[evlatests] More on VLA/EVLA phase behavior

[Rick Perley]

It was discovered yesterday that VLA-VLA phases were for many antennas 
badly behaved, in the sense that many antenna-IFs did not return to the 
prior phase after change of band.  Ken and I used the dynamic time last 
evening to repeat the experiment which revealed this behavior, adding a 
number of different modes to see if any particular combination caused 
the problem. 

Observational setup:

Alternating between C and X-band, one minute at each band. 
A single source -- 3C454.3 (10 Jy).
Two different bandwidths:  12.5 and 50 MHz
Two different correlator modes:  Continuum and spectral line mode '4' 
(giving RR and LL correlations in both IFs). 

There were thus 8 different combinations (two bands, two BW, two 
correlator modes), which were cycled thusly:
C-cont-50MHz
X-cont-50
C-line-50
X-line-50
C-cont-12.5
X-cont-12.5
C-line-12.5
X-lint-12.5

All six EVLA antennas were in the array, and give strong fringes at both 
bands, except 26, which is outfitted only at X-band, IFs A and C. 

Some EVLA antenna flagging was required -- see below for details. 

        A)   Phase stability results:

    - Both VLA-VLA and EVLA-EVLA baselines showed significant problems, 
as described next. 
    - There appears to be no dependence of these problems on bandwidth, 
band, or correlator mode. 

1) VLA-VLA baselines.  Using a central VLA antenna as phase reference, 
it is clear that many VLA antennas do not return to phase after a band 
change.  The worst offender by far is antenna 12 on the A and C IFs 
only, but there are many others which show the same problem although 
less frequency.  I can find no pattern in these phase changes. 

2) EVLA-EVLA baselines.  Using antenna 18 as a reference, I find that 
antenna 13, 14AC, 24, and 26 show good return to correct phase after a 
band change.  All four show strong long-term trends (as reported 
yesterday).   On the bad side, we have 14BD, which is a jumper (changes 
phase upon band change), and 16AC, which has two phase states seen 
within all 1-minute scans -- separated by 80 degrees, with the phase 
changing that much within that time interval.  Antenna 16BD also jumps, 
but much less frequently than 14BD.   These characteristics are exactly 
the same for all bands, modes, and BW. 

3) VLA-EVLA baselines.  With a VLA antenna as reference, the EVLA phases 
are terrible.  We have what appears to be multiple semi-stable phase 
states -- each 1minute scan is stable, but after a band change, the 
phase is somewhere else.  The two arrays have essentially no stable 
phase between them. 

    B) EVLA-specific flagging issues. 

Some special flagging was needed to remove some data:

1) The 10-seconds-at-the-end-of-scan problem occurs on about 10% of the 
scans.  In all cases *except 1*, all EVLA antennas go dead 
simultaneously.  In a single case, antenna 26 did not go dead, while the 
others did.  This oddity occurred on the last scan of the file -- but I 
can't see if there is any significance to this? 

2) We occasionally get no fringes from an EVLA antenna for an entire 
scan.  This is a rather common occurance -- perhaps 10% of the data from 
some antennas is lost.  The problem usually, but not always, affects all 
four IFs.  In a few cases, only one IF pair (e.g. A and C, or B and D) 
are affected.  Antennas 13, 14, and 16 are by far the most likely to 
have this happen.  Antenna 24 never lost any fringes.  Antenna 26 was 
dead only once (but, remember that it works only at X-band, in IFs A and 
C).   It seems clear that this is an LO setup problem.