C.8 Rate and Phase Calibration

Now that the data have calibrated amplitudes, the next step is to do the calibration of the antenna rates and phases. This section describes that process.

1.  Correct the antenna parallactic angles, if desired, using VLBAPANG. The RCP and LCP feeds on alt-az antennas will rotate in position angle with respect to the source during the course of the observation (all VLBA and VLA antennas are alt-az). Since this rotation is a simple geometric effect, it can be corrected by adjusting the phases without looking at the data. You must do this correction for polarization experiments and phase referencing experiments. Parallactic angles are important for phase referencing because the parallactic angle difference between calibrator and target is different at different stations which leads to an extra phase error which can be corrected. VLBAPANG copies the highest numbered CL table with TACOP and then runs CLCOR (OPCODE = ’PANG’; CLCORPRM = 1,0). VLBAPANG has no inputs that require discussion. Be sure to correct the parallactic angles before any of the following steps. Again keep track of which CL tables add which correction.
2.  Now you must remove global frequency- and time-dependent phase errors using FRING or one of the procedures which use this task, VLBAFRNG or VLBAFRGP. This cannot be done simply for spectral-line sources, so the practice here is to determine delay and rate solutions from the (continuum) phase-reference sources and interpolate them over the spectral line observations. The procedures run FRING along with CLCAL. VLBAFRNG and VLBAFRGP use FRING, with VLBAFRGP specifically for phase referencing. For all these procedures, if the SOURCES adverb is set, then CLCAL is run once for each source in SOURCES. For the phase-referencing procedure (VLBAFRGP), any source that is in the SOURCES list that is not in the CALSOUR list will be phase referenced to the first source in the CALSOUR list. These procedures will produce new (highest numbered) SN and CL tables. Since it is probably best to run CLCAL on each source separately, SOURCES should always be set. To use VLBAFRGP for a simple phase referencing experiment (remember that CAL-PHASE is the phase reference calibrator), set CALSOUR=’CAL-PHASE’,’CAL-BAND’,’CAL-AMP’,’CAL-POL’,’STRONG’; GAINUSE=highest CL table; REFANT=n; SEARCH 9 4 1 3 5 6 7 8 10; SOLINT=coherence time; DPARM(7)=1 (if a polarization experiment); SOURCES=’CAL-PHASE’, ’CAL-BAND’, ’CAL-AMP’, ’CAL-POL’, ’STRONG’,’TARGET’; INTERPOL=’SIMP’. For this example, FRING will be run on the sources in CALSOUR and then CLCAL will be run 6 times, with all of the sources except TARGET referenced to themselves and TARGET referenced to CAL-PHASE, using interpolation method SIMP. For a non-phase-referencing experiment you would use VLBAFRNG with inputs the same as above except for SOURCES, which would not contain TARGET. The results will be the highest SN and CL tables. The INTERPOL to use is a personal preference. You might want to restrict the channel range slightly using BCHAN and ECHAN, since the channels at the high end of each IF will have lower SNR, due to the cutoffs in the bandpass filters. For a data set with 16 channels per IF, numbered from 1 to 16, setting ECHAN to 14 or 15 may be worth trying. Note that some people like to run CALIB rather than FRING or KRING for this stage of phase-referencing observations, but fringe fitting is recommended, as it solves for rates.

   The above fringe fit may take a bit of time, depending on the computer and the spectral resolution. Then, use SNPLT or VLBASNPL to inspect the solutions in the SN table. It’s not totally out of the question that some data will be found that need flagging, which can be done with UVFLG. In that case, it’s a good idea to delete the last SN and CL table and re-run VLBAFRGP or VLBAFRNG.

   This fringe-fitting stage is the most likely place where things can go wrong, for reasons that are not immediately apparent to the observer. Below, a few common examples are listed.

   •  Many solutions failed. The source may be too weak, or the coherence time too short. Try increasing or decreasing SOLINT. Or narrow the search window. For most VLBA data, DPARM(2) = 400 and DPARM(3) = 60 should be a good first step, though the rate window specified in DPARM(3) is proportional to the observing frequency, and may need to be larger at 22 GHz and above. For more options you could try running FRING and reduce the SNR threshold with APARM(7) or averaging the RR and LL (APARM(3)=1). One last thing to try is to abandon FRING and solve for the phases in CALIB, obviously not ideal since you will not get rates or delays, but it is a hit worth taking if the the calibration can be salvaged.
   •  Some antenna has low SNR, and may cause an entire set of solutions to go bad. This typically happens because an antenna should have been flagged. A common cause is when OV is looking at the White Mountains, and neither the on-line system nor the astronomer has flagged the data. Then, you need to run UVFLG and re-run VLBAFRGP or VLBAFRNG.
   •  There are discrepant delay/rate solutions. Look at the solutions you believe, and try VLBAFRGP or VLBAFRNG again with DPARM(2) and DPARM(3) specified appropriately. Full widths are specified, so if the good solutions fall between +15 mHz and -15 mHz, use DPARM(3) = 30. (Actually, you should use a value somewhat larger to allow some margin.) It may be that an antenna is suffering from radio-frequency interference, so some channels and/or IFs will need to be flagged.
   •  Some solutions are outside the specified delay/rate range. This can happen because the initial coarse fringe search uses the range specified by DPARM(2) and DPARM(3), but the least-squares solution can take off from there and go elsewhere.
   •  Delays and rates for some station change rapidly near the beginning or end of the observation. This may be caused by low elevation at the relevant station. Depending on how desperate you are to include low-SNR data, you may wish to flag some time range, or flag all data at elevations below 5^∘, 10^∘ or even 20^∘ (particularly at high frequencies or for phase referencing) with UVFLG.
   •  Phases wrap rapidly, particularly on the phase-reference source, CAL-PHASE. There may not be a lot you can do about this initially, because it’s possible that the tropospheric delay just changed too fast for the cycle time used in the observation, especially at low elevation. However, you may wish to note the times and antennas when the phase connection is best (typically the southwestern antennas near transit). Later, when imaging the program source, it can be helpful to image with a subset of antennas and time ranges, then use that initial image to self-calibrate the rest of the data.
3.  Use SNPLT or VLBASNPL to inspect the interpolation of the phases in the CL table. When you inspect the CL table notice any phase wraps that seem out of place. The human eye is better at pattern matching than a computer and these phases may be in error. You can use SNBLP to examine the phase solutions on a baseline basis (as they actually affect the visibilities). This may make problems more apparent. If so you might want to run CLCAL independently and try another interpolation method or you might want to edit the CL table. Remember that this is your last calibration table; you want to get rid of any bad calibration now before applying it to the data. Getting rid of spurious wraps in the final CL table (SNEDT) or flagging the data associated with fast changing phases (SNFLG) will improve your final image more consistently than anything else, particularly for phase referencing.