C.7 Amplitude and Delay Calibration

Amplitude calibration uses measured antenna gains and system temperatures (T_sys), as well as finding a correction for voltage offsets in the samplers. Even though this is substantially still the case, we now recommend a somewhat different amplitude calibration procedure than in the past based on VLBA Scientific Memo #37 (Walker 2015).

Before amplitude calibration is done, there must be information for all antennas in the gain curve (GC), system temperature (TY), and weather (WX) tables (weather tables are needed for the opacity correction in APCAL). Missing T_sys and gain curve information can usually be obtained in ANTAB format and loaded with the  task ANTAB. §C.10 has information on including non-VLBA calibration information, if they are not already included in the data. §C.11 has details on how to incorporate the pre-EVLA VLA T_sys and gain curves. Otherwise consult §9.5.2.3.

1.  Correct sampler offsets and apply amplitude calibration by running VLBACCOR. The procedure VLBACCOR runs ACCOR, SNSMO, and CLCAL. ACCOR uses the autocorrelation to correct the sampler voltage offsets. After ACCOR creates an SN table, SNSMO smooths the table in order to remove any outlying points. Then the SN table is applied to the highest CL table using CLCAL (using INTERPOL=’2PT’), and a new CL table is created.
2.  Next, the instrumental delay residuals must be removed. These offsets or “instrumental single-band delays” are caused by the passage of the signal through the electronics of the VLBA baseband converters or MkIII/MkIV video converter units. There are two different methods to remove these instrumental delays, one for the case where you have pulse-cal information for some, but not necessarily all, of your antennas; and one for the case where you have no pulse-cal information at all. Note that the preferred method for continuum experiments is to use the pulse-cals, since they correct the instrumental delay over the whole experiment, rather than on a short scan. Spectral-line observers would have switched off the pulse-cals as they interfere with line observations, so they are forced to use the second (strong source) method. For VLBA continuum experiments before April 1999, you can load the pulse-cal data using PCLOD; consult the §9.5.2.
   •  For the case where you have some pulse-cal information, run VLBAPCOR. VLBAPCOR is another procedure which runs quite a few tasks, PCCOR, CLCAL, FRING (sometimes) and CLCAL again (sometimes). PCCOR extracts pulse-cal information from the PC table and creates an SN table. Then CLCAL is run to apply that SN table to the highest CL table, creating a new CL table. If there are antennas that do not have information in the PC table, or their PC entries are wrong, then VLBAPCOR can run FRING on a short calibrator scan (input TIMERANG). The SN table from FRING contains corrections for the antennas left out of the PC table, and is applied to CL table without corrections from PCCOR, and added to the CL table with the PC corrections. For the simplest case of all VLBA antennas, the inputs for VLBAPCOR should be TIMER=time range on CAL-BAND with good fringes for all baselines; REFANT=n; SUBARRAY=0; CALSOUR=’CAL-BAND’’; GAINUSE=0; OPCODE=’’; ANTENNAS=0. For the case where you have the VLA (in this example antenna 11), which does not have pulse-cals, your inputs should be the same as above except OPCODE=’CALP’; ANTENNAS=11,0. For the second case it is important that there are no “Failed” solutions from FRING; if there are failed solutions, then you should delete the tables that were created and find another TIMERANG with good fringes to the REFANT. Also see EXPLAIN VLBAPCOR for a detailed description of the steps involved with using pulse-cals and FRING without using VLBAPCOR.
   •  The alternate method is to solve for the phase cals “manually” with VLBAMPCL. This method uses the fringes on a strong source to compute the delay and phase residuals for each antenna and IF. VLBAMPCL runs FRING to find the corrections and then CLCAL to apply them. If there is no calibrator scan that includes all antennas then there is an option to run FRING and CLCAL again on another source and/or time range in order to correct the antenna(s) not corrected by the first scan. For the simplest case where all antennas have strong fringes to CAL-BAND, set TIMERANG = time range of scan on calibrator with strong fringes to all antennas; REFANT = n; CALSOUR = ’CAL-BAND’; GAINUSE = highest CL table; OPCODE = ’’. There must be no “Failed” solutions from FRING, if there are any failed solutions the data from that antenna/IF will be completely deleted from the data set.
   •  Now you must check the results of correcting your instrumental delays using VLBACRPL or POSSM. Set GAINUSE=highest CL table, and plot cross-correlations (VLBACRPL will do this for you). The plotted cross-correlations should show the phase slope removed from each IF and there should no longer be a phase jump between IFs, although the phase does not have to be at 0^∘. If you do a “manual” instrumental delay correction (i.e., you used FRING, not PCCOR); then the phases far in time from the scan on which FRING was performed may have a small slope and a small phase jump between the IFs. Also non-zero phase slopes still may be seen at low elevations, where the atmosphere causes additional delay residuals, or for low-frequency observations where the ionospheric delay varies. If you see significant phase slopes, or phase jumps between IFs on any baseline, then the instrumental phase corrections have not worked and you need to figure out why and start again.
3.  Next you must calibrate the bandpass shapes. To do this, run VLBABPSS on the bandpass calibrator, CAL-BAND. Make sure that the spectral line data for the bandpass calibrator are clean and devoid of high points, using UVPLT or SPFLG. Inputs for VLBABPSS are CALSOUR = ’CAL-BAND’ and a model for your calibrator if you have one. Then you must examine the BP table using POSSM by setting APARM(8)=2.
4.  Now it is time to finish the amplitude calibration by running VLBAAMP. This procedure runs several tasks: ACSCL, SNSMO, CLCAL, APCAL and CLCAL. After the previous steps the calibrated autocorrelation amplitudes will be offset from unity; ACSCL corrects this offset. ACSCL creates an SN which is then smoothed by SNSMO and CLCAL is run to apply the calibration to the next CL table. Then to finalize the amplitude calibration, APCAL is run on the highest TY and GC tables, and a new SN table is created. Adverb DOFIT controls whether APCAL also uses the weather tables to fit and correct for opacity. It is desirable to perform an atmospheric opacity correction at high frequencies, particularly if very accurate source fluxes are needed. See §9.5.4.6 for a more detailed discussion of APCAL. Lastly, VLBAAMP runs CLCAL to apply the amplitude calibration SN table to the CL created by the last run on CLCAL. VLBAAMP will print messages telling you about the new tables it has created. To keep track of your tables, it is important to copy these messages.
5.  At this point it is a very good idea to examine your calibration.
   •  Run the task SNPLT or procedure VLBASNPL (which is a very simplified SNPLT) to examine the tables created by ACCOR. Use INEXT=’CL’; OPTYPE= ’AMP’; INVERS=CL-table-with-sampler-offsets; DOTV=1 (to display to the TV; for a hardcopy use DOTV=-1 and LWPLA to print the plot files). The solutions that SNPLT plots should be close to 1000 milligain or 1 gain.
   •  Run SNPLT or VLBASNPL to examine the amplitude calibration. This time look at the SN table that APCAL created. Use INEXT=’SN’; OPTYPE= ’AMP’; INVERS=highest SN table; DOTV=1 for VLBASNPL; or to inspect IF m, use BIF = m; EIF = m; OPTYP = ’AMP’; INVER = 0; INEXT = ’SN’; OPCODE = ’ ’; NPLOT = 10; DOTV = 1; GO SNPLT. For a hard copy, use DOTV = -1; GO SNPLT; GO LWPLA. Plotted amplitudes are the square-roots of the system-equivalent flux densities (SEFDs), in Jansky, where the SEFD is the flux density of a source that would double the system temperature. (Low numbers are good!) At centimeter wavelengths, VLBA antennas have SEFDs near 300 Jy, so gains above 30^∘ elevation should be near 17–18 and should vary slowly and smoothly with time (i.e., change in elevation) for an individual source. To look at the input system temperatures, run SNPLT with OPTYP = ’TSYS’; INEXT = ’TY’; INVER = 0. On rare occasions, you might find clearly discrepant points that have leaked in from a different frequency band. In that case, you can use task SNEDT, or the clipping option of SNSMO, to get rid of the bad points. You may notice that at low elevations the gains on individual antennas are high. All data below a given elevation can be flagged by running UVFLG; e.g., to flag all data below 10^∘, run UVFLG with APARM(4)=0 and APARM(5) = 10. Note that FG tables are not applied to tables, so flagged data still may have points plotted by SNPLT. The T_sys measurements are also a very good diagnostic of bad data from poor weather, equipment failures, etc.. If there are time ranges of unusually high or low T_sys you may consider flagging those time ranges using UVFLG or EDITA. Be particularly suspicious of patches of unusual gains at only one IF or STOKES of an antenna. Remember, one of the best things you can do for your final result is to get rid of bad data.
   •  You may wish to use your favorite method of inspecting data for flagging (e.g., EDITR, SPFLG, TVFLG, IBLED). On-line flags are already included in FG 1 unless they were applied as the data were split into separate frequencies. For example, run EDITR with inputs SOURCES= ’CAL-BAND’,’’ (do each source separately); DOCAL=1; GAINUSE=highest CL table; FLAGVER=0; OUTFGVER=0; DOTWO=1; ANTUSE=1,2,3,4,5,6,7,8,9,10. Once you gain experience you might want to set CROWDED=2 ; DO3COL=1 which allows 3-color displays of all polarizations and IFs in one plot with a faster method for the TV displays; this can speed up editing significantly. Look for anomalously high or low amplitudes, remember there can be a slow change in amplitude with time due to source structure. Some people do no additional flagging at this stage, but later use the results of fringe-fitting and visibility plots of calibrated data to point the way to bad data.
6.  For spectral-line experiments needing velocity accuracy better than 1 km/s, a Doppler correction should be performed. Use CVEL; see §9.5.4.5 and §9.5.5 for details.
7.  This is a useful time to run TASAV to save all your ancillary tables to another file. If you foul up the calibration, the relevant tables can be copied back using TACOP.