9.1 VLBI data calibration recipe

See Appendix C for simpler and shorter recipes suitable for straightforward observations.

• LOAD THE DATA For data from the VLBA correlator, run FITLD (§9.3.1.1); if needed, follow up with MSORT, USUBA, INDXR, VBGLU, VBMRG, and MERGECAL (§9.3.1.4–§9.3.1.7). For data from a MkIII correlator, see §O.2. Data from the Penticton correlator should be loaded using FITLD, sorted (MSORT, §9.3.1.4), and indexed (INDXR, §9.3.1.6).
• POLARIZATION: The combination of the VLBA correlator and FITLD incorrectly labels polarizations for dual parallel-hand correlation (RR and LL only), even if RR and LL are in different frequency bands (e.g., LL at 5 GHz and RR at 8.4 GHz). For these types of data, you must run FXPOL (§9.3.1.8).
• EXAMINE THE DATA It is important to familiarize yourself with the data set before proceeding further, especially if you have little experience with VLBI data. There are many  tasks for the examination of your data (see §9.4 for a fuller discussion). Minimally, you should at first run LISTR, IMHEAD, EDITR, POSSM, VPLOT, and PRTAN. At later stages you will probably find SNPLT, PRTAB, DTSUM, and SHOUV useful for examining data and calibration tables.
• SVLBI: Task OBPLT allows you to examine different aspects of the spacecraft orbit.
• PROCESS THE CALIBRATION FILES You will have either received calibration files, or instructions on where to obtain them. Some calibration files can be automatically processed into a form suitable for use within  using VLOG (§9.5.2). ANTAB is now the primary  task for loading calibration information from log files.
• VLBA CORRELATOR: The VLBA correlator will usually attach calibration information directly to your data for all VLBA and some other antennas. This obviates the need to run VLOG, ANTAB, PCLOD, and UVFLG to process your a priori calibration information for VLBA antennas. Some information for non-VLBA antennas must usually still be loaded from text files.
• POLARIZATION: Be careful to make sure that the polarization labeling of the IFs in the calibration text files is the same as the labeling in the data.
• CORRECT FOR THE IONOSPHERE For low frequency experiments (< 3GHz), TECOR should be run to remove at least part of the ionospheric contribution to the phase offsets. This should also be considered for higher frequencies (e.g., 8 GHz) for abnormal ionospheric conditions.
• CORRECT FOR THE EARTH ORIENTATION PARAMETERS For phase referencing experiments correlated at the VLBA correlator, particularly between 5-May-2003 and 2-August-2005, CLCOR (OPCODE=’EOPS’) should be run. This will correct the possibly inaccurate Earth Orientation Parameters used by the VLBA correlator. This is particularly important for astrometry experiments but can effect any phase referencing experiment including those correlated outside the above range of dates.
• EDIT THE DATA Identifying and editing bad data now can save you time later. Data should first be edited using UVFLG to apply editing information supplied with your calibration files (§9.5.3). Some useful tasks for examining and editing data are EDITR, UVFLG, TVFLG, SPFLG, EDITA, BPEDT, FLAGR, FINDR, VPLOT, and QUACK.
• POLARIZATION: You may want to edit the data consistently in all polarizations (select STOKES = ’IQUV’ within EDITR, TVFLG or SPFLG) — this can greatly simplify the imaging stage (see step 15).
• POLARIZATION: ADD PARALLACTIC ANGLE CORRECTION For alt-az mounted antennas, a parallactic angle correction for the rotating orientation of the antenna feeds with respect to the observed source must be performed as the first step in the phase calibration using CLCOR (§9.5.4.1). This step should be performed no later than immediately after the T_sys calibration.
• PHASE REFERENCING: You will want to perform the parallactic angle correction described above for phase referencing observations even if you only correlated the parallel hands (RR, LL).
• APPLY SAMPLER CORRECTION We now advocate a new amplitude calibration strategy based on VLBA Scientific Memo #37 (Walker 2015). This strategy interleaves the classic a priori calibration with instrumental delay and bandpass calibration to improve the calibration of data from the new Roach Digital Backend (RDBE) on the VLBA (see VLBA Observational Status Summary for a description of the RDBE and VLBA Scientific Memo #37 for a discussion of amplitude problems when using the RDBE). With data from before the RDBE you can use either the old or new strategy. First corrections for sampler biases should be applied using ACCOR (§9.5.4.2) for data from the VLBA and some other correlators. Some correlators apply this correction to the data before writing them out — notably the EVN JIVE correlator and correlators used for the Australian LBA. The VLBA hardware and software (DiFX) correlators do not apply this correction to the data. Therefore, ACCOR is required for the VLBA correlators and any others that do not apply the correction. ACCOR should be benign (do nothing) for those correlators that do apply the correction prior to reading the data into . Note that you can always run ACCOR and look at the SN table produced with PRTAB or SNPLT to see if it was benign or not.
• CALIBRATE THE INSTRUMENTAL DELAYS Phase-cals, or measured single-band and multi-band instrumental phase errors, should be applied using PCCOR (§9.5.4.3). You can manually perform a phase-cal by running FRING on a limited subset of your data to account for missing phase-cal information or to refine the reported phase-cal measurements (§9.5.4.4).
• SPECTRAL-LINE: Delay calibration should be carried out only on the continuum sources at this stage. Since there should be no pulse-cals, the “manual” phase-cal method should be used.
• POLARIZATION: If running FRING to determine the instrumental delays, be certain to solve for independent left- and right-polarization delay solutions APARM(3) = 0. Run RLDLY after calibrating the instrumental delays, to determine a single delay offset between left and right polarization (§9.5.7.11).
• CALIBRATE THE COMPLEX BANDPASS RESPONSE FUNCTION Run BPASS or CPASS to determine the bandpass response function using the cross-power spectra (§9.5.4.5). The normalization should be over the full bandwidth, be careful because the default channel selection is the inner 75% of the band. If you use the inner 75%, it can lead to amplitude errors of up to 15%.
• SPECTRAL-LINE: The bandpass response function should be determined using only the continuum calibrator sources.
• APPLY ONE MORE AUTOCORRELATION CORRECTION AND THEN FINISH A PRIORI CALIBRATION After determining the bandpass calibration the autocorrelations are probably different from unity by a few percent. To correct this run the task ACSCL, applying the previous calibration and bandpass correction. Then finally use APCAL to complete the a priori amplitude calibration (§9.5.4.6) — this is called the T_sys method of amplitude calibration. APCAL can also be used to perform opacity corrections.
• SPECTRAL-LINE: Unless the line emission is very weak, you may wish to defer amplitude calibration of your line sources only until step 16 below. The template method described there is much more accurate than the T_sys method.
• FRINGE FIT THE DATA Estimate and remove residual delays, rates and phases using FRING or BLING and CLCAL (§9.5.7.7–§9.5.7.8).
• SPECTRAL-LINE: Only fringe-fit the calibrator source at this stage. Check the coherence of the target source using the resulting solutions to decide whether or not to zero the rate solutions using the ’ZRAT’ option in SNCOR (§9.5.7.10). It is probably better to force the rates to be zero in FRING rather than zeroing them after the fitting.
• PHASE REFERENCING: You should not fringe-fit on the target, or phase-referenced source. Rather, you should fringe-fit on the cal, or phase-reference calibrator. When you apply the solution, be sure to set the CALSOUR and SOURCES adverbs in CLCAL appropriately to interpolate the solutions for the cal source onto the target source (see §9.5.1.2). If you are not interested in astrometric calibration and your target source is strong enough, you may wish to consider fringe-fitting on it to further refine the phase calibration (§9.5.7.4).
• POLARIZATION: ESTIMATE THE INSTRUMENTAL POLARIZATION . Correct for the instrumental polarization terms, commonly known as ‘D-terms’ using PCAL, LPCAL, or SPCAL on the polarization calibrator (§9.5.7.13). This polarization calibrator should first be fully calibrated and imaged before this step can be performed.
• POLARIZATION: CALIBRATE THE POLARIZATION POSITION ANGLE If a calibration source with known polarization orientation is available, use CLCOR to make a final correction to adjust the polarization angles of the target source data (§9.5.7.11).
• SPECTRAL-LINE or POLARIZATION: The bandpass response function should be determined using only the calibrator source. Unlike step 7, this step cannot be skipped.
• APPLY THE DOPPLER CORRECTION (spectral-line data only). Run CVEL to compensate for the changing Doppler shifts of the antennas with respect to the source during the observation and between the different observations (§9.5.5).
• SPECTRAL-LINE: REFINE THE AMPLITUDE CALIBRATION Run ACFIT to amplitude calibrate the program source using the template spectra method (§9.5.6). Note that the traditional T_sys method (§9.5.4.6) can also be used if the line emission is too weak for the template method to work successfully.
• SPECTRAL-LINE:DETERMINE RESIDUAL RATES Now estimate the residual rates ONLY by running FRING or BLING on one or a few spectral points on the target source (§9.5.7.10).
• APPLY CALIBRATION, AVERAGE, AND INSPECT THE FINAL DATA Run SPLIT or SPLAT to apply the calibration solutions and to average the data in frequency if appropriate (§9.6.1), and UVAVG to average the data in time (§9.6.2). You can also run SPLAT to combine these three operations into a single step. It is recommended that you take the time to inspect the calibrated data to see if more editing is needed, and to check that no gross calibration errors remain in the data (§9.6.3).
• SELF-CALIBRATE/IMAGE OR SELF-CALIBRATE/MODEL-FIT THE DATA The final complex gain corrections are determined by iterating self-calibration with imaging of the resultant data set. This is called hybrid-mapping. Alternatively, self-calibration can be iterated while fitting models directly to the data — the goal is to self-calibrate using the best model possible. The options are outlined in §9.7.
• SPECTRAL-LINE: One final distinction remains between continuum and spectral-line data. Only one or a few spectral points are used to determine final complex gain corrections which are then applied to all spectral points in the line data. After applying these gains, the line source data can be imaged to form an image cube.
• POLARIZATION: While the Stokes I and Stokes V images formed using the RR and LL visibilities will be real-valued, the Stokes Q and Stokes U images formed using LR and RL visibilities can, in principle, be complex-valued. You must use a fully complex imaging and deconvolution technique (see the HELP files for CXPOL, CXCLN, and CXIMAGR) or you can simply edit the LR and RL visibilities to enforce the condition that whenever you have a RL visibility on a baseline, you also have the LR visibility on the same baseline; this ensures that the Stokes Q and U images are real-valued and allows you to use the standard imaging tasks.