[evlatests] PDif issues at L-band

[Rick Perley]

    I'm carefully reviewing the switched power data from the flux 
densities run.  For most bands, the switched power data are of excellent 
quality, although the seven-or-so correlator restarts make 
interpretation 'interesting'...

    Some extraordinary events are found -- these will be summarized once 
I get through all the data. 

    At L-band, there are a few issues ready for distribution:

    1) The PDif values are zero (within the noise) of ea23C (only), and 
ea25C and D.  I'm puzzled as to how a single IF can have zero PDif  -- I 
would have expected that ea23D should also be zero, but the data are 
fine.  In all cases, the PSum values are fine. 

    2) Something is not right with ea19, on all four IFs.  The system 
temperature is reported to be between 70 and 150 K, and the PDif values 
are weak (expected, due to high Tsys), and of very poor quality.  Poor 
quality means:  the noisy spread in the data are nearly 50% of the mean 
value -- this is far higher than expected just from thermal noise. 

    3) At L-band, most of the expected troubles when the system restarts 
-- due to insufficient time on any source to 'remember' the new gain 
setting -- should not be present:  The antenna slews between sources 
were always done at L-band, so the requisite 40 seconds or so will 
nearly always be met.  And the data show this, as the PDif data 
following each of the seven computer restarts were (almost) always 
stable (until the next crash).  However, things are not all good:

        3-a)  The 'new' power levels in most cases were within 1 or 2 dB 
of the old.  This is probably reasonable given changing spillover power 
and the effects of RFI within the 1 GHz-wide band.  (I'm only reviewing 
PDif data within RFI-free subbands).  But in some cases the new values 
were very different than the old ones.  In the worst case (ea12A), at 
least a factor of 32!  (In this case, the new power level is that much 
higher than the old, suggesting probably that *all* the attenuation was 
taken out.  Buy why would that be?  No other antenna (including the 
other IFs of ea12) did this.   In some cases, the change after the 
correlator restart was 3 dB -- rather larger than I think it should be. 

    4)  A truly extraordinary event occurred near 13h (local time), when 
the power levels of six antennas dropped by typically 3 to 6 dB, and 
stayed low until the next correlator restart (over an hour).  The 
characteristics are  unique:
     a)  Only six antennas were affected:  ea08, ea11, ea14, ea15, ea16, 
and ea27.
     b)  All four IFs were affected, but not equally:  the drop for A 
and B are the same in ratio.  Also the drop for C and D -- but the RCP 
drop is very different than the LCP.   Some of the drops are extremely 
large -- over a factor of 100 for ea16 (on all four IFs) at ea27 (in LCP). 
     c)  The time at which the drop occurred was different for different 
antennas.     The time(s) at which this occurred are not connected to 
the corrrelator troubles -- this drop occurred on its own, for other 
reasons. 
     d) Although I haven't gone through all the other bands, I have 
noted that at least some of the other bands showed a similar drop at the 
same time. 
     e) These represent a real gain change -- both PSum and PDif show 
the effect (i.e., the reported Tsys is unaffected, except for those 
antennas where the drop was very large -- clearly we're in a non-linear 
range).