Multi-Plane Models

When the model contains multiple planes, it should be noted that the self-calibration tasks do not perform any interpolation of models along the spectral dimension. The self-calibration tasks assume that there is a one-to-one correspondence between a model plane and some average of the channels. The correspondence is determined in a rather imperfect fashion -- the self-calibration tasks use the standard line parameter (see Section 5.4). This parameter determines the channels selected from the visibility dataset, and the averaging performed on them. Unfortunately, the self-calibration tasks only perform rudimentary checks on whether the line parameter makes sense ( i.e. whether the selected channels have the same frequencies as the model planes). Despite this, in many instances, the self-calibration tasks will provide appropriate defaults for the line parameter. Below we outline those instances where you can or should allow the line parameter to default, and those instances where you should not.

• MIRIAD tasks normally maintain a record of the line parameter used to form an image. The imaging task, invert , records the line parameter it uses, and the deconvolution tasks propagate it to their outputs. The default line parameter used by the self-calibration tasks is that which was used to form the model. This is generally quite appropriate if the model was formed from the data being self-calibrated. However if the model was formed by some other means, then the default will be quite inappropriate, and you should set the line parameter explicitly.

• selfcal can cope with the case of several models being derived from different channel ranges from one visibility dataset. This might occur, for example, if multiple lines were being observed simultaneously (as often happens at millimeter wavelengths, but rarely with the ATCA). In this case, the different models were formed with different line parameters. selfcal must rely on the record of the line parameter stored in the model.

• If a point-source model is used ( i.e. no input model dataset), then you should give the line parameter explicitly.

• Similarly, if using the mfs option, you should also set the line parameter explicitly -- see the next subsection.