C.6 Data Loading and Inspection
- Load the data using VLBALOAD (which is a very simplified FITLD). CLINT should be set so that
there are several CL table entries for each self-calibration or fringe-fitting interval anticipated; this
will minimize interpolation error during the calibration process. However, setting CLINT too short
will result in a needlessly large table. Somewhere between CLINT = 0.25 and CLINT = 1.0 is about
right. A FITLD parameter that is set automatically in VLBALOAD is WTTHRESH = 0.7, which results in
irrevocable discarding of all data with playback weight less than 0.7. The only way around this is to
use FITLD explicitly.
- Correct data with VLBAFIX. If necessary, VLBAFIX sorts (with MSORT), splits into different frequencies
(with UVCOP), fixes the polarization structure (with FXPOL), and indexes (with INDXR) the data. VLBAFIX
will also correct for subarrays (with USUBA), but you must tell it to do so. There are only 2 inputs
of interest in VLBAFIX, CLINT, the CL table interval, and SUBARRAY which should be set to 1 if there
are subarrays and 0 if not. This is a very benign procedure, it can be run on every data set read
into and will only perform the necessary fixes. Note that, if the data are split into different
frequencies, the flag table is applied and deleted.
- At this point it is a good idea to get a listing of the antennas and scans in your data by running
VLBASUMM. VLBASUMM runs PRTAN over all antenna tables and LISTR with OPTYPE=’SCAN’ and gives a
choice of writing a text file to disk or sending the listing to a printer.
- Apply corrections to the Earth Orientation Parameters (EOPs). VLBI correlators must use
measurements of the Earth Orientation Parameters (EOPs) to take them out of the observations.
These change slowly with time and therefore the EOPs used by the correlator must be continually
updated, and they are generally best two weeks or more after the observation. Since we try to correlate
observations as quickly as possible, it is likely that the EOPs used are not the most accurate. Therefore
it is recommended that all phase-referencing experiments be corrected for this possible error. The
procedureVLBAEOPS will do this correction. VLBAEOPS automatically downloads a file with correct
EOPs and runs CLCOR to correct the EOPs.
- Apply ionospheric corrections, if desired, with or VLBATECR. This procedure automatically
downloads Global Positioning System (GPS) models of the electron content in the ionosphere to and
uses them to correct the dispersive delays caused by the ionosphere. It is particularly important for
phase referencing experiments at low frequency. We recommend VLBATECR for all experiments at
8 GHz or lower. VLBATCOR is only as good as the ionospheric model, so it is a very good idea to
compare the corrected and uncorrected phases using VPLOT. To inspect the phases using VPLOT, use
options BPARM = 0, 2; APARM=0; DOCAL=1; GAINUSE=highest CL table. The phases should not wind
as much (although they will probably not be completely flattened), when the corrected CL table is
applied. To see the corrections themselves, use SNPLT on the new CL table setting OPTYPE = ’DDLY’.
- For a simple spectral-line data set, or any data set with high spectral resolution, it is a very good
idea to average the data set smaller number of channels before deriving the calibration parameters.
Otherwise, the calibration tasks may take forever to run. It is recommended that you quickly inspect
the channels of interest for your line data (e.g., with UVPLT) for high points. Remove obviously high
amplitudes with CLIP (or e.g., UVMLN) before averaging. Inspect the full resolution data also for high
delays and fringe rates. Spectral averaging in such cases may not be acceptable. Continue calibration
on the averaged data set as if it were a continuum set. There is a better method to calibrate spectral
line data described in §9.5.6 and §18.104.22.168, but the one used here is simpler and will usually give
acceptable results. To reduce the data-set size, run the task AVSPC with AVOPTION = ’SUBS’. For
example, to average IFs with Nchan down to 32 channels, set the adverb CHANNEL = Nchan∕32 (e.g., to
average from 2048 to 32 channels, use CHANNEL = 64).