Information regarding GEO data quality flags

Automatically inserted flags (near real-time): (ascii list)
(John Zweizig, Duncan Brown)

Manually inserted flags: (mailing list for insertion updates) (ascii list)
(Keith Riles)

Details:

• The SegGener online DMT monitor (John Zweizig) monitors a variety of channels and creates data quality flags ("segment types") that are inserted directly into the database in near real-time. More information can be found at John's Data Quality Investigation Page. (Note that the ASI_CORR_OVERFLOW flag was not implemented online until June 2006. Flagged segments before GPS 835483849 were inserted manually.)
  Flag(s): ASI_CORR_OVERFLOW H1_Not_Locked H2_Not_Locked PD_Overflow Wind_Over_30MPH
  
  
• Keith Riles has compiled triggers from the PlaneMon DMT monitor (author: Evan Goetz) to identify 2-minute intervals when there is strong acoustic evidence for overflying aircraft. Triggers are classified as "likely" or "very likely". Here are the results:

  AIRCRAFT_LIKELY	AIRCRAFT_VERY_LIKELY
  LHO (32691)	LHO (29697)
  LLO (12561)	LLO (10641)

  Flag(s): AIRCRAFT_LIKELY AIRCRAFT_VERY_LIKELY
  
  
• Under certain conditions of saturation, the IFO's output beam to the antisymmetric port's four photodiodes is temporarily diverted to a fifth photodiode, to avoid lock loss and resulting prolonged downtime. The triggering of this diverted light automatically takes the interferometer out of science mode. Unfortunately, there was a period during which this latter mechanism wasn't enabled for L1, as found by the Glitch Group. Gaby Gonzalez, Dan Hoak, and Steve Fairhurst compiled a list of these intervals (glitch notebook entry). In September 2006 it was discovered empirically that one can go into science mode when in the AS_TRIGGER mode! John Zweizig checked the data and found a handful of L1 incidents where this had happened.In March 2007, John updated his list through February 8, 2007 for H1, H2, and L1. In August 2007 John refined these intervals to account for a difference in online-derived and offline-derived segment lists to obtain: H1, H2, and L1. John provided a final full-S5 update in January 2008 for H1, H2, and L1.
  Flag(s): AS_TRIGGER
  
  
• Erik Katsavounidis has studied periodicity of KleineWelle triggers to identify L1 intervals when the online autoburt process was saving controls settings and causing glitches in the gravitational wave channel (mechanism not understood yet).
  Flag(s): AUTOBURT_GLITCHES
  
  
• Gaby Gonzalez identified a period in February 2006 when the notching of an L1 end mass suspension was incorrect, leading to a bad calibration of L1 near the drumhead resonance frequency of 7 kHz. Most analyses should not be affected by this problem.
  Flag(s): BAD_CALIBRATION_NEAR_7KHZ
  
  
• In February 2006 a lowering of air conditioning airflow at LLO (attempt to reduce inflow air turbulence leading to seismic noise) led to increased temperatures and shifting of a mechanical resonance frequency that had previously been notched. As a result, the DARM loop sensing function was unpredictably distorted, making calibration unreliable. Gaby Gonzalez determined that L1 segments 961-968 were affected (see her note and accompanying figures 1, 2 and 3).
  Flag(s): BAD_SENSING
  
  
• This flag is meant to cover known cases of servo loop troubles, such as a front-end processor malfunction. Shantanu Desai found a period in July 2006 when there were unphysical dips in apparent light power in L1's arms that coincided with a known problem with the processor l1iscex. Shantanu's estimate is that the interval 838424724-838426097 was affected.
  Flag(s): BAD_SERVO
  
  
• This flag is motivated by a period of high glitchiness in June 2007, as seen by the Block-Normal burst algorithm and reported by Shantanu Desai.
  Flag(s): BN_GLITCHINESS
  
  
• Shantanu Desai has identified glitches in the L1 gravitational wave channel due to glitches in the beam splitter (BS) servo at these times. John Zweizig has identified intervals in August/September 2007 during which the variance in the BS servo channel was larger than normal (see John's notes 1, 2 and plots), part of which overlap with segments found independently by Gaby Gonzalez characterized by a 3.6 Hz oscillation. Cristina Torres and Gaby Gonzalez have carried out a spectrogram study of all S5 data, looking for gain peaking in the this servo, yielding these intervals.
  Flag(s): BS_OPTLEVER BS_OPTLEVER_GAINPEAKING BS_OPTLEVER_HIGHRMS
  
  
• When the gravitational wave servo gain fluctuates a large amount (e.g., via uncompensated optical gain loss from mirror wobble), the servo can become less stable, leading to increased noise, including saturation. Injected calibration lines allow monitoring of the loop gain. Myungkee Sung has identified intervals for H1, H2, and L1 when the resulting calibration coefficients are outside the desirable range, when averaged over 60-second intervals. Myungkee has provided an update (April 30, 2007) for the version 3 calibration release that covers from the start of S5 to January 22, 2007 for H1, H2, and L1. Myungkee has also provided an update (May 8, 2007) covering the 50 hour period centered on GRB070201 for H1 and H2 (see GRB070201 calibration page for more info).
  Flag(s): CALIB_BAD_COEFFS_60
  
  
• For reasons not yet understood, the arbitrary waveform generator (awg) used for injecting calibration lines and astrophysical signals occasionally malfunctions (hence the importance of recording these channels). John Zweizig monitors these injection channels online with the ZGlitch DMT monitor, looking for glitches that can indicate a single-sample dropout, a 1-second (or more) dropout, a "stuck" injection (repeated 1-second sequences), or other large disturbances. He has compiled these intervals for H1, H2 and L1. More information can be found on his Data Quality Investigation Page. The ZGlitch monitor did not track these glitches, however, early in the run for L1. Shantanu Desai has provided an interim list of large such glitches found offline by the BlockNormal burst search algorithm. In addition, John later went back and generated the early L1 segments which have now been inserted. In addition, Peter Shawhan discovered some cases in L1 data during which the online ZGlitch monitor wasn't running and missed dropouts.
  To avoid confusion on version numbers, all of the ZGlitch-derived calibration dropout / glitch intervals were consolidated into version 4 for all three interferometers in February 2007. Here are the relevant merged files as of February 27, 2007:
  H1: 1-sample 1-second AWG-stuck Glitch
  H2: 1-sample 1-second AWG-stuck Glitch
  L1: 1-sample 1-second AWG-stuck Glitch
  John updated his intervals through October 15, 2007. Here are the corresponding consolidate interval files:
  H1: 1-sample 1-second AWG-stuck Glitch
  H2: 1-sample 1-second AWG-stuck Glitch
  L1: 1-sample 1-second AWG-stuck Glitch
  Flag(s): CALIB_DROPOUT_1SAMPLE CALIB_DROPOUT_1SEC CALIB_DROPOUT_AWG_STUCK CALIB_DROPOUT_BN CALIB_GLITCH_ZG
  
  
• Duncan Brown has provided the following information (more information can be obtained upon request). In order to perform a test of the inspiral analysis pipeline, the inspiral group has created a new data quality flag CBC_1YR_INJ_CHALLENGE which vetoes all except two of the hardware injections during the data being used for the 1 calendar year S5 run (Starting at Nov 04 2005 16:00:00 UTC and ending at Nov 14 2006 16:05:56 UTC.)
  Flag(s): CBC_1YR_INJ_CHALLENGE
  
  
• It was discovered in March 2006 that occasionally LHO data channels had repeated segments with an integer number of 1/16 seconds in length. More investigation revealed that the frame builders were occasionally overloaded, causing some data channels written to disk to "fall behind". A good diagnostic of these failures was found to be mismatches between the data copies written by the two framebuilder machines (fb0 and fb1). Ben Johnson has scanned all of the LHO level 0 data to look for a difference in the checksums produced by fb0 and fb1 for each 32-second frame. More information can be found here (see README file). The affected 32-second segments for the period November 4, 2005 to April 18, 2006 are here.
  Flag(s): CHECKSUM_MISMATCH
  
  
• Keith Riles used the Hanford conlog engine to identify epochs when the operator-flagged channels Hn:IFO-ACTIVITY_TYPE were not set to "No Activity". Other possible values for these flags during S5 were "Vehicular", "Other, see elog", "Anthropogenic", "Machine", and "Lab". Slightly modified flag names were defined for these epochs, given below:
  HUMAN_ACTIVITY: H1, H2
  LAB_ACTIVITY: H1
  MACHINE_ACTIVITY: H1, H2
  OTHER_ACTIVITY: H1, H2
  VEHICULAR_ACTIVITY: H1, H2
  Flag(s): HUMAN_ACTIVITY LAB_ACTIVITY MACHINE_ACTIVITY OTHER_ACTIVITY VEHICULAR_ACTIVITY
  
  
• There was a period of about two weeks in November 2006 when the conlog program was not working. During that period changes in relevant controls could occur without science mode being dropped, including AS triggers. In principle, this data is usable, but requires some care, e.g., explicit searching for control changes in the vicinity of a putative astrophysical candidate. The period that has been flagged here is Monday November 13 12:00 PST to Tuesday November 28 16:00 PST.
  Flag(s): CONLOG_SICK
  
  
• Note from Xavi Siemens (author of h(t) generation program) on corrupted, unrecoverable strain data. One more report. Keita Kawabe (chair of calibration review committee) provided an update on corrupted h(t) for version 3 ("C03") calibration on January 1, 2008. Update from Greg Mendell on January 14, 2008 concerning missing h(t) level 2 files, with a note on a corrupted interval for H1.
  Flag(s): CORRUPTED_RDS_C02_LX CORRUPTED_RDS_C03_LX
  
  
• Ray Frey and collaborators have set up a scintillator-based cosmic ray detector under the H1 ITMY mirror and collected data on showers since early S5. Since only strong showers have been flagged in the database, both H1 and H2 are marked. The first entries in the database include data from three epochs: 2006A (Feb 21, 2006 - July 25, 2006), 2006D (Sep 12, 2006 - Dec 31, 2006), and 2007A (Jan 1, 2007 - July 2, 2007). A December update from Ray provides data for three additional epochs: 2006B (July 25, 2006 - Sep 12, 2006), 2007B (July 2, 2007 - Aug 31, 2007), and 2007C (Sep 1, 2007 - Sep 30, 2007).
  Flag(s): COSMIC_RAY
  
  
• A data-valid flag is written for each archived data frame for each channel. If the data-valid flag is found to be invalid for the gravitational wave data channel, then a DAQ_ERROR DQ flag is set. As of September 2006, only a period of a few days in mid-September has been examined closely (known to have a problem). To be conservative, any 256-second frame in level 3 RDS showing a data-valid error for DARM_ERR is flagged. However, the presence of a data-valid error in an auxiliary channel is not flagged at this time (see sample LLO list from Ben Johnson). S
  Flag(s): DAQ_ERROR
  
  
• Justin Garofoli has studied upconversion of low-frequency noise in H1 DARM_ERR to the 70-110 Hz band of DARM_ERR and produced intervals of high noise in four bands using three different thresholds:
  0.9-1.1 Hz: Low, Medium, High threshold
  1.1-1.3 Hz: Low, Medium, High threshold
  1.8-2.4 Hz: Low, Medium, High threshold
  5.0-7.0 Hz: Low, Medium, High threshold
  which cover from the start of S5 to the end of April 2007. Justin's recommendation is that analysts pay most attention to the 5.0-7.0 Hz flags and to the high-threshold flag for 0.9-1.1 Hz. (Note that the version 1 flags should be ignored; they were replaced by version 2 on May 14.) Justin provided an update for the full S5 data in February 2008 (version 11):
  0.9-1.1 Hz: Low, Medium, High threshold
  1.1-1.3 Hz: Low, Medium, High threshold
  1.8-2.4 Hz: Low, Medium, High threshold
  5.0-7.0 Hz: Low, Medium, High threshold
  Flag(s): DARM_09_11_dHz_LOWTHRESH DARM_09_11_dHz_MEDTHRESH DARM_09_11_dHz_HIGHTHRESH DARM_11_13_dHz_LOWTHRESH DARM_11_13_dHz_MEDTHRESH DARM_11_13_dHz_HIGHTHRESH DARM_18_24_dHz_LOWTHRESH DARM_18_24_dHz_MEDTHRESH DARM_18_24_dHz_HIGHTHRESH DARM_50_70_dHz_LOWTHRESH DARM_50_70_dHz_MEDTHRESH DARM_50_70_dHz_HIGHTHRESH
  
  
• Emelie Harstad has studied dewar glitches found from her DewarMon DMT monitor program. Her first list covers the period December 12, 2006 to February 23, 2007.
  Flag(s): DEWAR_GLITCH
  
  
• The Glitch Group has set up elog-scanning shifts to go through the S5 LHO and LLO elog entries to identify periods when operators or scientists on shift noted a problem that should be flagged in the database. As those shifts are completed and reports written, the information here will be updated. In the tables below are 1) epoch; 2) links to the shift reports which in turn have links to original elog entries; and 3) links to distillations of the reports with explicit DQ flag names and segment lists; and 4) investigators:

  Livingston Observatory Elog Scans
  Epoch	Shift Report	Distillation	Investigator(s)
  Nov 2005 - Feb 2006			Sarah Caudill, Christina Torres
  Mar 2006 - Jun 2006			Natalia Zotov
  Jul 2006 - Oct 2006			(Syracuse)
  Nov 2006 - Feb 2007			(MIT)
  Mar 2007 - Jun 2007	Report		Laura Cadonati
  Jul 2007 - Sep 2007	Report		(UTB)

  Hanford Observatory Elog Scans
  Epoch	Shift Report	Distillation	Investigator(s)
  Nov 2005 - Jan 2006			Keith Riles
  Feb 2006 - Apr 2006	Report		Nelson Christensen
  May 2006 - Jul 2006			Shourov Chatterji
  Aug 2006 - Oct 2006			John Zweizig
  Nov 2006 - Jan 2007	Report	Distilled	Shantanu Desai
  Feb 2007 - Apr 2007			Robert Schofield
  May 2007 - Jul 2007			Michele Zanolin
  Aug 2007 - Sep 2007			Warren Johnson, Myungkee Sung

  
  Flag(s): COMMISSIONING_ELOG EARTHQUAKE_ELOG NONSTAND_CONFIG_ELOG NONSTAND_TCS_ELOG PROCESSOR_TROUBLE_ELOG RFID_READER_ON_ELOG TIMING_ERROR_ELOG
  
  

• John Zweizig created online DQ flags to mark intervals with overflow in end test mass coil currents, but they were found not to be useful. These are not recommended for use in analysis.
  Flag(s): ETMX_COIL_OVERFLOW ETMY_COIL_OVERFLOW
  
  
• Masahiro Ito has studied excitations of H2 folding-mirror roll-modes due to seismic noise from a nearby compressor motor and identified periods when the compressor motor frequency drifted into three narrow resonant band of the roll modes (in the range 17-18 Hz), greatly increasing excitations of the fundamental and higher harmonics.
  Flag(s): FM_ROLLMODE_EXCITED
  
  
• Gaby Gonzalez has examined intervals of L1 data characterized by large variance in the inspiral range and identified epochs for which the glitchiness merits flagging, for a variety of reasons documented here. Specific flag assignments can be found here. Gaby updated this study in February 2008 for L1 for all of S5, adding several new flags (distillation of segments). In March 2008 Shantanu Desai followed up on the railed-RBS-pzt glitches for L1 to produce a refined list of intervals with finer time resolution. Gaby carried out a similar study for H2 in March 2008, leading to many new flags (included below).
  Flag(s): ASC_TROUBLE BALING_ACTIVITY BADMONTH BADRANGE_GLITCHINESS BSOPLEV_3p6HZOSCILLATIONS EARTHQUAKE_GLITCHINESS ELEVATED_GLITCHINESS EX_LOGGING HURRICANE_GLITCHINESS ISCT10_TROUBLE LVEA_NOISY MMT3_GLITCHINESS MOVED_LVEA_SEIS RAILED_RBS_PZT ROUTE10_TRAFFIC SEC_LOGGING SENSEMON_RANGE_STEP SEVERE_GLITCHINESS SICK SPOB_GLITCHINESS STACK_RESONANCE TCS_TROUBLE
  
  
• Lindy Blackburn has examined intervals of H1 data containing H2 lock losses and acquisitions and vice-versa (H1 <--> H1). He sees an enhanced rate of KleineWelle glitches in the stable IFO when the other loses or acquires lock. Comparing these rates with control-sample rates, he flags the following intervals:
  H1 glitchiness when H2 loses lock (-10 to +100 seconds)
  H1 glitchiness when H2 acquires lock (-300 to 0 seconds)
  H1 glitchiness when H2 loses lock (-20 to +300 seconds)
  H1 glitchiness when H2 acquires lock (-400 to 0 seconds).
  Lindy updated his results in December 2007 through early August 2007 to produce the following intervals:
  H1 glitchiness when H2 loses lock (-10 to +100 seconds)
  H1 glitchiness when H2 acquires lock (-300 to 0 seconds)
  H1 glitchiness when H2 loses lock (-20 to +300 seconds)
  H1 glitchiness when H2 acquires lock (-400 to 0 seconds).
  Lindy revised his results in March 2008 to cover the rest of S5 through the end of September 2007 and to fix a bug in the original flag generation that incorrectly vetoed some intervals that spanned 00:00 UTC:
  H1 glitchiness when H2 loses lock (-10 to +100 seconds)
  H1 glitchiness when H2 acquires lock (-300 to 0 seconds)
  H1 glitchiness when H2 loses lock (-20 to +300 seconds)
  H1 glitchiness when H2 acquires lock (-400 to 0 seconds).
  
  Flag(s): H1_LOCKLOSS H1_LOCKGAIN H2_LOCKLOSS H2_LOCKGAIN
  
  
• Robert Schofield has compiled intervals during which the chance of light scattering from H1 to H2 and back again or vice-versa is enhanced, based on a study of minute trends of band-limited RMS (1-4 Hz) in the H1 and H2 DARM_CTRL channels.
  Flag(s): H1H2_SCATTERING
  
  
• Hardware signal injections are flagged automatically in the publishing of science segments, based on the current state vector. Occasionally malfunctions cause injections not to be flagged automatically. In April 2007 Peter Shawhan and Vuk Mandic provided updated injection intervals for H1, H2, and L1.
  Flag(s): Injection
  
  
• A number of "blind" injections have been carried out in 2007 in order to test the LSC's ability to detect a gravitational wave signal without bias. As of September 13, 2007, a number of the the blind injections made for testing have been revealed (unblinded) by the injecters.
  Flag(s): INJECTION_UNBLINDED
  
  
• John Zweizig has refined the algorithm for defining the original DAQ_ERROR DQ flag to pin down with finer time precision what second(s) of data are affected by a data-valid error. Level-3 RDS has been used to find data valid errors for DARM_ERR in the first year of S5 running for H1, H2, and L1. John updated the segments through the end of S5 in January 2008 for H1, H2, and L1 (only DARM_ERR segments are flagged).
  Flag(s): INVALID_DARMERR
  
  
• Shantanu Desai has identified glitches seen simultaneously in H2's gravitational wave channel and in accelerometers at these times on the output dark-port optical table ISCT10 during a period of reduced air pressure for the floating of that table. Shantanu updated this study in May 2008 to flag milder H2 glitches found by Nelson Christensen's group to be effective vetoes for inspiral triggers.
  Flag(s): ISCT10_TABLE_GLITCH ISCT10_TABLE_GLITCH_MILD
  
  
• Occasionally the light power in the interferometer arms dips (e.g., from a wobbling mirror), leading to a brief but significant drop in optical gain, leading to brief servo instability and increased noise. These dips in light power can be too brief to affect significantly the minute-averaged calibration coefficients. Peter Shawhan has looked at second-trends of arm power and compiled lists of such dips for many different percentage thresholds, to allow further investigation:
  Feb 2006 update:
  H1: 2% 3% 4% 5% 6% 8% 10% 12% 15% 20%
  H2: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  L1: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  May 2006 update:
  H1: 2% 3% 4% 5% 6% 8% 10% 12% 15% 20%
  H2: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  L1: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  Nov 2006 update:
  H1: 2% 3% 4% 5% 6% 8% 10% 12% 15% 20%
  H2: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  L1: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  May 2007 update:
  H1: 2% 3% 4% 5% 6% 8% 10% 12% 15% 20%
  H2: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  L1: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  Dec 2007 update:
  H1: 2% 3% 4% 5% 6% 8% 10% 12% 15% 20%
  H2: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  L1: 4% 5% 6% 7% 8% 10% 12% 15% 20%
  Flag(s): LIGHTDIP_02_PERCENT LIGHTDIP_03_PERCENT LIGHTDIP_04_PERCENT LIGHTDIP_05_PERCENT LIGHTDIP_06_PERCENT LIGHTDIP_07_PERCENT LIGHTDIP_08_PERCENT LIGHTDIP_10_PERCENT LIGHTDIP_12_PERCENT LIGHTDIP_15_PERCENT LIGHTDIP_20_PERCENT
  
  
• The front-end processor code keeps track of overflows detected in ADC's and in calculations and increments various "master counters" of such overflows. Peter Shawhan has examined second-trends for instances where the master overflow counters for different subsystems have incremented. Such overflows are likely to be important in the LSC (Length Sensing and Control) and in the ASC (Alignment Sensing and Control) subsystems. They may also be important in the SUS (Suspensions) subsystem (but see Peter's caution concerning SUS-MC2 overflows for L1). Note that in the December 2007 update a new overflow flag was added for all three IFO's: MASTER_OVERFLOW_IOO
  Feb 2006 update:
  H1: LSC ASC SUS-MC2 SUS-RM
  H2: LSC ASC SUS-MC2
  L1: LSC ASC SUS-MC2 SUS-RM
  May 2006 update:
  H1: LSC ASC SUS-MC2 SUS-RM
  H2: LSC ASC SUS-MC2
  L1: LSC ASC SUS-MC2 SUS-RM
  Nov 2006 update:
  H1: LSC ASC SUS-MC2 SUS-RM
  H2: LSC ASC SUS-MC2
  L1: LSC ASC SUS-MC2 SUS-RM
  May 2007 update:
  H1: LSC ASC SUS-MC2 SUS-RM
  H2: LSC ASC SUS-MC2
  L1: LSC ASC SUS-MC2 SUS-RM
  Dec 2007 update:
  H1: LSC ASC SUS-MC2 SUS-RM IOO
  H2: LSC ASC SUS-MC2 IOO
  L1: LSC ASC SUS-MC2 SUS-RM IOO
  Flag(s): MASTER_OVERFLOW_ASC MASTER_OVERFLOW_LSC MASTER_OVERFLOW_SUS_MC2 MASTER_OVERFLOW_SUS_RM MASTER_OVERFLOW_IOO
  
  
• Greg Mendell periodically runs scripts to look for missing raw data and missing reduced data sets by scanning frame files in the archive.
  Greg's original compilation from October 25, 2006
  Greg's update on November 25, 2006
  Greg's update on January 4, 2007
  Greg's update on February 1, 2007
  Greg's update on May 30, 2007
  Greg's update on January 31, 2008
  Flag(s): MISSING_RAW MISSING_RDS_LEVEL_1 MISSING_RDS_LEVEL_3 MISSING_RDS_LEVEL_4
  
  
• The program to generate calibrated strain data, i.e., h(t), uses finite-impulse-response filters that require a settling time of up to 8 seconds (H2 and L1) or 16 seconds (H1), depending on notch filters used in the interferometer actuation control, leading to short gaps at the starts and stops of many data segments (the filters run both forward and backward). Xavi Siemens (author of h(t) generation program) has provided the first lists of missing h(t) segments in S5 RDS frames through May 27, 2006 (H1, H2, L1) where most of the gaps come from the filtering effects. There are a few pathological cases, however, documented here by Shourov Chatterji, due to other problems. [For future updates, Greg Mendell has set up an automated program for keeping track of gaps in both the raw data and h(t) reduced data.] Update from Xavi on February 8, 2007. Missing segment reported by Drew Keppel. Another update from Xavi on March 7, 2007. Update on H1/L1 version 3 intervals from Xavi on June 13, 2007. Update on H2 version 3 intervals from Xavi on July 5, 2007. Update on H1/H2/L1 version 3 intervals created online after April 23, 2007 from Greg on June 8, 2007. H2 correction from Xavi on August 8, 2007. Update from Greg for all of S5 level 2 version 3 h(t). Update from Greg for all of S5 level 1 version 3 h(t).
  Flag(s): MISSING_RDS_C02_LX MISSING_RDS_C03_L1 MISSING_RDS_C03_L2
  
  
• During the period January 11-20, 2006 the Mode Matching Telescope mirror 3's (MMT3) optical lever servo for H2 was prone to glitches that fed into the gravitational wave channel. Gaby Gonzalez compiled a list of 1-second intervals containing such glitches in the period 821000000-822000000. To be conservative, these intervals have been lengthened to 3-second intevals (+/-1 second w.r.t. Gaby's intervals). There are more than 14,000 intervals flagged; analysts will likely want to use these for post-analysis vetoes, not for data selection (unless it proves prudent to veto the entire affected range: 821025287-821806675). See Gaby's e-notebook entry for more details.
  Flag(s): MMT3_OPTLEVER
  
  
• While scanning the elogs Shantanu Desai found an H2 epoch (846398109-846441633) during which saturation of the MMT3 yaw control servo led to degradation in inspiral range.
  Flag(s): MMT3_YAW_SATURATION
  
  
• Although standard operating procedure requires calibration lines to be turned on before entering science mode, occasionally science mode is entered without those lines, making the calibration of that data difficult to determine accurately. Myungkee Sung has compiled intervals of data from the start of S5 through July 12, 2007 for which the calibration excitation channel is zero, indicating no injection. Details can be found here. These lists were updated September 28, 2007 to include intervals previously omitted because they were already flagged as having bad calibration coefficients: H1, H2 and H1.
  Flag(s): NO_CALIB_LINE
  
  
• The local damping of the mirror motions depends on photodiode sensor channels that occasionally glitch. Aggressive low-pass filtering is meant to ameliorate the effects of these glitches, but early in S5 the H2 filtering was less aggressive than intended, leading to occasional glitches visible in the gravitational wave channel. John Zweizig has looked at periods of large rms deviation on the OSEM (Optical Sensing Electro-Magnets) channels that are correlated with glitches in the gravitational wave channel. For more information and sample plots, see his Data Quality Investigation Page. Intervals for H2 have been compiled for the following sensor channels:
  ITMX_LL ITMX_UL ITMY_LL ITMY_UL
  Flag(s): OSEM_GLITCH
  
  
• The published science segments are derived from the slow EPICS state vector channel which is delayed by 1-2 seconds in responding to lock loss. As a result, many science segments include 1-2 seconds of unlocked data at their ends. Peter Shawhan has examined 1-second trends of arm power light to determine more precisely the second in which lock is lost to compile the following intervals:
  Feb 2006 update: H1, H2, L1
  May 2006 update: H1, H2, L1
  Nov 2006 update: H1, H2, L1
  May 2007 update: H1, H2, L1
  Dec 2007 update: H1, H2, L1
  Flag(s): OUT_OF_LOCK
  
  
• A large variety of non-standard running conditions are flagged as "out of science mode", based on elog entries by operators or scimons.
  Flag(s): OUT_OF_SCIENCE_MODE
  
  
• John Zweizig has examined trigger rates from the PDnMon DMT monitor (looks at asymmetries in response from the four photodiodes at the antisymmetric port) and has used KleineWelle triggers to assess their relevance to burst searches. An initial pilot study for the latter part of S5 yields the following preliminary intervals (1-minute discreteness): H1, H2 and L1. This same study yielded a couple of L1 intervals when a single photodiode was not responding.
  Flag(s): PHOTODIODE_GLITCH PHOTODIODE_OFF
  
  
• Gaby Gonzalez reported an H2 interval with PEM injections for which the state vector was not properly marked (but was noted in the elog).
  Flag(s): PEM_INJECTION
  
  
• Robert Schofield has collected information from the electric power utility near Hanford on known glitches and disruptions to power in the first year of S5 and converted the information to GPS intervals. Two categories were reported: 1) short, intentional glitches caused by insertion of capacitor banks; and 2) unintentional disruptions on the transmission lines the exact times of which are much less well known. Robert provided an update in January 2008 for the second year of S5 for glitches and disruptions.
  Flag(s): POWERMAINS_DISRUPTION POWERMAINS_GLITCH
  
  
• Laura Cadonati has compiled a list of glitches on multiple magnetometers at Hanford or Livingston. These glitches are sometimes associated with glitches in the gravitational wave channel. Many coincide with disturbances on the power mains. Laura provided an update on May 13, 2007 with a slight refinement of selection criteria and extending the period covered up to March 15, 2007. Laura provided another update on November 5, 2007 to cover the remainder of S5.
  Flag(s): POWMAG
  
  
• In S5 (and in previous science runs) it has been found that the period immediately preceding lock loss is marked by increased noise, particularly glitchiness. Peter Shawhan has compiled intervals for the periods preceding lock loss by 10 seconds, 30 seconds, 1 minute, and 2 minutes:
  Feb 2006 update:
  H1: 10 secs 30 secs 60 secs 120 secs
  H2: 10 secs 30 secs 60 secs 120 secs
  L1: 10 secs 30 secs 60 secs 120 secs
  May 2006 update:
  H1: 10 secs 30 secs 60 secs 120 secs
  H2: 10 secs 30 secs 60 secs 120 secs
  L1: 10 secs 30 secs 60 secs 120 secs
  Nov 2006 update:
  H1: 10 secs 30 secs 60 secs 120 secs
  H2: 10 secs 30 secs 60 secs 120 secs
  L1: 10 secs 30 secs 60 secs 120 secs
  May 2007 update:
  H1: 10 secs 30 secs 60 secs 120 secs
  H2: 10 secs 30 secs 60 secs 120 secs
  L1: 10 secs 30 secs 60 secs 120 secs
  Dec 2007 update:
  H1: 10 secs 30 secs 60 secs 120 secs
  H2: 10 secs 30 secs 60 secs 120 secs
  L1: 10 secs 30 secs 60 secs 120 secs
  Flag(s): PRE_LOCKLOSS_10_SEC PRE_LOCKLOSS_30_SEC PRE_LOCKLOSS_60_SEC PRE_LOCKLOSS_120_SEC
  
  
• Pulsar hardware signals of low amplitudes were injected for a large fraction of the S5 run. Keith Riles has compared the injection logs in the blue web directories with excitation channel minute trends and taken the logical and of those time intervals. Although the minute trends give a discreteness of 60 seconds, in most cases the flags should have an accuracy of seconds. Because different sets of pulsar injections (test2, test4, s5try1, s5try2, s5try3, posts5binary) were carried out with different amplitudes, each set has its own DQ flag, to make pulsar search validation easier.
  Flag(s): PULSAR_INJECTION_test2 PULSAR_INJECTION_test4 PULSAR_INJECTION_s5try1 PULSAR_INJECTION_s5try2 PULSAR_INJECTION_s5try3 PULSAR_INJECTION_posts5binary
  
  
• As discussed below, The H1 interferometer has been found to be sensitive to seismic noise at the End-Y station, which is relatively close to Highway 240. In addition to the seismic-only SEISMIC_EY_99PCTL_3_10HZ flag defined below, Masahiro Ito has also studied and defined intervals based on simultaneity of high seismic noise in the 3-10 Hz band and high noise in the 5-7 Hz band of the gravitational wave channel itself. (Normally, one would be cautious about applying a DQ based on the GW channel itself, for fear of self-veto of an astrophysical signal, but that mechanism is not plausible here.) Please note that the coupling of the 6 Hz seismic noise into DARM_ERR was greatly mitigated on September 18, 2006 with the introduction of new servo filtering. Masahiro updated these studies in February 2008 for the 2nd year of S5, using an algorithm based on high actuation coil currents in low-frequency bands ((0.1-0.3 Hz, 1.0-1.6 Hz, 1.8-2.6 Hz, 5.0-7.0 Hz, 10.0-13.0 Hz and 17.0-19.0 Hz) found to correlate with upconversion in the GW band of 70-110 Hz, to obtain these H1 intervals.
  Flag(s): COIL_UPCONVERSION SEIS_DARMERR_5_7HZ
  
  
• The H1 interferometer has been found to be sensitive to seismic noise at the End-Y station, which is relatively close to Highway 240. Truck traffic, in particular, creates seismic spikes that correlate with degradation in the H1 gravitational wave channels. As of October 4, 2006, intervals have been identified by Masahiro Ito, based on the seismic noise at EY in the 3-10 Hz band. Seismic noise in that band has been found to vary systematically during the run, leading Masahiro to define 99 percentile levels for different epochs. It is not yet understood why there are distinctly different epochs. The intervals flagged so far are described here, and more information on ongoing LHO seismic studies can be found in reports Masahiro or Emelie Harstad have made at the following weekly detector characterization teleconferences (Sept 1 Sept 8 Sept 15 Sept 22, Sept 29). Masahiro updated his list (March 13, 2007) of intervals, but note that the start/stop times of intervals were refined, going back to the start of S5. Masahiro provided an update through the end of S5 on December 7, 2007.
  Flag(s): SEISMIC_EY_99PCTL_3_10HZ
  
  
• Alistair Hayden has compiled initial sets of intervals of elevated low-frequency seismic noise (0.03-0.10 Hz) at Livingston (Dec 2005-June 2007), based on minute trends of band-limited RMS from seismometer channels at the end stations. Intervals are defined separately for X and Y motion, although there is much natural overlap. Three different thresholds have been chosen, as described here.
  LLO X motion thresholds: Low, Medium, High
  LLO Y motion thresholds: Low, Medium, High
  Alistair updated the study in February 2008 to complete the S5 segments for LLO and to include available segments for LHO (Feb-Oct 2007):
  LLO X motion thresholds: Low, Medium, High
  LLO Y motion thresholds: Low, Medium, High
  LHO X motion thresholds: Low, Medium, High
  LLO Y motion thresholds: Low, Medium, High
  Flag(s): SEISMIC_X_30_100_mHz_LOWTHRESH SEISMIC_X_30_100_mHz_MEDTHRESH SEISMIC_X_30_100_mHz_HIGHTHRESH SEISMIC_Y_30_100_mHz_LOWTHRESH SEISMIC_Y_30_100_mHz_MEDTHRESH SEISMIC_Y_30_100_mHz_HIGHTHRESH
  
  
• Saturation or near-saturation of the side coil currents on the H1 end mirrors can lead to glitches in the gravitational wave channel. Laura Cadonati has studied this effect and produced intervals for H1 ETMX and ETMY covering the first year of S5 running. Laura provided an update covering up to February 19, 2007 on May 23, 2007. John Zweizig has studied the problem further and found that band-limited rms around 6 Hz in H1 side coils is a good indicator of glitchiness in ETMX and ETMY. John provided an update through October 4, 2007 for ETMX and ETMY. Laura provided an update through October 10, 2007 for H1 ETMX and ETMY. John provided another update for all of S5 using an algorithm with much less dead time for ETMX and ETMY.
  Flag: SIDECOIL_ETMX SIDECOIL_ETMY SIDECOIL_ETMX_RMS_6HZ SIDECOIL_ETMY_RMS_6HZ
  
  
• Gaby Gonzalez has identified a period when a sledgehammer was in use at LHO, affecting these 1-second intervals, based on followups of KleineWelle triggers found by Nelson Christensen's group to be effective inspiral vetoes. In addition, Gaby recommends that the entire epoch 16:00-17:50 UTC on July 20, 2006 be marked for concrete work.
  Flag: CONCRETE_WORK SLEDGEHAMMER
  
  
• Vuk Mandic has checked excitation and state vector channel histories to compile lists of intervals for H1, H2, and L1 when stochastic signal injections were present.
  Flag: STOCHASTIC_INJECTION
  
  
• The thermal compensation system (TCS) uses lasers that sometime glitch quite strongly, disturbing the test-mass mirrors upon which they shine. Shantanu Desai has studied this problem and provided lists of loud glitches (based on KleineWelle trigger significance) for H1 and H2. Shantanu has provided an update (April 20, 2007) for H1 and H2 and that covers up through February 1, 2007. Shantanu provided in October 2007 a full-S5 update with the following intervals flagged:
  H1 ITMX PD1, H1 ITMY PD1, H2 ITMY PD1, H2 ITMY PD2.
  In November 2007, Shantanu was able to update a list for L1 triggers from March 18, 2007 onward. L1 triggers from before that date will require scanning of second trends. Shantanu provided an update in June 2008 based on second-trend scanning from Nov 14, 2006 onward for loud glitches, in addition to new intervals marking light dips in the L1 ITMX TCS laser of 10% and 20%.
  Flag: TCS_GLITCH_LOUD TCS_ITMX_LIGHTDIP_10_PERCENT TCS_ITMX_LIGHTDIP_20_PERCENT
  
  
• Glitches are sometimes seen when the H1 tidal servo comes out of saturation. Shantanu Desai has studied this problem and provided lists of segments with different tidal saturation thresholds for the Y arm. He updated this study for the X arm in October 2007 and added intervals with evidence of glitchiness preceding saturation. Shantanu provided another update on January 4, 2008.
  Flag: TIDAL_SERVO_DESATURATION TIDAL_SERVO_PRESATURATION
  
  
• Passing trains south of the Livingston observatory create high seismic noise, especially at the Y end station. Jake Slutsky has flagged periods early in S5 for which seismic noise at End-Y exceeds a threshold guided by binary neutron star inspiral triggers.
  August 2006 update (Nov 14 - Apr 3): L1
  December 2006 update (Nov 14, 2005 - Nov 21, 2006) with refined windows: L1
  November 2007 update for rest of S5: L1
  Flag: TRAIN_LIKELY