[daip] visibility time stamps

[Aaron Cohen]

Hi Eric,

The ionospheric shifts are what I'm trying to measure.  I agree that the
rotation would be small compared to any single ionospheric shift, but
I'm trying to get statistics on this and would like to remove any
systematic errors that would affect the mean values.  I'm using the
message files from the data reduction of the VLSS (roughly 90 separate
observations), so going back and changing things is not an option.  If
there is a rotation, I will need to calculate it and remove it from the
analysis.

The data were all taken and prior to 2006, and all reduced before June
2007.  I believe that is before anything changed in AIPS.  Do you think
there would be any rotation in the fields caused simply by the fact that
times are measured at the end of the scans and u,v,w at the center?  If
so, is there a way to calculate it?

Thanks,
Aaron

On Tue, 2008-09-23 at 08:24 -0600, Eric Greisen wrote:
> Aaron Cohen wrote:
> > Hi Eric and Leonia,
> > 
> > Thanks for these explanations.  I think I understand things a bit better
> > now.  
> > 
> > Unfortunately the rotation could be a problem for me.  I'm working with
> > 74 MHz data, and so the fields are about 12 degrees across.  I'm looking
> > into statistics on the relative positions of sources in that field of
> > view.  So any rotations will affect this.  
> > 
> > Is there a way to calculate what the rotation would be for a certain
> > time offset?  I assume it would be dependent on RA and DEC, and perhaps
> > the altitude and azimuth.  Leonia, when you say that 5 sec. of time
> > shift equals ~4E-4 radians of "rotation", I assume you just took the
> > fractional rotation equal to 5 sec/ 24 hours?  Would that be the actual
> > rotation or just an estimate?
> 
> 74 MHz is affected with gross position uncertainties due to the 
> ionosphere so I doubt that you would be able to see any rotation due to 
> erroneous u,v recorded by the on-line system.  The "5 sec" business is a 
> red herring due to an error one could have made with UVFIX - but it is 
> not relevant if you did not run UVFIX.  Ken Sowinski was willing to 
> consider a 10-sec error in the u,v,w - i.e. that the updating of that 
> parameter got one cycle off - but he doubted that it would occur.  The 
> actual numbers we "measured" were +2 sec at one epoch and about -5 sec 
> at another.  The group we were working with seemed unable to follow the 
> UVFIX instructions for the longest time and so we got all sorts of 
> analysis until they finally did it right.  Then the difference between 
> the 2 epochs went away.  But I do not think that 74 MHz can yield 
> astrometric position accuracy and that is the level of error here.
> 
> If you are really worried - load the data again for one pointing with 
> the modern FILLM.  Run UVFIX with UVFIXPRM=0.  Apply your previous flag 
> and CL tables and reimage.
> 
> Eric Greisen