Online Inspiral Search

Introduction

The Online Binary Inspiral Serach FOM displays the signal-to-noise ratio of the loudest trigger generated by the online inspiral search every 60 seconds.

The online inspiral search is run by onasys on the observatory Condor pools. Onasys waits until there is 2048 seconds (35 min) of data to analyze and then runs lalapps_inspiral_online_pipe to perform a templated binary inspiral search on the data. The search takes approximately 5 min if sufficent compute resources are available on the cluster. The inspiral triggers generated by the serach are then sumarized and displayed in the control room via the InspiralMon DMT program.

Check the update time of the FOM plot

The Online Inspiral FOM should update regularly. If the time stamp in the lower left corner is more than one hour old, then the InspiralMon DMT program is dead. Contact Duncan Brown <dbrown@ligo.caltech.edu> to restart the monitor. (Note: at LHO you can do this yourself by logging into gdsdmt1 as ops, finding the process id of the dead InspiralMon and sending it a kill (9) signal. If you don't know how to do this, contact Duncan Brown or John Zweizig and ask them to restart the monitor.

Check that the Inspiral FOM is displaying data

The last 40 minutes of the inspiral FOM plot will be blank since it takes 35 min to aquire the data + 5 min to analyze it. If the detector is in science mode and there no triggers are displayed for more than two hours there may be an error. At LHO you should check that both H1 (red) and H2 (blue) triggers are being displayed. If there are no triggers for over 2 hours, either the cluster is overloaded and the search is not getting resources or onasys is no longer running the inspiral search. If this is the case follow the link to either the

and check that the colored lights for the inspiral analyses. If the light is green, the search is running. If the light is yellow, the search is running, but is falling behind. In both these cases with 30 mins for the search to catch up. If it has not caught up after 30 min, contact Duncan Brown <dbrown@ligo.caltech.edu> to diagnose the problem. If the light is red, then the search is dead and you should contact Duncan Brown <dbrown@ligo.caltech.edu> immediately.

Since the search requires 35 min of data, science segments shorter than 35 min will not be analyzed. This is not an error.


Things to watch for in the Online Inspiral FOM

The signal-to-noise (SNR) of the loudest trigger every 60 seconds for each IFO is displayed on a log scale. The y axis auto-scales to very loud triggers can be displayed. Watch the values of the y axis when interpreting triggers!

As a rough guide, the lower the SNR of the loudest trigger per 60 sec, the better the instrument is performing. No vetoes are applied to the online inspiral triggers (these are applied after coincidence off site), so the search is a very sensitive glitch monitor, as well as an inspiral search. In general, the values of SNR displayed in the FOM should be below SNR of 8.

The figure below shows a guide to interpreting the Online Inspiral FOM using an example from LHO. The points to note are

  1. The low maximum value of the SNR axis (the y-axis) shows that there are no very loud triggers in this time interval.
  2. For most of the time the maximum SNR is below 8, however there are a few triggers with SNR larger than 10, but smaller than 20. This is not unreasonable, but if you can do something to make these go awat, great! They are probably noise.
  3. At T = -9.6 hours there are several points that show triggers with SNR between 10 and 30 in both instruments. This typically indicates a coincident noise glitch which the scimon should investigate.

Watch for loud triggers due to glitches!

Loud transient noise sources will cause very loud inspiral triggers to be generated. The figure below shows an example of loud glitches at LLO. Note that the maximum value on the SNR axis is now > 2000. There are several 60 second bins where the maximum value of snr is greater than 100. These should be investigated to determine the source of the glitches.

Another example of loud glitches

The example below is from LHO. Note that at T = -4.6 hours, there are several very loud triggers (the SNR of one is at least 1000). This is very abnormal and should be investigated. In this case, we compare the InspiralMon FOM to the AS_Q band power FOM. Note that the loud inspiral triggers are coincident with sharp spikes in the AS_Q band power in both H1 and H2. This suggests some coincident transient which should be investigated.


Additional Information Generated by the Search and InspiralMon

As well as the maximum SNR every 60 seconds, InspiralMon also generates

  1. the number of triggers above threshold every 60 seconds
  2. time series of the analyzed data
  3. histograms of the square of the trigger SNR for the last 2, 6 and 12 hours
  4. the min, mean and max of the triggers every 60 seconds

Note that the chi-squared histograms contain no data, as the chi-squared veto is not being computed online in S4.

These channels are available via the DMT viewer and may be used to further investigate online inspiral triggers by scimons.


Page create by Duncan Brown.

$Id: InspiralMon.html,v 1.1 2006/07/06 18:04:58 jzweizig Exp $