Correcting Data - The Quick Recipe

The steps required for correction of EOP data using the new EOPS option in CLCOR in AIPS are listed here with minimal explanation. If confused about something, try reading the following sections. Note that the EOP errors are almost never large enough to cause significant loss of coherence. The fringe rate and delay offsets are very small compared to any windows that are used for VLBA correlation. To make corrections, CLCOR in a version of AIPS updated since 2005 September 29 is required.

To correct for EOP errors, CLCOR obtains the EOP used for correlation from the AIPS CT table, gets new, presumably better, EOP values from an external file provided by the user, calculates total delays using both sets of EOP and using a simple model that is otherwise the same for both calculations, takes the difference, and derives CL table corrections from that difference. The EOP values are provided for UT midnight each day and are interpolated to the time of each CL table entry. CLCOR must accurately reproduce the correlator's interpolation algorithm which it does for the VLBA. The CLCOR EOP correction has not been designed with other correlators in mind nor has it been tested with data from other correlators.

The steps to correct data are:

• Download http://gemini.gsfc.nasa.gov/solve_save/usno_finals.erp to get good EOP values adjusted slightly from the IERS values to work best with the source and station catalogs used in VLBI.
• Choose the UV file to correct. Most multi-source files from FITLD will be OK. In order of preference:
  • A UV data file loaded with a version of FITLD more recent than mid-September 2005. It will have a new style CT table that includes a column with the time range over which that row of EOP data was used in correlation.
  • Any file whose CT table consists of multiple sets of identical rows. This will be fine, even if loaded with the older FITLD. CLCOR will use the first 5 rows, which is what the correlator did. The CT table can be examined with PRTAB. Note that the boundary between row sets is distinguished by a drop in the day of the EOP data.
  • A UV file loaded with the older FITLD and having CT table row sets that vary, usually by gaining or losing a day. For this case, CLCOR needs to relate CT rows to time ranges. It does this by reading the history records written by FITLD as it loaded the distribution file from each job.
  • A UV file with an old style CT table, that does not have times associated with each row, and that has been through DBCON, VBGLU, UVCOP or other such tasks that combine, or pare down, data sets, should probably be avoided on the grounds that the attempt to assign times to CT rows might not work properly.
• Choose the CL table to correct. It is best to select one that does not contain the results of any self-calibration-like steps. These include pulse cal corrections (PCCOR), fringe fitting (FRING), and actual self-calibration (CALIB). It can have correlator corrections (ACCOR), gain and Tsys calibration (APCAL), parallactic angle correction and other geometry changes (CLCOR), and ionospheric calibration (TECOR) already applied since they are not based on the data phases. Bandpass correction is also best done after the EOP corrections. If a CL table that contains any self-calibration results is used, that type of processing should be repeated after the EOP correction. For a fresh data set, the EOP corrections are best made any time before the pulse cal correction (PCCOR) or first fringe fit.
• Run CLCOR to correct for the EOP errors. Use OPCODE='EOPS' and the GAINVER (CL table version) chosen above. Point INFILE to the usno_finals.erp file that was downloaded. Be careful to process all frequency ID's -- each has to be done in a separate pass of CLCOR, setting SELBAND and SELFREQ as required (it apparently is not possible to do all in one pass). An alternative is to separate the different FQID's using UVCOP before running CLCOR. Note that, if multiple runs of CLCOR with data selection are required, using GAINVER=GAINUSE after the first pass can incrementally modify the output file so that you have one final output CL table that applies to all data.
• Then redo all of the self-calibration type calibrations. Final phases for self-calibrated (including fringe fitted) sources should be the same as without CLCOR so images, bandpasses etc based on such sources do not absolutely need to be remade. The changes will be in the phases of phase referenced sources.

A script called VLBAEOPS is being added to AIPS. It optionally downloads usno_finals.erp and then runs CLCOR. It is one of the scripts loaded with RUN VLBAUTILS. It can use an input file already on a local disk.

Note that one way to check how much of a problem the data set has is to run CLCOR using CL version 1 as input and checking the delay and phase gains generated to see if they are large enough to be a concern. Remember that for phase referencing, the real concern is the double difference between stations and between target and calibrator.

To determine the quality of the EOP used on any particular project, see references in Section 4, especially the tables by date and project available from anonymous ftp at ftp.aoc.nrao.edu in directory pub/staff/cwalker/VLBA.

The rest of the memo gives a lot of detail for those interested.