WSCleanAndSelfcal

WSClean can be used to perform self-cal. This can especially be useful when e.g. combined with the ability to do (almost-)full-sky imaging, since in some situations MWA's field of view might require including highly off-axis sources in the calibration model.

To perform self-cal, you need to use major iterations, as these will fill/update the MODEL_DATA column. When using Casa's tasks for calibration, the calibration tasks will use the MODEL_DATA and calibrate the CORRECTED_DATA using this column. Other calibration utilities, like MWA's mitchcal or LOFAR's DPPP can similarly be instructed to calibrate using the MODEL_DATA column.

Filling the MODEL_DATA requires a setting of mgain < 1, e.g. an mgain of 0.9. As long as mgain is not 1, WSClean will end with a major iteration, and the MODEL_DATA column will be set to the "best" model from the cleaning model.

You can self-cal on Stokes I or on multiple polarizations, where the latter is more accurate (at least in the case of the MWA), but takes more time. If you run WSClean on the desired polarizations one by one, e.g. on XX and then on YY, or joinedly clean them, the MODEL_DATA column will have all imaged polarizations correctly filled in. The first run will create the MODEL_DATA column, set all values to zero and then fill the XX column, the second run will notice the MODEL_DATA already exists, and only update the YY column. B. McKinley has used this method and did a few self-cal loops to create very deep and well-calibrated Fornax A images. You can run all polarizations at once with -pol xx,xy,yx,yy.

Certain CASA commands (e.g. ft) will put keywords in a measurement set that turn on the "on-the-fly" (otf) mode. In OTF mode, CASA will ignore the MODEL_DATA column and use other keywords to determine the model data. To make use of the MODEL_DATA after, you can use the delmod CASA command to disable OTF mode:

delmod(vis='myobs.ms',otf=True,scr=False)

WSClean will never use or change OTF keywords in the measurement set.

Since WSClean version 1.2, WSClean understands the '-predict' option. This will fill the MODEL_DATA column with a prediction from an existing image. It's also possible to image the model by running a normal WSClean run and specify '-datacolumn MODEL_DATA'.

Next chapter: Image weighting