[Gb-ccb] Updated circuit spec

[Martin Shepherd]

I have updated my external hardware interfaces document to accomodate
recently discussed proposals for changes. Although the final design
will now be done by Randy, I wanted to make sure that we were starting
from the same point. The updated document can be found at:

  http://www.astro.caltech.edu/~mcs/GBT/ccb_external_hardware_interfaces.pdf

The changes, which mostly involve the proposed analog input circuit,
are as follows:

A. Following Randy's suggestion, whereas before I planned to power the
   AD8138 from dual +5V and -5V power supplies, I have now changed the
   circuit to power it from a single +5V supply. This changes a number
   of things:

    1. As Randy pointed out, it means that no clamping diodes are
       needed at the input to the ADC. This also has the benefit of
       greatly reducing the maximum current handling capacity needed
       from the series resistors at the AD8138 outputs.

    2. With the clamp diodes eliminated, each of the pairs of
       high-precision 20 and 30 ohm resistors at the AD8138 outputs,
       can be replaced by individual high-precision 49.9 ohm
       resistors.

    3. Since the removal of the clamping diodes reduces the shunt
       capacitance across the inputs to the AD9240, I have increased
       the values of the explicit shunt capacitors to keep the noise
       bandwidth the same.

    4. Since none of the rest of the analog electronics needs a -5V
       power supply, we can eliminate the -5V power-supply altogether.

    5. My original power-management circuit used a monitoring chip
       that required that one of its inputs be a negative supply
       voltage <= -5V (or it would shutdown the CCB). I have thus
       modified the power monitoring circuit to use a different
       version of the same chip, which doesn't require a -ve
       input. Unfortunately, whereas Newark previously listed this
       chip in their catalog, they now don't, and none of DigiKey,
       Allied, Arrow, or Mouser carry it.  Although Maxim appears to
       have a minimum order of 1000, they also have a link to request
       samples, and we only need 4, including spares, so this probably
       won't be a show-stopper.

B. I have changed the grounding scheme used by the ADC to comply with
   the recommendations in the ADC data-sheet, which are clarified in
   Analog Devices app-note, AN-280, "Mixed Signal Circuit
   Techniques". In the revised circuit, all of the ADC ground pins are
   connected together to the analog ground plane, and all power-supply
   pins, including the digital power-supply pin, are decoupled to
   analog ground. The digital ground plane is not used at all for the
   ADC circuit. Note that I described the rational for this in my
   email of Nov 25 2003, based on the contents of the above-mentioned
   app-note.

C. The range of allowable common-mode voltage levels for the
   differential inputs have been changed to accomodate both the change
   to single supply operation in the backend AD8138 stages, and to
   correct an error that I previously made.

   In fact the data-sheet claims slightly different numbers in two
   parts of the specifications table for single +5V supply operation,
   so I have adopted the more conservative limits of the two claims.
   The acceptible range of common-mode voltages at the input to the
   CCB is thus now, 1V to 3.2V.

D. Whereas I had changed section 2.1.2 to show the previously revised
   -2.5V to 2.5V differential input level requirements, the summary of
   input signal requirements in table 1.2 still showed the earlier -5V
   to +5V differential input range. I have updated the table
   accordingly. Randy's existing circuit complies with the correct
   values.

E. As suggested by John Ford, I have replaced the 1K feedback
   resistors of the AD8138 with a pair of 499 Ohm resistors. This
   allows both these resistors, and the 499 Ohm series resistors at
   the inputs to the AD8138 to be part of a single 6-element
   high-precision resistor array.

F. I added clamp diodes to the inputs of the AD8138.

G. I have rewritten the noise analysis section.

Martin