Description
This function is a specialization of the setsolve method
which should be used to arrange for bandpass solving when polynomial
solutions for B are desired, e.g., when SNR on calibrators is very
low. Currently, this solving mode is available only for single
polarization data.
Prior to the solution, the visibility data are averaged in time, and
if multiple spectral windows are present (after running setdata), they are concatenated in frequency and one polynomial
solution is delivered.
The solution is performed for both phase and amplitude, to the
polynomial degrees specified in degamp and degphase
(they can be different), and the result will be stored in the same
table. The frequency domain of the solutions is limited to only the
range of frequencies selected in setdata. When correcting
data with these solutions (for other solves or with correct),
only data within this domain will be corrected. Data outside (e.g.,
edge channels avoided in setdata for the solve), will not be
corrected. Therefore, the same (or narrower) channel selection is
recommended for all operations using solutions produced by this
function and solve().
Like other solve types, the solve() function actually
performs the solve according to the parameters set in this function,
and for this type, it generates a plot file which shows the baseline
spectra delivered to the solver (i.e., corrected according to any
setapply() invocations, and time-averaged), with the derived
solution (products of antenna-based bandpass polynomials) overlaid.
After solving, these solutions should be applied using type='BPOLY'
in setapply.
The visnorm parameter is used to normalize the assembled
spectral data, in a per baseline manner. If visnorm=T is
used, this will have the effect of removing any
non-frequency-dependent closure errors (e.g., as caused by source
structure, or introduced by the instrument) from the data, and should
be used with caution. The resulting solutions will be effectively
normalized as well. When visnorm=F is used, closure errors
in the data (as supplied to the solver) may be visible in the form of
offsets between the data and solutions. For bandpass calibration,
this is usually ok, as the shape of the bandpass is the most
important aspect of the solution.
The bpnorm parameter controls normalization of the resulting
solutions after the solve. Note that the baseline spectra plot
does not show normalized solutions, even if bpnorm=T (the
plot shows the data as provided to the core solver, and is generated
before the solutions are normalized).
Note that use of bpnorm=T or visnorm=T will remove or
prevent the storage of scaling information in the bandpass solutions.
Sometimes it is desirable to store absolute scaling information in the
bandpass solution; in these cases, bpnorm and visnorm
must both be F.
The maskcenter and maskedge parameters control how
many channels are ignored on-the-fly, at the center and edges of each
input spectral window, respectively. To avoid edge channels, it is
almost always better to flag these channels directly, or select
against them in setdata. Aggressive use of maskedge (large
values), will yield polynomial solutions which will tend to diverge at
the edges (especially when the polynomial degree is also high),
because maskedge does not change the frequency domain of the
solutions. Such solutions should be used with caution in subsequent
operations. (It is best to avoid use of maskedge.)