[daip] IMAGR position shifting

[Eric Greisen]

Mark Reid wrote:

> 
> AIPS has some issues, but overall it works very well and we get a lot 
> done with it.  Certainly we'd be dead if we had waited for AIPS++/CASA.
> Indeed, we have yet to see if CASA becomes a robust and useful package.

I am told that Fred, whenever the word AIPS is uttered in his presence, 
says "why are we still supporting that?".  Fortunately, management out
here is not completely CASAized - although the first day of the 
tutorials in this year's summer school will be only CASA (pre-canned 
scripts!).

> For example, consider a source observed offset from the phase center by 
> say RASHIFT=10" (in J2000 catalog coordinates).  Assume it has zero 
> proper motion.   If one observes at two different times (eg, separated 
> by years), would you expect the offset source to slowly rotate around 
> the phase center owing to precession (ie differential precession).   I 
> think that by imaging in epoch J2000.0 coordinates, this effect vanishes.
> 
> But, I expect that anything that changes coordinates between "mean" and 
> "apparent" would cause the offset source to change position between the 
> two observing dates.  So, we should see differential nutation and 
> aberration in the position, since J2000.0 coordinates are "mean" 
> coordinates.
> 
> Does this make sense?

Yes - the u,v are J2000 (by a simple rotation) so the images should be 
the same at that level.  When non rotation effects become important then
all bets are off.
> 
> Mark
> 
> By the way, I compared shifting in UVFIX with shifting in IMAGR and I 
> got agreement at the 1 mas level for VLA A-config data at 43 GHz and a 
> position shift of about 40".  Given that the beam was about 50 mas, I am 
> not surprised to see a 1 mas difference in fitted positions when the 
> data are handled differently. So, it appears that UVFIX and IMAGR do the 
> same thing when shifting, at least to the level tested.

IMAGR recomputes u,v,w to the facet center for the separate sources 
which should be about like UVFIX although less accurate since it is 
solely a geometric 3x3 rotation matrix in IMAGR.

> I tried the same test, but using CLCOR shifting instead of UVFIX and got 
> seriously distorted images.   It looks like CLCOR doesn't like shifting 
> VLA data or has a problem with a large shift (40"), as I mentioned a 
> while ago.  However, we do know that CLCOR works very well for shifting 
> VLBA data by ~few arcseconds.

I wonder if u,v,w needed to be recomputed after the shift with UVFIX. 
Your image here was made on the assumption that the u,v,w were correct 
which they no longer were...

Eric Greisen

> 
> -------------------------------------------------------------
> Mark J. Reid                 Phone: 617-495-7470
> Harvard-Smithsonian CfA      Fax  : 617-495-7345
> 60 Garden Street             Email: reid at cfa.harvard.edu
> Cambridge, MA 02138, USA     Web  : www.cfa.harvard.edu/~reid
> -------------------------------------------------------------