AIPS quarterly report, 1 April 1997 =================================== system ------ The first quarter of 1997 saw lots of activity related to the VSOP launch in February. All major elements needed to support space based VLBI were added to AIPS. We expect the first scientific data to be available in early April. Since we are reluctant to release a version of AIPS that has not undergone at least one pass of real data, we are seriously contemplating delaying the release date by one month. This will give us the time to correct some of the errors that without doubt will become evident once we use AIPS on true VSOP data. Institutes that need new AIPS before that later date already have the latest version of AIPS available through the 'midnight job'. A total of six non-NRAO sites is running the AIPS midnight job. Another is currently inactive, and two more sites have expressed interest. Midnight job sites are located in the US, Japan, and Europe. The AIPS DDT was used to test new SGI workstations considered for purchase by the VSOP project. The four-processor SGI 'Origin 200' set a new record at 13.7 for a single processor, or 12.0 per processor when all four processors were used in parallel. This enormous power is required for the computationally demanding steps of fringe-fitting and imaging space VLBI data. At the AOC, we installed a PC with the Pentium Pro 200 processor to test AIPS under Linux for an extended period for day-to-day data reduction. A serious problem in the tape accessing routines that did not allow Exabyte drives to be used under Linux, was fixed. As described in more detail below, we started porting several selected pieces of code from CVX (the Charlottesville based experimental version of AIPS) to classic AIPS. For this particular release, crucial for space VLBI processing, we deliberately restricted ourselves to code that is considered very useful and would not interfere too much with the integrity of AIPS. We plan to port more pieces of code after the upcoming release. software -------- 1) The key task FRING was further improved by adding the adverb SEARCH, which specifies alternative reference antennas in the initial FFT search. We added the 'exhaustive fringe searching' and 'subset solve' options. The latter option was implemented in CALIB as well. 2) Some of the more fundamental code was changed that cause uv scratch files to be much smaller than before when baseline selection is used. Especially for large space VLBI datasets this is a major improvement. 3) BLING now uses quadratic interpolation in delay, rate, and acceleration, which speeds up the task considerably. 4) A beginning was made porting some of the features found in the experimental version of AIPS running in Charlottesville (CVX). New tasks EDITA and EDITR were introduced. EDITA edits uv data based on calibration tables, and EDITR allows interactive editing - based on a secondary dataset - of a number of baselines. In the existing task SCMAP, the 'Edit Data' menu command from CVX was enabled in AIPS. 5) The VLBA correlator can change correlation modes with great flexibility. This can lead lead to time-variable rate and delay decorrelation corrections which depend on the type of frequency and time filtering performed in the correlator. In order to solve this problem we implemented a new correlation_id random parameter for VLBA datasets. This parameter points to the recorded correlation modes stored in the existing CQ table. A new form of decorrelation corrections using this new random parameter was implemented. 6) The new orbit table (OB) was introduced, and a new task was written to fill this table with additional information. The FITS reading task FITLD was modified to support both the new OB table and the new random parameter. 7) A new uv sorting task MSORT was written. It reproduces the functionality of UVSRT. MSORT does an in-memory sort of a UV-data file. At best, MSORT should be about 3 times faster than UVSRT (in case when the data is only weakly mis-sorted - as is the case for normal VLA and VLBA data). At worst MSORT should take no more than 50% longer. In either case, MSORT requires no ancillary disk space - making this the preferred sorting method for large data files. 8) Another new task is RESEQ, which will renumber antennas in a UV file. This is the first step for the space VLBI requirement that stations may need to be aliased together at some point in the data reduction stream. 9) Numerous smaller improvements were made and bugs were fixed. Many of these improvements were prompted by the 'designated AIP' program that continues to be popular with users around the world.