mike and dan,
i have a couple concerns about adding the switch into the
drawing at all. whether you use a switch or move cables or
use a quad BERT or multiple single BERTs should be up to the
implementer. i think the drawing should show the most
straightforward and illustrative way to make the
measurement.
1) if you include it, you need it in three places:
between the attenuator and PMD (as mike suggests), between
the PMA and BERT input (as dan suggests), and between the
aggressor optical sources and lanes not under test.
2) it makes the switch feel mandatory or at least
suggested, rather than letting that be up to the
implementer's discretion.
would it be more... elegant, to use a numbering scheme to
show that the BER method is looking at one lane at a time?
e.g. the lane under test's connection is TP3<i { 0:3>
(use the "exists on" e symbol), while the aggressor
connections to the PMD could be TP3< k : 0:3, k != i>.
or something similar.
i also think bit error ratio and symbol error ratio should
have no box since they are measurements and not equipment.
cheers,
patrick