Calibration essentially involves solving Eq. 1 for s.
This can be done using a non-linear least square minimization
algorithm. Since the right hand side of Eq. 1 in the
bootstrapping scheme involves the model visibilities
(which in turn are equal to Fourier transform of
where
is the PSF), solving Eq. 1 needs the model image. In
the case of an unresolved source,
is trivial. We
refer to the algorithm used to solve Eq. 1 for an
unresolved source as antsol (see reference 2). A robust
implementation of antsol is implemented in the program rantsol and in practice has proved to be robust to the presence of
bad/corrupted data and have been regularly used to automate various
tasks (e.g. GMRT phased array mode, beam shape measurements,
polarization purity estimates, baseline measurements, etc.). The
program calibbp uses the rantsol algorithm to calibrate
LTA data in time and frequency using an unresolved source.
It is easy to modify these programs to work for resolved sources by accepting a model image. With that modification, it appears that the step of calibration can also be reasonably automated.