error correction coding for passive optical networks · 2019. 3. 1. · poninnovation 2,005 2,010...

Error Correction Coding for Passive Optical Networks

Rainer Strobel

2019-02-27

1 OWHTC2019 2019-02-27 Intel Connected Home Division

Overview

1 Introduction

2 Error Correction in PON

3 Improved Receiver Architectures

4 Conclusion and Outlook


Passive Optical Networks

Fiber

ONU

OLT

CPE

Splitter ONU

DPU

CPECopper

PON Characteristics

• Fiber access network, fiber to the home (FTTH), but alsofiber to the building/distirbution point (FTTB, FTTdp)

• Point-to-multipoint (P2MP) link between optical linetermination (OLT) and multiple optical network units (ONU)

• No active components between OLT and ONU

• Typical reach 20 km, typical split factor 32

• Low cost electrical/optical components


PON Innovation

2,005 2,010 2,015 2,020100

101

102

NRZEDB?PAM4?

Year

Agg.rate/G

bit/s PON Standards

Modulation

• NRZ used instandard PONs

• EDB, PAM-4discussed forG.HSP

Multiplexing

• TDMA used instandard PONs

• TWDM usedin more recentPONs

Coding

• RS code used inlegacy PONs

• LDPC introduced inIEEE 802.3ca andconsidered for G.HSP


PON Innovation

2,005 2,010 2,015 2,020100

101

102

TDMA TWDMNRZ

Year

Agg.rate/G

bit/s PON Standards

Modulation



Multiplexing



Coding




PON Innovation

2,005 2,010 2,015 2,020100

101

102

TDMA TWDM

NRZ

Reed-Solomon LDPC

Year

Agg.rate/G

bit/s PON Standards

Modulation



Multiplexing



Coding




Overview

1 Introduction





ONU Receiver Block Diagram

CDR

PhaseError

LDPCDecoder

Decode

PLL

RXClock

LLRCalculation

APD TIA LA

PMD PMA PCS

PON Physical Layer

PCS: Physical Coding Sublayer (LDPC, scrambling, control channel)

PMA: Physical Medium Attachment (digital signal recovery)

PMD: Physical Medium Dependent (optoelectronic conversion)

LDPC Requirements (25G)

• Output bit error rate < 10−12 at 10−2 receive bit error rate

• Low latency ≈ 10µs for wireless front-haul


Binary Receiver Error Probabilities

21 22 23 24 25

10−11

10−9

10−7

10−5

10−3

10−1

Fiber length/km

Biterrorrate

Raw

Coded

0 0.5 1 1.5 2

Unit interval

0

0.5

1

1.5

2

2.5

3

3.5

Am

plitu

de

×10-5

0 1 2

PDF ×105

0

0.5

1

1.5

2

2.5

3

3.5×10-5

• Optical channel cannot bedescribed by a simple binarysymmetric channel (BSC) oradditive white Gaussiannoise (AWGN) channel

• Definition of codingrequirements with raw BERis not convenient


Receiver Optimization - Binary Receiver

0.8 1 1.2 1.4

·10−5

10−3

10−2

10−1

100

Threshold

Biterrorrate

pe(x = 0)

pe(x = 1)pe

py(y = 0) = py(y = 1)

Min. BER

Max. I(x, y)

0.8 1 1.2 1.4

·10−5

0

0.2

0.4

0.6

0.8

1

Threshold

I(x,y)

Threshold Control

• Minimum input BERpoint can be differentfrom optimal LDPCoperating point


Statistical Properties

Errors per Codeword

150 200

10−6

10−4

10−2

100

Errors per codeword

Probab

ility

25G PON

Poisson

• Distribution of errors percodeword influences LDPCperformance

Burst Errors

1 2 3 4 510−10

10−7

10−4

10−1

Number of consecutive errorsProbab

ility

25G PON

Uncorrelated

• Knowledge of burst errorscan be used for decoding


Binary Error Models

px(1) =12

px(0) =12

py(1) =1−pe(1)+pe(0)

2

px|y(1|1) = 1− pe(1)

px|y(0|0) = 1− pe(0)

px|y(0|1) = pe(0)

px|y(1|0) = pe(1)

py(0) =1−pe(0)+pe(1)

2

Memoryless Channel

x ∈ {0, 1}: transmitted symbol;y ∈ {0, 1}: received symbolpe(0), pe(1): Error probability of1 and 0 transmitted

p(01|11)

px(1)

py(0)

px(1)

py(1)

px(0)

py(1)

px(0)

py(0)p(00|01)

p(11|00)p(00|11)

p(11|11)

Binary Channel with Memory

Transition matrix T and statevector p = [pxy(0, 0),pxy(0, 1), pxy(1, 0), pxy(1, 1)]

T

such that pt+1 = Tpt.


Overview

1 Introduction





ONU Receiver Implementation

CDR

PhaseError

PLL

RXClock

LLRCalculation

APD TIA LA

Hard Decision Receiver

• Low cost components

• Low power consumption

• Lower FEC performance

ClockRecovery

PhaseError

ADC

PLL

RXClock

LLRCalculation

APD TIA

Soft Decision Receiver

• Analog-to-digital converterrequired

• Higher FEC performance

Binary Receiver/Soft Input

• Derive soft input for LDPCfrom CDR

• Compromise betweencomplexity and powerconsumption


Receiver Optimization - Soft Information

CDR State Informationt+ 12

time

CDRstate s[t]

tt− 2

y[t+1

2]

RXSignal

RXClock

t− 12t− 1t− 3

y[t]y[t−1

2]y[t−1]y[t−

3

2]y[t−2]

−5 0 50

0.2

0.4

LLR

Probab

ility Default

Shifted

• Derive LLR from CDR states

• Optimize sample points

Analog-to-Digital Converter

Amplitude

Zero

ProbDensity

One

−10 −5 0 50

0.2

0.4

LLR

Probab

ility

• Optimize quantizationintervals of ADC


Overview

1 Introduction





Conclusion and Outlook

Conclusion

• Coding is a key component for evolving fiber accesstechnologies

• Optimized channel coding requires

good understanding and modeling of the channelinnovative, cost-efficient transceiver designs

Outlook

• Coding in combination with improved modulation formats

• Digital signal processing and error correction coding

• Energy efficient designs for high throughput


IntroductionError Correction in PONImproved Receiver ArchitecturesConclusion and Outlook

error correction coding for passive optical networks · 2019. 3. 1. · poninnovation 2,005 2,010...

Documents