The VLBA antennas generate too much astronomical data to be dealt with in
real-time (up to 512Mbits/second -- and NRAO is working on 1Gbits/second)
so all of the radio signals received from the observed sources are digitized
and recorded on magnetic tape at the antennas. The glass reels in the photograph
above are 15 inches in diameter, and hold 18,000 feet of tape. The tape is
about 16 microns thick. Your hair is about 100 microns in diameter. The "normal"
recording speed is 80 inches of tape per second, but astronomers, of course,
always want more.
The recording head records up to 32 tracks simultaneously, and can be moved back and forth across the one inch width of the tape so that 14 passes can be made. The 24 dark-colored objects along the wall back in the photograph of the control room are the playback drives where the data tapes from the VLBA telescopes, and radio telescopes around the world, are played back to the correlator. Each VLBA antenna has two tape recorders that look just like the playback drives in the photograph above. At the standard recording rate a tape will last about 10 hours, but can be filled in under three hours at high data rates. When the tapes are full the site technicians at the antenna sites package them up and ship them to the AOC. When all of the tapes for a particular experiment arrive at the AOC the project can be correlated. When an experiment is scheduled for correlation all of the tapes are mounted on the playback drives, synchronized, and then played back in unison just like they were when they were recording at the different locations. The correlator mathematically recombines all of the data from each antenna with the data from every other antenna in the experiment to simulate a single antenna up to 8000 miles in diameter -- the diameter of the Earth. Clever, huh?
The correlation of an experiment is controlled by a "job script." This script recreates the observation and controls everything from the configuration of the correlator circuitry to the positioning of the tapes. Above is a photograph of the correlator control console. This bank of computer screens is used for various aspects of correlation. The small monitor on the far left, the checker display, displays error and status messages regarding the system hardware, such as the state of the correlator and of the playback drives. Statements about job start and end time, warnings of tape drive vacuum loss and other system errors are also posted there. Next to the checker message display is the console display for a workstation called Reber, the heart of correlator operations. Reber requires two monitors because there is so much information to keep track of. The first one, with the black background, is where the operator interface runs. The interface has a number of options and a couple of dozen different windows, each showing a unique aspect of the system. Correlator jobs, the archiving of data, the positioning of tapes, and the operation of the playback drives are controlled from there. The next screen, of predominately white windows, is where the system software is loaded. There is also a running log of events and conditions during correlation, and a view of the array operator's logs that were written when the observations were actually performed. Other useful information, such as the job scripts, tape location, and maintenance status of the hardware, can be conveniently accessed from there. The boxes in the middle, and the computer on the far right are the distribution system which is explained further down the page.
The photograph above is the correlator. It is a special purpose supercomputer designed to perform the one special task of recombining all of the astronomical data fed to it by the playback drives. It is capable of performing about 36 billion floating point computations per second. Most of the work of the correlator takes place in specially designed VLSI chips that are very good at performing mathematical functions called fast Fourier transforms. The correlator is full of them.
Above is a photograph of the the final product. All of the antennas, the sensitive radio receivers, the complex control and monitoring software, the sophisticated digital hardware, the delicate reels of magnetic tape, and all of the manpower produce a Digital Audio Tape (DAT) about the size of a cassette tape that contains the correlated output from the correlator. Archive copies of all correlation jobs done at the Array Operations Center are kept. A copy of the data is made for the investigating astronomer. The computer on the far right in the console photograph above, named Gardenia, is used to control the production of the distribution tapes. The astronomers then take that data and feed it into another computer -- at the AOC or at their home institution -- capable of running the Astronomical Image Processing Software (AIPS) which, after a few days of crunching, will produce an image, or a "map", of the object that the telescopes were pointing towards when the observation was conducted.
DIRECT LINKS TO PAGES:
WHERE I USED TO WORK
THE ARRAY SIDE
THE CORRELATOR SIDE