[evlatests] More wideband results

[Rick Perley]

    Ken continues to run 3-bit full bandwidth (8 GHz) tests.  Last night 
he ran a 30-minute test on 3C286 and a nearby noise field.  There are 64 
subband pairs, each 128 MHz wide, with full polarization, and 2.0 MHz 
spectral resolution. 
    The tuning ranges chosen are:

    A1/C1:  11.476 -- 13.524 GHz
    A2/C2:  12.476 -- 14.524 GHz
    B1/D1:  14.476 -- 16.542 GHz
    B2/D2:  16.476 -- 18.524 GHz

    Here are some basic characteristics -- a more complete report will 
come later (since Ken has just taken another test, with wider tuning 
range, which I must look at ...)

    1) The first scan on 3C286 was a 'warm-up' adjustment scan, of some 
5 minutes duration.  Sure enough, bandpasses are not normal, nor are 
fringe levels.  The following 1-minute scan was stable. 

    2) The basebands I reported yesterday that had odd ~12 MHz 
oscillations were still misbehaving.  The effect is certainly of digital 
origin -- misbehaving samplers?  The culprits are:  ea19 C2, and ea23 A1. 

    3) ea13 D2 has a hugely sloping bandpass, and very low spectral 
power overall.  This looks like a bad FE response, as the spectral power 
is close to flat for the lowest ~500 MHz (16.5 to 17.0 GHz), then 
steeply slopes away above that.  But the overall fringe power is far too 
low overall, so perhaps there's more trouble here.    This problem was 
also present yesterday. 

    4) The 16th subband in many -- but not all, basebands -- is very low 
(typically 5 to 10 dB).  This is presumably the 'DC offset' problem.  
But why are some antenna/basebands not affected? 

    5) In general, the spectral power levels and slopes are very good.  
But some basebands are definitely lower than they should be -- either 
due to mis-set power level, or to unusually poor SNR (i.e., high SEFD), 
so the spectral power (which I determine through bandpass calculations, 
using the visibilities) is too small a fraction of the total power 
(which is what the system used to set power levels).  Those notably 
deviant are:
    13D2 (as noted above) -- low by at least 10 dB overall
    25A2 and C2:  low by 5 dB
    25B2 and D2:  low by 5 dB. 
    All three of these have been noted before.