[evlatests] Time offset is baaack

[Vivek Dhawan]

Ken noticed a pointing offset in the pointing run on Friday night,
2009-02-21, UT 07:05 to 12:08. Sure enough, there is the time offset
again, in the same sense(sign) and same magnitude as its last known
occurrence, 0.85s, on Feb 14th.

Trying to pin it down to 16 or 17 WG cycles is not possible, 0.833s
and 0.885s seem to straddle it. Atmospheric phase errors limit the
accuracy.

I see no significant change over the 5hr run (or within previous runs)
So it changes at project boundaries, not within an observing file.

Next time we have the array idle, I suggest the operators run several
1-hr pointing or baseline files back to back to see if it happens.

Going back to the pointing data, I averaged up the EW tilt and get
0.24arcmin, or a time offset of 0.96s, not 0.85 as above. The NS tilt
is about +_ 0.05' and so is the avg EW tilt when the time offset is
absent. On Feb 14 the pointing offset was 0.28' but the time offset was
still 0.85s. Thus the pointing offset is reliable at about 5-sigma,
whereas the phases in the pointing or baselines should show up a time
offset at ~100 sigma but are limited by weather to ~20 sigma.


Vivek.