[daip] frequency dependent correction in IMAGR

[Eric Greisen]

On 10/14/2014 04:15 PM, Nicholas Wrigley wrote:
> Many thanks for this, it is much appreciated. I shall try what you recommend by setting a FLUX and a high NITER, within POPS and then again using ParselTongue to check whether it is causing a problem. I notice that you still have IM2PARM 2 set in the attached inputs.
> I was led to believe that some baseline averaging had already been carried out. The uvrange was simply our attempt to eliminate some noisy data on the shortest baselines that looked suspect in uvplt. We are using 31DEC14, but the MNJ is not ordinarily set up on our machines - I will ask IT to get it added if you think it may make a difference.

\

I set IM2PARM(1) to 2 to let it make new boxes for me.  The time 
averaging and other things got set in the one run that failed because I 
did not re-read my inputs and had said im2p = 2 which sets all 40 im2p 
values to 2.  "Noisy" data on short spacings?  Or real and complicated 
signals???  It might be wise to figure out which or to do self-cals with 
your UVRANGE model but applying to all data or something.  The VLA is 
designed with lots of short spacings for good imaging reasons.  The 
short spacings may have more RFI and if that is the case leave them out. 
  I was less interested in baseline-length time averaging (so none on 
the long spacings and a fair amount on short spacings) which is 
something you do not do to the input data if it is to be self-cal'd but 
you can do on the fly in IMAGR to affect the data weighting and the 
volume of data entering into every model subtraction and FFT.

The MNJ should be run once in a while, not every day like we do. 
Otherwise you are really running a version that is nearly 31DEC13 but 
without the patches that have been released for that otherwise frozen 
version and none of the improvements that have happened in the last 10.5 
months.

ERic Greisen