Self-calibration can improve the “dynamic range” of your images significantly but only if there is sufficiently good signal-to-noise in the uv database. It works by comparing the input uv data with the predicted visibilities from a model of the source; from this a set of complex gain (amplitude, phase) corrections are generated for each antenna in the array as a function of time. Before engaging in a potentially long and useless exercise, it is wise to look at the continuum or single-channel data set you have created to determine whether there is sufficient signal (with respect to the noise) to enable detailed editing and/or self-calibration. Even if the continuum data would profit from these things, you must also decide whether there is sufficient signal-to-noise in the line data to benefit from the improved editing and calibration. Consult the sections on editing (§4.3.5 and §4.3.13) and on self-calibration (§5.4) before deciding to continue with this section.
In fact, there is very little to be written here. The editing and self-calibration of the continuum or single-channel data set are precisely those described in Chapter 4 and Chapter 5. The goal is to create and fill an FG table (if flagging is needed) and an SN table (if self-calibration is needed) attached to the single-source, single-channel data set. Note that, for single-source data, we use an SN table containing the accumulated calibration while we use, for multi-source files, a number of SN tables with incremental calibrations and a single CL table with the net corrections. When all editing and self-calibration are done, you use TACOP (or TAPPE if the flag table must be appended to an existing flag table) to copy the two tables to the multi-channel data set and then use SPLIT to apply both tables to the multi-channel data (and to the single-channel data set too if desired).
In 31DEC22, task CLOSE may be used to plot the rms closure phase or amplitude as a function of channel. Use this on your calibration source or your target data after division by the continuum model. Values should be near zero but will probably rise at the edges of spectral windows (IFs). See §5.4.4.
In general, you may keep the flagging information in a flag table (specified with adverb FLAGVER) and add to and apply the table whenever it is needed. Be aware that flag tables are applied only by those tasks that have the FLAGVER adverb, but a number of tasks have been upgraded to understand calibration and flagging tables. In fact, data that are not in time order may have flag tables applied (if they are small enough) and time-independent calibrations (e.g., DOBAND=1 and DOPOL > 0) may also be done.