Next: Frequency Handling
Up: Computation
Previous: Self-Calibrating with Multiple
Models to the self-calibration tasks should generally be Stokes parameters
(I, Q, U or V).
The biggest difference between selfcal
and gpscal
is in the
handling of polarisation
and dual feeds.
Though selfcal
can handle multiple pointings and frequency ranges,
it is quite simplistic in its approach to polarisations.
- selfcal
assumes a single
gain for each antenna. If the antenna has dual feeds, they are assumed to
have equal gains.
This may be a reasonable assumption if phase errors are the dominant
ones, and if the XY
phase is constant and has been well determined.
It is somewhat less likely that the amplitude gain errors will be closely
matched between the feeds.
- Generally selfcal
is unable to convert between raw polarisations
and Stokes parameters. This means that the visibility dataset must contain
a correlation which matches the Stokes or polarisation type of each model.
For example, if selfcal
is given a Stokes Q model, then the dataset
being calibrated must contain Stokes Q visibilities.
- selfcal
generally assumes that a source is unpolarised. In this
case
it treats parallel-hand polarisation parameters (XX, YY, RR and LL)
as if they were measures of Stokes I. It will also refuse to
accept models of Stokes Q, U and V. Giving options=polarized will
allow selfcal
to accept Stokes Q, U and V models, but it will no longer
assume that parallel-hand correlations are measures of Stokes I.
- selfcal
will accept multiple input models which represent the
data at different polarisations, different pointing centres and different
frequency or velocity ranges. As the gains are assumed to be independent
of these, all the information is used in a simultaneous solution.
Task gpscal
is intended for telescopes with dual linear feeds,
such as the ATCA. Indeed, it will fail for other sorts of data.
- gpscal
solves for the two feeds either totally independently, or
with the constraint that the XY
phase of an antenna must be constant with
time.
- gpscal
converts between Stokes parameters and raw polarisations.
It can take multiple input models for different Stokes parameters. However,
if a model of a particular Stokes parameter is not given, it is assumed to
be zero. For example, if you give it models of I, Q and U, but not V, then
the source is assumed to have no circular polarisation.
- gpscal
uses the antenna polarisation leakage table
if this information is present. If it is not present, leakages are assumed
to be zero.
- Unlike selfcal
, gpscal
cannot take models from
multiple pointings or from different frequency ranges. All
models must correspond to the same pointing and same frequency range.
Next: Frequency Handling
Up: Computation
Previous: Self-Calibrating with Multiple
Last generated by rsault@atnf.csiro.au on 16 Jan 1996