[alma-config] Calculating reconstruction errors

[Steven]

In calculating the reconstruction errors I have to convolve the model images
with the (gaussian) CLEAN beams used to in the simulated and deconvolved
images. In my prototype attempts so far I have let the BMAX, BMIN and BPA
parameters in IMAGR assume their defaults, and had hoped to extract them
from the header of the simulated image...but they aren't there. So what I
propose instead is that I should set BMAX and BMIN at a good approximation
to the FWHM measured for the dirty beams sliced in the appropriate
direction. Having looked at the dirty beams the good approximation in
practice seems to dictate that I should set them as follows:

BMAX = 5 * (resolution in arcseconds per pixel)
BMIN = 5 * (resolution in arcseconds per pixel) / (cos (declination +
23degrees))
BPA = 0
i.e. the FWHM of the dirty beams are approx five pixels across for each of
the A, C or E arrays,
with the resolution in arcseconds per pixel being 0.01 for the A arrays,
0.044 for the C arrays and 0.2 for the E arrays.

Thus e.g for the A array at -23 declination, BMAX=BMIN=0.05arcseconds
for the C array at -23 declination BMAX=BMIN= 0.22arcseconds
for the E array at -23 declination BMAX=BMIN= 1 arcsecond
for the A array at +25 declination BMIN becomes 0.0747arcseconds
for the A array at -70 declination BMIN becomes 0.0733 arcseconds

etc.

Note that BMIN is greater than or equal to BMAX- will this work or should I
interchange them and set BPA to 90 degrees?

Does this seem reasonable? I could calculate a restoring beam for each array
and declination specific to each of the ring or spiral types based on a more
complicated fitting method, although the numbers would turn out pretty much
the same.
In the meantime I propose to use this approach on one array type (K), to get
some results out.

Cheers,
    Steven