[daip] Conv size vs. Coord incr

[Eric Greisen]

Your coverage is rather sparce and seeemingly rather elliptical as
well with a p.a of -36 or so (by eyeball).  The fitting routine may
have some trouble and the results may well depend on how many of the
beam pixels are included.  I suggest plotting the beam (use SLICE at
several angles and then TVSLICE).  My bet is that you see a beam in
the -36 p.a. that is centrally peaked but with a broad plateau under
it - i.e. not really a Gaussian.

Weighting affects the beam shape.  The cell in the uv plane is 
    1 / (Npixel * cellsize)
so changing the cellsize changes the uv pixel size and - with the
defaults anyway - the uniform weighting in IMAGR then changes.  The
ROBUST parameter especially as well as all of the parameters affecting
how uniform weigting is done (UVWTFN, UVSIZE, ROBUST, UVBOX, UVBXFN,
UVTAPER) affect the beam shape.  Also you can force the BMAJ, BMIN,
BPA with adverbs if you decide the fitting is not going well.  The
problem really is that beams from sparce arrays are not well
represented by a single elliptical Gaussian.

Eric Greisen