Sorting and Indexing

            

1. Hopefully, you used ATLOD so that it loaded your data in time (and possibly baseline order). The time order is essential, but not the baseline order. Use IMHEAD and see if the visibility file header shows TB order. If not, then you must sort the data. Use the task UVSRT and set the adverb sort='tb'. The input and output file names can be the same in UVSRT if you are running short of disk space. But if the machine crashes part way through, you will have to reload your data with ATLOD. Note that UVSRT creates two scratch files as well as the output file, so that you need plenty of space anyway.

   Make sure that you don't allow UVSRT to access any network mounted disks (see § 2.3.2) by setting baddisk (control of scratch files) and outdisk (control of disk for sorted data base).

   sort='tb' Sort into time-baseline order

   outdisk Use these to avoid putting output

   baddisk on NFS mounted disk

        

2. If you needed to run UVSRT you must also run INDXR to create the initial index (NX) and calibration (CL) tables (ATLOD usually creates these for you). The defaults for INDXR with regard to the NX table should be adequate, but make sure of this. For the CL table, you need to specify the interval at which entries are to be written. The default of 5 minutes is generally adequate (i.e., significant gain variations don't occur much faster than this). More detail on CL tables will be provided later. This is all you need to know for now.

cparm=0 Most of the defaults OK

cparm(3)=3 Gain table entries every 3 mins