[10GMMF] TP3: Request for suggested stressor levels
Jim and others,
Thank you for the call for input.
Prior to giving my input on NEW stressor levels for TP3, I would like to
clear up what I perceived to be confusion among some over why we are going
through a new exercise on TP3 stressors now.
It is not the case that concern about either a) finite equalizer penalties
or b) the January adoption of the option for center launch require
reconsideration of the TP3 IPRs adopted in Vancouver by >75% vote. Although
Sudeep's 12/04 sieve predated adoption of the joint launch and did not
consider finite DFE penalties, lingle_1_0305 (presented by Prof. Ralph in
Atlanta) showed that the stressors in the current draft lie between the 96th
and 99%tile of Gen67yy for a relevant range of finite tap DFE architectures.
The only reason to change them now is if the LRM Task Force has decided that
they are too tough to be equalized with acceptable cost or power consumption
by practical circuits.
1. One might argue for change based on the fact that 2 out of 3 of the
current stressors have PIE-D ~5.1dB, that being above the 4.55dB 99%tile
coverage point for Gen67yy. However, I suggest that the fact that the
FINITE DFE penalties for these are between the 96th and 99%tile range of
finite DFE coverage renders the PIE-D %tile largely irrelevant. However
arrived at, the stressors voted into the draft of the standard in Vancouver,
with >75% vote, do in fact adequately test the relevant DFE architectures
against a goal of 99% coverage, just touching 99.0%tile point for a 12+3 DFE
for the pre-cursor case.
Furthermore, I would argue that starting with a stressor PIE-D of ~5.1dB is
reasonable, since the 1998 OFS/GaTech fiber model, based on a large set of
measured fiber DMD data (balemarthy_1_0105), places the 99th %tile point at
5.2dB. The Corning model (abbott_1_0305) places it at 5.1dB. It is
explained in lingle_1_0305 that Gen67yy has an unrealistically weak
correlation between center and offset launch, leading to optimistic joint
launch coverage (see also abbott_2_0305).
2. The other reason one might argue for change is the number of commercial
voices in Atlanta stating that it may be acceptable for LRM to aim at a
95%tile coverage goal rather than a 99%tile goal. I personally do not have
the experience in 802.3 to properly judge the advisability of that.
However, making that choice and applying it properly to finite equalizer
coverage (NOT PIE-D coverage) will be the determining factor for picking new
stressors (if in fact we must). If LRM over 300m is too difficult to
accomplish with 99%tile coverage, then we should probably state in the
standard that we are aiming at 95% coverage of the installed base.
I also understand there was some discussion in Atlanta with the Chair about
whether a 99%tile (or 95th) coverage goal applies to a two-way link, rather
than one-way. In that case of 99%tile (or 95th) for a two-way link, then
the required one-way coverage becomes 99.5%tile (or 97.5%tile). That point
should be clarified by David before getting too far along with picking new
1) My first choice is to leave the TP3 pulses as they were adopted in
Vancouver by >75% vote.
2) My second choice, for 99%tile coverage, is to run John Ewen's model
aiming at DFE coverage between 96 and 99%tile for three stressor cases,
expecting the precursor penalties to be nearer 99%tile while postcursor and
quasi-symmetric penalties will be lower in the given range.
3) If we must reduce the coverage goal to 95%tile, and if David tells us to
enforce that on two-way link coverage, then my third choice would be to aim
at DFE coverage at the 95 to 97.5%tile range.
4) If we must reduce the coverage goal to 95%tile, and if David tells us to
enforce that on one-way link coverage, then my fourth choice would be to aim
at DFE coverage of 94 to 95%tile. It should get easier to hit a more narrow
range of penalties with the three IPR categories at those reduced %tile
Robert Lingle, Jr, Manager
Fiber Design and Transmission Simulation
OFS R&D, Atlanta, GA