[evlatests] An interesting correlation in PDif Compression with power to the samplers

[Rick Perley]

    Ken ran a special script on Saturday night whose purpose was to 
provide clues to solving our 'PDif Compression' problem. 

    The script was set up thusly:

    We observed sequentially:  1) Cold Sky, 2) A calibrator, and 3) 
Cygnus A. 
    The default *input* attenuator setting was set, and remembered 
throughout the experiment. 
    The output attenuator was then changed in 3 dB steps, such that the 
power to the sampler was:
    a) -12 dB from nominal for the first round of three sources,
    b) -9 dB from nominal for the second
     ... etc.
    e) at the nominal level
    .... etc.
    h) +9dB for the last. 

    (Ken intended to have a +12 dB setting, but for whatever reason, it 
didn't happen).

    Each of these 'lettered' states contained three rounds of Cold Sky, 
Calibrator, and Cygnus A:  the first was with noise diode switching on, 
the second with the diodes fully off, the third with the diodes fully on 
(and not swtiching). 

    This experiment makes for quite a lot of data, which I'm still 
sorting through.  But one very striking effect has been found, and 
deserves circulating now.

    Changing the output attenuators in the manner described above 
revealed a curiosity -- for a number of antennas there was no change in 
output power for the first three or even four commanded changes in the 
T304 output attenuator (steps 'a' through 'd', listed above).  So, there 
was no change in either PSum or PDif for these commanded attenuator 
settings.  Ken explains that for these antennas, the power level in the 
T304 module is so high that the output attenuators are 'maxed out' for 
the first 3, or even 4, steps. 

    However, for a few antennas, the expected changes in output power 
were seen for all steps.  Apparently, for these antennas, the internal 
power level is much lower, so we then have a wider range in output 
attenuator settings (from the nominal setting point). 

In looking more closely at these antennas, it was found that:

*** The antennas which are 'maxed out' with the highest output 
attenuators  -- those for which there was no change in output power for 
the first few attenuator settings -- all have maximum PDif compression. 

*** The antennas which show the expected change in output power for the 
whole range of the experiment all have minimal -- and usually no -- PDif 
compression!  That is, those antennas whose power level within the T304 
is low give no, or small, PDif compression. 

    The power levels to the samplers, at the nominal setting (setting 
'e' in the list above), were all about the same, as they should be. 
    The compression effect would thus appear to be related to power 
levels within the T304. 

    A few other checks were made: 
    1) The compression is the same whether the diodes are switching, or 
not.  The digital system which demodulates the switched signal is thus 
exonerated.
    2) The level of compression does not appear to be a function of the 
power to the samplers.  This is most easily seen in plotting the 
apparent system temperature against the output attenuator setting:  no 
strong relationship is seen.  That is, the apparent Tsys is not affected 
by the power level to the sampler.  This says that the compression 
equally affects PSum and PDif (since it is the ratio which generates 
Tsys). 


    I suggest this experiment be re-run, but this time with the *input* 
attenuators turned down by perhaps 10 dB.  Or, we repeat it but change 
in input attenuators over the 8 steps listed above, rather than the 
output attenuators.