[daip] visibility time stamps

[Eric Greisen]

Aaron Cohen wrote:
> Hi Eric,
> 
> I have a quick questions about the different time stamp methods for VLA
> data.  As I understand, before last year, times were for the beginnings
> of scans.  For the previous AIPS (before 2007) did the imaging routines
> (such as IMAGR) take this into account when calculating the u,v,w of
> visibilities and calculate these for the central times?  If not, then
> would there be a timing error of half the time interval?  Thus, if the
> time interval was 10s, the RA of sources would be shifted by 5s or 75
> arcminutes times cos(dec).  Is this true?

Your understanding is a bit wrong.  The times were for the END of the 
interval, but the u,v,w recorded from the on-line system was for the 
center of the interval.  IMAGR (and MX, UVMAP, et al) simply use the 
u,v,w recorded with the data so there was and is no systematic error 
unless the on-line system has a problem (and there is evidence that 
there were such problems).  The AIPS task UVFIX allows one to recompute 
u,v,w and offers the option of correcting the time.  For data from FILLM 
before July 6, 2007 the times are the end times.  For a version of AIPS 
current or later than July 6, 2007, the times are the center of the 
interval and no time correction should be made in UVFIX.  The revised 
FILLM prints a message near the start about the change in the recorded 
times.

Note that the shift of sources is nothing like what you suggest.  The 
errors we found comparing two very deep wide-field integrations on the 
GOODs field taken several years apart showed a rotation of the image by 
a very few degrees and could only be told from the outermost sources.
This error tested out to be something like 7 seconds of time in the 
u,v,w calculated online.  UVFIX - when run correctly - fixed the 
discrepancy entirely.

Eric Greisen