RE: Equalization
Dear colleagues,
Okay, so where are we? Let me review what I think I have heard so
far.
Yes, the idea of 10G Serial on installed MMF is interesting, but
there are a couple of holes in the proposal, including bad timing.
Here is the summary, followed by holes.
850 nm Serial solution:
-----------------------
- Will benefit from TIA FO 2.2 Encircled Flux and Restricted Mode
Launch work.
- Offset Launch jumper not required.
- Start with 385 MHz-Km bandwidth assumption on installed MMF.
- Add 6 dB equalization to support 100 meters operation, additional
9.54 dB to support 300 meters. Equalization has to be adaptive, in
the sense that impulse response will be different for each link. It
must overcome severe DMD in some cases. With Encircled Flux launch,
for a given link, the impulse response will not vary significantly
with time, so it can be assumed as time-invariant or very slowly
varying in time.
- Potentially, end up with a total of 3 PMDs that meet all 5
Objectives.
1310 nm Serial solution:
-----------------------
- Will benefit from EMB work done with 802.3z.
- Offset Launch jumper is required.
- Start with 500 MHz-Km bandwidth assumption on installed MMF.
- Add 4 dB equalization to support 100 meters operation, additional
9.54 dB to support 300 meters. Equalization has to be adaptive in
the sense that impulse response will be different for each link. It
must overcome severe DMD in some cases. With offset launch, for a
given link, the impulse response will not vary significantly with
time, so it can be assumed as time-invariant or very slowly varying
in time.
- Potentially, end up with a total of 2 PMDs that meet all 5
Objectives.
List of holes:
-------------
1. It hasn't been established that Encircled Flux over a randomly
selected fiber from installed base will ensure 385 MHz-Km bandwidth
(850 nm) with a high degree of statistical confidence. Gair's
suggestion of tagging an RML-compliance condition is one possible
solution. We need to know if that suggestion will be acceptable to
802.3ae end users and system integrators.
2. It hasn't been established that 10G equalization is feasible. By
feasible, I mean something like - a demonstrable solution before the
Working Group Ballot, capable of overcoming severe DMD, consuming
less than 3 watts, with a cost comparable to that of other
components, backed by technical presentations in September that
instill a high degree of confidence in the 802.3ae members.
3. Perhaps it is too late. There is a high likelihood that at the
September Interim, motions to adopt other PMDs that meet Objectives
1 and 2 will pass.
Hole 1 can be bypassed by adopting a 1310 nm Serial solution. Hole 2
can't be plugged until we have heard presentations from DSP experts
in September. Hole 3 is the most regrettable. I don't know how to
plug it. Jonathan, Walt, do you have any comments or suggestions?
Thanks,
Vipul