[evlatests] 3-bit update

[Michael Rupen]

We spent most of the day on 3-bit testing.

1- We confirmed that RCP from the antenna is being labeled LCP by the time
   the data reach the correlator.
   - Rich will go out tomorrow (Friday) morning and switch the A1/C1 cables
     into the 3-bit samplers to fix this.

2- Hichem & Mike sorted out a version confusion for the deformatters,
   so the monitor points now correctly report the rms from the 3-bit samplers.

3- We tested the Station Board software which fits a line to the
   wideband correlator (full baseband) output, and found that
   (1) the data to which that line is fit do not agree with those
     reported by the correlator's Real Time Data Display (RTDD), and
     by Todd's bandpass tool.
   (2) the fitting routine correctly fits those incorrect data and reports
     accurate parameters for that fit.
   Bruce is looking into the data problem.  The aim is to enable auto-setting
   of the gain-slope equalizers.

4- Barry showed Michael how to keep the attenuators steady within a 3-bit
   observing script.

5- The sampler for ea12 was returned to the array.
   - We now have fringes for all 3-bit baselines and IFs.  We didn't
     bother to set the 3-bit delays.

   - The 8-bit sampler for IF A now looks fine.

   - The 3-bit samplers for ea12 are a bit odd:
     * the input power required to achieve the desired rms (~1.7) of the
       state counts is by far the highest for A1 (-22 dBm, vs. next highest
       3-bit sampler is -33 dBm), while that required for C1 is slightly
       lower than for any other 3-bit sampler (-42 dBm, vs. -40 dBm for the
       next lowest).
     * the gain slope equalizer can be set to give a flat bandpass for all
       3-bit samplers on the array *except* ea12-A1.

6- Keith & Michael spent the afternoon testing whether the gain slope
   equalizers can flatten the bandpass at a range of "problematic"
   frequencies:
     C  band  5-7 GHz
     K  band 19-21 GHz
     Q  band 44-46 GHz
     Ka band 35-37 GHz, 38-40 GHz
   I will send out a detailed summary this evening, but in general
   we were able to flatten the bandpass for all samplers except ea12-A1,
   with gain slope equalizer settings between 3 and 12.  The *best* setting
   for ea12-A1 still left a 12-15dB slope across the 2 GHz band.

   - The input power required to achieve the desired rms from the 3-bit
     samplers (~1.7) ranged from -33 to -40 dBm (excluding ea12; see above),
     compared to the standard 8-bit setting of -30 dBm.  Keith has some
     theories which the detailed summary I'll send this evening may help
     sort out.
   - Based on this 15 dB attenuators will be installed before the 3-bit
     samplers, to allow us to have enough power going into the T304s.
     If this works that would allow us to remove one of the 15 dB amplifiers
     from the 3-bit sampler boards, which would make many of the engineers
     happy.

More details (including state-count histograms) later --

           Michael