Further Clarification of FEC Performance over PAM4 links with Bit-multiplexing

Similar documents
Further Investigation of Bit Multiplexing in 400GbE PMA

100Gb/s Single-lane SERDES Discussion. Phil Sun, Credo Semiconductor IEEE New Ethernet Applications Ad Hoc May 24, 2017

Analysis on Feasibility to Support a 40km Objective in 50/200/400GbE. Xinyuan Wang, Yu Xu Huawei Technologies

Optical transmission feasibility for 400GbE extended reach PMD. Yoshiaki Sone NTT IEEE802.3 Industry Connections NG-ECDC Ad hoc, Whistler, May 2016

802.3bj FEC Overview and Status IEEE P802.3bm

50 Gb/s per lane MMF baseline proposals. P802.3cd, Whistler, BC 21 st May 2016 Jonathan King, Finisar Jonathan Ingham, FIT

CDAUI-8 Chip-to-Module (C2M) System Analysis #3. Ben Smith and Stephane Dallaire, Inphi Corporation IEEE 802.3bs, Bonita Springs, September 2015

Toward Convergence of FEC Interleaving Schemes for 400GE

802.3bj FEC Overview and Status. 400GbE PCS Baseline Proposal DRAFT. IEEE P802.3bs 400 Gb/s Ethernet Task Force

Summary of NRZ CDAUI proposals

64G Fibre Channel strawman update. 6 th Dec 2016, rv1 Jonathan King, Finisar

Further information on PAM4 error performance and power budget considerations

Investigation on Technical Feasibility of Stronger RS FEC for 400GbE

Ali Ghiasi. Jan 23, 2011 IEEE GNGOPTX Study Group Newport Beach

Ali Ghiasi. Nov 8, 2011 IEEE GNGOPTX Study Group Atlanta

System Evolution with 100G Serial IO

Achieving BER/FLR targets with clause 74 FEC. Phil Sun, Marvell Adee Ran, Intel Venugopal Balasubramonian, Marvell Zhenyu Liu, Marvell

A Way to Evaluate post-fec BER based on IBIS-AMI Model

Proposed reference equalizer change in Clause 124 (TDECQ/SECQ. methodologies).

Updated Considerations on 400Gb/s Ethernet SMF PMDs

Thoughts about adaptive transmitter FFE for 802.3ck Chip-to-Module. Adee Ran, Intel Phil Sun, Credo Adam Healey, Broadcom

Technical Feasibility of Single Wavelength 400GbE 2km &10km application

100G PSM4 & RS(528, 514, 7, 10) FEC. John Petrilla: Avago Technologies September 2012

The Case of the Closing Eyes: Is PAM the Answer? Is NRZ dead?

802.3bj FEC Overview and Status. PCS, FEC and PMA Sublayer Baseline Proposal DRAFT. IEEE P802.3ck

Impact of Clock Content on the CDR with Propose Resolution

PAM8 Baseline Proposal

100GBASE-FR2, -LR2 Baseline Proposal

CDAUI-8 Chip-to-Module (C2M) System Analysis. Stephane Dallaire and Ben Smith, September 2, 2015

An Approach To 25GbE SMF 10km Specification IEEE Plenary (Macau) Kohichi Tamura

Development of an oscilloscope based TDP metric

DataCom: Practical PAM4 Test Methods for Electrical CDAUI8/VSR-PAM4, Optical 400G-BASE LR8/FR8/DR4

SECQ Test Method and Calibration Improvements

Component BW requirement of 56Gbaud Modulations for 400GbE 2 & 10km PMD

Application Space of CAUI-4/ OIF-VSR and cppi-4

Improving the Performance of Advanced Modulation Scheme. Yoshiaki Sone NTT IEEE802.3bs 400 Gb/s Ethernet Task Force, San Antonio, Novenver 2014.

Development of an oscilloscope based TDP metric

100GBASE-SR4 Extinction Ratio Requirement. John Petrilla: Avago Technologies September 2013

50 Gb/s per lane MMF objectives. IEEE 50G & NGOATH Study Group January 2016, Atlanta, GA Jonathan King, Finisar

Comparison of NRZ, PR-2, and PR-4 signaling. Qasim Chaudry Adam Healey Greg Sheets

100G CWDM Link Model for DM DFB Lasers. John Petrilla: Avago Technologies May 2013

Measurements Results of GBd VCSEL Over OM3 with and without Equalization

Clause 74 FEC and MLD Interactions. Magesh Valliappan Broadcom Mark Gustlin - Cisco

PAM4 signals for 400 Gbps: acquisition for measurement and signal processing

FEC Codes for 400 Gbps 802.3bs. Sudeep Bhoja, Inphi Vasu Parthasarathy, Broadcom Zhongfeng Wang, Broadcom

Baseline Proposal for 200 Gb/s Ethernet 40 km SMF 200GBASE-ER4 in 802.3cn

Problems of high DFE coefficients

Draft Baseline Proposal for CDAUI-8 Chipto-Module (C2M) Electrical Interface (NRZ)

Need for FEC-protected chip-to-module CAUI-4 specification. Piers Dawe Mellanox Technologies

50GbE and NG 100GbE Logic Baseline Proposal

On Figure of Merit in PAM4 Optical Transmitter Evaluation, Particularly TDECQ

CAUI-4 Application Requirements

Comment #147, #169: Problems of high DFE coefficients

Proposal for 400GE Optical PMD for 2km SMF Objective based on 4 x 100G PAM4

100G SR4 Link Model Update & TDP. John Petrilla: Avago Technologies January 2013

100GBASE-DR2: A Baseline Proposal for the 100G 500m Two Lane Objective. Brian Welch (Luxtera)

40/100 GbE PCS/PMA Testing

Baseline proposal update

400GbE AMs and PAM4 test pattern characteristics

Update on FEC Proposal for 10GbE Backplane Ethernet. Andrey Belegolovy Andrey Ovchinnikov Ilango. Ganga Fulvio Spagna Luke Chang

Practical Receiver Equalization Tradeoffs Applicable to Next- Generation 28 Gb/s Links with db Loss Channels

Reducing input dynamic range of SOA-preamplifier for 100G-EPON upstream

FEC Selection for 25G/50G/100G EPON

Refining TDECQ. Piers Dawe Mellanox

Investigation of PAM-4/6/8 Signaling and FEC for 100 Gb/s Serial Transmission

Comparison of options for 40 Gb/s PMD for 10 km duplex SMF and recommendations

Brian Holden Kandou Bus, S.A. IEEE GE Study Group September 2, 2013 York, United Kingdom

40G SWDM4 MSA Technical Specifications Optical Specifications

CAUI-4 Chip to Chip and Chip to Module Applications

Thoughts on 25G cable/host configurations. Mike Dudek QLogic. 11/18/14 Presented to 25GE architecture ad hoc 11/19/14.

EEE ALERT signal for 100GBASE-KP4

PRE-QSFP-LR4L 100G QSFP 28 Dual Range Optical Transceiver, 10km. Product Features: General Product Description:

MR Interface Analysis including Chord Signaling Options

New Results on QAM-Based 1000BASE-T Transceiver

40G SWDM4 MSA Technical Specifications Optical Specifications

10GBASE-LRM Interoperability & Technical Feasibility Report

MIGRATION TO FULL DIGITAL CHANNEL LOADING ON A CABLE SYSTEM. Marc Ryba Motorola Broadband Communications Sector

400G-FR4 Technical Specification

Multi Core fibers and other fibers for the future.

100G MMF 20m & 100m Link Model Comparison. John Petrilla: Avago Technologies March 2013

Architectural Consideration for 100 Gb/s/lane Systems

More Insights of IEEE 802.3ck Baseline Reference Receivers

10G-BASE-T. Jaime E. Kardontchik Stefan Wurster Carlos Laber. Idaho - June

FEC Applications for 25Gb/s Serial Link Systems

Transmission Strategies for 10GBase-T over CAT- 6 Copper Wiring. IEEE Meeting November 2003

100G-FR and 100G-LR Technical Specifications

Architectural Considera1on for 100 Gb/s/lane Systems

Issues for fair comparison of PAM4 and DMT

PAM8 Gearbox issues Andre Szczepanek. PAM8 gearbox issues 1

Click to edit Master title style

COM Study for db Channels of CAUI-4 Chip-to-Chip Link

100G EDR and QSFP+ Cable Test Solutions

ATSC compliance and tuner design implications

Cost Effective High Split Ratios for EPON. Hal Roberts, Mike Rude, Jeff Solum July, 2001

CAUI-4 Chip to Chip Simulations

TP2 and TP3 Parameter Measurement Test Readiness

Transmitter Preemphasis: An Easier Path to 99% Coverage at 300m?

FEC Architectural Considerations

Draft 100G SR4 TxVEC - TDP Update. John Petrilla: Avago Technologies February 2014

Recommended Changes to Optical PMD Proposal

Transcription:

Further Clarification of FEC Performance over PAM4 links with Bit-multiplexing Xinyuan Wang-Huawei Ali Ghiasi- Ghiasi Quantum Tongtong Wang-Huawei

Background and Introduction KP4 FEC performance is influenced by PMA multiplexing scheme, error model and BER in physical link. The following contributions are presented in Pittsburgh: FEC Performance over PAM4 links with Bit-multiplexing http://www.ieee802.org/3/bs/public/15_05/wang_t_3bs_01_0515.pdf FEC performance with PAM4 on multi-part links http://www.ieee802.org/3/bs/public/15_05/anslow_3bs_03_0515.pdf In this contribution, following considerations are investigated to address FEC performance in 400GbE Burst error in optical links Performance on Bursty+Bursty link Error floor Issue Influence from worst FEC Lanes 1X400G FEC with Non-FOM bit mux will limit current and all future implementations with DFE tap >= 2 and/or MLSE! Page 2

Potential Source of Burst Error Infrequent pattern dependent event. DC blocking caps: low frequency cut off coupled with data wonder. Long data transition charging effect coupled with non-linear response of O/E devices. VCSEL slow turn off (see page 9) has similar error floor as seen commonly in 802.3bs SMF contributions as result of top 3 effects. http://www.ieee802.org/3/100gngoptx/public/mar12/plenary/ghiasi_02_0312_ng10 0GOPTX.pdf Page 3

DC Block Penalty with PRBS31 http://www.tek.com/dl/65w_26043_0_letter.pdf Page 4

Non Linear Response Coupled with Long Data Patterns Long tail associated with PRBS31 like pattern coupled with nonlinear response of EA can result in infrequent degradation period which may result in an error burst. Driving Pattern http://pdfserv.maximintegrated.com/en/an/an292.pdf http://optoelectronics.ece.ucsb.edu/sites/default/files/publications/shim04ptl.pdf Page 5

Don t Assume Optical Receiver Won t Have DFE or MLSE! Traditional NRZ optical link operates with open eye with simple slicer CRU where noise is dominated The 50G and 100G/lane PAM4 links may have significant transmitter and receiver impairments where simple CRU slicer is no longer viable. A CTLE receiver where high frequency is emphasized has limited benefit plus noise enhancement penalty An FFE receiver can better equalize an optical link including fiber dispersion but FFE also has noise enhancement penalty A DFE receiver is very effective to equalize bandwidth limited component of an optical link without noise enhancement at expense of burst error Implementing long DFE with PAM4 signaling is complex but we shouldn't rule out a short 2-3 taps DFE and an MLSE and assume optical links have absolutely no burst error! Page 6

Burst Error From DFE/MLSE Usage of PAM4 Optical link In wang_t_3bs_01a_0315 : From silicon vendors with 2X50G PAM4 Transceiver: DFE/MLSE is included in line (Optical) side interface. Page 7

KP4 FEC Performance on PAM4 Links Error propagation parameter a of DFE will significantly shift error floor of Non-FOM Bit Mux even from 0.75 to 0.5. Considering burst error from optical physical link, FEC performance by Non-FOM Bit Mux will be further degraded. Architecture with Non-FOM Bit Mux can t effective benefit from burst error correct capability of RS FEC. a=0.75 a=0.5 Page 8

Questions Remained in FEC Performance with Non-FOM Bit Mux Error floor@1e-16 equivalent is showed for Non-FOM Bit Mux in multi part link performance for NRZ signaling. How about error floor in PAM4 links? What is the impact? For PAM4 signaling link, even 0.16dB borrowed from KP4 FEC(3.2dB), the 2.9E-6 is still challenge to cover up 4 electrical link as illustrative in gustlin_3bs_02a_0515. anslow_3bs_02_1114 anslow_3bs_03_0515 Page 9

Error Floor Issue from FEC Performance In wang_t_3bs_01_0515, the following figure shows FEC performance for Random + Bursty link with a=0.75. Error floor@~1e-16 is shown in Non-FOM Bit mux scheme, starting from SNR=~12.2dB (BER=2.3E-5 on optical). Error floor@~1e-27 also exists for FOM bit mux, but much lower. Page 10

What Causes RS FEC Error Floor? In 802.3bj Project, cideciyan_02a_1111 for symbol mux in PMA to face burst error Performance of Non-FOM bit mux is poor than symbol mux as in this figure. FEC performance in Non-FOM bit mux is much degraded and it can not be significantly improved even with only lower BER/DFE impact of electrical link. Page 11

Error Floor Issue from System Perspective To improve system robustness and interoperating capability, sufficient BER floor for optical physical link is required to ~1E-6 as refer to stassar_3bs_01_0515, even when 3E-4 is enough from KP4 FEC correct capability perspective. Penalty of DFE in electrical link by Non-FOM Bit Mux invalidates the margin below BER 2.3E-5 from optical link Due to severe error floor from SNR=12.2dB aligning to BER =2.3E-5 in the above figure, the benefit of further lower BER in optical link is cancelled off by Non-FOM bit mux. This error floor with Post-BER at ~1E-16 can t provide sufficient margin for stable system operating, even assuming only random error from optical link. If some burst error in second part optical/electrical link are considered, the BER/MTTFPA will expect to fail the objective requirement of 802.3bs project. Page 12

Dilute Errors from Worst Lanes There is a speculation that averaging BER across multiple lanes can help add margin to the physical links with poorer performance in anslow_3bs_03_0515. This statement is based on the following assumption: To average the bad physical lanes, good lanes need to have better performance than the spec. For example, to compensate a corrupted physical link with BER > 2E- 4, it requires the other lanes are operating with BER < 2E-4 Requiring adjacent link have better BER than specification is additional level of constrain, if one to take advantage requires clear definition in the standard How many lanes with inferior BER or improved BER are allowed on each link? How to account for interaction between bad/good optical and bad/good electrical lanes which may not be constructive What happens when multiple lanes are working at BER limit Benefits of dilution from 1x400G FEC decrease by multiple corrupted lanes or in case of fewer optical lanes. Page 13

Further Analysis for diluting the errors from the worst lane How many inferior lanes operating at BER limit is legal? 1,2,3,4, FEC Lanes? If more than 4 Lanes, 1X400G and 4X100G FEC is same even from this proposal. What is exactly the worst lanes in 400GbE project? Is it CDAUI-16 electrical interface? The BER of CAUI-4 is 1E-15, much lower than Pre-BER level of KP4 FEC. Or CDAUI-8 electrical interface or 8X optical solution, errors from physical lanes already be split to TWO FEC lanes. For 4X optical solution, errors from physical lanes already be split to FOUR FEC lanes. Even from above error split perspective, diluting error is no better than FOM as it can t solve burst error. 100G Optical Lanes FOM 4X25G FEC Lanes To Each 100G FEC Page 14

FEC Performance Enhanced by Pre-interleave Pre-interleave option was presented in wang_t_3bs_01a_0115 without the complexity of wire crossing and with more consistent FEC performance Better breakout support without limiting the architecture to 1 tap DFE or not supporting MLSE Pre-interleave also averages BER across multiple physical lanes. Page 15

Conclusion 1X400G FEC with bit mux even after constraining the link it may not deliver the required BER objective as result of potential burst error Complex electrical-optical link BER interaction are difficult to isolate and may result in shipping products that do not robust interoperate 4x100G FEC with FOM bit mux Offer consistent FEC performance with the need to constrain electrical or optical link Can deliver 1E-15 Post-BER naturally Supports DFE/MLSE likely required for future CR/KR and potentially optical PMDs Why risk or limit the architecture when FOM offer burst protection and ease of breakout! Page 16

Thank you

BACKUP Page 18

Error Floor Issue from FEC Performance (Cont d) For further investigate the error floor and how to improve, we add the following two result for comparing. For relax BER of up to 4 electrical interface, we use 5E-7 as target BER, the Post-BER is ~1E-17. For relax error propagation by DFE with a=0.6 as an extreme case, the error floor is ~1E-18 to ~1E-19. Page 19

In Joint Slides wang_x_3bs_01a_0115 Page 20

In wang_t_3bs_01_0514 Page 21

Post-BER improved by FOM Bit Mux As in wang_t_3bs_01_0515, even in the following figure show FEC performance for Random + Bursty link with a=0.75. If based on BER=2E-4 in optical link to get Post BER=1E-13 objective, with the help of FOM Bit Mux, it can reach Post-BER=1E-15 to improve system robust. Page 22