[evlatests] evla timing

[Ken Sowinski]

Walter reported:
> Barry thinks the 13,14,16 phase flips, amp drops are caused by the EVLA 
> and VLA antennas becoming desynchronized by 1 waveguide cycle.  This is 
> consistent with what I've seen. EVLA-EVLA baselines are stronger than 
> EVLA-VLA baselines during these times.  Ant 16 showed notably lower 
> amplitudes during these periods than 13 or 14, probably a consequence of 
> the particular walsh codes used.

We have found a problem in the L302 mib which could lead to problems
consistent with this and with other symptoms seen over the last few 
weeks.  The problem is that the if the mib's idea of time changes,
even slightly, it may decide to think it is the wrong WG cycle for 
the purposes of calculating the Walsh switching state.  The mib has
always done this, but it has become more obvious of late because 
congestion in the T1 link to Socorro has made NTP time at the VLA
less stable than it used to be.

The Walsh sequence generating code in the mib should be able to 
tolerate time errors which approach half a WG cycle, but there was 
an error which causes it be in one state or the other according to 
whether time in the mib was ahead or behind the "true" time 
associated with a WG cycle tick.  The code has been corrected, and
once installed the L302s should be much more stable with respect to
small time errors in the mib.

James has modified the routers to the VLA so that NTP packets (there
are not very many) have higher priority than other traffic.  Pete
has fixed the L302 mib and will install the new images Monday morning.
In the mean time the IAT-UTC in the EVLA data base has been shifted
by half a WG cycle to make the L302s operate in their area of stability
in order to allow observations during the weekend.

There is no doubt that there was an error in the L302 mib and that
time at the VLA has been 'jittery' of late.  It remains to be seen
if curing these defects will make the EVLA antennas more stable.