[daip] Re: Problem with the `IMAG' option of UVSUB and CALIB in 31DEC03 AIPS

[Eric Greisen]

Subhashis Roy writes:

 > I notice that while using the CMODEL `IMAG' option in UVSUB with
 > FACTOR=-1, the flux density which is added with the CLEAN components is
 > exhorbitently high than expected.

     With CMODEL 'IMAG' the Clean components are entirely ignored and
do not have to exist.

 > 
 > It appears that UVSUB is taking Fourier Transform of flux densities of each
 > pixel, without normalising by the `square of the number of pixels per Beam' 
 > (at least the extra flux density can be expressed closely by the
 > `(number of pixels per Beam)**2').

     The help file for UVSUB makes it clear that, if the image is a
normal Clean image, that you must set FACTOR to correct for the beam
size in pixels.  I do not know why you are talking about beam size
SQUARED - it is a simple linear scaling.

 > 
 > Same problem exists with CALIB (i.e., with CMODEL `IMAG' and SOLMODE
 > 'A&P', the UV data of the output file has flux densities multiplied by
 > the `(number of pixels per Beam)**2'.

I agree that the problem also affects CALIB and there is no FACTOR to
correct things by.  On the other hand, it is an error to use a Cleaned
image as a model for calibration.  Cleaned images restored with a
Clean Beam do not match the UV data but instead the UV data with a
serious taper.  I would not want to use gains determined in this way.

For a numerical example, to test this problem, I ran UVSUB in division
by an image mode and in division by the CC file.  The latter produced
an output with amplitude 1 with only a few points down to 1/4 and up
to 4.  The former when scaled by 11 (my beam area) produced 1's at
very short spacing with a scattering of points to 4 but then rose
rapidly to have the scattered points around 130 at the longer
spacings.  By 1/2 of the maximum spacing the fluxes were already over
2 on average.

I will add to these programs a scaling for Clean beam area, but I will
also add a serious warning about their use in this mode.

Eric Greisen