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

[802.3_B400G] 答复: [802.3_B400G] BER Discussion



Hi Mark and All,

 

I understand that there are concerns for potential power/latency introduced by FEC. Hope this email could further clarify why we consider 10-14 is feasible. Taking the following three FEC approaches as examples,

 

1, For 100G per lane solution in B400GE based on E2E interleaved RS(544,514) FEC, in order to achieve 1E-14 in slide 13 of our contribution, our point is that we can reuse current 100G AUI specification from Annex 120G.1.1 without modifications, just define new optical link BER of 2.2E-4 for 800G/1.6TE instead of 2.4E-4 for 802.3bs/cu. The question is if this RAW BER improvement of optical link is a big challenge for module manufacturers? During 802.3bs discussion, we agreed two orders of magnitude margin required at ~1E-6 to validate the PAM4 based solution. This also helps module and system vendors to ship products based on sufficient margin at ~1E-6 to reach manufacturer’s yield requirement. So we believe 2.2E-4 should be an easy target for optical modules, which could lead to the 1E-14 post FEC BER. There is no change needed for the current RS(544,514) FEC and AUI interface specifications.

 

2, For segment by segment FEC approach, dedicate FEC protects each AUI segment and the optical link, similar to 802.3ct architecture. It uses RS(544,514) FEC for each AUI segment to improve the 1E-5 BER of 100Gb/s Lane or 1E-4/1E-5 BER of 200Gb/s lane to the range of 1/3 of the targeted 1E-13/-14 BER, which I don’t think is challenging for RS(544,514).  For optical link, similar methodology as mentioned item #1, using SC-FEC of 802.3ct as an example, in order to lower post FEC BER from 1E-13 to 1E-14, the pre FEC BER requires only a little improvement ? no more than 1% improvement with from 4.58E-3 to 4.54E-3 due to the steep waterfall curve of the SC-FEC ? which is already covered by manufacture margin.

 

3, The unclear and more concerning one is concatenated FEC. The exact technical solution and capability of 200G per lane optical link is still unknown. On slide 14 of our contribution, we listed two key aspects to be investigated. For AUI we could just let it evolve based on RS(544, 514) and possibly additional light BCH FEC depending on the link capability. The more challenging question is how to improve RAW BER of ~2E-3 on the optical link to ~1E-4 range by inner code so that the outer RS(544,514) code could ensure to meet the BER objective. This needs more work during task force period. We need to compare the additional area/power/latency of the inner code required to improve BER objective from 1E-13 to 1E-14. In our opinion it should not be challenging if we are looking at the ratified time around 2025 when more advanced process technology is available with consider the similar waterfall curve of inner FEC as in item #1 and 2.

 

It is not our purpose to propose a very challenging target. We would like the group to consider BER such as 1E-13, 5E-14 or 1E-14 etc, from more aspects, including latency, feasibility, power, area cost and different applications.

 

Thanks

Xinyuan

 

发件人: Mark Gustlin (mgustlin) [mailto:00000ca015ef9343-dmarc-request@xxxxxxxxxxxxxxxxx]
发送时间: 2021319 5:47
收件人: STDS-802-3-B400G@xxxxxxxxxxxxxxxxx
主题: Re: [802.3_B400G] BER Discussion

 

John,

 

One path forward in the study group for the BER objective is to adopt the same objective that we had in 802.3bs:

Support a BER of better than or equal to 10-13 at the MAC/PLS service interface (or the frame loss ratio equivalent)

 

With the intention to investigate if we can further improve the BER objective without a high cost (power/complexity) once we are into the details of the FEC options, modulation options etc. (in task force).

 

Thanks, Mark

 

 

 

From: John D'Ambrosia <jdambrosia@xxxxxxxxx>
Sent: Wednesday, March 17, 2021 1:48 PM
To: STDS-802-3-B400G@xxxxxxxxxxxxxxxxx
Subject: [802.3_B400G] BER Discussion

 

All,

This week’s Study Group included some very good technical discussion on BER.  In my opinion these discussions ventured into a number of different areas that included not just the BER discussion, but started to go into the FEC architecture, application needs, and discussion of all of the necessary analysis.

 

While I recognize this discussion needs to happen, we do need to remember the process.  It is not the role of the Study Group to define the actual future standard.  Instead we define the project that will work on the development of the actual standard.

 

Therefore, performing the full analysis to resolve the BER issue to everyone’s satisfaction is something that likely won’t happen until we are in Task Force.  We need to identify the Ethernet rate or rates to be considered, the physical layer specifications, the specific FEC architecture, etc.  Please keep this in mind.

 

Also, as a reminder, the Study Group has a 6 month life with one 6 month extension.  So time is clearly an issue. 

 

Therefore, I will note to everyone, we will need to identify objectives that the group can live with, knowing that objectives can always be exceeded or changed (assuming the necessary consensus is achieved), in order for the actual project to progress. 

 

Please keep this in mind, as we look to define this project and move forward. 

 

I welcome everyone’s feedback and would encourage the Study Group to leverage using the reflector.

 

Regards,

 

John D’Ambrosia

Chair, IEEE 802.3 Beyond 400 Gb/s Ethernet Study Group

 

 


To unsubscribe from the STDS-802-3-B400G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B400G&A=1


To unsubscribe from the STDS-802-3-B400G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B400G&A=1


To unsubscribe from the STDS-802-3-B400G list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-B400G&A=1