO-G Z LSC Nov 5, Current performance is currently getting 84 Tflops

LIGO-G Z LSC Nov 5, Status of S3 Analysis Finished: -Final S3 analysis and writeup have been reviewed and approved by the CW Review Committee and the LSC Executive Committee. -Results are posted on the web site. -We didn’t find any CW sources

LIGO-G Z Overview of S4 analysis Coherently analyzed 30-hour data stretches (10 LHO, 7 LLO) 540 hours total. Spanned times vary, but all < 40 hours. Searched 50 Hz Hz in 6,731,410 work units from to Near optimal grid (within ~2) on the sky and in frequency and df/dt Explicit search over spindowns (df/dt) corresponding to pulsars older than a few thousand years. Previous searches had |df/dt|<1/(integration time) 2. Each host searches the entire sky and fdot range and a variable-sized region of frequency df ~ f -3 and one stretch of data. It then returns list of ‘top 13,000 events’. Range of frequency that decreases with increasing frequency as f^-3 f 1,000 years f  300 Hz: mismatch 0.5, spin-down ages > 10,000 years Each workunit produces a compressed data file that is about 150kB in size. The total data volume to post-process is 1 TB (compressed) or 4 TB (uncompressed). Many hardware improvements “behind the scenes” to handle this data volume and permit postprocessing on Nemo cluster.

LIGO-G Z LSC Nov 5, S4 post-processing coincidence strategy Search for signals that appear in each of the 17 different data stretches, with consistent parameters Steps: -Shift candidate frequencies to a fixed fiducial time -‘Bin’ candidates in four dimensions (alpha, delta, f, fdot) -Search for bins which have candidates from many of the 17 data stretches Span entire sky, entire frequency band, entire fdot band. Bins are chosen to be as small as possible, consistent with: -Sky bin size > largest grid separations (use Gaussian model in delta) -Frequency bin size > frequency resolution + (grid spacing in fdot) x T -Fdot bin size > Fdot grid spacing Bins are also shifted by 1/2 the bin width in all 2^4 combinations, so as not to miss any candidates on opposite sides of cell faces.

LIGO-G Z LSC Nov 5, How many events to keep? Goal: constant false alarm probability per data stretch per coincidence cell. Why: in the coincidence analysis, this makes it easier to interpret the results and to predict false alarm probability. How: in a given frequency band, keep the same number of events from each data stretch. How many events to keep: to get a false alarm probability of 0.1% to find 7 or more coincidences (out of 17) in random noise in a 1/2 Hz band. Example: band Hz: -Data stretch 1 (10 workunits): keep top 1000 events per workunit -Data stretch 2 (5 workunits): keep top 2000 events per workunit -Data stretch 3 (8 workunits): keep top 1250 events per workunit … -Data stretch 17 (20 workunits): keep top 500 events per workunit

LIGO-G Z LSC Nov 5, Coincidence Analysis Grids Typical sky grid has points separated more broadly near the equator. Each of the 17 data segments has a different grid In doing the coincidence analysis we use a Gaussian fit to the declination differences to ensure that we don’t miss correlated events.

LIGO-G Z LSC Nov 5, Sample results ( Hz) Number of events: 54,841 x 17 per 1/2 Hz Number of coincidence cells per 1/2 Hz: 5,398,250 Max coincidences found: 5

LIGO-G Z LSC Nov 5, Sample results ( Hz) Number of events: 166,033 x 17 per 1/2 Hz Number of coincidence cells per 1/2 Hz: 19,747,000 Max coincidences found: 7 (Line: “Yousuke’s 8 Hz comb”, only in L1)

LIGO-G Z LSC Nov 5, Sample results ( Hz) Number of events: 40,364 x 17 per 1/2 Hz Number of coincidence cells per 1/2 Hz: 5,398,250 Max coincidences found: 11 This is fake pulsar #3. Injection was off for 5 of the 17 data stretches.

LIGO-G Z LSC Nov 5, Complete S4 results ( Hz) More work still needed: -Postprocessing being repeated to fix some mistakes made the first time (wrong counting of coincidence cells led to wrong false alarm rate). -After removal of hardware injections, need to follow up outliers

LIGO-G Z LSC Nov 5, Determining S4 sensitivity Holger Pletsch has developed methods to estimate the sensitivity and has tested this using software injections. For a given signal, did analytic estimate the 2F values in each of the 17 data segments to determine how many times it falls in the top list of candidates. This method correctly predicts how many coincidences would be observed. Repeat for 500 randomly placed and oriented simulated signals per 0.5 Hz band Note: v1 calibration

LIGO-G Z LSC Nov 5, Current S5 run Very similar to previous S4 run, but more data (22 x 30 hours) which is also more sensitive Postprocessing not even started yet Search has been underway for about 140 days About 100 days of work left

LIGO-G Z LSC Nov 5, Next S5 Search The CW group is planning to start running the first true hierarchical search in about 3 months! All-sky, TBD: f years A new search code (union of multi-detector Fstat and Hough). A stack- slide incoherent option is also “in the works”. This will use approximately 96 x 20 hours of coincident H1/L1 data (adding strains to gain factor of sqrt(2) in strain sensitivity) Combines coherent Fstat method with incoherent Hough method (48 25-hour stacks) Many CW group members are working very hard on this: Krishnan, Prix, Machenschalk, Siemens, Hammer, Mendell, Sintes, Papa, and others. Should permit a search that extends hundreds of pc into the Galaxy This should become the most sensitive blind CW search possible with current knowledge and technology As soon as we can: a coherent follow-up integration stage.

LIGO-G Z LSC Nov 5, Next S5 search: data selection T data = N stack T coherent and h min ~ T coherent -1/2 N stack -1/4 Constraints and tradeoffs -Practical: smaller data T data volume is good for users; larger data volume gives more sensitivity & confidence -Sensitivity: larger T coherent gives more sensitivity but less data. Coincident data increases sensitivity by sqrt(2) but yields less data -CPU cost: larger T coherent increases cost of coherent stage relative to incoherent Hough stage Study carried out by Siemens, Prix, and Krishnan -Determine much data would be available as a function of T coherent with and without 2-detector coincidence -Extrapolate to January 2007, and estimate –sensitivity h min –total data volume T data –computational cost = sum of the two different search stages

LIGO-G Z LSC Nov 5, Next S5 search: data selection Tdata Tstack Nstack hmin CPU_inc CPU_coh CPU data_volume 12h 16h h 20h h 25h … LHO LLO … … TdataTstack LHO LLO Tdata Tstack Nstack hmin CPU_inc CPU_coh CPU data_volume 12h 16h h 20h h 25h

LIGO-G Z LSC Nov 5, Coherent + incoherent “merge” FFT … Keep 1/5 of peaks (lots of data!) Make Hough maps Fstat … Keep 1/4 of peaks Too much data to store or transmit Make Hough maps For the multi-IFO Fstat (Code 1) and Hough (Code 2) had to be merged into a single stand-alone executable that could pass the needed data between the two stages. This has been done by Krishnan. Also planned: integrate the Stack-Slide incoherent step as an alternative to Hough. Current Hough analysis Integrated code for analysis Code 1Code 2 Detector Frequency Source Frequency Time

LIGO-G Z LSC Nov 5, Conclusions is healthy and growing, and providing 10x more CPU cycles than other LSC resources. S4 analysis postprocessing is making good progess. First reliable estimates of the sensitivity. Similar S5 analysis should be finished in months. Ambitions plans to run the first true hierarchical search under on S5 data in just a few months.