[daip] CALIB with model

[Eric Greisen]

Olaf Wucknitz writes:

 > sorry to bother you again. I have a problem with CALIB when using a CC
 > model. The HELP page says:
 > 
 >              For a multi-source file, the flux of the clean components
 >              selected for the model are summed and scaled to the source
 >              flux found in the SU table.  If that flux is zero, no
 >              sacling is done.
 > 
 > Since I experienced strange effects with modelled calibrators, I tried
 > to SETJY the flux of the primary to different values before running
 > CALIB and found that the final GETJY values obtained for secondary
 > calibrators did not change at all. This seems to indicate that the
 > model flux is not scaled at all but left as it is in the CC table.
 > 
 > I have a similar effect when using models for secondary calibrators.
 > The final GETJY value is always very close to the initial SETJY value
 > of the secondary. This would also be explained by the fact that the CC
 > model of the secondary is not scaled. Because its total flux is close
 > to the one in my data, GETJY scales by only very small amounts and
 > thus does not change the flux of the secondary very much.
 > 
 > Maybe I misinterpreted the help file. If so, is there any source of
 > more detailed information about what the calibration task actually do
 > to the data?
 > 

In this case you have misinterpreted the help file.  When Cleaning,
especially in the days of slow computers, one often does not Clean to
the point where the Clean Components contain all of the (known) flux
of a source.  The CCs are summed and if their sum differs from a real
(non-zero) flux in the SU table, then all the CCs entering the model
are re-scaled so that their sum matches the flux in the SU table.  In
other words, the SU table is assumed correct and the CCs rescaled to
match.

I ran across this just today when reviewing the code to try to see
what it does when using an image for a model.  The final source of
info about what the programs do is the programs themselves.  I do not
trust anything else since documentation always lags code when it is
done at all.  For example, I just fixed UVFLG to allow 40000 entries
in an external file from 10000.  In various places in the help file, I
found this limit quoted as anywhere from 100 to 1000 - nowhere was the
former 10000 limit correctly stated.

Eric Greisen