[evlatests] Flux Redux Results

[Rick Perley]

    'Flux Redux' refers to another lengthy (~18 hour) test run last 
week, primarily to check on system robustness, and to confirm whether 
thermal isolation of the Ku-band noise diodes is sufficient to remove 
the notable (~6%) temperature sensitivity of the array's gain. 

    The methodology is the same as for the bigger 'Flux Densities' run:  
Rapid observations at every band (typically 25 seconds at each band) for 
a large number of sources -- in this case, 61 different objects, each 
observed two or three times. 

    The first (and possibly most important) result is that there were 
*no* correlator failures.  The timing problem which plagued the longer 
earlier run has been repaired, and the system seems immune to the worst 
I can make it do.  :)  .  A quick inspection indicates all intended 
observations were made. 
    Perhaps even better -- no other failure modes were uncovered.   From 
past experience, we might have anticipated that another, unsuspected 
failure mode would surface -- none did. 

    Because there were no correlator crashes, and as I carefully ensured 
that 'set and remember' worked at the beginning, the 'PDif' values in 
nearly all IFs are now smooth curves.  This permits useful statements to 
be made about switch stability, and other system characteristics. 

I've completed calibration of the X-band portion of the data.  In 
general, the data quality is outstanding.  But there are some issues 
with the switched power, and other system characteristics, that I 
summarize below.

    1) ea02 has very low switched power -- undoubtedly caused by the 
high reported system temperature:  about 500 -- 800 K.   

    2) ea05 dropped out for about an hour (at about 8AM), and when it 
came back, the power levels were at very different values (typically a 
factor of 2 different).  I presume the antenna was taken out of 
service.  But in doing this, are the attenuator values then forgotten? 

    3) ea11B and ea12A shows highly variable (~30%) PDif values, not 
shared by the other three IFs.  Bad switch? 
   
    4) ea14, in all four IFs, showed an *extradorinary* time variability 
in its PDif and PSum.  PDif smoothly dropped by a factor of about seven 
over a duration of about two hours.  The gain slowly returned to normal 
over the succeeding 9 hours.   The visibilities also changed, but are 
not fully corrected by application of PDif. 

    5) ea23 had some sort of DTS issue on all four IFs -- each had 
something unusual in the bandpass.  All four were different.  IFs A and 
D were most affected, and had to be flagged. 

    Finally:  Following the usual calibration regimen, I have plotted 
the residual gains vs. time.  There is no detectable signature of 
time-variable gain at X-band, to a level of better than 1%.  Note 
however, that:  the temperature range was not so great this time (about 
16C, compared to 22C in the earlier run), and that the duration of this 
run was much less than 24 hours.  In the previous, longer run, the 
effect was no more than 1% in power, and was easily visible on only a 
few antennas.   The current result is consistent with the older one. 

    More critical is the Ku-band test.  These data are being reduced next.