AIPS report, 6 November 1996 ============================ software -------- A large scale clean-up of unused or oversized buffers in February and March had caused several tasks to fail. After a prolonged effort, now all tasks affected are back in working order, with the additional advantage of smaller memory requirements. Most of the effort went into further debugging of VLBA and VSOP related software, and into adding new functionalities to these tasks. A few illustrative examples: 1 - VPLOT. The possibility of specifying the subarray was added; SUBARRAY = 0 defaults to plotting all subarrays. VPLOT can plot a model based on a configuration with one or several VLBI antennas in orbit. It now also can plot a model based on many fields (up to 16), e.g. resulting from a multi-field clean. 2 - PCCOR. It now supports multiple subarrays, with SUBARRAY=0 defaulting to all subarrays. The new adverb CUTOFF allows to switch off the cable delay, which is useful for geodesy applications. 3 - APCAL. GAIN (GC) table buffers were increased to 16384 to accommodate Mark-3 datasets with 28 IF channels, null characters were stripped from antenna names to increase the robustness of the calls to the KEYIN routines, and handling of source selection of the type "-SOURCE" was improved. 4 - CALIB. An error in the way CALIB dealt with dual polarization data was corrected. It sometimes occurred when one polarization was missing throughout a solution interval. In the fit, polarization-independent weights were used, which at times led to very high gains in the missing sense of polarization. This in turn biased the mean gain modulus (MGM), if computed. Errors in the MGM were cumulative over several iterations. 5 - BLING It is now possible to divide a model into the data before searching fringes and can use the Schwab-Cotton FFT algorithm to add in data from indirect baselines for greater sensitivity. Off-center window handling has been changed allowing greater padding to be used in the FFTs: this has removed the need to refine fringe positions using non-linear least squares. This has also greatly decreased the chances of obtaining wild solutions. The Mark-3 mode search was modified to look for multi-band delay and rate before single-band delay, which should improve BLING's chances of finding some fringes in this search mode. 6 - BLAPP Added missing tests for bad BLING solutions in the input. Without these tests divide-by-zero errors were almost guaranteed since the acceleration error is set to zero when acceleration is not solved for. This was not discovered earlier since BLAPP worked fine under the Solaris operating system. Another modification deals with the misaligned time-stamps for different baselines that will arise if different solution intervals are used on different baselines. 7 - SNCOR Two new features were added: 1) OPCODE='PCOP' allows solutions to be copied from one polarization to another with the copy direction controlled by SNCORPRM(1); 2) OPCODE='PNEG' will flip the sign of the gain phase for all selected solutions. Both features are sometimes required in line polarization calibration. Also added the capability of SNCOR to work on single source files. 8 - USUBA A major re-write of this task was prompted primarily to accommodate the new subarray features allowed by the VLBA correlator. The task now works in three possible modes: i) automatic subarray identification; in this case an algorithm is used which minimizes the total number of subarrays and maximizes subarray continuity; ii) multiple subarray definition in an external text file; and iii) selection of an individual subarray through the input adverbs. Work continued on the VLBA DDT. This is a sequence of tasks to mimic a typical VLBA data reduction session. The purpose of this is twofold: 1) it compares the outcome with the results of an earlier - standard - run, and as such measures the effects caused by intermediate software changes. 2) it serves as a benchmark, useful to test the actual AIPS speed of workstations during a typical VLBA session. This is a welcome addition to the existing VLA DDT. The error analysis in the image fitting routines was completely overhauled; these tasks now yield a far better error estimate. A new task was written that reads GPS ionospheric data from an ASCII file and writes it to the newly created table type 'GP' attached to a uv data file. This table is used by the other new task GPSDL; it fits a simple, local model of the ionosphere to GPS data loaded by LDGPS and uses this model to correct phases for the excess path in the ionosphere and to calculate ionospheric Faraday rotation. The experimental pseudoverb XHELP was added to AIPS. It provides direct access to HTML versions of cookbook-level help. When invoked, it will automatically start a Netscape process (if not already running) and loads the specified HTML file. personnel --------- In August, 1996, David Adler left the AOC AIPS group. This means that the AIPS group has lost two FTE's alone in the current year, including Eric Greisen's - hopefully temporary - departure from the AIPS group earlier this year. The 5 FTE's that are left share the responsibility for i) AIPS local and global system support, including adapting AIPS for new architectures and more recent versions of operating systems, ii) AIPS application software user support, iii) management of AIPS releases and shipping requests, iv) development of new software, especially to support new instruments (VSOP). Bug-fixing has not received quite the attention that it should have, and if AIPS is to maintain its good level of robustness of the past years, the AIPS group would need at least one of the two lost positions back.