[evlatests] temperature-induced gain variation

[Vivek Dhawan]

For a linear system, Pdif is independent of Psum. Pdif vs Psum was
measured yesterday for all 8-bit signal paths. This was done with
tipping scans at K band, 27 ants * 4 basebands, ea11 not available.

Most of the 104 examples show mild compression, i.e. a 5~10% Pdif
reduction when Psum is doubled. Exceptions are ea14 and 16 with
gain expansion of 1-2%.  Worst are ea10C,D and ea20C,D, with 15~
20% compression.

This effect could be caused by an actual non-linearity, or by the
change of diode noise output with antenna elevation (via a change
of temperature in the rack modules with elevation, which we have
seen on the LO phase, before that was fixed by rendering the LO
insensitive to temperature. The temperature changes are probably
still present.)

I claim that the change of noise diode output with temperature is not
the sole cause, because R and L compression on an antenna often differ,
but share the same diode. Occasionally the 2 basebands from the same
polarization differ in compression as well.

Two possible locations in the signal chain are the T304's or the
samplers themselves.

We are running the 8-bit samplers so as to fill 4 or 5 bits, and the
expected compression should be very small (under 1%, right??)

I seem to recall the T304's are used below the 1% compression point.

So, could someone please dig up the chart of headroom vs. signal
location?

Rick Perley wrote:
>     The power from the noise diodes used to monitor the receiver gains 
> are known to be temperature sensitive -- this has been demonstrated by 
> monitoring their output over a 24-hour cycle with the antennas parked.