[evlatests] 3bit test LL replicated

[Rick Perley]

    For this test, the T304 LCP signal was fed into both RCP and LCP 
sampler inputs. 

    If the noise matrix asymmetries are all due to the digital system, 
they should remain in this test. 

    And they do.  But there is one notable change -- 22RCP is not nearly 
so bad in this test as it was in the previous test  (where the RCP 
signal was replicated into RR and LL).  Perhaps somebody can explain 
this ...

    Here's the noise matrix:

          12          15             22           28
--------------------------------------------------------
12  |     X         .20            .25          .21
15  |    .17         X             .21          .18
22  |    .16        .15             X           .22
28  |    .18        .17             .16          X
------------------------------------------------------

The 'improvement' in 22R is clearly related to the power level.  Here 
are the power gains applied to the data:

Ant/IF        Power Gain
----------------------------------
12R                56
12L                42
15R                40
15L                38
22R                57          !!!!!!!!
22L                33
28R                39
28L                36
----------------------------------

-------- Original Message --------
Subject: 	3-bit test: *C1* fed to A1 and C1
Date: 	Fri, 10 Sep 2010 11:48:28 -0600 (MDT)
From: 	Michael Rupen <mrupen at aoc.nrao.edu>
To: 	Rick Perley <rperley at aoc.nrao.edu>, Vivek Dhawan 
<vdhawan at aoc.nrao.edu>
CC: 	Ken Sowinski <ksowinsk at aoc.nrao.edu>, Robert Long 
<rlong at aoc.nrao.edu>, Doug Gerrard <dgerrard at aoc.nrao.edu>
References: 	<Pine.LNX.4.64.1009101027280.29329 at scamper>



Hi folks --

  Doug & Rob moved the splitter, such that the signal from *C1* is now fed
to A1 and C1.  Ken adjusted the attenuators at each antenna to get the
optimal rms for one of the samplers, without over-driving the other sampler:
i.e., in each case he set the highest-rms sampler to give an rms of 1.7 (the
theoretically desired value for a Gaussian distribution of state counts).

 The measured power levels at the output of the downconverters and
corresponding rms values of the samplers were as follows:

ea12  initial   -30.3  0.75  1.12
      final     -25.1  1.27  1.69

ea15  initial   -29.3  1.23  1.23
      final     -25.1  1.27  1.65

ea22  initial   -30.3  1.07  1.28
      intermed. -25.1  1.62  1.82
      final     -27.3  1.38  1.59

ea28  initial   -30.7  1.00  1.24
      intermed. -22.5  1.68  1.95
      final     -26.8  1.34  1.61

  "initial"   --> as set automatically ("set-and-remember")
  "intermed." --> what happens if we set attenuators to get the weaker
                  sampler to have rms 1.7
  "final"     --> final setting used for the rest of the observation
                  (stronger sampler has rms 1.7)

After setting the attenuators to the "final" state given above, Ken
ajdusted the filter scalings & requantizer gains to give an rms of 2.7
after re-quantization to 4-bits.

The final attenautor settings & filter scalings & gains were all in place by
1714 (scan 13), and remained undisturbed for the following ~20 minutes
(through the end of the observation [scan 32]).

The attached figure shows the distribution of state counts as seen at the
Input FPGAs on the Station Boards (i.e., before subband filtering &
requantization) for the 3-bit antennas, after converging on the final
attenuator settings.  The RED (top, Path 0) subplots are for A1; the BLUE
(bottom, Path 1) subplots are for C1; and the four plots are laid out as:

        ea12    ea15
        ea22    ea28

The observing script is our standard: one subband pair (128 MHz) at C band,
centered at 7056 MHz. The first scan is 5 minutes on 3C286, to let the
samplers stabilize.  After that we have pairs of 1-minute scans {3C286,
off-source}.

Please DISCARD all data before the attenuators & filters are set up (at
1714, or scan 13).

The data are available at
  /home/scamper2/WIDAR0/C3tuned_3bit128MHz_001.55449.706552766205

Enjoy --

      Michael