30 Hz HP - LIGO by linxiaoqin

VIEWS: 3 PAGES: 17

									           S2 Glitch Investigation Report

                        Laura Cadonati (MIT)
              on behalf of the Glitch Investigation Team
               LSC meeting, August 2003, Hannover



LIGO-G030412-00-D
                                 Glitch Investigation
     Partecipants                                    Goals
   MIT: Adikhari, Ballmer, Cadonati,             identification of S2 outliers (> 4-5 sigma)
    Desai, Katsavounidis                           starting from the online DataBase
                                                   population.
   SYRACUSE: Bernstein, Dalrymple,
    Di Credico, Frei, Magri, Saulson              statistical description of glitches (function
                                                   of channels, thresholds, filters…)
   CIT: Shawhan, Zweizig
                                                  correlation of glitches among channels
   LSU/LLO: Gonzalez, Franzen, Zotov              and with AS_Q
   OREGON: Ito, Schofield                        definition of segments with unacceptable
   CARLETON: Christensen, Steussy                 glitch rates
                                                  identification of a usable veto for the Burst
   FLORIDA: Klimenko
                                                   and Inspiral analysis groups
   PSU: Ashley


                         http://ligo.mit.edu/ldas/dc/index.html
     LIGO-G030412-00-D       Hannover LSC meeting - August 18,2003                        2
                                              H1 (MIT)
Analyzed 10 channels with most triggers in the DataBase (online production) for ALL S2
Compared to the glitches found in AS_Q by the same algorithm
glitchMon                       MICH_CTRL Tr_10_15 30 Hz HP           AS_I Tr_6_8 30 Hz HP
AS_Q Tr_6_8 30 Hz HP            efficiency=6.9%                       efficiency=6.7%
156,402 events in 1039.24 h     96.6% veto success                    71.4% veto success
Also interesting: POB_Q (90% overlap with AS_Q, 0.6% veto efficiency, peak in lag plot)
Others: AS_DC, POB_I, REFL_Q, PRC_CTRL (less than 30% overlap with AS_Q, but good lag plots)


       Lag plot and amplitude scatter plot for MICH_CTRL vs AS_Q
                                                                         DeadTimes < 0.1%


                                                                        Found ~ 4000
                                                                        simultaneous glitches
                                                                        on AS_Q, MICH_CTRL,
                                                                        AS_I (POB_Q?)


      LIGO-G030412-00-D       Hannover LSC meeting - August 18,2003                          3
                         H1 (MIT)

Next step: compare to Burst trigger
candidates in the S2 playground (single
IFO lock list).




NOTE the importance of frequency of burst
events!!!




ALSO: there is poor overlap (few %)
between Burst candidates (>100Hz) and
glitchMon triggers (>30Hz)


     LIGO-G030412-00-D      Hannover LSC meeting - August 18,2003   4
                                             H1 (MIT)
    Nothing interesting seen for TFCLUSTERS events below 1kHz
      »   TFCLUSTER’s thresholds might be too low for this study - now repeating with higher thr.
    For TFCLUSTERS events in 1-3kHz (broadband glitches):
      » Low efficiency (~0.2%) and deadtime (< 0.1%) but interesting lag plots
        (peaks at 0 lag, low statistics…):
          MICH_CTRL AS_I AS_DC POB_Q
    For the EXCESS POWER events in 160-670 Hz
      » Low efficiency (0.2-0.6%) and deadtime (0.01%), zero lag peak:
          AS_DC POB_Q

    In progress right now for H1…
    GlitchMon reruns on MICH_CTRL, AS_I, AS_DC, POB_Q at 3 threshold
    settings, with 30, 70, 100 Hz high pass and production of efficiency vs
    deadtime plots. More channels will follow.


For more details on the H1 investigation
see Erik Katsavounidis’ talk in Wednesday afternoon DC session
LIGO-G030412-00-D         Hannover LSC meeting - August 18,2003                               5
                                           H2 (MIT)
Analyzed 17 channels with most triggers in the DataBase (online production)
Compared to the glitches found in AS_Q by the same algorithm
glitchMon: 40,831 triggers in 818 hours


    MICH_CTRL: 0.4% dead time, 50.4% overlap with AS_Q and 44.1% veto efficiency
    Comparable performance in: AS_I PRC_CTRL POB_Q
    Also interesting (good lag plots!):
                                                 Lag plot for MICH_CTRL vs AS_Q
      »   REFL_Q
      »   POB_I
      »   ASC-WFS1_QP
      »   ASC-WFS1_QY
      »   ASC_QPDY_DC
      »   ASC_QPDX_DC
      »   AS_DC




     LIGO-G030412-00-D     Hannover LSC meeting - August 18,2003                  6
                         H2 (MIT)
   When comparing to Burst
    candidates in the playground:
    Low efficiency (<4%) and deadtime
    (< 0.1%) but interesting lag plots:
     »   AS_I
     »   MICH_CTRL
     »   AS_DC
     »   PRC_CTRL




For more detail on the H2 investigation, see
Erik Katsavounidis’ talk in Wednesday
afternoon DC session



     LIGO-G030412-00-D        Hannover LSC meeting - August 18,2003   7
                        L1 glitches in the playground
                                 (Syracuse)
   Analyzed coincidences between AS_Q signals and auxiliary channels as
    seen by zGlitch and glitchMon (using online production in the playground):
     »   Best candidates:
     »   LSC-MICH_CTRL (zGlitch: e = 46%, deadtime=0.2)
     »   LSC-AS_I (**) (zGlitch: e = 45%, deadtime=0.4%) (**)probably unsafe
     »   LSC-POB_Q (glitchMon: e = 9.6%, deadtime=0.2%; )



                                        Same analysis performed on ETG triggers
                                        (TFCLUSTER) against auxiliary channels as seen by
                                        zGlitch and glitchMon




    LIGO-G030412-00-D       Hannover LSC meeting - August 18,2003                    8
                                L1 glitches in the playground
                                         (Syracuse)
          Best candidates are still LSC-MICH_CTRL and LSC-AS_I for both zGlitch and glitchMon
    TFCLUSTER                Total             0-100Hz      100-400Hz       400-800Hz     800Hz +
    Singles                   414497           16311         20770              9736       374425
    Clusters                  101967           14694         14683              6922        82875
        EFFICIENCY
    zGlich vetos             Total            0 – 100 Hz   100 - 400 Hz    400 – 800 Hz   > 800 Hz
    LSC-MICH_CTRL               2.4%             1.4%           1.5%              4.7%       2.7%
    LSC-AS_I                    7.0%              4.8%          6.0%             11.7%       7.9%
    glitchMon                Total            0 – 100 Hz   100 - 400 Hz    400 – 800 Hz   > 800 Hz
    LSC-MICH_CTRL               2.1%              3.1%          3.7%              4.7%          1.9%
    LSC-AS_I                     1.9%             1.3%          1.4%              1.8%          2.1%
    ASC-QPDY-DC                  4.4%             3.6%          4.1%              5.1%          4.6%

     For more detail, see:
        Syracuse web page at: http://www.physics.syr.edu/research/relativity/ligo/
        Alessandra Di Credico’s talk in Wednesday afternoon DC session
         LIGO-G030412-00-D              Hannover LSC meeting - August 18,2003                       9
              Veto safety studies (Ito, Schofield)

                                         allow veto

                                                            The position of this cut
                                                            Is not decided yet!
                                           do not
                                           allow veto




                                                           Study veto safety
                                                        with hardware injections
          H1


LIGO-G030412-00-D   Hannover LSC meeting - August 18,2003                      10
                         WaveMon (S. Klimenko)
   WaveMon looks at coincidence of AS_Q channel and an auxiliary channels and
    produces a pair of triggers (VETO and ASQ) in all S2
                          MC_F , MICH_CTRL , REFL_I , REFL_Q
                         POB_I, POB_Q, LVEA_V1, MAG1X, SEISX

   WaveBurst web page at http://www.phys.ufl.edu/LIGO/bursts/wavemon/S2.html
   Sergei Klimenko’s talk in Wednesday afternoon DC session

                                    L1:LSC-MICH_CTRL


                                                       off-time
                                                       15 sec lag
                                                                      Dead times:
                                                                      H1: 2%
         VETO log(CL)                                                 H2: 1%
                                                                      L1: 2.5%
                             on-time

                              ASQ log(CL)               ASQ log(CL)
     LIGO-G030412-00-D   Hannover LSC meeting - August 18,2003                  11
                         WaveMon (S. Klimenko)
                                                                  L1 HW injections

   Used hardware injections to look for cross coupling
    between AS_Q and veto channels                            VETO

   Select qualified triggers by setting a threshold on
    strength of veto clusters.
     » Lost <4% waveMon AS_Q triggers with this cut

                                                                                             ASQ


In Progress:
 Re-run WaveMon on-line on larger number of
  channels and wider frequency band.
 Study of veto efficiency

                                                                         Time lag [s]

     LIGO-G030412-00-D    Hannover LSC meeting - August 18,2003                         12
                              Coherence studies
                         (N. Christensen, D. Steussy)
    Looking at coincidences between inspiral candidates and aux channels
                        L1:SUS-ITMX_OPLEV_POUT
                        L1:SUS-ITMX_OPLEV_YOUT

Observed coincident glitches AND coherence increase between 50 and 200 Hz



        80 sec before the glitch:                     During the glitch:
        Coherence plot between ASQ                    Coherence plot between ASQ
        and ITMX_YOUT                                 and ITMX_YOUT (16 sec)




     LIGO-G030412-00-D     Hannover LSC meeting - August 18,2003                   13
                         Coherence studies
                    (N. Christensen, D. Steussy)

                                                 Coherence increase observed
                                                 during hardware injections too!
                                                  Worry about safety…




                For more information:
                http://physics.carleton.edu/Research/ligo/7-10/S2noiseb.html
                Nelson Christensen’s talk in Wednesday afternoon DC session

LIGO-G030412-00-D        Hannover LSC meeting - August 18,2003                 14
                         Glitch tracking (Zotov)

   Tracking foot stomps in the control room through the interferometer
     » Glitches appear on AS_Q clearly at low frequencies, and possibly
       in 100-300Hz as well (bleeding from lower frequencies)
      » http://www.ligo-la.caltech.edu/~nzotov/glitch_investigations/glitchTrack.html



   Also checking 10 loudest inpiral candidates in the S2 playground
    against various aux channels – nothing big found.
      » http://www.ligo-la.caltech.edu/~nzotov/glitch_investigations/glitch.html




LIGO-G030412-00-D       Hannover LSC meeting - August 18,2003                      15
                            LineAmp (PSU)

                                                  LineAmp - (Ashley, Finn)

                                                  Tracks   the calibration line
                                                  amplitude and phase by
                                                  heterodyning the AS_Q against a
                                                  local (software) oscillator operating
                                                  at the line frequency.

                                                  Samples    are averaged over a
                                                  specified interval.

                                                  Coded   in MATLAB and produces
                                                  eps plots of the amplitude and
                                                  phase and saves trend files
                                                  (available online).


                            http://gravity.psu.edu/~ashley/LSC/LineAmp
LIGO-G030412-00-D   Hannover LSC meeting - August 18,2003                      16
                                       Summary

   The pursue of glitches is ongoing – so far analyzed only
    interferometer channels
      » MICH_CTRL, AS_I, POB_Q, AS_DC
      » Testing validity of these channels as vetos for Burst and Inspiral analysis

   Performing reruns of AS_Q and AUX channel glitch search
      » different methods

   Veto safety tested with hardware injections
   Next: PEM channels

      »    lots of work still needed …

LIGO-G030412-00-D       Hannover LSC meeting - August 18,2003                         17

								
To top