[evlatests] Better T304 attenuator settings

[Walter Brisken]

I wrote a program that sweeps the T304 attenuators.  First the RF 
attenuator is held fixed at 5 dB while the Ch1 output attenuator is swept 
over full range.  Then the reverse occurs.  During this time I monitor 
power as seen by the deformatter.  It turns out that for low values of the 
swept attenuator the curves for the two attenuators match quite well.  At 
high attenuation values they do not.  From the various measurements I 
conclude that the mixing portion of the T304 adds an RMS noise of 4 to 12 
(expressed in units of < (sampler value)^2 >).  For comparison, in setting 
the attenuators we have been aiming for 64 on this scale.  The RMS noise 
added in the T304 is attenuated by the output attenuator.  Thus for output 
attenuator settings of 10dB or more this added noise is insignificant. 
However, several of the output attenuator settings in the parameter 
database have been low enough for substantial T304-induced noise to get 
digitized.  I have derived a new set of attenuator values that I plan to 
test either this weekend if possible (in an EVLA-only subarray) or next 
week.  The values have not yet been applied to the database.

It is hoped that the new attenuator values will result in better overall 
sensitivity, by perhaps as much as 15 or 20% in extreme cases.

-Walter