Deconvolution is no different to conventional observing. When restoring, you may wish to use the same restoring beam size for all fields. Otherwise treat each pointing as a separate observation.
Much of the discussion in
the self-calibration section of the `joint approach' applies equally
well here. The major difference is, of course, that because the
deconvolution step has been performed on the individual fields, the
demos
step is not required. If you have split the visibility
data into separate single-pointing datasets, you will not need to
use options=mosaic (although it will not hurt). Also for single-pointing
datasets, if you wish to use
just one or a few of the stronger fields for self-calibration, you
will need to copy the resultant antenna gain tables across to the
other datasets, using gpcopy
.