Input Parameters to the Self-Calibration Tasks

We now describe the input parameters to the self-calibration tasks. Given the similarity of the two tasks, we will describe them together, noting any differences as we go.

• vis: This gives the name of the input dataset to be self-calibrated.
• model: This gives the names of the input models. Several models can be given. Task selfcal can take models from different pointings, frequency ranges and polarisations. Task gpscal can take models from different polarisations only.

• interval: This gives the self-calibration solution interval time, in minutes. This time must be long enough to integrate an appreciable signal-to-noise ratio (5 or more) on visibilities. On the other hand, it should be short enough so that the gains can be approximated as being constant during this interval. The default is 5 minutes.

• select: This selects the u-v data to be used in the self-calibration process. For selfcal , you need not perform source, dra, or ddec selection if you use the selradec option (see below).
• minants: The minimum number of antenna that must be present before the tasks will attempt to find a solution. The default is 3 for phase self-calibration, or 4 for amplitude self-calibration.
• refant: The reference antenna ( i.e. the antenna which has zero phase) to be used in the solution process. If this is not specified, selfcal chooses a default for each solution. gpscal , however, uses antenna 3 as the default reference antenna.
• line: The standard line parameter, as described in Section 5.4. This parameter is described in some detail above.
• clip: This parameter allows you to set the limit of the faintest believable flux density in the model. For Stokes I or parallel-hand correlations, this is a maximum limit -- flux densities below this limit are assumed to be noise or error. For Stokes Q, U and V, this is an absolute limit -- flux densities less that are assumed to be noise or error.
• offset: The self-calibration tasks can be run assuming a point-source model. A point-source model will be assumed if no model parameter is given. In this case, offset gives the offset, in arcsec, from the observation centre, of the point source. The default is to assume that the point source is at the observation centre. Note that you should generally also give the line parameter when self-calibrating multi-channel data with a point-source model.

• flux: For selfcal , when using a point-source model, this gives the Stokes I flux density. The default is 1. For gpscal , when using a point-source model, this parameter gives the flux density of I, Q, U and V. The default is 1,0,0,0. Also see the noscale option below.
• options: The options parameter, as usual, gives a number of extra processing options. Options common to selfcal and gpscal are:

  amplitude

  and phase: This determines whether the self-calibration tasks solve for just phase ( options=phase, which is the default) or amplitude and phase solution ( options=amplitude).

  mfs:

  This option is described in the section on frequency handling (see above).

  noscale:

  Normally the self-calibration tasks scale the gains so that their rms amplitude is 1. This preserves your primary flux density calibration. The noscale option prevents this scaling, thus allowing the absolute flux density scale to be changed by the self-calibration process. You will want to allow the flux density scale to change if you are self-calibrating several datasets of the same source (i.e. using the same model) separately. For example, you may have several datasets containing different configurations or frequencies of the one source, and yet all these datasets are used to produce the one output image. Using noscale ensures that the flux density scales resulting from the different self-calibrations are the same.

  Options specific to selfcal are:

  polarized:

  This option is described in the section on polarisation handling (see above).

  smooth:

  The smooth option constrains the solution, at one time, to be related to the solutions at adjacent times. This results in a somewhat smoother variation of gains with time. This option may be helpful when the signal-to-noise ratio is poor.

  relax:

  This relaxes the convergence criteria of gain solutions somewhat, and may be helpful if there are a large number of solutions which fail to converge. This option will be useful when the model is particularly poor. This will be so in the early self-calibration iterations when the primary calibration was quite poor.

  apriori:

  The source is one of a handful of known sources. selfcal uses its knowledge of this source to determine the model. Unfortunately the known sources (northern hemisphere sources at millimeter wavelengths) are not likely to be thosed observed with the ATCA.

  selradec:

  This causes selfcal tp select only those visibilities whose observing centre is within plus or minus three pixels of the model reference pixel. This is needed if the input visibility dataset contains multiple pointings or sources. By default, no observing centre selection is performed.

  Task gpscal takes a few options to dictate how it handles XY phases:

  noxy:

  Do not attempt to solve for XY phase. By default gpscal solves for a constant XY phase on all antennas except the reference antenna.

  xyref:

  Solve for the XY phase of the reference antenna (as well as the XY phase of the other antennas). This cannot be used together with the noxy option. To use this option, the source should be strongly polarised (at least 5%), and Stokes Q and/or Stokes U models should be given.

  xyvary:

  Solve for a variable XY phase. It is strongly recommended that you use this option. If xyref is given, then gpscal additionally solves for a variable XY phase on the reference antenna.