>Richard:
>
>  Just done a quick skim of the CMX document and noticed incoming
>  tracks being designed for 65 ohm impedance (page 14). Wasn't
>  the backplane and CPMs etc built with 60 ohm traces
>  (from Section 2.5 Backplane spec)?


  We very much welcome guidance as to the best trace Zo
  to use for the 400 backplane processor traces on the CMX
  card.  So far I have used section 3.2 of the Backplane
  document for information about the potential range of Zo
  in the actual backplane traces.

  We are very concerned about making the processor card to
  CMX Base Function FPGA  signal path as clean as possible.
  Error free reception of these 400 signals every 6 nsec
  is one of the riskier parts of this overall design.

  My understanding is that the impedance along this full
  signal path will not be perfectly controlled, e.g. going
  through the backplane connectors.  To the extent that
  there is a range of impedances along the signal path then
  one design strategy is to make the final run on the CMX
  card near the high side of this range,  i.e. don't have
  a negative reflection coefficient coming onto the CMX.

  We have made a rather careful simulation of this signal
  path and I'm encouraged  (and almost a little surprised)
  that it looks like it will work with 6 nsec signals.
  As expected, a feature that we can easily see in the
  simulation is the importance of getting the back
  termination about right.

  Thank you again.  I'm very interested in everyone's
  knowledge and recommendations in this matter.



>Uli:
>
>  the JEM has definitely been designed for 60 Ohms,
>  and the series termination is made at 60 Ohms...
>
>Sam:
>
>  Thanks for catching that, Richard. The processor backplane was
>  indeed specified at 60 ohm single-ended impedance. In practice,
>  the manufacturer was given a 10 percent tolerance limit.
>  All backplane layers were tested to be within 10 percent,
>  some much better than that. That said, the backplane pcbs
>  came in multiple batches, so the exact impedances vary
>  from backplane to backplane.


  I'm happy to make these 60 Ohm traces.  That is an easier
  board to build than 65 Ohm.  The number in section 3.2 of
  the Backplane document that had caught my eye was 67.5 Ohm.

  If the range on the actual Backplane traces was as wide as
  to include 67.5 Ohm then one has to think carefully about
  the best compromise design for CMX traces.  That's where my
  65 Ohm number came from - avoiding a significant negative
  reflection coming onto the CMX.


>Sam:
>
>  Yes, section 3.2 is a bit confusing. To begin with, the measurements
>  here were performed on one of the prototype backplanes, which were
>  produced by a different company.
>
>  Second, there were two independent tests, one by myself in Stockholm
>  and the other by Weiming at RAL. As I wrote in the document,
>  my technique was to use a pulse generator with ~2 ns rise time
>  on one end of the trace, and a small trimpot on the far end that
>  I adjusted until the reflected signal was cancelled out.
>  With this technique I got results that were very similar
>  to the specifications.
>
>  Weiming tested reflection at the interface between a 50 ohm coaxial
>  cable and the unterminated backplane traces for different types
>  of lines. I am attaching a PDF of his 2003 talk. It was
>  a good measurement, and in my writeup I deferred to his results.
>  In retrospect, the error bars Weiming's measurements would have been
>  large enough that it's not clear that even the prototype trace
>  impedances were off by more than 10%.
>
>  On the last page Weiming also pointed out (correctly) that the 6 mil
>  traces on the backplane are much narrower than standard VME,
>  so current capacity is reduced. Fortunately this hasn't been a problem.
>  Fun fact:  one of the main reasons for choosing 60 ohms over 50 was
>  to increase the termination resistance (and thus reduce current draw).