Since a major concern in calibrating cross hand data is that the right-left phase difference remain constant at the reference antenna, plot the phase difference of each dual polarization antenna from the resultant solution table. If the difference is not constant for all stations, check your solution intervals, delay and rate windows, or phase cals and try again. Sometimes, large phase differences are unavoidable, but jumps in phase difference should be as few as possible. The variation of R-L phase is a somewhat subtle problem. If the R-L phase at a station other than the reference is varying, this will manifest itself in the plot for that station, but no other. Global fringe fitting as done here will correct for this variation at the station, and the calibrated polarization data are usable. However, if the variation is at the reference station, all antennas will show the same fluctuations. In this case, a new reference station is required, and the fringe fitting must be run again. In short, if the fluctuations are systematic, choose a new reference station and perform the fringe fitting again. Otherwise, accept the solutions.
In some experiments, the reference antenna may not be present in all scans. One must control the sequence of antennas used as the reference antenna to ensure that only those with stable R-L phases are used. Starting with the 15APR97 release of , the adverb SEARCH is provided which does precisely this. Set APARM(9)=1 and fill in the SEARCH array with a sequence of antennas to be used as reference antenna, in order of preference. For earlier releases of , the ANTWT parameter may be used to control the sequence of reference antenna, with the higher weight antennas being preferred. The messages produced by FRING will report which antenna is being used as the reference.
The task SNSMO can be used to smooth and clip the derived delays and rates. This is especially important for VLBA data, where the first few visibilities on each baseline could be noise, resulting in a large rate which will be interpolated to adjacent points, giving large slopes of phase versus time where there hadn't been before fringe fitting. Smoothing should be done for each source separately because adjacent solutions on different sources could be very different.
To apply these solutions to the data, use the task CLCAL. The task CLCAL will interpolate between solutions to find values to put in each entry in the CL table. This could mean interpolating between solutions on different sources, which could have wildly different delays and rates. This will corrupt data at the beginning and end of scans. To avoid this, use INTERPOL='SELF', which forces to only use solutions on the source in question for the interpolation.
Once the solutions are properly smoothed, apply them to the data and plot the phase of the parallel hand data for all scans, making sure that the phases are nearly constant over moderate (say, a few minutes) time and across the band. Figure 3b shows the same data as before with the fringe fitting solution applied. The intent of fringe fitting is to make the phases vary slowly enough that the visibilities can be averaged in time and frequency to give a smaller, more manageable data set. The purpose is not to do detailed phase calibration. Soon the data will be averaged across each IF (not across the band until after the instrumental calibration has been applied) and for a short time (say, 10 seconds).