PAM8 & FEC Options

Vipul Bhatt – Cisco

Sudeep Bhoja – Inphi

Arash Farhood – Cortina

Gary Nicholl – Cisco

IEEE P802.3bm 40 Gb/s and 100 Gb/s Fiber Optic Task Force, Nov 2012

2

Supporters

Brain Welch, Luxtera

Mark Nowell, Cisco

Beck Mason, JDSU

3

PAM8 Introduction

http://www.ieee802.org/3/100GNGOPTX/public/jul12/schell_01_0712_optx.pdf

10 GBaud

28 GBaud 32 Gbaud, 8 PAM electrical eye, using DAC

http://www.ieee802.org/3/bm/public/sep12/lewis_01_0912_optx.pdf

• PAM-8 measurements results have been presented at .bm

• EML is used as light source and external modulator

• In this presentation, we propose a mapping and FEC options for PAM8

Follow on to presentation from bhatt_01a_0912

4

Option 1: PAM8 Coded Modulation

000 001 010 011 100 101 110 111

• PAM8 is partitioned into 2 subsets, Blue & Green

• High rate code is applied on the 2 MSB’s.

.bj RS FEC for 2 out of 3 bits. Throughput 76Gb/s

Lower power than .bj FEC running at 100Gb/s

2 MSB’s gain 6dB in distance after set partitioning.

.bj FEC provides additional 5.7dB in addition to 6dB set

partitioning gain

• LSB is coded with a low rate strong hard decision FEC

11.2dB FEC code is applied at 38Gb/s

5

PAM8 Coded Modulation Example

PAM8

Mapping

Strong FEC

R = 0.68

38.16GBaud

• Combined rate: (2 * 514/528 + 0.6836) / 3 = 0.8769

• Ethernet Rate = 100/3 * 257/256 * 1/0.8769 = 38.16 GBaud

LSB

RS(528, 514)

802.3bj 2 MSBs

6

Option 1: Coded Modulation Sim Results

16 18 20 22 24 26 28 30 3210

-18

10-16

10-14

10-12

10-10

10-8

10-6

10-4

10-2

SNR

BE

R

PAM8 Coded Modulation

PAM8 Uncoded

PAM4 Uncoded

PAM4 with .bj RS FEC

• PAM8 coded modulation FEC delivers 11.2dB coding gain

11.2dB

7

Option 2: PAM8 with low latency

PAM8

Gray

Mapping

BCH(3018, 2646, t=31)

3 38.17GBaud

• Gray Mapping

• Total FEC latency is 115ns. BCH targets 2E-3 input BER

• Ethernet Rate = 3018 / 2646 * 100 * 257/256 / 3 = 38.17G

8

Option 2: PAM8 with BCH

20 22 24 26 28 30 3210

-18

10-16

10-14

10-12

10-10

10-8

10-6

10-4

10-2

Slicer SNR (dB)

BE

R

PAM8 Uncoded

PAM8 with t = 31 BCH

1E-15

• PAM8 with BCH FEC delivers 9.2dB coding gain

9

PAM8 Slicer SNR Summary

PAM8 Option1

(Coded Mod)

PAM8 Option 2

with BCH

PAM4 with .bj

RS

Baud rate 38.16G 38.17G 51.56G

FEC Target BERi

for 1e-15 BERo 9E-3 2E-3 2.2E-5

FEC Latency 600ns 115ns 100ns

Ideal Required

Slicer SNR 19.9dB 21.9dB 19.1dB

Relative SNR

penalty -0.8dB -2.8dB Ref

Relative Noise

BW penalty Ref Ref -1.27dB

Total Penalty -0.8dB -2.8dB -1.27dB

Relative Margin Ref Ref-2dB Ref-0.47dB

10

PAM Options Comparison

PAM8 Coded

Modulation

PAM8 BCH PAM4 .bj RS

Baud rate 38.16G 38.17G 51.56G

FEC Target BERi for

1e-15 BERo 9E-3 2E-3 2.2E-5

Latency Target <600ns <115ns <100ns

Coding Overhead 14.04% 14.04%

2.72%

Mapping / Coding gain 11.2dB 9.2dB 5.8dB

11

Aug, 2012

12

Simulation Parameters

Tx_RJ = 150fS

Tx_BW = 22GHz

ER = 6dB

Tx_SNR = 30dB

RIN = -144 dB/Hz

Resp = 0.85 A/W

NEP = 15 pA/sqrt(Hz)

Rx_BW = 18GHz

Rx_RJ = 150fS

OMA = 0dBm

PJ = 2ps @ 100MHz

13

PAM8 Simulation Results

-9 -8 -7 -6 -5 -4 -3 -2 -1 0 16

7

8

9

10

11

12

13

14

OMA [dBm]

Segm

ent

SN

R (

dB

)

________________________________________________________________________________________________________________________________________

Segment SNR is defined in Backup

14

-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1

10-4

10-3

10-2

10-1

TP3 OMA [dBm]

BE

R

PAM8 Simulation Results

Allocate 1dB MPI penalty allocation for -35dB RL (See farhood_01_1112_optx)

3dB Optical margin for Option 1 Coded Modulation FEC

1.5dB Optical margin for Option 2 BCH FEC

Option 1: Coded Modulation

Option 2: BCH

15

Summary

We have proposed two options for coding and mapping of 38G PAM8

PAM8 coded modulation with 600ns latency solution provides a robust optical link with 3dB margin

PAM8 and BCH with 115ns latency offers a lower latency solution with 1.5dB margin

We have a high degree of confidence in technical feasibility for -35dB Return Loss cable plant

16

Backup

17 17

SNR Channel Model

FEC

Encoder

Laser

Optical

Modulator Driver

PIN

TIA CDR FEC

Decoder

Data Encoded

Data Fiber

FEC

Encoder

Driver &

Optical

Modulator Encoded

Data

h(t) R |E-field |2 + +

h(-t) FEC

Decoder

PIN

N(t)

y(t)

y(t) _

E field,

Optical power = |E-field|2

= {0, 1, …, M-1}D

TIA output levels

= {-(M-1)/2, -(M-3)/2, …, -1/2,

1/2,…(M-1)/2} RD

TIA

CDR

Current levels

= {0, 1, …, M-1}RD

Fiber Data

Physical Channel

Equivalent AWGN Channel Model

- +

18 18

General Case SNR

N(t) is a white Gaussian random process which has both signal independent thermal noise

and signal dependent components due to laser RIN and shot noise. For optimum receiver

thresholds, the symbol error probability is:

DD

D

D

2

0

2102

2

@,

@,

,

1

)(1

1

2

)()(

bandwidth)Rx W ()102(

,

)1(

12

)12)(1(

6

1

2

2-M0,...,k),(/ R Q(k)

M

k

s

RIN

Thk

TIAoutputelectricalavg

TIAinputelectricalavg

opticalavg

kk

kPM

P

kQerfckP

WRkqRkN

M

P

MM

P

M

RP

R

A single sigma is not valid

Each individual level has its own noise variance

Define Segment SNR for each eye, Q(k)

-0.5 0 0.5-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

Time (UI)

Am

plit

ude

PAM4 @50G with BT4 25G BW

0

1

2

3

DR

DR

DR