| The LIGO
|Back to the Newsletter Front Page|
The Seventh Engineering Run - A Special ReportThe E7 Run - Taking the Interferometers Out for a Test Drive
In January, for the first time, LIGO operated all three laser interferometers at the Hanford and Livingston sites simultaneously during a 16-day engineering run. This was our seventh engineering run to date (dubbed with hip brevity "E7") and was the most extensive test conducted of the interferometers thus far. The run provided an opportunity to test interferometer performance, assess the reliability of the hardware and software during sustained operation, and to test the data handling and data analysis hardware, software, and methods. During the run, the ALLEGRO resonant bar detector at Louisiana State University and the GEO 600 laser interferometer near Hannover, Germany, were also in operation.
Below: The control room at the LIGO Hanford Observatory, 4:24 am.
LIGO engineering runs serve the same purpose for interferometers that test-drives serve in developing a new car. They confirm those aspects of the new equipment that work well, and those needing a bit of fine-tuning to optimize performance. Unlike previous runs, which focused mainly on characterizing each of the interferometers and improving their reliability, the E7 generated large amounts of data from sustained operation that will now be used to develop and exercise the LIGO data analysis pipeline. Of course, work is still underway to tune up the LIGO interferometers to their design sensitivities, and much remains to do. Still, the data recorded during E7 is vital in that it will provide LIGO Scientific Collaboration (LSC) scientists with real interferometer data needed for perfecting and tuning gravitational wave search algorithms.
Above: The control room at the LIGO Livingston Observatory, about 6:00 am.
Even though the instruments were operated well short of their design sensitivities, we are treating the data in the same way we will at full design sensitivity. To isolate potential gravitational wave signals from environmental or instrumental noise, we require that candidate signals must be observed in coincidence by multiple detectors. The joint analysis of data from LIGO's three interferometers, as well as from the GEO 600 interferometer and the ALLEGRO resonant-bar gravitational wave detector, will be a dress rehearsal for future observations using an international detector network. Particularly powerful can be the observation of astronomical events by more conventional astronomical telescopes and satellites in coincidence with gravitational wave detectors. Such observations will increase confidence in events detected and have the potential to revolutionize the way we understand energetic processes. Joint astronomical observations can provide, for example, new windows on compact and difficult to study astronomical objects such as stellar cores. To gain experience toward fully exploiting these possibilities, we received and recorded triggers from gamma ray burst observatories for comparison with the gravitational wave detector.
The LSC has formed four working groups to analyze the data from the E7 as well as from upcoming scientific runs slated for later this year. Each working group will search for a different source of gravity wave signals. The binary inspiral group will search for the distinctive gravitational wave signals emitted by two massive compact objects, such as black holes or neutron stars, as they spiral into one another. Inspirals have well-characterized waveforms. Meanwhile, the burst group will search for explosive events like supernovae or gamma-ray bursts. These events may well produce a short but intense gravitational wave emission that could be detected by LIGO even if the exact waveform cannot be reliably predicted. The third group will search for periodic sources from known pulsars or from as-yet unidentified periodic sources. Finally, the stochastic working group will search for randomly emitted gravitational waves generated in the early evolution of the universe.
The LIGO interferometer commissioning team, together with the LIGO data analysis system group, the LSC data analysis groups, and the LSC detector characterization group, all worked intently over much of last year to prepare the hardware, the monitoring, and the analysis software for this run.
The four kilometer long interferometer at the LIGO Livingston Observatory (LLO) was operated in a recombined, but not recycled, configuration. While we have demonstrated the operation of the LLO interferometer in a power recycled configuration, we are able to operate for longer time periods in the non-recycled mode. Seismic noise is significantly higher at Livingston than at Hanford, and use of the non-recycled mode reduces seismic sensitivity. In addition, a newly installed active compensation system was used to filter the large microseism present at LLO (see our Winter 2001 LIGO newsletter for details).
At Hanford, the two kilometer long interferometer was operated in a power-recycled configuration, while the site's second instrument, the four kilometer long interferometer, was operated in a recombined, but not recycled, configuration. A new system to compensate for the tidal distortion of the earth's crust by the sun and moon was used to allow for extended continuous locking of the instruments. Large levels of vibration were present at Hanford each weekday from approximately 7:00 am to 3:30 pm, due to a building construction approximately 700 feet from the corner station. Our machines were unavailable during these periods as expected. There were also some high-wind conditions and when wind speeds exceeded 25-30 miles per hour we were able to observe effects in our data. The four-kilometer interferometer used the new digital suspension controllers that will eventually be retrofitted into the other interferometers. Microseism compensation for the Hanford interferometers was not yet available.
Figure 3 below shows the calibrated strain sensitivity of the three LIGO interferometers at the time of the E7 run. Commissioning continues on all the interferometers and further improvements in noise performance are already observed.
The response of each interferometer was calibrated three separate times during the engineering run by injecting sine waves into the mirror actuators while sweeping the frequency.
The data acquisition system continuously sampled data on 6544 channels from the two interferometers at Hanford, and 1348 channels from the interferometer at Livingston. Although a gravitational wave signal will be seen on a single channel, many other channels are needed to supply information on the operation of interferometer components as well as its physical environment. Each channel was sampled at a rate between 16 Hz and 16 kHz, depending on the frequency range of interest for the given channel. The resulting data rate was 4.7 MB/s at Hanford, and 2.7 MB/s at Livingston. The acquired data were stored on local disk caches of 8.5 TB at Hanford, and 4.8 TB at Livingston, and were made available for online data analysis. The data were also copied to tape for archiving at the Caltech Center for Advanced Computing Research.
Monitoring programs ran continuously to assist the operators and scientists in keeping tabs on the current interferometer status and to record state transitions and noise transients for use in later analysis. Monitors run during the E7 included a summary of seismic activity, searches for transients in various channels, a lock statistics monitor, and two measurements of timing stability. Summary data and status reports calculated by the monitor programs were retrieved using local programs with graphical user interfaces and web-based status pages. The monitor programs were written primarily by members of the LSC Detector Characterization working group and used the Data Monitoring Tool infrastructure, which provided the monitors with the acquired data in real time via a dedicated link to the data acquisition system. Other control room tools made it possible to plot arbitrary data channels versus time or frequency.
Preliminary analysis of the data collected during the E7 run took place on site using the LIGO Data Analysis Systems (LDAS) at the observatories. The data analysis system at each observatory consists of a complement of four symmetric multiprocessor (SMP) servers and a set of 16 Linux PCs forming a so-called "Beowulf" cluster. Each node in the cluster houses 512 megabytes of memory and a CPU capable of over one billion floating point operations per second. Similar systems exist at Caltech and MIT. Information about in-lock interferometer data segments, and triggers from the data-monitoring tool, were ingested into the on-site relational databases (IBM DB2). These database servers stand at the center of data analysis activities. Astrophysical and detector characterization search codes polled the database to identify interesting data segments prior to launching analysis tasks on the Beowulf cluster. In total, close to 114,000 individual jobs were processed--more than 95 percent completed successfully without generating an error condition. Three quarters of the jobs were the performing of astrophysical and detector characterization tasks. The remaining quarter were preparing the database with necessary trigger and data segment information for use in the subsequent searches.
The E7 run, like previous runs, was an occasion for large numbers of LSC scientists to participate actively in the operation of the interferometers and to perform other scientific activities at the observatories. One or two operators and two LSC scientists typically manned each control room. While the operators adeptly maintained the interferometers locked and running, the scientists checked the quality of the resulting data using the on-line monitors and standard control room tools.
A representative plot of single interferometer lock performance during the E7 run is shown in Figure 4 below.
In this plot, the length of each bar represents the duration of a lock segment, while its position marks the start time of the segment. In spite of the occasional long lock, the majority of the lock segments were shorter than 40 minutes. The longest lock stretch of more than seven hours was recorded with the Hanford 2-Km detector. Very short lock segments are not particularly useful for most of the test data analysis. The consensus of the analysis groups is that locks of 15 minutes or longer are adequate. We collected 140 hours (~34%) of triple-coincidence lock segments, out of which 71 hours (~18%) represent segments longer than 15 minutes. Since this was our first attempt at a triple-coincidence run we are rather pleased with this performance. The earth tide and microseism compensators performed well in their first extended tests. And when these compensators and other stability enhancements, like wavefront stabilization, are fully developed on all interferometers, we expect to quadruple the triple-coincidence duty cycle.
Meanwhile, commissioning of all three interferometers continues around the clock. The interferometer sensitivities described in Figure 3 have already been significantly improved upon by the implementation of additional features of the interferometer control system, such as the common mode servo system and further tuning of the optical alignment and electronic gains in the control electronics. As our first real test-drive of all three interferometers together, the success of the E7 was not measured by any race to a finish line but by the number of laps driven, and the rich data and experience furnished during the course. Fueled by this knowledge, we now return to our labs and "garages" to tune up for the next big drive, a LIGO milestone--the first scientific data taking run, scheduled later this year.
|Onwards!||to LIGO Hanford||to LIGO Livingston||to LIGO MIT||to LIGO Caltech|