[evlatests] More results from 3-bit tests

[Rick Perley]

    A 3 hour run using the seven 3-bit antennas was taken just around 
sundown.  There were three point sources (strong, medium, weak), plus a 
noise field.  All seven antennas fringed stably -- the data are 
excellent.   All SY table values are good -- virtually no editing was 
required, either in the visibilities or in the SY table. 

    For this run, we tuned to the range 14 -- 16 GHz (with the 8-bit 
path tuned to 15 -- 16).  This avoids the nasty super-birdie coming from 
ea07 (at 13894 MHz --- NB  it would be very good to understand the 
origin of this!  Even better would be to get rid of it!).  Examination 
of the autocorrelation spectra show -- as expected -- that each 
antenna/polarization has the resonance phenomenon, which occurs at a 
different frequency for each. The peak intensity is typically 10 times 
the noise in a single 2 MHz channel.  Faint images of these are found in 
the cross-power spectra -- most likely due only to the increase of noise 
due to the power in the affected channels. 

    Bandpass solutions were made for both 3-bit paths and the 8-bit 
path.  The stability is disappointing for all three.  I'll comment more 
on this later. 

    I've already commented on the variation of the PDif values with both 
source strength and elevation.  It is absolutely clear that the PDifs 
are responding to total system power.   The mechanism by which this is 
happening remains elusive. 

    Comparison of the PDif values with Visibilities shows that the 
visibilities are not affected by this compression problem -- or at 
least, not to the same degree. 

    Since the data which I used two days ago to derive the system 
sensitivities -- and the degradation of this compared to the 'golden' 
8-bit path -- had the PDif values applied, I made sure that for this new 
database, the calibration was done against the source whose system 
temperature was closest to that of the blank field.  When this is done, 
I find that the noise degradation factor caused by the 3-bit system are 
rather worse than reported earlier.  Although I have made histograms for 
each polarization and each antenna pair for each of the three paths 
(tedious!), the 'bottom line' is the noise in a sky image.  So I've done 
this, and the result is that, in the A1C1 path, the noise is 10% higher 
than in the B0D0.  In the A2C2 path, the noise is 14% higher.  The 
difference in the A1C1 to A2C2 paths is easily seen in the histograms.  
ea06RCP-2 has a much higher degradation than any other 
antenna/polarization -- by 35%!  Something fishy there ... 

    Although there was only 3 hours of data, I got quite convincing 
cross-polarization solutions (remarkable, given that the parallactic 
angle rotated by only ~35 degrees).  Comparison of these solutions for 
all three paths shows great similarity (as there should be!),  
especially in phase.   But there are remarkable differences in the 
D-term amplitudes for some antennas -- as much as 1 to 2%.   This is 
quite a bit more than I had expected (and hoped!) to see.    I have 
plots for those interested in seeing them. 
    The possibility that the cross-polarizations are different is not 
actually important, so long as they are stable over time.  The duration 
over which this run was taken was not long enough to search reliably for 
*changes* in the cross polarization over this time.  We are planning to 
try this tonight, with a longer run which includes 3C147 -- a source 
whose polarization is well known.  (The run last night did not include a 
polarization calibrator, so the true R-L phase is unknown). 

    Stay tuned.  I intend this next run to be the last test!