[evlatests] Bandpass Stability -- 3bit vs. 8bit.

[Rick Perley]

    I reported yesterday that the bandpass stability from the 3-8 bit 
tests was 'disappointing'.  I've now had a chance to review this a 
little more closely. 

    The bandpass stability is indeed disappointing -- but it has nothing 
to do with the 3-bit samplers. 

    A bandpass solution was made on 3C84 (30 Jy) once every 8 minutes.  
The duration of the 3C84 scan was 2 minutes, so the SNR is excellent.  
Hanning smoothing was applied to prevent 'ringing' and improve SNR in 
the vicinity of the resonance features. 

    The AIPS program BPLOT is very useful in showing differential 
temporal variations.  Plotting the three paths for the frequency span 
common to all three (15.0 -- 15.896 GHz) shows the solutions are:
    1) Nearly identical for all three paths.  They are especially 
identical (!) for the A1C1 and A2C2 paths.  The 8-bit B0D0 path varies a 
little in detail, but shows all the same major features.  From this we 
can deduce that the path division within the T304, and the subsequent 
digital processing make very little difference to the bandpasses.
    2) There are distressingly large temporal variations over the 2.5 
hour of the experiment.  How large?  By 1 to 2% in amplitude (or up to 
4% in spectral power), on frequency scales varying from a ~10 MHz to 150 
MHz.  Some antennas are especially bad:  ea10 is easily the worst. 
    3) However -- there is hope.  All these changes are gradual over the 
observing window -- it's clear that an hourly calibration, or perhaps 
twice hourly, will nicely remove these changes. 
   
    In important question is the origin of these variations.  The 
observation was made during a period of very large temperature range -- 
but also a period of rapid elevation rise.  Which -- if either -- is 
responsible? 

    I have plots, for interested eyes ...