[evlatests] Another Linearity test...

[Rick Perley]

    Michael ran an experiment yesterday which -- although not what I had 
expected -- is quite instructive on the linearity and response of the 
system. 
   
    The experiment was to observe a strong calibrator (3C147) at C-band, 
and vary the power level over a wide range in 1 dB steps.  The intention 
was to do this at the input to the T304, but in fact the variation was 
made to the output attenuators.  Hence, the conclusions below apply to 
the samplers and beyond. 

    The nominal level for the sampler inputs is -30 dBw (1 mW).  Michael 
changed the inputs, for the AC IFs only, from -38 through -26 dBw (0.15 
mW through 2.5 mW), in 1 dB steps.  Due to a 'granularity' problem, not 
all steps were observed, with some being ignored so the changes are 
occasionally by 2dB.  The full range was observed for all antennas. 
    The B/D IFs were left fixed.  This provided a good  check on overall 
performance. 

    Results:

    1) The steps in both PSum and PDif are easy to see.  If the system 
is linear the ratio (Tsys) will be constant.  We might expect trouble at 
each end of the input range as we 'starve' and 'saturate' the samplers 
in turn, although with an 8-bit sampler, we should not see much 
non-linearity, as each bit is about a 4dB step in power.  And indeed 
there is virtually no change in Tsys over this input power range, with 
the maximum excursion being about 10%, and the typical excursion much 
less. 

    2) We should expect a greater effect on the corrected visibilities, 
as the subbands are requantized to 4 bits.  And there is a larger 
effect.  'Corrected Visibilities' means the cross-products after having 
applied the switched power monitoring.  The results show that the 
corrected visibilities begin to fall off (become too small) for all 
antennas, starting about 5dB down from the highest power level -- i.e., 
at about -31 dBm.  This is actually lower than the nominal input, 
indicating that for some antennas, we are running a bit 'hot'.

    3) As expected, the BD side shows no change in visibility amplitude 
over the time period. 

    4) An extra 'spin-off' of the experiment is another measure of 
bandpass stability.  I generated bandpass functions every 2 minutes.   
On the 'BD' side, where there were no attenuator changes, nearly all 
antennas showed dispersions in the bandpasses at the level of 0.01% or 
less.  They are noise limited.  Unhappily, four antennas showed coherent 
variations in time:  ea02, ea08, ea13, and ea21.  The characteristics 
are the same for all four:  B and D show the same variations of 1 to 2% 
in amplitude over a 30 minute span, with a frequency period of about 60 
MHz. 
    For the AC pair, where the output attenuators are changing every 2 
minutes or so, the well-known variations of ~2% (pk-pk in amplitude), 
with a frequency period of about 30 MHz.

    The next test will be to change in input attenuators to the T304 
over their entire range, but let the output attenuators adjust to the 
desired output level for each step.  This will let us test the T304 
response over the range of power.   I believe Ken has set up the script, 
which will run shortly.