[bananas] Second set of patches for 15APR98

[Patrick P. Murphy]

There are an additional three patches to 15APR98 AIPS available, as well
as the two previously announced here on April 23.  The new patches are:

   Task     Description of problem                             Date
   -------------------------------------------------------------------
   DBCON    reweights the wrong thing                          1998-04-28 
   BPASS    handles flagged channels and interpolation wrongly 1998-05-05 
   IMAGR    can get wrong frequency in output image            1998-05-05 

Complete information on these, including the modified source code and
instructions for rebuilding the affected tasks, can be obtained via:

	 http://www.cv.nrao.edu/aips/15APR98/patches.html

Descriptions of the individual problems are shown below.

				- Pat

------------------------------------------------------------------------
Task:         DBCON

Summary:      DBCON had an off-by-one error causing it to re-weight
              the imaginary part of the visibilities rather than the 
              weights when the two data sets are in the same sort order.

Documentor:   Eric Greisen
Date:         1998-04-28
------------------------------------------------------------------------
Task:         BPASS

Summary:      BPASS was found to have several errors in handling flagged
              channels and the interpolation option to deal with them.
              1. The worst was an addressing error which meant that the
                 address for weights did not get incremented when a
                 channel was flagged.  Then the weights for higher
                 numbered channels were put in for lower numbered ones.
                 This could lead to channels being eliminated, but did not
                 seem to cause flagged channels to acquire false
                 solutions.  The excess channel elimination did lead to a
                 message, so users should have noticed.
              2. The fit routine did not return an error code for no valid
                 samples so the task did not know to invoke interpolation
                 when a channel was fully flagged.
              3. The interpolation method used the wrong formula for
                 weighting channels so that the interpolation was not
                 quite right and it would fail to extrapolate a solution
                 if the last couple of channels or more were failed.

Documentor:   Eric Greisen
Date:         1998-05-05
------------------------------------------------------------------------
Task:         IMAGR

Summary:      The OOP utility routines had the usual OOP sort of error,
              this time in handling the reference pixel for spectral-line
              imaging.  A low-level routine partly corrected for channel
              selection and averaging unbeknownst to a higher level
              routine which did the correction again, fully this time.  If
              BCHAN and the frequency reference pixel in the input data
              set were both 1 (a common occurence) or if NPOINTS was 1 (no
              channel averaging) or large (all channels "averaged") then
              no error was made.

Documentor:   Eric Greisen
Date:         1998-05-05
------------------------------------------------------------------------

-- 
  \|/  ____  \|/   Pat Murphy; email http://orangutan.cv.nrao.edu/kippure.html
   @~ / oO \ ~@                  Division Head, NRAO/Charlottesville Computing
   /_( \__/ )_\       Plan, PGP, Pages: http://orangutan.cv.nrao.edu/plan.html
      \_U__/         "I don't believe in the no-win scenario" -- James T. Kirk