[evlatests] Referenced Pointing performance, over time

[Rick Perley]

    I've reviewed all the Mars data I've taken at K and Q bands (since 
1995) to try judge our referenced pointing performance over time. 

    A reasonable statistic is the scatter in the observed 
visibilities.   Because Mars is a resolved object, I looked only at the 
scatter in the visibilities at the shortest spacings.  (A fairer test 
will be to look at the calibrators, which I may try to do later today if 
time permits). 
    For reference, a 10 arcsecond offset results in a ~8% loss at 
Q-band, and 2% loss at K-band.   A 20 arcsecond offset gives a loss four 
times greater. 

    Note that in the table below, the expected four-times-worse 
performance at Q-band is not generally seen -- this is easily explained 
by my zealous editing of discrepant visibilities -- the 4X larger 
amplitude drop at Q-band is much more likely to get my attention! 

    In the table are given the fractional spread (peak -- peak) in the 
visibilities, in percent. 

Year                K            Q
------------------------------------------------
1995   Mar      5%         not enough data
1998   Feb     15           30
1999   Apr      7           22
2000   Oct       9           25
2001   Nov      6         not enough data
2003   Feb       6           13
2004   Aug     12          25
2006   Jan       10          10
2007   May     12          25
2008   Sep      13          30
2010   Jan         6          20   
2010   Dec      20          50 
---------------------------------------------------

Because of the editing process, it's best to concentrate on the K-band 
data.  It's hard to see any trends here, but it's certainly clear that 
the December 2010 data are much worse than the rest.

I checked the WX tables, hoping to see that the more degraded days were 
correlated with high winds -- except for the May 2007 data, all runs had 
low to moderate winds, less than 6 m/sec.  

A complicating factor is that I had to use difference reference pointing 
objects, at varying separations. 

    A much fairer test is to utilize the point-source calibrators for 
this survey.  I'll do this when time permits.