[daip] CLCOR vs IMAGR for shifts

[Eric Greisen]

Mark Reid wrote:
> Eric Greisen wrote:
>> I do not have completely suitable data to test your issue but I tried 
>> anyway - C array at L band.  I shifted north by 15 arc minutes.  This 
>> is about the same as yours - to the half power point of the beam.  
>> Then with DOCAL=1 and DECSH=-900, I ran IMAGR.  The point source 
>> looked sort of okay but it did not Clean down to the other sources in 
>> the field and was not "right".  I then ran UVFIX.  Running the same 
>> IMAGR, but now on the fixed data, produced normal - i.e. right - results.
>>
>> A array would have made all this worse.
>>
>> Note that UVFIX has a coordinate shift option and some discussion 
>> about its use in the help file.
>>
>> Eric Greisen
> 
> So, if I understand the import of your tests, UVFIX does large shifts 
> more accurately than CLCOR (as you thought it would).
> 
> I'll look at the documentation and do some tests here.

UVFIX is needed to recompute the u,v,w which IMAGR assumes are correct.
CLCOR cheats - the phases and the header coordinate change but the u,v,w 
do not.  I will add something to the CLCOR help file reminding about that.

See the UVFIX help file.  I think it suggests 1 uvfix with no shift 
followed by another with the shift.  Modern VLA data may not require 
this although there was a hint that the on-line system had an error and 
used u,v,w for a somewhat offset time(!) before the swith away from the 
ModComps.

I have copied your file and will now play with it.  Thanks

Eric