[802.3ae_Serial] From Serial PMD 26 June: ORL spec, LR/LW overload & link attenuation, 1550 sensitivity, Jitter, ...
Derivation and consistency of receiver stressed and unstressed sensitivity
(1550 nm)
----------------------------------------------------------------------------
--------
There was some discussion but we couldn't puzzle out whether the numbers
were right in the meeting. Those who understand it are asked to join the
email discussion. We'll discuss it again next week, having hopefully got to
an understanding via email.
ORL
---
The consensus was that the Optical Return Loss requirement was of value at
1550 nm to protect the transmitter from too much back reflection, and we
should add it to the standard as a cabling requirement. The value of -21 dB
used for the RIN test was worked out on the basis of three discrete
reflections at -26 dB near each the transmitter. We should double-check
this calculation and should use this value as the ORL spec at 1550 nm. This
compares with -24 dB for the same thing in the telecomms world so we are not
putting extra burden on installed inter-office cabling.
At 1310 nm, it was believed that an ORL requirement was superfluous by the
following argument:
The dominant reflection allowed is the receiver at -12 dB. If the
transmitter saw this without significant attenuation but with additional
reflection from e.g. connectors, it would have to be a very short link with
low attenuation in the forwards direction too, hence plenty of margin at the
receiver in case of extra reflection induced RIN as the transmitter sees
around -10 or -11 dB ORL.
On the other hand, there was a request for experience-based input to protect
ourselves.
Cabling experts please confirm or reject this analysis!
On a closely related subject, we should add an informative note in the
standard explaining why our connector reflection spec is -26 dB rather than
aligned with telecom's -27 dB. The reason is for backwards compatibility
with 1G Ethernet.
Receiver overload
-----------------
The concept of working out the Tx max. mean power from current likely Tx
min. mean power + 5 dB for tolerances was seen as acceptable; some
discussion of what "likely" or reasonable" minimum meant. 850 nm is a
special case, as eye safety limited. At 1310 nm, people did not want to
increase the overload requirement from +1 dBm; the question was whether
there is margin to decrease it. Mike Dudek and Piers would each try to draw
up linear programming charts to show what this tolerancing method meant in
practice.
1310 serial link attenuation
----------------------------
There was a brief discussion about whether the 7 dB attenuation we have was
overkill for a 10 km link (G.691 implies 5.5 dB). We may need to allocate
more to jitter penalty: we already have what is in effect a jitter penalty
in the model through DCD and the effective baud rate, but it is small at 0.2
to 0.4 dB. This led on to a long discussion about ...
Jitter
------
The discussion focussed on how much timing window the receiver needed. We
will continue the discussion next week: please join the call and find some
experimental evidence!
Eye mask for 850 nm
-------------------
This discussion was postponed to the following meeting.
Next meeting
------------
Is the last one before the comment deadline and the Portland meeting!
Agenda items to include:
850 nm mask proposals
Refining jitter DJ and RJ spec numbers
LR/LW Receiver overload
Derivation and consistency of receiver stressed and unstressed sensitivity
(1550 nm)
Usual coordinates:
15:15 GMT = 4:15 pm BST = 17:15 CET = 11:15 am EDT = 8:15 am PDT,
Tuesday
+1(816)650-0631 Access code 39209
Piers