[evlatests] power levels and system performance

[Ken Sowinski]

                    Signal Levelling Lessons
                          Nov 4, 2005

An attempt was made to explore the level setting parameter space
yesterday.  When I started only antenna 14 was available and I
chose to work with IF A because it provided the best response
of the three working IFs at that time.  3C273 was used at the
beginning, but the system decided to go to 3C345 while I ate 
lunch causing some confusion.  All the observations were at
X band.

There are two free parameters in the system, the power level into
the sampler as deduced by the RMS of the sampler values and the
'Gain' parameter in the deformatter.  The first is controlled by
attenuators in the downconverter, the second effectively chooses
which four bits in the output of the first FIR of the transition
module is passed to the second.  For values of these parameters
two parameters can be measured, backend sync detector voltage and 
correlation coefficient.  Backend ALC, total power, and transition 
module overflow flags were recorded as well.  The entire two dimensional
grid was traversed, but really only two cuts across the two dimensinoal
surface are of interest.  One with input power kept fixed and Gain
varied, another with Gain fixed and input power varying.

The major conclusions are:
1.  There is only little dependence of correlation coefficient on
sampler input power if Gain is allowed to vary for the best response.
2.  The best correlation coefficient response is broadly peaked and 
begins only after the  requantizer overflow flag comes on and and 
continues until the DAC ovefflow flag comes on.
3.  There is only little dependence of SD voltage on sampler input
power if Gain is allowed to vary for best response, but
4.  the best SD response is much more narrowly peaked and falls of
more sharply.  Higher power level to the sampler produces a broader
SD repsonse curve.
5.  We are under driving the T4 because the ALC does not begin to
level until after the requantizer overflows.

I have plots of correlation and coefficient and SD voltage as a
function of each parameter with the other held constant which illustrate
these conclusions.  Each plots SD volatage and correlation coefficient
against one of the free parameters.  Requantizer overflow, DAC overflow
and the point at which the ALC levels are all marked.  The first plots
these against downconverter attenuator setting with the Gain parameter
fixed at 85.  The second plots these against (10*log(Gain**2) - 41) with
the input to the sampler set to provide an RMS of 9.70.

The next step, if possible, is to provide more signal to the T4 by
reducing the pads on the deformatter output by no less than 3 dB. This 
will cause the ALC to level at a power level which optimiazes the 
correlation coefficient and will take better advantage of the T5 ALC to 
keep the SD volatage constant over a larger range of power.

As it is we cannot use the SD voltage to make a reliable system 
temperature correction to the correlated data over a power range of
mare than about 3-4 dB.