[daip] CLCOR vs IMAGR for shifts

[Eric Greisen]

Mark Reid wrote:
> Eric and Craig,
> 
>    I was working on some VLA A-config 43 GHz data (AR678:maser stars 
> near Sgr A*) and seem to have a problem shifting with CLCOR again.  Here 
> are the particulars:
> 
> I'm using 31Dec10 AIPS.
> The data has 2 polarizations (RCP & LCP)
> 
> We pointed at Sgr A* and used it to (self)calibrate the data.
> Then we image with several boxes, one on SgrA* at the origin and the 
> others shifted (IMAGR: RASH,DECSH) by ~10 arcsec to the positions of 
> masing stars.   This has worked well both now and in the past.
> 
> As a test, I tried shifting with CLCOR (ANTC: CLCORPRM(5),(6)) and then 
> imaging with IMAGR.  This caused problems:
> 
> 1) when I shifted a large amount (30" mostly in Dec) in CLCOR and then 
> by -30 arcsec in IMAGR, the image of Sgr A* was highly distorted. 
> Actually, the image resembled the VLA in shape!
> 
> 2) when I shifted a modest amount (5.6" mostly in Dec) in CLCOR and then 
> by -5.6 arcsec in IMAGR, the image of Sgr A* was fine.  HOWEVER, there 
> was a residual offset of about 10 mas in each axis.  This is about 1 
> part in 500 of the shift...pretty large.
> 
> I would think this could be simulated with almost any VLA-A data set 
> that has a source with good SNR.
> 
> Any ideas what is going on?
> 
> Mark
> 
> PS: the reason I was doing this test is that I'm worried that by 
> shifting in IMAGR (which I believe does only pure geometry changes), 
> that I will need to correct for differential precession, aberration, 
> refraction, etc.  I believe that CLCOR does the full calculation and 
> thus would show me the "truth" (if it worked).  Does this sound right to 
> you?

You give CLCOR way too much credit!!!  IMAGR and CLCOR do only pure 
geometric shifts.  IMAGR does correct u,v,w for shifted facets.  Doing a 
shift with CLCOR will change the recorded source position - but it does 
not recompute u,v,w in any way.  IMAGR assumes that the u,v,w recorded 
with the data are correct - which with a shifted source would not be 
true.  UVFIX is the only task in aips that considers differential 
precession, etc etc

30 arc sec seems pretty small for getting serious sky curvature 
problems.  But at such southern declinations W errors are very serious.

Eric Greisen