[evlatests] more tests, May 30 2006

[Ken Sowinski]

I resurrected my data from Tuesday afternoon as well as the
data that Walter took to examine it a little more carefully.
Below are comments based on that and other information.

On Tue, 30 May 2006, Walter Brisken wrote:

> Antenna 18BD has "double dropouts" where two consecutive 400 ms integrations 
> show reduced amplitude.  The first one is reduced to ~1/3 nominal value and 
> the second one is ~2/3 nominal.  Likely related to this are ~105 degree phase 
> jumps every _5_ seconds; alternate 5 second periods are at 0 and 105 degrees 
> from nominal.  The ~1/3 amplitude dropouts happen 3 10 second ticks in a row, 
> then skip several 10 second ticks, come back for 3 more, ...  The 2/3 
> amplitude samples are always dropped out.

Matt points out that 18BD has an L302 without the latest
firmware changes, but with the debilitating changes, we
discovered in antenna 13, so we should ignore it until it
is up to date.

> On Tue, 30 May 2006, Ken Sowinski wrote:
>
>> The major result today is that the L302 firmware fix to
>> address the once every ten second amplitude glitch seems
>> to work.  I can see no problem in either IF pair in antenna
>> 13 after the modules were installed.
>
> Still there, all IFs.  Drop is exactly one per 10 seconds down in amplitude 
> to ~70% nominal.  If this is related to the sawtooth (ie discontinuous phase 
> breaking LO lock) then this might not appear for sources in certain parts of 
> the sky where the fringe rate relative to array center is near enough to 
> zero.

After examining the data on disk, I can confirm that I saw no
no glitches and that Walter did.  Differences?
1.  Different executor.
2.  I had the source marked as a calibrator, but Walter did not.
3.  Modcomps were rebooted inbetween the observations.
4.  Different HA.
It is hard to see that any of these are relevant.

>> 2.  the sawtooth in phase with ten second period,
>
> On 13-15 AC there in X-band is no obvious sign of this, but I still claim to 
> see some phase behavior with 10 second intervals.  The phase is a continuous 
> function of time, but is 'lumpy' to coin a new term.  Is it possible that the 
> fix that went into the linear term should have gone into the quadratic term? 
> I claim the phases are still 'rampy' (ie discontinuous at 10sec boundaries) 
> on 13 BD, 14, 16, 18.

The change made was to the quadratic term in phase.  An error was
found in calculating the chirp parameter in BD that can explain
why a strong sawtoodth was seen in 13BD, but not AC.  This was
fixed in EvlaAntennaPhysical.java this afternoon.

> On closer look, at K-band where phases are moving a bit more, I claim the 
> phase(t) is not quite continuous -- there is a much reduced discontinuity at 
> the 10 second boundaries on AC compared to BD.  Curvature in phase(t) within 
> each 10 second interval is quite clear in AC.

I wonder if there is still something wrong with the "chirp"
term.  It is pretty easy to introduce a factor of two here
and there.