Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

[EFM] 10/9 Minutes:FEC Conf Call



FEC Conference Call
Hosted by EFMA
August 9, 2002, 10AM-11AM, PST

Attendees:

Eric Lynsky, UNH IOL
Jonathan Thatcher, Worldwide Packets
Lior Khermosh, Passave
Larry Rennie (FEC Group Chair), National Semiconductor

Now that the frame based FEC baseline has been selected for consideration, testing to resolve issues related to adding FEC is now the near term focus of the FEC group.  Besides the two tests identified already, Jonathan suggested additional FEC tests. The tests are:

a) CDR performance at high (10E-4) BER.
b) MPN vs distance for high (10E-4) BER .
c) (new) A test to determine the effectiveness of the FEC
d) (new) Test operation non-FEC PMD's with  FEC'd data.

Test plans and and test methodology are needed.  Larry will organize this test team which would consists of experts
in the field.  Eric (UNH) has agreed to do some initial low BER CDR testing in his UNH IOL optical lab right away. Spent some time discussing Jonathan's CDR lock time test idea that would compare the optical input signal (from laser turn on) to the recovered data.  This idea needs refinement wrt methodology and equipment needed.

We need help especially with MPN testing (methodology, equipment, etc.). None of the meeting attendees are knowledgeable enough about MPN and its characteristics to properly define this type of test.  All agreed we need some expert MPN help.

Jonathan suggested that a convincing FEC test to do would be to use real FEC'd data in the CDR and MPN tests.  This could
be done by using real packets and manually adding the FEC headers and parity bytes off line.  This data could then be used
in the test, gathered at the receiver and then FEC decoded off line to measure FEC improvement. Larry suggested that it
might be good enough to somehow establish the error distribution and then use FEC theory to prove FEC performance.

It would also be convincing to demonstrate operation of legacy equipment with FEC'd packets.  It was suggested that this
could be done by taking a small packet, adding the FEC overhead and the special FEC start and stop characters and then
continuously sending the packet through an 8B10B encoder and into a non-FEC legacy PMD/PMA/PCS Phy.

Next FEC conference call will be same time next week.  An announcement will be e-mail to the group.

Regards,

Larry