This section discusses some of the more complex areas of spectral-line processing. Some of the steps described here are pre-imaging, but some include imaging, and in those cases you should also refer to the imaging section (15) which follows.
The main differences between spectral-line imaging and continuum imaging are a) the amount of data you produce is much greater, b) you must remove any putative continuum emission, and c) you must try to keep track of velocities and frequencies more precisely.
Much of the processing discussed below can be done on either multi- or single-source files, and I discuss both options. However, it gets rather confusing as to what order things should be done in when there are many paths through the maze. I suggest that the clearest route following calibration and bandpass computation is generally SPLIT\ (applying editing, calibration and bandpass), UVLSF or UVLIN\ (visibility based continuum subtraction), CVEL (velocity alignment), DBCON (combine data), and then imaging. Bear this in mind when reading what follows.