S2 / S3 Investigation Reports

 
• Data quality   (John Zweizig)

John summarized a number of ongoing investigations of S3 data quality he and others are carrying out. His data quality investigations page has details, but in brief there are a number of DAQ/controls or operational issues under study right now or which have been studied:
• Detected channel overflows within the LSC system. If treated as vetoes, the periods of overflow give S3 deadtimes of 0.01%, 0.14% and 1.4% for H1, H2, and L1, respectively.
• DAQ errors in which lowest-order bytes of some channels were sometimes corrupted. This problem is potentially quite serious if AS_Q is affected. So far there is no strong evidence of AS_Q being affected, but the possibility cannot yet be ruled out. John is working on an offline emulation of the LSC front-end controller to cross-check AS_Q against DARM_CTRL. Meanwhile Dave Barker has made some progress with post-S3 invasive testing at LHO in isolating the malfunction. It is hoped that definitive statements on AS_Q corruption can be made soon.
• Preliminary studies indcate that glitches observed on some of the large optic OSEM sensor channels did not affect AS_Q.
• Because of various operational glitches, calibration lines were not always present. Resulting deadtimes are 0.29%, 0.64% and 1.04% for H1, H2, and L1, respectively, where the bulk of the L1 troubles came from a single night.
• Front-end synchronization errors in which processors failed to complete cycles in the nominal 61 us (1 / 16 KHz) are a potential worry. Both error flags and recorded processing times have been examined. Setting reasonable thresholds for vetoing data leads to deadtimes of 0%, 0% and 0.18% for H1, H2, and L1, respectively.

 
• Calibration    (Mike Landry)

Mike has been carrying out post-S3 calibration measurements: swepts sines with loop gains, cross-couplings, a variety of DC calibration techniques, and measurements of electronics chain transfer functions. A new technique for DC calibration in which one end mass and the other arm's input mass are used to form a Michelson seems quite promising as a more direct handle on end mass actuation. It is hoped to reduce the uncertainty in DC calibration by as much as a factor of five. Similar measurements are underway or already completed at Livingston by Brian O'Reilly. Point functions for L1 have been posted based on early-run calibrations. Point functions for H1 and H2 based on post-run calibrations will be available in the next day or so. Mike is also producing ascii versions of alpha and alpha*beta for the pulsar group. Patrick mentioned that he is starting to look at SenseMon results for these parameters too.
 
• Timing (Szabi Marka, Daniel Sigg)

A preliminary report on S3 timing indicates timing spreads no worse than 10 us over the run, sub-us precision for H1 and about 1 us precision for H2. L1 was subject to timing jumps of several us during the run. It is believed (but has not yet been verified) that all timing jumps are associated with DAQ reboots. At John's request, Szabi agreed to provide a list of means and standard deviations of timing for each S3 science mode segment for each IFO  (note: after the meeting Szabi provided a revised draft above with the requested lists in appendices). Much work remains to pin down the timing for S3, but work so far looks very promising. A report on the new atomic clock systems at the sites will be given at a future telecon.
 
• Glitches / Burst/Inspiral search group vetoes (Laura Cadonati, Erik Katsavounidis)

No update.
 
• Interchannel correlations   (Nelson Christensen)

Nelson was unable to make the telecon, but provided the link above as a snapshot of ongoing work. He will report at the next detchar telecon.
 
• Hardware signal injections (Peter Shawhan)

Peter gave a brief summary of injection work during the S3 run. Pulsars were injected for all of the run, except for a three-week hiatus December 1-22. The stochastic group did some injections at the start of the run. There were a large variety of burst/inspiral waveforms injected at various times during the run, with fewer injections on L1 because of limited livetime. Szabi has posted on the hardware signal injections home page a summary of all elog entries concerning injections. Peter is working on a list of locked segment numbers for all burst/inspiral injections and on a list of science-mode segments when the pulsars were turned on. A problem with the actuation function handling of a violin mode notch was fixed manually by editing the actuation function. Mike mentioned that in the future the calibration group will provide actuation functions with and without the notch present.
 
• Reduced data set generation - channel lists (Isabel Leonor)

Isabel confirmed that all level 1, 2, and 3 RDS sets have been generated at the sites. (Stuart mentioned that the last batch of level 1 RDS from each site will be mailed shortly to the other site.) The RDS generation at Caltech is nearly done. Greg Mendell has posted web pages listing RDS data gaps for LHO and LLO , where the RDS gaps arise from gaps in the raw data.
 
• Violin modes (S. Klimenko, J. Castiglione)

Sergey has not yet looked at the S3 results, but believes that the recorded trends from LineMon will be sufficient for his studies. He does not expect to rerun over the S3 data, as he had to do for S1 and S2.
 
• Bilinear couplings (S. Penn)

Steve ran a couple of different version of the new BicoMon background monitor during S3 and needs to go back and create a complete, consistent history. He also found and fixed a bug recently that may have affected the recorded trends. He plans to do a full rerun over the S3 level 1 RDS at the sites. As before, he would welcome assistance in analyzing the S3 results.
 
• Correlated inter-site environmental transients (R. Schofield)

The DMT program (based on glitchmon) that looks for excess noise in coincidence in inter-site channels is about halfway through processing the special S2 combined PEM reduced data set. Also, Tania Regimbau is running the stochastic group's coherence code on S2 PEM channels and is about halfway done. Greg Mendell will start production of combined PEM RDS for S3 today.
.
• Local environmental disturbances (R. Schofield)

PEM injections were carried out three times for S3 (one was half a set) at LHO, and twice at LLO (Josh Dalrymple is carrying out the second LLO set).  Preliminary results include:
• Acoustic coupling into AS_Q is worst for L1, best for H1, with H2 intermediate. For all three IFO's the worse acoustic coupling point is the REFL port. It appears that present acoustic noise in the 57-100 Hz range for L1 would prevent reaching design sensitivity in that range.
• Except for EY on L1 (where it is believed one actuation magnet has a flipped polarity), the higher-order harmonics of 60 Hz (120 Hz and higher) present in AS_Q appear to come from electronics noise, not from direct magnetic coupling to the test masses.
• The 60 Hz fundamental, however, does introduce substantial harmonics in the form of a ~1 Hz comb extending out 10 Hz to either side of 60 Hz. Fine structure in the comb permits identification of the test mass coupling sites.
• Gravel trucks on the highway near LHO seemed to cause some lock losses on both LHO IFO's during S3. The trucks appear to increase glitch rates on AS_Q correlated with seismic noise levels in the 3-10 Hz band. In particular, Robert recommends vetoing LHO data when LVEA_SEISZ is above 1000 counts.
• The infamous glitch "bubble" on H1 in the 100-300 Hz range was suspected to arise from dust stirred up by entries into the LVEA, but the implied settling time of several hours seemed too long. After the run, Robert installed a dust monitor and verified that the settling time is indeed several hours. He also verified that dust flashes visible with a video camera coincide with bubble-like glitches.
  
  
• Data Access (Peter Shawhan)

Daniel's latest version of DTT (vsn 2.3) permits data stored on tape in SAM-QFS to be retrieved as if on disk. One must be careful, however, not to provoke a regeneration of  directory file lists. Peter has written a script to make pre-indexed list of files (see instructions for use on the S3 data access page). The zero-th order S3 segment lists (H1, H2, L1) were created in real-time and needs some vetting, (although the segment numbers themselves will not change). Stuart asked whether it would be useful to give fortress and decatur direct fiber channel access to archived data (100 MB/s vs present 10 MB/s). John wasn't sure, but would look into it.

Gamma-Ray-Burst Alert Procedures [ e-mail from Rauha Rahkola and presentation ( ppt , pdf )]

• Rauha summarized the commissioning work during S3 of the GRB alert system in the control room. These alerts are triggered by e-mail generated by the HETE satellite. Although 30 alerts were issued in the last month of S3, only one was eventually confirmed by HETE to be a GRB. There was one outage of the e-mail alerting because of a power problem at Caltech which took down saiph at Caltech on Christmas day. Based on e-log entries, it appears that operators and scimons did follow the instructions to avoid unnecessary lock loss for an hour after an alert, which itself follows the GRB detection by up to six minutes (most less than three minutes). Rauha proposes that between now and S4 (toward the end of 2004),  the same procedure be followed (avoiding lock loss for an hour) if an IFO is already in common mode. No one on the line objected; so it will be tried to see how it will work in practice. Part of the procedure will involve operators following a web checklist. The initial "automation" of data archiving (+/- 1 hours about alert) will be based on an e-mail to Greg, Ben Johnson and Igor Yakushin). Stuart mentioned that he thought Ben has a script that could be adapted for truly automated archiving. Mike remarked that GEO is very interested in joining the GRB "watch", given their ~90% duty cycle when not commissioning.

• Next detector characterization telecon: Friday January 30