Comments on "Calibration, Imaging and Datasystems for AIPS++"
             from Green Bank Meeting 3-14 Feb., 1992
             Ed. D. L. Shone & T. J. Cornwell

                  by W. D. Cotton
                    23 Feb. 1992


   My principle comment on this document is that it not what is
required at the present time.  Instead of a detailed technical
description of the problem it is an attempt at a design.  I have
fundamental disagreements with this design but since these arguments
are irrelevant to the current situation I will not comment further on
the design.

   At the current stage of this project what is needed is a detailed
technical description of what it is required to do.  Due to the
extremely ambitious scope of the project, adequate design and
implementation of the system require a range of expertise and
experience far exceeding that of any individual or even a small number
of individuals.  Many issues are clouded by the use of different names
to describe the same thing or the same name for different things.  For
the design and implementation to proceed a document(s) is needed
explicitly detailing what needs to be done. 
  
   The way data is collected, organized, calibrated and imaged differs
a great deal among the instruments to be served by this project.
A major part of the design effort will be to determine what is done in
common among the various instruments and what is done differently.
It may be possible to increase the commonality by negotiations
among representatives of the various instruments.  This process
cannot even be started without a coherent and detailed descriptions of
the requirements of the different instruments.

   Many of these technical details were discussed in Green Bank but
were not included into the resulting document.  A description of the
data collected from each instrument includes the polarization
correlations measured, number of frequency channels, number of
simultaneous spectra and/or pointing positions and estimates of the
volume of data produced in typical and extreme cases.  If auxiliary
measurements, such as system temperature or instrumental delay, are
needed in the calibration process this should also be specified.
   Calibration is probably the area with the largest instrument to
instrument variations.  What is needed for each instrument is the
detailed model of the instrumental response and how the parameters are
determined and corrections applied.  It is not clear to me whether or
not calibration normally done by observatory staff such as antenna
pointing calibration is within the scope of this project.  The range
of calibration to be considered needs to be settled fairly soon.
   For each instrument a description of how the calibrated data is
converted to the image domain is necessary.  For example, in AIPS VLA
teh calibrated correlations are convolved with any one of a number of
convolving functions and sampled at a bounded set of regular grid
spacings on a single half plane of the uv plane.  A symmetric half
plane is Fourier transformed...
   The common methods of the deconvolution of the point spread
function from the "dirty" image (e.g. some variant of CLEAN and/or
MEM) need to be listed and described.  Finally, since different
variants of self calibration (iterating through calibration, imaging
and deconvolution to remove atmospheric variations) are commonly
practiced these should be analyzed as well.  The most advanced of
these, the "difference mapping" technique from Jodrell Bank needs to
be examined in some detail.

   A document giving the techincal description of the various
instruments could be constructed starting from the notes of the people
who gave the telescope presentations in Green Bank assisted by other
knowledgable people.  I feel that this must be done before any further
progress can be made.