Study&of&a&Scheme&of&Multiple&Access&in& … · iii Abstract Multiple access based on orthogonal...

!!!!!

!Study&of&a&Scheme&of&Multiple&Access&in&

OFDM8based&Next&Generation&Optical&Access&Networks&

!!!

Patrícia&Santos&Fidalgo&de&Sousa&!!!!!

Thesis!to!obtain!the!Master!of!Science!Degree!in!Electrical&and&Computer&Engineering&

!!!!

Examination&Committee&!

Chairperson:!Prof.!Fernando!Duarte!Nunes!Supervisor:!Prof.!Adolfo!da!Visitação!Tregeira!Cartaxo!

Members!of!the!Committee:!Prof.!Maria!do!Carmo!Raposo!de!Medeiros!!!!!

November,&2013

i

Acknowledgements

I would like to thank my supervisor, Professor Adolfo Cartaxo, for being always available to

clarify all my questions and for providing access to all the materials and literature I needed to

develop this work.

I would like to thank my family who always supported me and understood when I couldn’t

be present. In particular, I would like to thank my parents and grandparents who helped me

with their love, attention, support and patience in all the good and bad moments of this journey.

I would like to thank also Dr. Tiago Alves and the Ph.D. students Eng. Filipe Carvalho and

Eng. Pedro Cruz at the Group on Optical Fiber Telecommunications Systems of IT-Lisboa, for

their availability to answer my questions as well as in keeping an enjoyable working environment.

I would like to thank all of those who started at colleagues but ended as close friends,

especially Yao, Hugo and Raquel, with whom great memories I will carry of the years attending

IST.

Last but not least, I would like to thank my older friends outside of the university which

always understood when I wasn’t available due to work, particularly in the last year, but kept

supporting whenever it was needed. Sara, Nuno, Jorge, Hugo, Maria Joao, Tiago and Bernardo,

thank you for your love and friendship and the great moments you provided me with when I

needed the most.

iii

Abstract

Multiple access based on orthogonal frequency division multiplexing (OFDM) signals has been

proposed to be used in the next generation of optical access networks (NG-OANs). The objective

of this dissertation is to select and study a scheme of orthogonal frequency division multiple

access (OFDMA) and evaluate its performance and assess the capacity of the selected OFDMA-

based OAN.

In this dissertation, a survey of the OFDMA-based OANs is presented. An adaptive subcar-

rier allocation (ASA)-OFDMA-PON scheme, where the OFDM symbols’ subcarriers are prepro-

cessed at the OLT to carry specific data to each user, is selected and studied. The particularity

of this scheme is that each ONU is able to recover its assigned information, with lower sampling

rates and computation cost, since the received signal is sampled at a lower rate than the trans-

mitter of the OLT. Furthermore, data reception security is ensured at the transmission, since

the ONUs can only demodulate the information which was assigned to them.

The operation of the studied ASA-OFDMA-PON is demonstrated using MATLAB-based

numerical simulation, developed by the author. The performance of the ASA-OFDMA-PON,

when transmitting OFDM signals at 10 Gb/s, is assessed for different M - QAM mappings,

number of users, capacities assigned to each user, PON reaches, electrical noise power spectral

density (PSD) levels and equalization at the receiver. The results show that, with the studied

ASA-OFDMA-PON, it is possible to assign the system capacity and transmit specific data to the

different users of the PON. For an electrical noise PSD level of 10−24 A2/Hz, network reaches

of 60, 50 and 35 km are achieved for PONs serving 16, 32 and 64 clients, respectively. The

increase of the electrical noise PSD level causes a remarkable degradation of the ASA-OFDM-

PON performance.

Keywords: Passive optical network, orthogonal frequency division multiplexing, multiple

access, next generation of optical access networks, subcarrier allocation

v

ResumoRecentemente, esquemas de acesso multiplo utilizando sinais com multiplexagem por divisao

ortogonal de frequencia (OFDM) foram propostos para as redes opticas da nova geracao (NG-

OANs). O objectivo desta dissertacao e identificar e estudar um esquema de acesso multiplo para

ser utilizado nas NG-OANs e avaliar o desempenho e capacidade destas redes quando utilizam

o esquema proposto.

Esta dissertacao compreende o levantamento das solucoes apresentadas para as redes opticas

de acesso recorrendo a sinais OFDM. E selecionado e estudado um esquema de acesso multiplo

com alocacao dinamica de portadoras de sinais OFDM para redes opticas passivas (ASA-

OFDMA-PON), em que as portadoras dos sımbolos OFDM sao pre-processadas no terminal

de linha optico (OLT) a fim de transportar informacao especıfica destinada a cada unidade de

rede optica (ONU). A particularidade do ASA-OFDMA-PON e a capacidade de cada ONU

receber a informacao que lhe e destinada utilizando ritmos de amostragem e complexidade com-

putacional reduzidos. Foi tambem que, uma vez que cada ONU apenas consegue desmodular a

informacao que lhe e destinada, e assegurada a seguranca e privacidade da transmissao.

E demonstrado o funcionamento do esquema escolhido utilizando simulacao numerica, desen-

volvida pela autora, em MATLAB. O desempenho da rede optica do ASA-OFDMA-PON para

o sentido de transmissao descendente quando transmitidos sinais OFDM a 10 Gb/s, e aferido

para diferentes mapeamentos M -QAM, numero de subportadoras, capacidades atribuıdas a

cada utilizador, alcances da PON, nıveis de densidade espectral de potencia (PSD) de ruıdo

electrico e igualacao do sinal no receptor. Os resultados mostram que, utilizando o esquema

ASA-OFDMA-PON, e possıvel alocar capacidade e transmitir informacao individualmente para

diferentes utilizadores da PON. Alcances de 60, 50 e 35 km foram atingidos para PONs a servir

16, 32 e 64 clientes, respectivamente. Estes resultados foram obtidos para os nıveis mais baixos

de PSD do ruıdo electrico considerados, 10−24 A2/Hz. O aumento da PSD do ruıdo electrico

provoca uma degradacao notoria no desempenho do sistema.

Palavras-chave: Redes opticas passivas, multiplexagem por divisao ortogonal na frequencia,

alocacao de portadoras, redes opticas da nova geracao

vii

Table of Contents

List of Figures xiii

List of Tables xvi

List of Symbols xvii

List of Acronyms xxi

1 Introduction 11.1 Scope of the work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 Orthogonal frequency division multiplexing . . . . . . . . . . . . . . . . . 11.2 Optical access networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.2.1 Passive optical networks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2.2 Next generation optical access networks . . . . . . . . . . . . . . . . . . . 51.2.3 OFDMA-based next generation passive optical networks . . . . . . . . . . 6

1.3 Objectives and structure of the dissertation . . . . . . . . . . . . . . . . . . . . . 121.4 Main contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2 OFDM fundamentals 152.1 Mathematical formulation of OFDM signals . . . . . . . . . . . . . . . . . . . . . 152.2 Channel estimation, guard interval and cyclic prefix . . . . . . . . . . . . . . . . 162.3 OFDM signal characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.4 Architecture of OFDM system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.4.1 Electric components of OFDM transmission system . . . . . . . . . . . . . 212.4.2 Optical components of OFDM system . . . . . . . . . . . . . . . . . . . . 23

2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3 Adaptive subcarrier allocation OFDMA-PON 273.1 ASA-OFDMA-PON scheme architecture for downstream transmission . . . . . . 283.2 Mathematical formulation of the ASA-OFDMA-PON scheme operation . . . . . 313.3 Characteristics of the PPM operation . . . . . . . . . . . . . . . . . . . . . . . . 33

3.3.1 Requirements for the subcarrier assignment . . . . . . . . . . . . . . . . . 333.3.2 Requirements for the sub-sampling and FFT operation . . . . . . . . . . . 343.3.3 Training sequence and channel estimation . . . . . . . . . . . . . . . . . . 35

3.4 Example of ASA-OFDMA-PON signal processing operation . . . . . . . . . . . . 363.5 ASA-OFDMA-PON advantages and impairments . . . . . . . . . . . . . . . . . . 393.6 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

viii TABLE OF CONTENTS

4 Performance analysis of the ASA-OFDMA-PON scheme 41

4.1 System parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

4.2 Subcarrier allocation schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4.3 Optimal modulation index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

4.4 Performance of the preprocessing matrix as channel characteristic estimator . . . 47

4.5 ASA-OFDMA-PON performance with all ONUs at the same distance from theOLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.5.1 Performance results when the system capacity is uniformely distributedby ONUs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

4.5.2 Performance results when the capacity is non-uniformely distributed byONUs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.6 Performance evaluation of the PPM with ONUs at different distances from theOLT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.6.1 Performance results when the capacity is uniformely distributed by ONUs 56

4.6.2 Performance results when the capacity is non-uniformely distributed byONUs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.7 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

5 Conclusions and future work 67

5.1 Final Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68

Appendix A OFDM system details 75

A.1 Electrical OFDM transceiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75

A.1.1 Constellation mapping and demapping . . . . . . . . . . . . . . . . . . . . 75

A.1.2 DAC and ADC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76

A.1.3 Up-converter and down-converter . . . . . . . . . . . . . . . . . . . . . . . 77

A.1.4 Equalizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

A.1.5 Electrical Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

A.2 Optical components of the transmission system . . . . . . . . . . . . . . . . . . . 79

A.2.1 Optical modulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

A.2.2 Optical power splitter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

A.2.3 Optical fiber . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81

A.2.4 Optical receiver . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

A.3 Optical DSB OFDM signals dispersive time delay spread computation . . . . . . 83

Appendix B Performance analysis 85

B.1 Error vector magnitude . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

B.2 BER . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86

Appendix C Subcarrier allocation schemes 89

C.1 Scheme 1 - System capacity uniformely distributed by the ONUs . . . . . . . . . 90

C.2 Scheme 2 - System capacity non uniformely distributed by the ONUs . . . . . . . 93

Appendix D Distributions of the ONU-OLT distance in the ASA-OFDMA-PONnetwork 97

D.1 ASA-OFDMA-PON with 16 ONUs . . . . . . . . . . . . . . . . . . . . . . . . . . 97



TABLE OF CONTENTS ix

Appendix E Additional Results 101E.1 ASA-OFDMA-PON performance when the ONUs have the same network reach . 101E.2 ASA-OFDMA-PON performance when the ONUs have different network reaches 104

x TABLE OF CONTENTS

xi

List of Figures

1.1 Architecture of passive optical network. . . . . . . . . . . . . . . . . . . . . . . . 41.2 OFDMA-PON architecture and upstream data flow. . . . . . . . . . . . . . . . . . . . 71.3 Different OFDM-based PON implementations for multi-user access. a) OFDMA with dif-

ferent users assigned for different subcarriers. b) OFDMA-TDMA hybrid with different

users assigned to different subcarriers at different time slots. c) OFDMA-TDMA-WDMA

different users assigned to different subcarriers at different timeslots on different wave-

lengths. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81.4 Conceptual architecture of a multi-band WDM-OFDMA-PON. i) illustrates the

multi-band OFDMA signal generated at the multi-band OFDMA transmitter andii) illustrates the sub-band 3 selected by the sub-band OFDMA receiver in ONU 1. 10

1.5 Adaptive subcarrier allocation scheme. i) illustrates the subcarrier allocationscheme generated at the OLT. ii) illustrates the subcarriers selected by ONU-1,iii) illustrates the subcarriers selected by ONU-2, iv) illustrates the subcarriersselected by ONU-N . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1 Electric components of an OFDM: a) transmitter and b) receiver . . . . . . . . . 222.2 Optical components of OFDM architecture. . . . . . . . . . . . . . . . . . . . . . 242.3 Spectra of 4-QAM modulated DSB OFDM signal: (a) after the electrical modulator, (b)

after the DC block and 0 km SSMF, (c) after the DC block and 50 km SSMF. . . . . . 242.4 Spectra of 16-QAM modulated DSB OFDM signal: (a) after the electrical modulator,

(b) after the DC block and 0 km SSMF, (c) after the DC block and 50 km SSMF. . . . 25

3.1 Architecture of the studied OFDMA-based adaptive subcarrier allocation scheme. 283.2 Concept of the signal preprocessing operation of the studied ASA-OFDMA-PON

scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.3 Illustration of the subcarrier assignment for ONUs with a different number of

subcarriers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.4 Spacing between samples (a) at the OLT at DAC input and (b) at the ADC

output of an ONU which is assigned 122 of the total bandwidth. . . . . . . . . . . 35

4.1 EVM as a function of the modulation index when transmitting a 16-QAM OFDMsignal to 1 ONU when d=20 km and Sc = 10−23 A2/Hz. Results for setup SI. . . 45

4.2 EVM as a function of the modulation index when transmitting a 16-QAM OFDMsignal to 16-ONUs when d=20 km and Sc = 10−23 A2/Hz. Results for setup SII. 45

4.3 EVM as a function of the modulation index when a 16 QAM OFDM signal istransmitted to 1 ONU, considering an electrical noise PSD of Sc = 10−23 A2/Hz,for different network reaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46

4.4 EVM as a function of the distance when transmitting a 16 QAM OFDM signalwhen Pout = 6dBm and Sc=10−24A2/Hz. . . . . . . . . . . . . . . . . . . . . . . . 48

xii LIST OF FIGURES

4.5 EVM as a function of the distance when transmitting 16-QAM OFDM signalsubcarriers over (a) 16-ONUs, (b) 32-ONUs and (c) 64-ONUs according to thesubcarrier allocation scheme 1 when all the ONUs are at the same distrance fromthe OLT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.6 EVM as a function of the distance for different M -QAM OFDM signals trans-mitted to (a) 16-ONUs, (b) 32-ONUs and (c) 64-ONUs ASA-OFDMA-PON, ac-cording to the subcarrier allocation scheme 1 when all the ONUs are at the samedistance from the OLT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50

4.7 EVM as a function of the distance when transmitting a 16-QAM OFDM signalsubcarriers following the subcarrier allocations scheme 2, for a network with 16-ONUs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.8 Constellations of the 16-QAM OFDM signal received in a ONU which is assigned16 subcarriers and is at 20 km from the OLT, for different electrical noise PSDlevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

4.9 EVM results as a function of the distance when transmitting a 16-QAM OFDMsignal to the 16-ONUs of the ASA-OFDMA-PON which are distributed accordingto D1 (a), D2 (b) and D3 (c). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

4.10 EVM results as a function of the distance when transmitting a 16-QAM OFDMsignal to the 16-ONUs of the ASA-OFDMA-PON which are distributed accordingto D4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

4.11 EVM as a function of the distance when transmitting a 4-QAM OFDM signal tothe 16-ONUs of the ASA-OFDMA-PON, which are distributed according to D1. 59

4.12 EVM as a function of the distance when transmitting a 16-QAM OFDM signalto the 16-ONUs of the ASA-OFDMA-PON, which are distributed according to D2. 60




4.16 EVM as a function of the distance when transmitting a 4-QAM OFDM signal tothe 16-ONUs of the ASA-OFDMA-PON, which are distributed according to D5. 64


A.1 Constellation mapping of 32-QAM using gray coding. . . . . . . . . . . . . . . . 76A.2 Spectrum of the signal sOFDM after the up-conversion is performed. . . . . . . . 78A.3 Spectrum of the signal sr(t) after the down-conversion is performed. . . . . . . . 78A.4 Transfer functions of the optical intensity and the optical field against the drive

voltage. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80

B.1 EVM concept. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85B.2 BER as a function of the EVM for different M -QAM constellations. . . . . . . . 87

C.1 Distribution of the OFDM subcarrier frequencies in the frequency spectrum. . . . 89C.2 Correspondence between the OFDM subcarriers indexes k and their position in

the frequency spectrum. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

E.1 EVM as a function of the distance for an 16-QAMOFDM signal being transmittedover 32-ONUs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102

E.2 EVM as a function of the distance for an 16-QAMOFDM signal being transmittedover 64-ONUs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

LIST OF FIGURES xiii

E.3 EVM as a function of the distance when transmitting a 16-QAM OFDM signal tothe 32-ONUs of the ASA-OFDMA-PON which are distributed according to (a)D1, (b) D2 and (c) D3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

E.4 EVM as a function of the distance when transmitting a 16-QAM OFDM signalto the 32-ONUs of the ASA-OFDMA-PON which are distributed according to D4.106

E.5 EVM as a function of the distance when transmitting a 16-QAM OFDM signal tothe 64-ONUs of the ASA-OFDMA-PON which are distributed according to (a)D1, (b) D2 and (c) D3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107






xiv LIST OF FIGURES

xv

List of Tables

2.1 Characterisation of several OFDM signals for a system capacity of 10 Gb/s. . . . 19

4.1 Splitter insertion losses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 424.2 Insertion losses considered in the downstream transmission. . . . . . . . . . . . . 424.3 Parameters used to obtain the optimal modulation indexes for the different ASA-

OFDMA-PON setups studied in this work. . . . . . . . . . . . . . . . . . . . . . 444.4 Optimal modulation indexes obtained for the different setups of the studied ASA-

OFDMA-PON. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 454.5 Maximum network reach of the studied ASA-OFDMA-PON when all the ONUs

are at the same distance from the OLT and the subcarrier allocation scheme 1 isused. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

4.6 Maximum network reach of the studied ASA-OFDMA-PON when all the ONUsare at the same distance from the OLT and the subcarrier allocation scheme 1 isused. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

4.7 Maximum network reach of the studied ASA-OFDMA-PON when all the ONUsare at the same distance from the OLT and the subcarrier allocation scheme 2 isused for PONs with 16, 32 and 64 ONUs. . . . . . . . . . . . . . . . . . . . . . . 53

A.1 Bit mapping for 4-QAM using Gray coding. . . . . . . . . . . . . . . . . . . . . . 76A.2 Bit mapping for 16-QAM using Gray coding. . . . . . . . . . . . . . . . . . . . . 76

B.1 EVM limits when BER is 10−3 and 10−12. . . . . . . . . . . . . . . . . . . . . . . 88

C.1 Subcarriers assignment scheme 1 for 16 ONUs. . . . . . . . . . . . . . . . . . . . 90C.2 Subcarriers assignment scheme 1 for 32 ONUs. . . . . . . . . . . . . . . . . . . . 91C.3 Subcarriers assignment scheme 1 for 64 ONUs. . . . . . . . . . . . . . . . . . . . 92C.4 Subcarriers assignment scheme 2 for 16 ONUs. . . . . . . . . . . . . . . . . . . . 93C.5 Subcarriers assignment scheme 2 for 32 ONUs. . . . . . . . . . . . . . . . . . . . 94C.6 Subcarriers assignment scheme 2 for 64 ONUs. . . . . . . . . . . . . . . . . . . . 95

D.1 ONUs distances to the OLT. Scenario D1: ONUs with long and short networkreaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

D.2 ONUs distances to the OLT. Scenario D4: Same as D1 but for different ONUs. . 98D.3 ONUs distances to the OLT. Scenario D2: Short network reaches for every ONU. 98D.4 ONUs distances to the OLT. Scenario D3: Long network reaches for every ONU. 98D.5 ONUs distances to the OLT. Scenario D5: ONUs assigned the same number of

subcarriers at different distances from the OLT. . . . . . . . . . . . . . . . . . . . 98D.6 ONUs distances to the OLT. Scenario D1: ONUs with long and short network

reaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98

xvi LIST OF TABLES

D.7 ONUs distances to the OLT. Scenario D4: Same as D1 but for different ONUs. . 99D.8 ONUs distances to the OLT. Scenario D2: Short network reaches for every ONU. 99D.9 ONUs distances to the OLT. Scenario D3: Long network reaches for every ONU. 99D.10 ONUs distances to the OLT. Scenario D5: ONUs assigned the same number of

subcarriers at different distances from the OLT. . . . . . . . . . . . . . . . . . . . 99D.11 ONUs distances to the OLT. Scenario D1: ONUs with long and short network

reaches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99D.12 ONUs distances to the OLT. Scenario D4: Same as D1 but for different ONUs. . 100D.13 ONUs distances to the OLT. Scenario D2: Short network reaches for every ONU. 100D.14 ONUs distances to the OLT. Scenario D3: Long network reaches for every ONU. 100D.15 ONUs distances to the OLT. Scenario D5: ONUs assigned the same number of

subcarriers at different distances from the OLT. . . . . . . . . . . . . . . . . . . . 100

xvii

List of Symbols

α fiber atenuation coefficientbtotal total number of bits transmitted in each OFDM symbolβ propagation constantβ0 zero frequency term of the Taylor series expansion of ββ1 inverse of group velocityβ2 group velocity dispersion parameterβ3 second order of group velocity dispersion parameter

c speed of light in vacuumck transmitted information symbol at the kth subcarrierc′k received information symbol at the kth subcarriercki ith information symbol at the kth subcarrier

d distance

η spectral efficiency

fRF RF carrier frequencyfk frequency of the subcarrier kth

fn,e noise figurefT end-to-end multi-band bandwidthφLD(t) phase of the source laser

kb Boltzmann constant

λ wavelengthλn nth wavelengthlog2M Number of bits transmitted per subcarrier

m integer multiplesmi modulation index

ν optical frequencyν0 central optical frequencyn integer multiples

ωLD angular frequency of the source laser

rm mth sample of the received signal

xviii LIST OF SYMBOLS

r(t) received signal in the time domain

s(t) continuous transmitted signalsOFDM (t) bandpass OFDM signalsTS received training sequencesk(t) waveform for the kth subcarrierscONUx subcarriers assigned to ONU xsm mth sample of transmitted OFDM signalsr(t) complex baseband signal at the downconverter outputs(t) OFDM signals vector of complex amplitudes at each ONU FFT output

ts effective OFDM symbol durationtd dispersive delay spread

B sub-band bandwidthBw OFDM signal bandwidth

C Nsc ×Nsc matrix

∆B fraction of the band B∆DAC spacing between samples of the signal at the DAC input∆ADC spacing between samples of the signal at the ADC output∆λ spectral width of the transmitted signal∆fSB inter-band separationDλ dispersion parameter of the optical fiber∆φ differential phase noise∆G frequency guard interval∆f subcarrier frequency spacing

ELD(t) optical field amplitude of the source laser diodeEs(t) optical signal at the MZM outputEr(t) received optical field

F Fourier transform

GE equalizer transfer function

HE channel frequency responseHfib(ν) optical fiber response

I Identity matrixIout photo-detected signalIl,k in-phase measured constellation symbol

Irefl,k In-phase ideal constellation symbol

L length of the fiberLs splitter insertion lossesLe excess losses

List of Symbols xix

N number of ONUsNsc number of subcarriersNc number of discrete samples zero-valuedNscONUx number of subcarriers assigned to ONU xNtotal total number of transmitted OFDM symbolsNI number of information symbolsNT number of training symbols

Ω equivalent baseband angular frequencyONUx ONU with index x

PCP fraction of the signal duration corresponding to the CPPout laser output powerΠ(t) pulse shaping function

Ql,k Quadrature measured constellation symbol

Qrefl,k Quadrature ideal constellation symbol

Rb OFDM signal bit rateRbONUx system capacity assigned to each userRbias bias resistanceRλ PIN responsivity

Ts Total OFDM symbol duration, ts +∆G

Tr room temperatureT Preprocessed symbols at the preprocessing module output

S vector of original subcarriers of each ONUST Training symbols

Sreceived complex-valued symbols carried by the NscONUx subcarriers as-signed to the ONUx.

SRDAC rate of the samples at the DAC inputSRADC rate of the samples at the ADC outputSc(f) PSD of electrical noiseSλ0 Dispersion Slope of the fibersT kx kth training symbol obtained at the FFT outputsTONUx

kth training symbol obtained at the FFT output at each ONU

VRMS root mean square voltageVDC DC bias voltage of the MZMVπ half wave switching voltage

Y signal at the IFFT outputY signal at the at the serial-to-parallel converter outputYTm mth sample of the training sequence at the IFFT outputYTmDAC mth sample of the training sequence at the DAC inputY TONUx training sequence at the ONUx ADC output

X Unknown Nsc ×Nsc matrix

xx LIST OF SYMBOLS

xxi

List of Acronyms

ADC Analog-to-Digital Converter

ASA-OFDMA-PON Adaptive Subcarrier Allocation Orthogonal Frequency Division Multi-ple Access Passive Optical Network

ATM Asynchronous Transfer Mode

APON Asynchronous Transfer Mode Passive Optical Network

AWG Arrayed Waveguide Grating

BER Bit Error Rate

BPON Broadband Passive Optical Network

CDMA Code Division Multiple Access

CO-OFDM Coherent Optical Orthogonal Frequency Division Multiplexing

CP Cyclic Prefix

CPE Common Phase Error

CW Continuous Wavelength

DAC Digital-to-Analog Converter

DC Direct Current

DDO-OFDM Direct Detection Optical Orthogonal Frequency Division Multiplexing

DFT Discrete Fourier Transform

DSB Double Side Band

DSP Digital Signal Processing

EPON Ethernet Passive Optical Network

EVM Error Vector Magnitude

xxii LIST OF ACRONYMS

E/O Electrical-to-Optical

FDM Frequency Division Multiplexing

FEC Forward Error Correction

FFT Fast Fourier Transform

FTTH Fiber To The Home

GEPON Gigabit Ethernet Passive Optical Network

GPON Gigabit Passive Optical Network

GVD Group Velocity Dispersion

ICI Inter Carrier Interference

IDFT Inverse Discrete Fourier Transform

IFFT Inverse Fast Fourier Transform

ISI Inter Symbol Interference

I/Q In-Phase/Quadrature

LD Laser Diode

LO Local Oscillator

LPF Low Pass Filter

MCM Multi Carrier Modulation

MZM Mach-Zehnder Modulator

NG-OAN Next Generation-Optical Access Network

NG-PON Next Generation-Passive Optical Network

OAN Optical Access Network

ODN Optical Distribution Network

OFDM Orthogonal Frequency Division Multiplexing

OFDMA Orthogonal Frequency Division Multiple Access

OLT Optical Line Terminal

ONU Optical Network Unit

OSNR Optical Signal to Noise Ratio

List of Acronyms xxiii

O/E Optical-to-Electrical

PAPR Peak to Average Power Ratio

PIN Positive-Intrinsic-Negative

PON Passive Optical Network

PSK Phase Shift Keying

PPM Pre-Processing Matrix

PSD Power Spectral Density

QAM Quadrature Amplitude Modulation

RF Radio Frequency

RMS Root-Mean-Square

SC Single Carrier

SCMA Subcarrier Division Multiple Access

SDOT Software Defined Optical Transmission

SER Symbol Error Rate

SMF Single-Mode Fiber

SNR Signal to Noise Ratio

SSB-OFDM Single Side Band Orthogonal Frequency Division Multiplexing

SSMF Standard Single-Mode Fiber

TS Training Symbols

TDM Time Division Multiplexing

TDMA Time Division Multiple Access

WDM Wavelength Division Multiplexing

WDMA Wavelength Division Multiple Access

xxiv LIST OF ACRONYMS

1

Chapter 1

Introduction

This chapter aims to present the scope of the work and provide a frame of the developed work.

Therefore, in section 1.1, the scope of the work is described along with an introduction to

optical (OFDM) systems. In section 1.2, optical access networks are presented. Passive optical

networks (PONs) are described in section 1.2.1 and the next generation (OFDMA)-based PONs

are presented in section 1.2.2. The objectives and organization of this dissertation are described

in section 1.3, and the main contributions of this work in section 1.4.

1.1 Scope of the work

The scope of this work is to study and select a multiple access scheme for OFDM-based next

generation optical access networks (NG-OANs). The performance and capacity of the selected

scheme is assessed in this work.

1.1.1 Orthogonal frequency division multiplexing

Orthogonal frequency division multiplexing (OFDM) has been the center of several studies re-

garding high-speed optical communications for some years to this part. Previously implemented

in radio frequency (RF) systems, OFDM has become of interest in the optical domain due to

the multi-media driven growth of internet traffic.

Many advances have been made for the use of single carrier (SC) modulation for optical

networks. In SC, as the name conveys, the information is modulated through a single carrier.

However, multi-carrier modulation (MCM) presents some interesting characteristics that

have spiked the debate on which modulation format is better for optical communications. MCM’s

main virtues comprehend a scalable spectrum partitioning from individual subcarriers to a sub-

band, providing enormous flexibility in system design compared to single-carrier transmission.

2 INTRODUCTION

The adaptation of pilot subcarriers simultaneously with the data carriers, enables rapid and

convenient ways for channel and phase estimation [1]. Belonging to MCM modulation class,

OFDM inherited these attributes.

However, OFDM presents some disadvantages, primarily, high peak to average power ratio

(PAPR) and sensitivity to frequency and phase noise. The advances in computation capability

improved the applications of digital signal processing (DSP) to optical transmission and fed

the interest of applying OFDM to optical systems. While the SC systems present a simple

architecture, OFDM includes drastic modifications, in which an electronic DSP module and

a digital to analog converter (DAC) are required for complex OFDM signal generation at the

transmitter’s end [1].

With the progress heading for highly reconfigurable networks at channel speed over 100 Gb/s

and SDOT, optical OFDM does bring some assets for next-generation optical networks, namely

in what concerns flexibility in the system, network design and high scalability over ultra fast

transmission line rate [1].

There are two types of optical OFDM systems, coherent optical (CO)-OFDM and direct

detection optical (DDO)-OFDM. In DDO-OFDM, the optical carrier is transmitted along with

the OFDM band so that direct detection can be performed at the receiver resorting to a single

photodiode to convert the optical field into the electrical domain. In CO-OFDM, since the optical

carrier is not transmitted along with the signal, coherent detection is required at the receiver side

in order to perform the optical to electrical (O/E) conversion. For this effect a local oscillator

is used. CO-OFDM performs better regarding optical signal to noise ratio (OSNR), difference

between the power of the transmitted signal and the noise generated by the optical amplifiers,

and spectral efficiency. Therefore, it is appropriated for long-haul transmission systems with

high bitrates [1], where the signal is more sensitive to disturbances. DDO-OFDM systems offer

a more cost-efficient and less complex solution which is advantageous for networks where is

desirable to reduce optical network unit (ONU)-side costs, as it is the case with the future PON

systems.

1.2 Optical access networks

The growing demand for the capacity of access networks as well as the increase of the fiber-to-

the-home (FTTH) subscribers, which is expected to reach 100 million by 2013 [2], justify the

necessity for an improvement of the optical access network (OAN) capacity.

Multiple access networks for fiber-optic networks arrived as a response to the need to increase

the potential transport capacity offered by PONs [3]. These networks are characterised by the

Optical access networks 3

connection of nodes through passive optical components so that no opto-electric (or vice versa)

conversion between the transmitter and the receiver’s end of the connection, is required.

Multiple access networks operate at much lower rates than the aggregate throughput of metro

or core networks and offer a very high capacity [4]. Nevertheless, these networks require multiple

access techniques, and in this work, OFDM-based multiple access next generation networks, are

investigated. Previous studies have, however, explored other multi access techniques based on

wavelength, subcarrier, time and coding.

Point-to-multipoint PONs, responsible for the majority of the FTTH deployments around

the globe, are expected to play an important role in future optical access [2].

In the last years, OFDM has been regarded as a modulation scheme that can be used to

implement these multiple access networks. OFDMA, where each OFDM subcarrier is considered

an independent bandwidth resource which can be assigned to different users [2] [5] [6], has been

studied as a multiple access technique for future PON-based multiple access, for its dynamic

bandwidth assignment and hybrid configurations aptitude.

OFDM promises, for the next generation optical access, specially in PON-based access, to

conciliate an efficient spectrum use and equalization, with resources for dynamic, multi-user

bandwidth access [2]. OFDMA is expected to provide dynamic bandwidth assignment and deal

with huge amounts of traffic [2]. OFDMA topology distinguishes itself from the other proposed

by having its backbone in DSP. The other topologies proposed have been: wavelength division

multiple access (WDMA), subcarrier division multiple access (SCMA), code division multiple

access (CDMA), time division multiple access (TDMA) and hybrid solutions composed of the

mentioned above [2] [3].

This section starts with a brief description of PONs. Afterwards, optical OFDMA principles,

advantages, impairments and its usage in PONs is discussed, followed by OFDMA based multiple

access schemes proposed for the next generation optical access networks (NG-OAN).

1.2.1 Passive optical networks

PONs have emerged as a promising candidate to respond to the steadily increasing capacity,

bandwidth and advanced network services demand, for residential and enterprise customers [7].

The architecture of a PON, presented in Fig. 1.1, is based on a physical point-to-multipoint

structure where there is an optical line terminal (OLT) in the central office, which is connected

and shared by multiple ONUs via a 1:N passive optical splitter [7]. There are only active

components in the central office and in the user premises. The connection elements between the

OLT and the ONUs are known as optical distribution network (ODN).

4 INTRODUCTION

Due to the PON point-to-multipoint structure, collisions occur in the upstream transmis-

sions. In order to avoid it, multiple access techniques are used.

ONUs can be located in homes, buildings, a curbside or even a cabinet [7]. When the ONUs

are located in buldings, curbsides or cabinets, it is usually necessary an additional network

distribution, which can be either physical or wireless, to connect the ONUs to the end users [7].

OLTCentral Office

SSMF

Feeder Fiber

1:NSplitter

ONUONUONUONUONU...

SSMF

Distribution Fiber

OLT : Optical line terminalONU : Optical network unitSSMF : Standard single mode fiber

Figure 1.1: Architecture of passive optical network.

Both ITU-T and IEEE have created standards for PONs. The first ITU standard was based

on asynchronous transfer mode (ATM) and has been referred as APON. These networks support

622.08 Mb/s for downstream and 155.52 Mb/s for upstream traffic [7].

Broadband PON (BPON) was deployed afterwards as an improvement of the APON. BPON

has transmission data rates equal to those of APON but included APON, Ethernet and video

transport.

To respond to the growing demand for higher bandwidth in access networks, ITU proposed

new standards with higher capacities than BPON. Gigabit PON (GPON) standard represents a

boost, compared to BPON, in both total bandwidth and bandwidth efficiency through the use

of larger, variable-length packets. The maximum GPON data rate is 2.488 Gb/s for downstream

transmission and 1.244 Gb/s for upstream transmission [19].

IEEE also developed a standard for PONs, but based on Ethernet (EPON). The EPON

standard supports a maximum of 1.25 Gb/s downstream/upstream traffic. Therefore, it is also

known as Gigabit EPON (GEPON). Since Ethernet has been used widely in local area networks,

GEPON became an attractive access technology [7].

The same wavelength plan is used by BPON, GPON and EPON, where the OLT downstream

data transmission wavelength is 1490 nm and video is broadcast at a 1550 nm wavelength,

whereas the upstream wavelength is 1310 nm.

The maximum transmission distance obtained by the mentioned standards is 20 km and

the number of users served by these PONs is, depending on the network power limitations, set

between 2 and 128 [7].


In the last few years, other standards, promising either longer transmission reaches or higher

transmission rates, have been presented. Comprehended in these new standards are: i) extended

reach GPON which increases the reach of GPON from 20 to 40 km and ii) 10G-EPON which

can provide up to 10 Gb/s transmission rates for both downstream and upstream traffic.

All these standards are based on time division multiplexing (TDM). As the name suggests,

in TDM-PONs, data distribution for the different users is done using specific time slots for

each user. In the downstream link, the ONU selects and extracts its preassigned data. In the

upstream link, different ONUs cannot transmit simultaneously. Each ONU transmits in its

appointed time slot. The line bit rate is shared by the ONUs and therefore, the maximum bit

rate per user depends on the splitter ratio. Due to their aggregate bandwidth and power budget

limitations, TDM-PONs are not expected to be able to handle the future network requirements.

Wavelength division multiplexing (WDM)-PONs were proposed later. In this architecture,

a wavelenght is attributed to each ONU that works at an individual bit-rate. WDM-PONs

provided a solution for the bit-rate capacity limitations that TDM-PONs presented, since they

enable individual ONU bit-rates up to 10 Gb/s. WDM-PONs also provide transparency to

several services.

Different types of WDM-PON architectures have been proposed. In the architecture broad-

cast and select, the entire band containing all wavelengths is broadcast by the OLT in the

downstream transmission and each ONU selects its own wavelength through optical filtering. In

upstream transmission, each ONU transmits its own wavelength and these are combined at the

passive optical splitter. In this architecture, either the ONUs have different configurations, or

tunable optical filters and transmitters are required.

A wavelength routing solution using arrayed waveguide gratings (AWG) is a WDM-PON

architecture where the AWG routes the different wavelengths to different ONUs. The AWG has

considerably lower costs and insertion losses than the power splitters used in the previous con-

figurations and can support more than 40 ONUs working at 10 Gb/s each. In this architecture,

while there are no wavelength selective receivers, each ONU has to either transmit at different

wavelengths, which either requires ONUs with different components or tunable transmitters.

Both presented WDM configurations are restricted from a cost efficiency standpoint since

they require expensive components to be implemented on the ONU-side of the network.

1.2.2 Next generation optical access networks

The next generation of multiple access networks are expected to achieve aggregate data rates

equal or higher than 40 Gb/s and 10 Gb/s for the downstream and upstream, respectively [4][6].

6 INTRODUCTION

Sustained per-user data rates of at least 1 Gb/s for up to 1000 ONUs and longer transmission

reach (> 20 km) have also been envisioned to be available for NG-OANs [6]. To satisfy this

requirements several multiple access alternative implementations have been proposed, including

TDMA-PON, WDM-PONs, OFDMA-PON and multiple hybrid solutions, built combining the

referred multiple access architectures.

The next generation of PONs (NG-PONs) emerged as a response to the continuous growth of

user-side bandwidth demand applications [2]. The NG-PONs are expected to provide dynamic

bandwidth assignment, and address the requirements for higher data rates, in both downstream

and upstream transmission, longer transmission reach and the increasing number of users.

Both TDM-PONs and WDM-PONs presented several constraints when considered to achieve

the NG-PON requirements. From a cost efficiency viewpoint it was necessary to protect the

legacy network investments even if a new technology was being introduced. Ideally, this implied

that new technologies would have to be deployed over TDM-based NG-PONs. However, to

update the existing configurations, (10GPON, 10G-EPON) expensive components and complex

scheduling algorithms, would be required [8].

Implementing WDM-based configurations would require fundamental changes in the ODN

of the PON, replacing passive optical components with WDM-MUXs or develop stable, cost-

efficient wavelength ONU-side technology. Moreover, WDM technology is not yet able to dynam-

ically allocate bandwidth with sub-λ granularity and enable bandwidth sharing among different

services [2].

As an alternative, OFDM based multiple access configurations, envisioned to respond to the

the demands of the NG-PONs with reduced costs, were proposed.

1.2.3 OFDMA-based next generation passive optical networks

Using OFDM signals, it is possible to consider the OFDM signal subcarriers as transparent,

finely granular resources to dynamically allocate the bandwidth to multiple users and services.

The basic principle of OFDMA-PON consists in dynamically assigning bandwidth resources

to different users and services. The generic configuration of an OFDMA-PON is depicted in Fig.

1.2.

For the downstream traffic, the OLT modulates the data in OFDM subcarriers according to

the bandwidth allocation defined for each ONU. The complete OFDMA frame which contains

the scheduled signal is transmitted through the optical channel. Each ONU demodulates the

OFDMA frame and selects its designated subcarriers in DSP.

In the upstream transmission, each ONU maps the data to the given subcarrier(s) and an


OLTOLT: Optical line terminalONU: Optical network unit

ONU 1

ONU 2

ONU 3

1:N Splitter

...

f

f

...

f

ONU 1

ONU 2

ONU 3

OFDMA-framef

f

Figure 1.2: OFDMA-PON architecture and upstream data flow.

OFDMA frame, organized following the allocation pre-decided by the OLT, is generated. The

OFDMA frame is then converted into optical OFDM symbols. The optical OFDM symbols,

incoming from multiple ONUs, are combined at the passive optical splitter, forming a single

OFDM frame, which is detected by a single photo-detector at the OLT receiver [9].

A differentiating feature of OFDMA-based PON is its reliance on electronic digital signal

processing [8] to overcome performance and cost demands [2]. Through OFDMA-PON, the

trend of software defined optical communications is extended to next-generation optical access.

The performance is improved due to the efficient algorithms designed to the PON environments,

and the cost-efficiency is ensured by the reuse of the legacy fiber, mature optics and silicon DSP

platform that can be mass produced [8]. Furthermore, OFDMA reliance in DSP conveys greater

flexibility to the system since the bandwidth assignment can be redistributed without requiring

changes in the network components [2].

The OFDMA-based schedule for both downstream and upstream transmissions is generated

at the OLT, therefore it is crucial to ensure accurate synchronization to enable a multiple access

configuration.

In the downstream link, due to the high stability of the fiber channel, synchronization can

be ensured through the broadcast of a common clock signal from the OLT to the ONUs [2], [10].

In the upstream transmission, the synchronization presents a more complex issue since the

OFDMA frame has passed through different paths when combined in the passive optical com-

biner. If synchronization is not established, signals arriving from different ONUs will be decor-

related which will prevent the OFDMA frame from being assembled correctly at the splitter and

the OLT will not be able to properly demodulate the received signal.

8 INTRODUCTION

Some solutions, as hybrid WDM-OFDMA transmission or coherent OLT-side detection have

been proposed to solve the synchronization problems in the OFDMA upstream transmission

[11], [8].

Multiple alternatives for optical OFDMA bandwidth assignment at the OLT have been

proposed. The simplest form consists of assigning different subcarriers from the same OFDM

band to different users. This form is presented in Fig.1.3 a), where the arrows illustrating

the subcarriers allocation are constant along time, varying only in frequency. It is possible

to dynamically allocate bandwidth by changing the number of subcarriers for a given user

depending on real time traffic demand [2].

TimeTDM distribution of OFDMA frame

FrequencySubcarrier distribution

λ1


λ2

TDM distribution of OFDMA frame

Time


TDM distribution of OFDMA frame

Time

λn

...





a) b)

c)

Figure 1.3: Different OFDM-based PON implementations for multi-user access. a) OFDMA with dif-ferent users assigned for different subcarriers. b) OFDMA-TDMA hybrid with different users assignedto different subcarriers at different time slots. c) OFDMA-TDMA-WDMA different users assigned todifferent subcarriers at different timeslots on different wavelengths.

In Fig. 1.3 b) and c), hybrid alternatives combining OFDMA, TDMA and WDM are pre-

sented. As opposed to Fig. 1.3 a), in these Figures, the arrows illustrating the subcarrier alloca-

tion, vary along time and frequency. The topology presented in Fig. 1.3 b) is a OFDMA+TDMA

implementation which consists on changing the subcarrier attribution to users along time, al-

lowing multiple users to access the same OFDM subcarrier in different time-slots.

The last implementation suggested is presented in Fig. 1.3 c) and it is a combination of

OFDM+TDM+WDM. This scheme consists of implementing the previous topology on each

possible WDM wavelength. This method requires colourless ONU-side optics and if there is a

dynamic wavelength assignment, tunable optical devices at the ONUs are required [2].

The choice between implementations lies on the requirements on bandwidth flexibility, the


distribution network complexity, desired aggregate capacity and cost-efficiency.

Despite its dynamic bandwidth allocation flexibility, in OFDMA-PONs the entire OFDMA

frame has to be processed by each ONU demading high digital capacity from the ONUs.

The first OFDMA-PON concept to be proposed, in [12], presented the first bidirectional

experimental demonstration of a 10 Gb/s OFDMA-PON in a WDM-OFDMA-PON architecture.

Due to its promising features to NG-OANs, OFDMA-PON has been studied by several groups

and the first terabit OFDMA-PON featuring a 90 km reach and support up to 800 ONUs with

1.25/10 Gb/s dedicated/peak data rates, was demonstrated [11].

At the present, the studies regarding OFDMA-based configurations for the next generation

optical access networks are branched into two specific paths, multi-band OFDMA and adaptive

subcarrier allocation [3].

Several studies have been conducted to achieve high speed, long reach transmission, and

terabit per second aggregate transport capacity [13] [6], through OFDMA-based architectures,

has already been demonstrated. In the next sections, two OFDMA-based multiple access con-

figurations, multi-band OFDMA and adaptive subcarrier allocation, are described.

1.2.3.1 Multi-band OFDMA

One of the proposed OFDMA-PON implementations is multi-band OFDMA. In this scheme, a

multi-band OFDM signal, composed by several sub-bands, is generated and transmitted in the

optical channel and its sub-bands are assigned and distributed to the different users according

to the ONU-side demand.

A generic multi-band architecture based on WDM-OFDMA-PON is presented in Fig. 1.4.

In the downstream transmission, a multi-band OFDMA signal is generated at the OLT

transmitter. A simplified example of a multi-band signal can be observed in Fig. 1.4 i), where

B is the sub-band bandwidth, ∆fSB is the inter-band separation and fT is the end-to-end

multi-band bandwidth.

Since this is a WDM-based implementation, multi-band signals are generated for each trans-

mission wavelength. These signals are combined using an AWG and the aggregated signal is

transmitted through the fiber. In the ODN, the signals are routed using another AWG into the

pairs of upstream/downstream wavelengths assigned for each set of ONUs. The signal from each

wavelength is splitted, using a 1:N passive optical splitter, to the N ONUs.

At the ONU-side, one or more sub-bands can be selected. Through reconfigurable tuning

and low pass filtering, each ONU can frequency-select its designated sub-band [10]. This is

illustrated in Fig. 1.4 ii), where only the sub-band 3 is selected by ONU 1. Therefore, the entire

10 INTRODUCTION

AWG

AWG

...Multi-band OFDMA Tx λ1

λn

λ2...

...

1:N splitterλ1

ONU 1

ONU 2

ONU N

...

λ11:N

ONU 1

ONU 2

ONU N

...

AWG

OLTUpstream

Downstream

λn

1 2 3 4ΔfSB ΔfSB ΔfSB

fT

B i)

3

B + ΔB

ii)

λn

Sub-band OFDMA DSP Tx

Sub-band selection Tx

AWG

Multi-band OFDMA Rx

DS

US

DS

DS

US

US

...

DS: DownstreamUS: Upstream

Sub-band OFDMA DSP Rx

Sub-band selection Rx

Figure 1.4: Conceptual architecture of a multi-band WDM-OFDMA-PON. i) illustrates themulti-band OFDMA signal generated at the multi-band OFDMA transmitter and ii) illustratesthe sub-band 3 selected by the sub-band OFDMA receiver in ONU 1.

OFDMA band, fT , does not need to be processed. The receiver from each ONU can operate at

the bandwidth B +∆B, where ∆B is a fraction of the band B.

For upstream transmission, the output of each ONU sub-band OFDMA transmitter is

mapped into its preassigned place in the multi-band OFDM RF spectrum [10] using a sub-

band selective transmitter. At the passive combiner the optical signal is passively combined

with the signals incoming from the other ONUs, generating a multi-band OFDMA signal on

each upstream wavelength.

The signal arriving from every wavelength is aggregated using an AWG and sent to the OLT

where it is routed to the corresponding wavelength multi-band receiver.

Multi-band OFDMA warrants each ONU the capacity to select and process lower-rate bands,

which reduces computation complexity, energy consumptions and power budget constraints. The

downside of this approach is the limited achievable peak per-ONU rate and the dependence of

an accurate synchronization in the upstream transmission to ensure the correct aggregation of

the bands arriving from each ONU into a single multi-band to be detected and processed by the

OLT [6].


1.2.3.2 Adaptive subcarrier allocation scheme

This scheme proposes, as the name suggests, the allocation of capacity to the users through the

OFDM signal subcarriers.

For the downstream transmission, as depicted in Fig. 1.5, an OFDMA subcarrier assignment

schedule, illustrated in Fig. 1.5 i), is defined in the OLT. The data is then mapped and modulated

into the subcarriers pre-assigned to each user and the OFDM signal is launched into the fiber.

The signal is routed to each ONU using a 1:N passive optical power splitter. The ONUs select

and only process their assigned subcarriers, illustrated in Fig. 1.5 ii)-iv), thus simplifying and

reducing the costs of the ONU-side components.

In the upstream transmission, each ONU modulates the information into its allocated sub-

carriers according to the assignment schedule. All the signals are combined at the 1:N passive

optical splitter and the combined signal is sent at the appropriate upstream wavelength to the

OLT where it is detected and demodulated.

f

... 1 2 3 4 5 6 7 8 9 10 11 ... N-1 N

OFDM symbol subcarriers

ONU 2ONU 1i)

Data allocation schedule

DSP

OLT

i)Electrical modulator

Optical signal

modulationFeeder Fiber

SSMF d km

Power Splitter

1:N

...

ii)

Optical signal detection

+O/E

conversion

ONU 1

ADC FFT

Electrical demodulator

....

f

1 2 3 4 5 6 7 8 9 10 11 ...N-1 N

OFDM symbol subcarriers assigned to ONU 1

Demap and

Decode

iii)

ONU N

d1 kmSSMF

d2 kmSSMF

dN kmSSMF


+O/E

conversion

ONU 2

ADC FFT


....

f

1 2 3 4 5 6 7 8 9 10 11 ...N-1 N

OFDM symbol subcarriers assigned to ONU 2

Demap and

Decode


+O/E

conversion

ONU N

ADC FFT


....

f

1 2 3 4 5 6 7 8 9 10 11 ...N-1 N

OFDM symbol subcarriers assigned to ONU N

Demap and

Decodeiv)

iv)

ii)

iii)

Figure 1.5: Adaptive subcarrier allocation scheme. i) illustrates the subcarrier allocation schemegenerated at the OLT. ii) illustrates the subcarriers selected by ONU-1, iii) illustrates the sub-carriers selected by ONU-2, iv) illustrates the subcarriers selected by ONU-N

Several subcarrier allocation algorithms have been proposed [14]. Since the subcarrier allo-

12 INTRODUCTION

cation is software defined, the subcarrier distribution can be adapted according to the desired

requirements for the system whether it is to maximize the spectral efficiency, to ensure data

security or cost efficiency. The fine granularity of the subcarriers, combined with efficient dis-

tribution algorithms, enable a highly flexible resources assignment which can be implemented

in the existing networks.

The main impairments of this OFDMA-PON dynamic bandwidth sharing implementation

lies in the precise synchronization required in both downstream and upstream transmissions.

In the downstream transmission, synchronization is necessary to so that each ONU recovers

its preassigned subcarriers when sampling the signal. In the upstream transmission, precise

synchronization is required to ensure that an OFDM signal, containing all the transmitter sub-

carriers, is formed when the OFDM signals containing the modulated subcarriers arriving from

different ONUs are passively combined.

The OFDMA-based multiple access scheme selected to be studied in this work is the adap-

tive subcarrier allocation (ASA)-OFDMA-PON. A subcarrier allocation scheme using signal

preprocessing was firstly proposed in [3]. In this article, it is indicated that the preprocessing

is performed based on channel characteristics division. The proposed scheme is experimentally

demonstrated using optical single side band (OSSB) signals transmitted over 25 km of opti-

cal fiber where 128 OFDM subcarriers, mapped using a 32-quadrature amplitude modulation

(QAM) format, are transmitted over with a capacity of 40 Gb/s to a PON with 32 ONUs.

The ASA-OFDMA-PON scheme studied in this work is also a subcarrier allocation scheme

using signal preprocessing. However, the preprocessing was developed considering that optical

double side band (ODSB) OFDM signals are transmitted. This is preferred from a cost reduction

point of view. The maximum reach of the network for the investigated ASA-OFDMA-PON is

assessed for a different number of clients being served, different M -QAM symbol mappings and

different subcarrier distribution schemes to the ONUs.

1.3 Objectives and structure of the dissertation

The main objective of this dissertation is to select a multiple access scheme for OFDM-based

NG-OANs, evaluate its performance and identify its advantages and its limitations.

For complexity reasons, and since different techniques can be used in the upstream and

downstream links in a multiple access OFDMA transmission system, it was only studied and

evaluated the operation of the ASA-OFDMA-PON scheme for the downstream transmission, in

this work.

The objectives for this dissertation are i) study and characterisation of OFDM signals, in

Main contributions 13

the time and frequency domain, ii) survey of OFDMA-based NG-OANs, iii) selection of multiple

access scheme for OFDM-based NG-OANs, iv) identification of the architecture of the selected

OFDMA-based NG-OAN and main impairments for downstream direction, v) demonstration of

operation of the selected scheme of OFDMA-based NG-OAN using MATLAB-based numerical

simulation developed by the author and vi) evaluation of performance and assessment of the

OFDMA-based NG-OAN.

This report is structured as follows.

In Chapter 2, the main characteristics of OFDM signal and system, and the mathematical

formulation of an OFDM signal are presented.

In Chapter 3, the architecture, the mathematical formulation, and operation of the ASA-

OFDMA-PON scheme studied in this work are detailed.

In Chapter 4, the performance evaluation and assessment of the OFDMA-based NG-OAN

capacity for the different M -QAM symbol mappings, number of clients and capacity assigned

to each user are presented.

In Chapter 5, a summary of the main conclusions of the study developed in this dissertation

and suggestions for future work on this subject are presented.

1.4 Main contributions

The main original contributions of the work developed in this dissertation are:

• Mathematical formulation of the signal preprocessing operation which enables that the

ONUs are able to select and receive the information carried by their assigned subcarriers,

at lower ADC sampling rates and with smaller sized fast Fourier transforms (FFTs). The

preprocessing was developed considering that ODSB OFDM signals are transmitted.

• Demonstration and study of the ASA-OFDMA-PON scheme operation, using a numerical

simulator developed by the author in MATLAB.

• Assessment of the performance, advantages and limitations of the optical network when

the studied ASA-OFDMA-PON scheme is used for the transmission of OFDM signals

with a total capacity of Rb = 10 Gb/s. The assessment is done for different network

configurations, namely:

– PON reaches;

– M -QAM symbol mappings (4, 16 and 32-QAM);

– Number of clients being served by the network;

14 INTRODUCTION

– Number of subcarriers assigned to each client;

– MZM modulation indexes

– Using, or not, signal equalization at the receiver.

15

Chapter 2

OFDM fundamentals

In this chapter, the OFDM system fundamentals are presented. The mathematical formulation

of OFDM signals is described in section 2.1. The insertion of guard interval and cyclic prefix

in an OFDM signal is presented in section 2.2. The characteristics of the OFDM signals, as

capacity, bandwidth and spectral efficiency, are analysed in section 2.3. In section 2.4, it is

depicted the architecture with the basic elements of an OFDM transmission system and it is

presented an overview of the electrical and optical components which are part of the transmission

system as well as the transmission process. Finally, some conclusions are drawn and presented

in section 2.5.

2.1 Mathematical formulation of OFDM signals

OFDM signals belong to the class of the MCM signals. The high rate information in MCM signals

is separated into lower rate streams using frequency division multiplexing (FDM). Separating

high rated streams of data into lower ones protects the signal from vulnerabilities, such as linear

distortions which increase with the symbol rate [1]. This is achieved through the insertion of

spectral guard bands which separate the subcarriers and prevent the overlap of frequency-domain

subcarriers. This usage of extra band to avoid the overlapping results in a reduction of spectral

efficiency.

In OFDM signals, the binary input data is mapped into orthogonal information subcarriers

using a simple modulation format as M - phase shift keying (PSK) or M -QAM. An OFDM

symbol is composed by Nsc subcarriers which correspond to the Nsc symbols modulated using

the desired format. The OFDM signal is composed by a sequence of OFDM symbols.

A continuous time representation of the OFDM signal, s(t), belonging to the class of MCM

16 OFDM FUNDAMENTALS

signals can be described using [1] :

s(t) =∞∑

i=−∞

Nsc∑

k=1

ckisk(t− its) (2.1)

where cki is the information symbol in the kth subcarrier belonging to the ith OFDM symbol,

Nsc is the number of subcarriers, ts is the OFDM symbol duration and sk(t) represents the

waveform of the kth subcarrier. For discrete tones, this waveform can be written as:

sk(t) = Π(t)ej2πfkt (2.2)

where fk is the frequency of the kth subcarrier and Π(t) is the pulse shaping function of the

OFDM symbol. For a rectangular pulse shaping function, Π(t) can be described as [1] :

Π(t) =

⎧

⎪

⎨

⎪

⎩

1, (0 < t ≤ ts)

0, (t ≤ 0, t > ts)(2.3)

MCM signals use non-overlapping signals, and require the channel spacing between carriers

to be a multiple of the symbol rate to ensure the cost efficiency of the filters and oscillators

used in both transmitter and receiver’s ends. The orthogonal configuration, characteristic of

OFDM signals, avoids the carrier overlap without compromising the spectral efficiency. The

orthogonality is ensured if the correlation between any two consecutive subcarriers, k and k+1,

is zero [1]. Null correlation is verified when the frequency spacing, fk+1−fk, is a multiple of the

inverse of the OFDM symbol duration. Mathematically, null correlation is when the equality:

fk+1 − fk = m1

ts,m ∈ N

+ (2.4)

is fullfilled. This expression translates the assignment of the OFDM signal subcarriers’ frequen-

cies.

2.2 Channel estimation, guard interval and cyclic prefix

The OFDM signal is built at the electrical OFDM transmitter, it is transmitted along an optical

link which adds distortion and time delay to the signal, and it is received and demodulated at

the electrical OFDM receiver.

Different techniques are used to estimate the optical channel and mitigate the degradation

effects that the OFDM signal will suffer in the channel.

Channel estimation, guard interval and cyclic prefix 17

In this work, training symbols (TS) are used to measure the transmission channel to improve

the demodulation at the receiver. The TS, that are used at the receiver exclusively to estimate

the transmission channel and are discarded afterwards, do not contain information symbols.

When considering this technique, the total number of transmitted OFDM symbols in a trans-

mission, Ntotal, comprehends both the NI information symbols and the NT training symbols:

Ntotal = NI +NT .

The OFDM signal transmitted in the optical channel experiences a time delay spread. This

spread leads to the situation where symbols cross the boundaries of the adjacent symbols creating

interference between them, known as inter-symbol interference (ISI). Additionally, due to this

spread, the signal OFDM waveform is incomplete, resulting in the loss of the orthogonality

between subcarriers and consequently, inter-carrier interference (ICI).

The dispersive delay spread of the channel, td, is given by:

td = Dλ · L ·∆λ [ps] (2.5)

where Dλ [ps/nm/km] is the dispersion parameter of the optical fiber, L [km] is the length of

the fiber and ∆λ [nm] is the spectral width of the transmitted signal.

ISI can be prevented using a guard interval, ∆G, inserted between symbols. In order to

ensure an ISI-free transmission, the guard interval should be as long as the dispersive spread,

td, satisfying the following condition:

td ≤ ∆G (2.6)

Combining the condition 2.6 with the expression 2.5, it is possible to calculate the duration

of ∆G necessary to ensure an ISI-free transmission. Therefore, the minimum duration of ∆G

has to be:

Dλ · L ·∆λ ≥ ∆G. (2.7)

A technique used in OFDM systems to tackle the loss of orthogonality between subcarriers,

resultant of the channel dispersion, is the insertion on the cyclic prefix (CP). The CP is a cyclic

extension of the symbol waveform into the guard interval ∆G. This extension contains a copy

of part of each OFDM symbol which is placed in the beginning of each symbol frame.

Calculating the percentage of signal duration to which corresponds the guard interval, a

perspective of amount of CP required and if it is of important concern. This percentage is

expressed as:

PCP [%] =∆G

Ts× 100%. (2.8)


where the total OFDM symbol duration Ts, containing the effective OFDM symbol duration,

ts, and the guard interval, is given by Ts = ts +∆G. The evaluation concerning the need to use

CP insertion in the OFDM system used in this work is presented in the next section where the

OFDM signal parameters are evaluated.

2.3 OFDM signal characteristics

Several parameters characterize an OFDM signal. The capacity of an OFDM system is usually

defined by its transmission data rate. The bit rate of an OFDM system is expressed as:

Rb =Nsc · log2M

Ts·

NI

Ntotal[bit/s] (2.9)

where log2M is the number of bits transmitted in each modulated subcarrier, NI is the number

of information symbols and Ntotal is the total number of OFDM symbols (information+training)

transmitted in an OFDM frame.

The total number of bits transmitted in each OFDM symbol is:

btotal = Nsc · log2M. (2.10)

where Nsc is the number of subcarriers in each OFDM symbol.

Considering ∆f = fk+1 − fk as the subcarrier frequency spacing between two consecutive

subcarriers, and combining with the Eq. 2.4, the bandwidth of an OFDM signal is given by:

Bw = Nsc∆f = Nscm

ts[Hz] (2.11)

with m ∈ N+. In order to use the narrowest bandwidth possible, the desirable value for m is 1.

Combining the Eqs. 2.9 and 2.11, assuming m = 1 and a sufficiently large number of

subcarriers, Nsc ≫ 2, the spectral efficiency of the OFDM signal, η, is expressed as:

η =Rb

Bw=

Nsc · log2Mts +∆G

·NI

Ntotal·tsNsc

=tsTs

·NI

Ntotal· log2M. (2.12)

As can be observed from Eq. 2.12, the spectral efficiency translates the relation between the

bit rate and the used bandwidth and it can be enhanced using higher-order QAM modulations.

In this work, the chosen bit rate, Rb, is 10 Gb/s, since it is one of the most common

values used nowadays in fiber optic transmission systems. Signals with different characteristics

were used in the study presented in this work. The bit mapping configurations used were 4,

OFDM signal characteristics 19

16 and 32-QAM. OFDM symbols with 128 and 256 subcarriers were considered. For channel

estimation purposes, 10 training symbols were used. In each OFDM frame, 128 OFDM symbols

are transmitted. Table 2.1 presents the results of some OFDM signals characteristics which were

referred in sections 2.2 and 2.3.

Table 2.1: Characterisation of several OFDM signals for a system capacity of 10 Gb/s.

M -QAMRb = 10 Gb/s

Nsc btotal Ts [ns] Bw [GHz] η [bit/s/Hz]

M = 4128 256 25.6 5 1.84256 512 51.2 5 1.84

M = 16128 512 51.2 2.5 3.69256 1024 102.4 2.5 3.69

M = 32128 640 64 2 4.61256 1280 128 2 4.61

For a fixed system capacity, higher-order QAM modulations occupy less bandwidth and have

a longer symbol period. Nevertheless, there is a compromise in the system performance when

the spectral efficiency is improved. This is expected due to the spacing between the mapped

symbols in the constellation. When the distance between symbols is shorter, the probability

of error when detecting the symbol is higher, thus decreasing the system performance. The

analysis of each M -QAM performance is done in section B.2.

As expected, the spectral efficiency of the signal increases for higher-order QAM modulations

due to the increase in the number of bits transmitted per modulated symbol. The results of the

spectral efficiency were calculated assuming that no CP is inserted and that 10 out of the 128

OFDM symbols transmitted are used as training symbols.

The minimum dispersive delay spread, td, is computed using the equations presented in

section 2.2 and Appendix A.3. It takes smaller values for higher-order QAM modulations. This

implies that the percentage of symbol duration corresponding to the guard interval, PCP , where

the CP is inserted decreases for higher-order QAM modulations. This percentage is, for the

signals with the characteristics presented in Tab. 2.1 and a fiber length of 50 km, always less

than 1%. When the fiber length increases, from Eq. 2.7 it is conclusive that the value of ∆G

will increase, and consequently the need for the CP insertion to avoid ISI will increase as well.

When no guard interval is inserted the spectral efficiency has higher values since the entire

OFDM symbol is used to transmit information symbols (∆G = 0 and Ts = ts).

In this work, the OFDM transmission system configurations considered have a maximum

reach of 50 km. Therefore, due to the low values that PCP presents for this fiber length it is not

used the CP insertion in this work.


2.4 Architecture of OFDM system

The architecture of an OFDM system is constituted by optical and electrical components. Ac-

cording to the desired system implementation requirements (cost-efficiency, capacity, reach),

there are several possible configurations to implement an optical OFDM transmission system.

The functional blocks which characterise an OFDM system are i) RF-OFDM transmitter, ii)

electrical-to-optical (E/O) converter, iii) optical channel, iv) optical-to-electrical (O/E) con-

verter and v) RF-OFDM receiver [15].

The large number of subcarriers required for the transmission channel to process each carrier

as a flat channel presents a challenge when implementing an OFDM system, since several syn-

chronized analogue oscillators are required, at both ends of the transmission link, to physically

implement all the non-interfering subcarriers [1] [2]. The solution encountered to tackle these

system limitations is to implement the modulation and demodulation of OFDM signals resort-

ing to the inverse discrete Fourier transform (IDFT) and discrete Fourier transform (DFT),

respectively. Due to the existence of an efficient IFFT/FFT algorithm, it is possible to reduce

the number of complex multiplications [1], and to generate and demodulate a large number of

orthogonal subcarriers without requiring more oscillators.

In order to describe the modulation/demodulation process employing the IFFT/FFT al-

gorithm, it is considered an OFDM signal, s(t), sampled at every time interval of TsNsc

. The

resulting OFDM discrete signal, sm, is a periodical function of fk with a duration of TsNsc

and is

given by:

sm =1

Nsc

Nsc∑

k=1

ckej2π· (k−1)(m−1)

Nsc = IFFTck (2.13)

where m ∈ [1, Nsc] and sm corresponds to an IFFT of the transmitted information symbol ck,

having as many points as the number of subcarriers. At the receiver side, this relation implies

that the received information symbol, c′k, is a Nsc-point DFT of the received signal and it is

given by:

c′k =Nsc∑

m=1

rme−j2π· (k−1)(m−1)Nsc = FFTrm (2.14)

The implementation of a DFT/IDFT based architecture requires, in the transmitter side,

the conversion of the discrete values of sm to the transmitted continuous analog signal, s(t).

Correspondingly, in order to demodulate the received signal, the continuous received signal,

r(t), has to be converted to a discrete signal, rm. These operations require the usage of a

digital-to-analog converters (DAC) and analog-to-digital converters (ADC) and are described in

Architecture of OFDM system 21

section 2.4.1.

In this work, as in most of the current OFDM transmission systems, a DFT/IDFT based

architecture is implemented, where the OFDM signal is processed in the digital domain using

FFTs and IFFTs.

In the following sections, the fundamental electrical and optical components of the OFDM

transmission system used in this work are described.

2.4.1 Electric components of OFDM transmission system

The electric components of the OFDM transmitter’s and receiver’s architecture are depicted in

Figs. 2.1 a) and b) respectively, and are described as follows.

On the transmitter’s side the electrical signal modulator (Fig.2.1 a)):

• Input serial data bits are converted into several parallel data arrays.

• These data arrays are mapped onto corresponding information symbols according to the

desired digital modulation scheme. There is a mapped symbol for each subcarrier.

• Digital time domain signal is obtained through a Nsc-point IFFT. The resulting signal,

sm, is given by equation 2.13.

• Afterwards, if required, the CP is introduced as described in section 2.2.

• The Nsc parallel arrays are then converted by the P/S block into a single serial vector

containing the complete OFDM signal.

• The complete digital OFDM is converted to analogue using a DAC, resulting in a contin-

uous time waveform. A low pass filter (LPF) is used to attenuate the aliasing products

originated from the DAC conversion.

• The complex baseband signal is then up-converted to a suitable RF bandpass signal using

an In-Phase/Quadrature (I/Q) modulator.

The signals’ up- and down-conversion are necessary to convert the baseband complex valued

signal, used in both the transmitter and receiver’s ends of the system, into the bandpass real val-

ued signal which is transmitted through the optical channel. These complex-to-real and real-to-

complex conversions can be performed using electrical or optical I/Q modulators/demodulators.

Electrical I/Q modulators/demodulators were chosen to implement the OFDM system in this

work, since they are less expensive than the optical ones.


The up-conversion process is mathematically expressed as:

sOFDM (t) = Res(t)ej2πfRF t = Res(t) cos(2πfRF t)− Ims(t) sin(2πfRF t) (2.15)

where the baseband complex-valued OFDM signal, s(t), is converted to a real-valued signal with

the spectrum centered at the intermediate RF carrier frequency, fRF . This bandpass signal is

defined in 2.15. The down-conversion process is identical but in reverse order. Both up- and

down-converters using I/Q modulators are depicted in Figs. 2.1 a) and b), respectively.

(a)

(b)

Figure 2.1: Electric components of an OFDM: a) transmitter and b) receiver

On the receiver’s side, the electrical signal demodulator (Fig.2.1 b)):

• The bandpass sOFDM (t) signal is down-converted into a complex baseband signal using

an I/Q demodulator.

• After the down-conversion, the complex baseband signal r(t) is converted into the digital

domain using an ADC.

• The signal is then split into Nsc streams of complex symbols using the serial-to-parallel

converter, and the CP, if introduced, is removed.

• The Nsc streams go through a Nsc-point FFT in order to obtain the corresponding complex

amplitudes. From these amplitudes, it is possible to recover the transmitted information

symbols and, accordingly to the mapping process, the binary input.

Architecture of OFDM system 23

• These steps are repeated for each OFDM symbol.

2.4.2 Optical components of OFDM system

The optical components which compose an optical OFDM system comprehend all the compo-

nents responsible for the E/O and the O/E conversions and the optical channel.

There are several possible configurations to perform the E/O and O/E conversions. In the

next sections, the optical elements which compose the system used in this work are presented.

For a detailed description of theses components, see section A.2.

2.4.2.1 E/O and O/E converters

There are several possible implementations for both E/O and O/E conversions. The E/O con-

version can be done resorting to an optical I/Q modulator. This is the direct-conversion ar-

chitecture, which employs two Mach-Zehnder modulators (MZMs) to convert the signal from

complex baseband to real bandpass signal and converts it to the optical domain to be launched

in the fiber.

However, as MZMs are expensive components, it is frequently implemented the approach

presented in Fig. 2.1 where the I/Q modulation is performed in the electrical domain. The

signal is then converted to the optical domain using only one MZM.

On the receiver’s side, according to how the O/E conversion is implemented, optical OFDM

can be classified as CO-OFDM or DDO-OFDM.

The direct-detection is the simplest implementation and it has a lower cost than the coherent-

detection. In DDO-OFDM, the signal detection can be done using a single photodetector and

it is commonly used the positive-intrinsic-negative (PIN) photodiode to this effect.

The coherent-detection has a more complex architecture since it resorts to a local oscillator

(LO) to provide an unmodulated light source, known as reference light, which is combined with

the received optical field before going through the photodetector in order to be converted into

an electrical field. As the reference light adds power to the signal intensity, the performance of

the detection increases. There is, however, a compromise in the implementation cost due to the

usage of a LO, which is an expensive component. There are other possible implementations for

the coherent detection, namely the usage of an optical hybrid where the I and Q components of

the signal are linearly recovered suppressing the common mode noise [1].

In this work, the OFDM signal is E/O converted using a single MZM and a laser diode (LD)

as the light source. The O/E conversion is implemented resorting to a single PIN followed by a

filter to remove the direct current (DC) component of the detected signal.


These implementations, which are depicted in Fig. 2.2, were chosen due to their simplicity

and cost-related considerations.

2.4.2.2 Optical Channel

The transmission medium used in this work is the single-mode fiber (SMF) since it is the most

commonly used type of fiber deployed in telecommunication networks. The optical fiber model

used in this work is described in a detailed manner in section A.2.3.

A double side band (DSB) OFDM signal in the different stages of a transmission system can

be observed in Fig.2.3 and Fig.2.4 for 4-QAM and 16-QAM subcarrier modulation, respectively.

In the results present in Figs.2.3 and 2.4, a carrier frequency of fRF = 4GHz, and the MZM

modulated at a quadrature bias point (a further analysis regarding the MZM modulation can

be observed in the Appendix A.2.1.) were considered. Linear transmission along the SMF, with

a dispersion parameter of Dλ = 17 ps/nm/km and optical carrier wavelength λ0 = 1552.52 nm,

was considered.

OFDMElectrical modulator

MZM

LD

PIN DC block

OFDMElectrical

demodulator

SMF

TransmitterOpticalChannel Receiver

Figure 2.2: Optical components of OFDM architecture.

The optical components which were previously described and characterize the OFDM system

implemented in this work are depicted in Fig. 2.2. For further details concerning the OFDM

system architecture considered in this work see the Appendix A.2.

ï10 ï5 0 5 1010

15

20

25

30

35

Frequency [GHz]

PSD

[dBW

/Hz]

(a)

ï10 ï5 0 5 10ï10

ï5

0

5

10

Frequency [GHz]

PSD

[dBW

/Hz]

(b)

ï10 ï5 0 5 10ï95ï90ï85ï80ï75ï70ï65

Frequency [GHz]

PSD

[dBW

/Hz]

(c)

Figure 2.3: Spectra of 4-QAM modulated DSB OFDM signal: (a) after the electrical modulator, (b)after the DC block and 0 km SSMF, (c) after the DC block and 50 km SSMF.

As can be observed in Figs. 2.3(b), 2.3(c) and Figs. 2.4(b), 2.4(c), which correspond to the

signal after the DC block for different distances, the signal is affected by a power degradation

Conclusions 25

ï10 ï5 0 5 1010

15

20

25

30

Frequency [GHz]

PSD

[dBW

/Hz]

(a)

ï10 ï5 0 5 10ï15ï10ï5

05

1015

Frequency [GHz]

PSD

[dBW

/Hz]

(b)

ï10 ï5 0 5 10ï85

ï80

ï75

ï70

ï65

Frequency [GHz]

PSD

[dBW

/Hz]

(c)

Figure 2.4: Spectra of 16-QAM modulated DSB OFDM signal: (a) after the electrical modulator, (b)after the DC block and 0 km SSMF, (c) after the DC block and 50 km SSMF.

when the distance is 50 km. The observed power fading is caused by the fiber dispersion and it is

only observed after the photodetection and for DSB signals. This effect is described by a periodic

function which depends of the frequency for which the power fading is evaluated, indicating that

each carrier will be affected in a different way [16]. This can be perceived through Figs. 2.3(c)

and 2.4(c) where the signal with the larger bandwidth, as confirmed in the results presented

in Table 2.1, has a larger number of frequencies available, hence suffering a greater impact of

the power fading. Therefore, while lower order M-QAM modulations present in theory a better

performance for the same bit error rate (BER) levels (as evaluated in section B.2 and 2.3), by

having a larger bandwidth when the same capacity is considered, these signals are more prone

to suffer power fading degradations.

2.5 Conclusions

In this chapter, OFDM signal fundamentals were presented. The mathematical formulations

of an OFDM signal were shown and the OFDM signal parameters and characteristics were

analysed. Both the electrical and optical elements, which constitute the backbone of the OFDM

system architecture used in this work, were described.

Recapitulating the decisions presented in this chapter regarding the system studied in this

work, an architecture where the signals are processed in the digital domain using FFTs and

IFFTs, is implemented. Electrical I/Q modulation is performed, followed by the E/O conversion

using a single MZM and a LD. Direct detection with a PIN was the implementation selected

for the O/E conversion. The DSB OFDM signal is transmitted through an SMF where training

symbols are used to perform the channel estimation.

Several configurations regarding the subcarrier modulation and signal characteristics will be

further evaluated in the next chapters for the multiple access architecture implemented in this

work.

27

Chapter 3

Adaptive subcarrier allocation OFDMA-PON

Low building cost, high operation efficiency, friendly structure and adaptability to fast growing

service demands are expected from the next-generation optical access networks [2].

In this work, a multiple access scheme referred to as adaptive subcarrier allocation (ASA)-

OFDMA-PON, with signal preprocessing and channel estimation for downstream transmission,

is investigated for OFDM-based NG-OANs. This scheme uses the OFDM signal subcarriers to

distribute the bandwidth resources to the users according to their requests.

The distinguished feature of this technique, which was proposed in [3], is that the ONUs

are able to select their destined OFDM subcarriers by sampling the received signal with lower

rates than the ones used to generate the optical signal. Since each user is assigned part of

the total bandwidth, ADCs with lower rates and FFTs with lesser points can be used at the

receivers, decreasing the implementation costs at the ONUs and increasing the signal processing

efficiency. To achieve this goal, the signal must be rearranged before the transmission so that

the complexity of the process is centered at the OLT, thus reducing the costs at the receiver side.

The processing of the transmitted OFDM signal at the OLT, so it acquires the characteristics

that enable that each ONU is able to select and process only its assigned subcarriers, will be

named for the remain of this work as preprocessing.

Furthermore, since the signal is preprocessed according to the subcarriers assignment and

the channel characteristics, which are different for each user, transmission security is ensured

at a physical layer. Due to the fact that the characteristics of the channel at each receiver are

unique, different users cannot demodulate the OFDM symbol subcarriers which are destined to

others.

In this chapter, the architecture and operation of the ASA-OFDMA-PON investigated in

this work is explained. The most complex part of this work is the development of the signal

preprocessing module which grants that the OFDM signal, at each ONU receiver, is correctly

28 ADAPTIVE SUBCARRIER ALLOCATION OFDMA-PON

selected and demodulated using lower rate ADCs and FFTs with less points.

This chapter is organized as follows. The architecture used to demonstrate the operation of

the proposed ASA-OFDMA-PON is described in section 3.1 and the mathematical formulation

of the preprocessing module of the ASA-OFDMA-PON scheme is presented in section 3.2. The

characteristics and requirements of the preprocessing module are presented in section 3.3. For

better understanding of the preprocessing module, a short example of its operation is provided

in section 3.4. The advantages and impairments of the ASA-OFDMA-PON are enumerated in

the section 3.5, and an overview of the chapter and its conclusions are presented in section 3.6.

3.1 ASA-OFDMA-PON scheme architecture for downstream trans-

mission

Before explaining how the signal preprocessing is achieved, the architecture, used to demonstrate

the operation of the proposed ASA-OFDMA-PON scheme and evaluate its performance, is

presented.

The preprocessing module is represented by a functional block whose operation is explained in

sections 3.2-3.4. The downstream transmission architecture of the ASA-OFDMA-PON analysed

in this work is depicted in Fig. 3.1.

Figure 3.1: Architecture of the studied OFDMA-based adaptive subcarrier allocation scheme.

In this scheme, the preprocessed OFDM signal is built at the OLT transmitter as follows:

ASA-OFDMA-PON scheme architecture for downstream transmission 29

• The input data bitstream is split into Nsc parallel bitstreams, one for each subcarrier of

the transmitted OFDM symbols, using a serial-to-parallel converter.

• Each parallel bitstream is mapped using a constellation mapping technique as M-QAM.

In this work, 4, 16 and 32-QAM constellation mapping techniques are analysed. The Nsc

mapped symbols at the mapper output are represented by S.

• The Nsc parallel mapped symbols are preprocessed as described in sections 3.2 and 3.3.

• The Nsc preprocessed symbols at the preprocessing module output, represented by T

modulate the OFDM symbol subcarriers using a Nsc-IFFT, which is followed by a parallel-

to-serial converter, in order to obtain a serial stream of samples. This operation is repeated

for every OFDM symbol of the OFDM signal. The signal at the IFFT output is represented

by Y .

• The digital signal is converted into an analog waveform using a DAC. The rate of the

samples at the DAC input is SRDAC .

• A low-pass filter (LPF), consisting of an ideal rectangular filter with bandwidth equal to

the OFDM signal bandwidth, is used to suppress the aliasing products generated by the

DAC operation.

• The signal is up-converted to a real-valued bandpass signal using an oscillator which up-

converts the baseband signal to a carrier frequency, fRF . In this work, a carrier frequency

of 4 GHz was used for the signal transmission.

• A MZM, biased at the quadrature bias point with a continuous wave (CW) laser diode

(LD) are used to perform the E/O conversion. For the LD output power values of 4 and

2 mW, are considered. With a CW laser output power of 4 mW (6 dBm) and 2 mW (3

dBm) and a MZM biased at a quadrature bias point with low modulation index values,

the MZM output power is 2 mW (3 dBm) and 1 mW (0 dBm), respectively.

For the transmission channel, the following characteristics are considered:

• SSMF is used as the optical transmission medium. The model describing the fiber trans-

mission is presented in section A.2.3. The optical network performance is analysed for

different fiber lengths. The same properties are considered for the feeder and distribution

fibers.


• Circulators with insertion losses of 0.7 dB are considered to route the signal to the upstream

link. Two circulators are considered, one to route the signal at the ONU and one at the

OLT.

• A passive optical 1 : N optical splitter is used to separate the signal to the different ONUs.

16, 32 and 64 ONUs are considered for the PON studied in this work. The properties of

the optical splitter and the expression used to compute the insertion losses of the splitter

are presented in section A.2.2.

At the optical OFDM receiver at each ONU, the bits are retrieved from the OFDM signal

by performing the following steps:

• The direct detection of the signal at the distribution fiber output is performed using a

single PIN with a responsivity of 1 A/W.

• The photo-detected signal, Iout, is down-converted using a down-converter with the same

carrier frequency as the one of the up-converter, fRF . The signal phase shift, ∆φ, caused

by the channel dispersion is corrected to ensure the signal synchronization.

• After down-conversion, the baseband complex signal is filtered by a LPF to reduce the out-

of-band power. An ideal rectangular filter, with the same bandwidth as the LPF present

at the transmitter, is used for this purpose.

• The filtered signal is converted into discrete samples using an ADC. One of the require-

ments of the preprocessing module technique is that the sampling rate of the ADC, SRADC ,

at the ONU, has to be proportional to the bandwidth assigned to that ONU. This is further

explained in sections 3.3 and 3.4.

• A serial-to-parallel converter is used to split each OFDM symbol into NscONUx OFDM

subcarriers. The variable Y represents the sampled signal at the serial-to-parallel converter

output.

• A NscONUx-point FFT is used to recover the data carried by the NscONUx subcarriers

assigned to ONUx. Either an equalizer or a simple amplitude gain module is used to mit-

igate the distortion and the attenuation effects, respectively, and recover the transmitted

complex-valued symbols. The resulting vector, S, is composed by received complex-valued

symbols carried by the NscONUx subcarriers assigned to the ONUx.

• Finally, the received complex-valued symbols are demapped, and the binary data stream

is recovered.

Mathematical formulation of the ASA-OFDMA-PON scheme operation 31

3.2 Mathematical formulation of the ASA-OFDMA-PON scheme

operation

In this section, the theoretical formulation of the signal preprocessing at the OLT, is detailed.

Following the ASA-OFDMA-PON architecture depicted in Fig. 3.1, the concept of the signal

preprocessing operation is presented in Fig. 3.2.

Mathematically, the referred signals are represented by vectors, and matrices are used to

describe the transformations between signals.

Defining C as a matrix containing the frequency response at every k subcarrier frequency,

fk, from the M -QAM mapper output to the FFT output at each ONU receiver, the signal

preprocessing consists of applying a previously obtained matrix, X, to the mapped signal, S,

obtaining the preprocessed signal, T .

The signal S contains the complex-valued symbols carried in the original Nsc OFDM sym-

bol subcarriers which carry data destined to the different ONUs. The preprocessed signal, T ,

contains the complex-valued symbols carried in the subcarriers after the preprocessing.

When the preprocessed signal is transmitted along a transmission system with the response

characterized in C, at each transmission system output (ONU FFT output), the data carried in

the assigned OFDM signal subcarriers, S, is retrieved exactly as it was originally mapped.

... ... ...

Signal before pre-processing.At the M-QAM mapper output

Signal at the ONU Rx Equalizer

output.X C

SS T

Figure 3.2: Concept of the signal preprocessing operation of the studied ASA-OFDMA-PONscheme.

The preprocessing operation explained above and depicted in Fig. 3.2 is described by equa-

tions 3.1-3.5, presented in the following paragraphs.

To ensure that the data carried in the subcarriers is received at each ONU, as it was trans-

mitted by the OLT, the following equality must be verified:

S ≡ S (3.1)

meaning that, if a signal is built containing all the data carry subcarriers received by the ONUs,

in the same disposition in the frequency spectrum they had when the signal was transmitted,


the signal containing the data carried by all received subcarriers, S, is equivalent to the original

modulated signal, S.

Since each user’s optical transmission channel has different characteristics (noise, distance,

dispersion, attenuation), the signal detected by the receiver at each ONU is distinct from the

others, due to the impact of different optical channel responses on the signal waveform. Fur-

thermore, the data assigned to each user is carried by subcarriers at different frequencies, fk.

The signal preprocessing is only possible if the frequency response of all users’ transmission

system, at the frequency of the subcarriers which carry their destined data, is obtained. For

that purpose, the Nsc ×Nsc matrix C, is obtained.

The data carried by each subcarrier has to be preprocessed, according to the characteristics of

the user’s transmission channel, to which it is destined. As depicted in Fig. 3.2, the transmitted

preprocessed signal is described by:

T = X · S (3.2)

where X is a matrix that when applied to the original subcarriers, S, changes the subcarrier

structure, rearranging the information carried by the original subcarriers [3].

The signal containing the set of all the Nsc OFDM symbol subcarriers retrieved by every

ONU, S, given by:

S = C · T (3.3)

results from transmitting the preprocessed subcarriers, Tk, over the transmission system char-

acterized in C.

The matrix X is obtained through manipulation of the Eqs. 3.1, 3.2 and 3.3 using matrix

algebra. Combining 3.2 and 3.3, the relation between the data carried in the original and the

received Nsc subcarriers, is given by:

S = C ·X · S. (3.4)

In order to obtain the equality expressed in Eq. 3.1, the following equation must hold:

C ·X = I (3.5)

where I is the identity matrix. From Eq. 3.5, it can be concluded that X is obtained through

the inversion of the matrix which describes the frequency response of the transmission system,

Characteristics of the PPM operation 33

C, verifying:

X = C−1. (3.6)

Therefore, the relation between the data carried by the OFDM subcarriers at the M -QAM

mapper output and the preprocessed subcarriers which are transmitted, is given by:

T = C−1 · S. (3.7)

Due to its function, the matrix C−1 is named for the remain of this work as the preprocessing

matrix (PPM).

3.3 Characteristics of the PPM operation

The first step in the preprocessing is to define the subcarrier allocation scheme. The subcarrier

allocation scheme describes the distribution of the data destined to each user to the subcarriers

which are to carry it.

It was verified, when studying the operation of the ASA-OFDMA-PON using numerical

simulation, that certain requirements have to be met so the users can correctly sample and

demodulate the received signal. In this section, the requirements for the subcarrier allocation

scheme are presented, followed by the detailing of the formulations of the sub-sampling and

FFT operations at each ONU. Finally, the training sequence used to estimate the transmission

channel and obtain the matrix C is explained.

3.3.1 Requirements for the subcarrier assignment

The conditions presented next were either indicated in [3], where this technique was proposed, or

identified when studying the operation of the ASA-OFDMA-PON using numerical simulation.

The number of subcarriers assigned to an ONUx, NscONUx, has to verify the following equation:

NscONUx=

Nsc

2n, n ∈ N | n ≤ log2 (Nsc) (3.8)

where Nsc is the total number of subcarriers of an OFDM symbol. The capacity per subcarrier

corresponds to the total system capacity (in this work it is used a fixed total system capacity of

Rb = 10 Gb/s) divided by the total number of subcarriers. Therefore, the capacity assigned to

each user is:

RbONUx =Rb

2n(3.9)


The subcarrier assignment to each ONU is not arbitrary. The subcarriers are indexed with

indexes going from 1 to Nsc, where the correspondence of the subcarrier index k to the frequency

it occupies in the spectrum, fk, follows the scheme detailed in Appendix C.

The indexes k of the subcarriers assigned to a certain ONUx, according to the capacity

attributed to that user, are:

k = i : 2n : Nsc (3.10)

where i is the index of the subcarrier assigned to ONUx with the lowest index. The notation

presented in Eq. 3.10 represents all the subcarriers whose the indexes go from i, counting with

a step of 2n, to the last index ku, holding Nsc − 2n ≤ ku ≤ Nsc.

Fig. 3.3 illustrates an example of this distribution for 3 ONUs with the indexes 1, 5 and 9 and

a stream of Nsc = 64 subcarriers. These are assigned Nsc

23 , Nsc

24 and Nsc

25 subcarriers, respectively.

Figure 3.3: Illustration of the subcarrier assignment for ONUs with a different number of sub-carriers.

As indicated in [3], at the ONUs, the ADCs sample the detected electrical signals in different

sub-Nyquist rates1 to produce spectral aliasing. Therefore, the sampling rate as well as the

FFT size at each ONU is proportional to its preassigned bandwidth. To reduce the IFFT/FFT

computation complexity the number of data points in the FFT should be a power of 2. These

conditions explain the requirements described in Eqs. 3.8 and 3.10. These requirements limit

the ways the capacity of the system can be allocated to different users.

3.3.2 Requirements for the sub-sampling and FFT operation

As stated above, the sampling rate as well as the FFT size at each ONU have to be proportional

to its preassigned bandwidth.

If the ONU is preassigned 12n of the total signal bandwidth, the sampling rate of that ONU

will be 12n of the samples at the DAC input and the FFT will have Nsc

2n ponts at that ONU.

Fig. 3.4 illustrates this, using as example an ONU which is assigned 122 of the total bandwidth.

The samples at the DAC input have a spacing of ∆DAC between them. At the ADC output, if

the signal is sub-sampled with a 14 of the sampling rate at the DAC input, the spacing between

1Sub-Nyquist sampling rate: smallest integer sub-multiple of bandpass frequency fRF that meets the basebandNyquist criterion fs > 2B.

Characteristics of the PPM operation 35

samples of the signal will be 22 of the time interval between samples at the DAC input. Therefore,

the number of samples at the FFT input of that ONU will be Nsc4 .

(a) (b)

Figure 3.4: Spacing between samples (a) at the OLT at DAC input and (b) at the ADC outputof an ONU which is assigned 1

22 of the total bandwidth.

When the IFFT algorithm is applied to the signal, the information from each carrier fre-

quency of the OFDM symbol is spread along the entire OFDM symbol duration. The prepro-

cessing is necessary to ensure that when a FFT is applied over a time-domain sub-sampled signal

with less samples, the information carried by each carrier frequency is recovered. The sampling

process at the ADC must, however, start at a specific time instant. This is required to obtain

the necessary information to retrieve all the data carried by the subcarriers assigned to each

ONU, and to guarantee the the non-singularity of the matrix C, which is a condition to obtain

C−1. This was a requirement verified when simulating the ASA-OFDMA-PON operation.

The signal sampled at the ADC goes through a NscONUx-point FFT described by Eq. 3.11.

The signal at the FFT output is given by:

sk =

Nsc2n∑

m=1

Yme−j

2π(k−1)(m−1)Nsc2n (3.11)

where sk is a Nsc2n -length vector containing the complex-values carried in the ONUs’ selected

subcarriers and Ym is the sampled signal at the ADC output.

3.3.3 Training sequence and channel estimation

The matrix C is filled using a training sequence which is previously transmitted to all ONUs.

The feedback signal is retrieved at the OLT through the upstream channel.

To obtain the impact of the user’s transmission channel on the assigned kth subcarrier, a

training OFDM symbol with Nsc subcarriers, is transmitted. In the training symbol transmitted


to obtain the frequency response of the transmission channel at the kth subcarrier frequency, fk,

all the symbol subcarriers but the kth are transmitted with amplitude zero. The kth subcarrier

is transmitted with amplitude one. Each transmitted training symbol, ST , is given by:

ST (n) =

⎧

⎪

⎨

⎪

⎩

0, (n = k)

1, (n = k), 1 ≤ n ≤ Nsc (3.12)

where k is the subcarrier for which the frequency response of the transmission channel is being

obtained.

The amplitude of the training symbols received at each ONU are transmitted back to the

OLT. According to the subcarrier allocation scheme defined and the subcarriers’ indexing, the

amplitudes of the training symbols received at the OLT, sent by each ONU, are used to complete

the matrix C. The kth column of the Nsc ×Nsc C matrix is filled in with the amplitudes of the

training symbol received from the ONU to which the kth subcarrier was assigned.

After obtaining the matrix C, its inverse, C−1, is computed in the preprocessing module at

the OLT. The calculated PPM matrix, C−1, can be used for this transmission system for all

data slots transmitted with the same subcarrier allocation schedule used to obtain this matrix.

It is necessary to calculate the average of the electric noise PSD at the receiver when obtaining

the PPM. Otherwise, as the noise has a random behaviour, it is not be possible to estimate

adequately the channel characteristics when high electric noise PSD levels were considered.

It is also necessary to perform gain equalization at the receiver for the training symbols used

to obtain the PPM. When there is a distance longer than 20 km between different ONUs to the

OLT, the signal levels are quite different. If no equalization is done, the PPM, containing the

amplitudes of the training symbols received from all ONUs, when inverted and applied to the

signal, does not correctly rearrange the information of the signal subcarriers. This results in an

incongruence where ONUs at shorter distances from the OLT have worse performance than the

ones at longer distances from the OLT.

3.4 Example of ASA-OFDMA-PON signal processing operation

A brief example is used to illustrate the signal preprocessing and the operation of the investigated

ASA-OFDMA-PON. Consider an OFDM system serving 2 ONUs (in a back-to-back operation)

and 2 subcarriers per OFDM symbol. The system capacity is equally divided by the two ONUs.

An assignment schedule where, for each transmitted OFDM symbol, the first subcarrier, s1, is

assigned to the ONU-1 and the second subcarrier, s2, is assigned to ONU-2, is considered.

Example of ASA-OFDMA-PON signal processing operation 37

The following ST k =[

s1 s2

]

OFDM symbols:

ST1 =[

1 0]

ST2 =[

0 1]

(3.13)

are generated to be sent as training symbols. As can be seen, for each training symbol ST k, the

subcarrier being studied is set to 1.

A 2-point IFFT, expressed by Eq 2.13, is applied to each symbol. The signal, YTm, at the

IFFT output is:

YTm =

⎡

⎣

12

12

12 −1

2

⎤

⎦ (3.14)

Then, the signal is converted using a serial-parallel converter into a stream of samples. The

signal at the DAC input is:

YTmDAC=

[

12

12

12 −1

2

]

(3.15)

At the ONU side, a sampling rate, SRADC which is half of the rate of the samples at the

DAC input, SRADC=SRDAC

2 , is used by the ADC to sub-sample the received signal. At each

ADC output of each ONU, the signals are:

YTONU1=

⎡

⎣

YTmDAC(1)

YTmDAC(3)

⎤

⎦ =

⎡

⎣

12

12

⎤

⎦

YTONU2=

⎡

⎣

YTmDAC(2)

YTmDAC(4)

⎤

⎦ =

⎡

⎣

12

−12

⎤

⎦ (3.16)

The expression of the FFT applied to the signal at each ONU-side, according to Eq. 3.11,

is given by:

sT kx =

NscONUx∑

m=1

YTmONUx· e−j2π(k−1)(m−1). (3.17)

The recovered symbols, STk, at each ONU, after applying the FFT described in Eq. 3.17 to


the sampled signals, are:

sTONU1=

[

12

12

]

(3.18)

sTONU2=

[

12 −1

2

]

A signal containing the two received symbols, ST k, is sent to the OLT using an upstream

link feedback channel. The matrix C, is completed at the OLT using the signals received by

both ONUs:

C =

⎡

⎣

12

12

12 −1

2

⎤

⎦ (3.19)

The matrix C is inverted, and the PPM is obtained:

C−1 =

⎡

⎣

1 1

1 −1

⎤

⎦ . (3.20)

The matrix C−1 is the one to be used to preprocess the OFDM signals to be transmitted. If

the ADCs of both ONUs started the signal sampling at the same time instant (here described

by discrete samples), with the same sampling rate, the same samples would be obtained at the

output of each ADC (Eq. 3.16). If this was the case, the matrix C would be singular and not

invertible. Thus, it would not be possible to obtain the matrix C−1. This constraint, relative to

the time instant the ADC starts the signal sampling, was identified when studying the system

operation using numerical simulation. In future work, it is important to assess if this is an

ASA-OFDMA-PON system limitation, implying that the synchronization at each ADC input is

crucial for the operation of the scheme. Alternatively, it can be a numerical simulation limitation

since the continuous waveforms are represented by discrete samples. In either case, it is also

important to study the resilience of the studied ASA-OFDMA-PON to a de-synchronization of

the time instant the ADC starts the signal sampling.

Considering now the amplitudes of the subcarriers of an OFDM information symbol, S =[

a1 a2

]

, transmitted along the system defined above. The product between the PPM matrix,

C−1, and S is:

T =[

a1 + a2 a1 − a2

]

. (3.21)

ASA-OFDMA-PON advantages and impairments 39

The signal at the IFFT output, computed using Eq. 2.13, is given by:

Ym =[

a112 [(a1 + a2) + (a1 − a2)ejπ]

]

. (3.22)

As an electrical back to back system is considered, the next step is to apply the FFT described

in Eq. 3.17 to the signals sampled by the ADC of each ONU, which are given by:

YONU1 =a1 (3.23)

YONU2 =1

2[(a1 + a2) + (a1 − a2)e

jπ].

Therefore, the signals, sk, at the FFT output for each ONU are:

sONU1 = a1 (3.24)

sONU2 = a2

It can be observed that the original subcarriers are recovered by the ONUs according to the

assignment schedule.

While this example was presented for a system with no optical components, in an optical

transmission system, the received signal used to obtain PPM contains the frequency response

of all the system components. Ideally, with the frequency response of all the link, it would be

possible to compensate the channel and components dispersion and attenuation effects in the

transmitted signal.

3.5 ASA-OFDMA-PON advantages and impairments

The studied ASA-OFDMA-PON scheme presents a flexible OFDMA-based dynamic bandwidth

assignment solution to the NG-OANs. The resources allocation is defined on the digital domain

and a change on the assignment schedule has no impact on the network configuration.

Since each ONU has only to process its assigned subcarriers, the ADCs have lower sampling

rates and the FFTs smaller sizes. Therefore, the components complexy at the ONU-side, is

reduced. The simplification of the elements on the user end is very advantageous from a cost-

efficiency standpoint.

Furthermore, since every ONU has a channel with different characteristics, different wave-

forms arrive at each ONU. This implies that each user is only able to recover its assigned sub-

carriers. As a consequence, the signal transmission security and information privacy is ensured

on a physical level [3].


One of the limitations of this scheme is the austerity of the sampling process. The ADC

at each ONU has to start sampling the signal at a specific time instant. This is necessary to

guarantee that the information carried by the assigned subcarriers are recovered by the ONUs,

and the matrix C with the frequency response at the frequency of each subcarrier is not singular.

This implies that the ONUs must be synchronized so that the ADC is able to start the sampling

process at that time instant. The resilience of the ASA-OFDMA-PON to a de-synchronized

sampling time instant must be evaluated on an experimental setup where it is possible to sample

continuous waveforms.

It is also predictable that the signal preprocessing will not be able to mitigate any of the

noise-related degradation in the signal, due to its arbitrary nature.

When the inversion process of the matrix C and the application of the PPM to the signal takes

place, the signals destined to the different ONUs become correlated. Therefore, any numerical

errors or problems with the channel frequency response are propagated to the entire system.

This makes the system vulnerable to strange behaviours caused by the PPM calculation.

3.6 Conclusions

In this chapter, the investigated ASA-OFDMA-PON scheme operation was analysed. This

scheme, through its subcarrier assignment scheduling, allows for a dynamic digitally controlled

bandwidth allocation. The studied ASA-OFDMA-PON system is based on a preprocessing

matrix which is used to rearrange the information carried by the OFDM symbols’ subcarriers

so the users are able to select and recover their information.

The architecture of the studied ASA-OFDMA-PON was presented in this chapter, followed

by the mathematical formulation of ASA-OFDMA-PON operation. The characteristics of the

PPM operation were detailed.

A numerical example, where the validity of the preprocessing operation is demonstrated, was

presented. Finally, the identified advantages and disadvantages of the studied ASA-OFDMA-

PON scheme, were described.

It was concluded that, while the preprocessing works as intended, there are limitations which

compromise the applicability of the proposed ASA-OFDMA-PON in the NG-OANs. Namely,

the subcarrier allocation scheme imposed by the sampling process, as well as the synchronization

required of each ONU so that each ADC starts the signal sampling at a specific instant.

41

Chapter 4

Performance analysis of the

ASA-OFDMA-PON scheme

Noise and distortion are the main contributors to the system performance degradation in most

optical OFDM systems.

In the studied ASA-OFDMA-PON configuration, no optical amplification is used because it

is considered that it falls out of the scope of the work which is to analyse the capacity and reach

of the proposed OFDMA-based OAN. Therefore, only electrical noise is considered in this work.

The error vector magnitude (EVM) as a function of the root-mean-square (RMS) voltage

of the OFDM signal, is used to quantify the quality of the transmission, in this work. The

performance evaluation method and the limits considered for an acceptable performance are

detailed in Appendix B.

In this chapter, the performance of the ASA-OFDMA-PON, when a transmission capacity

of Rb = 10 Gb/s is required, is evaluated.

This chapter is organized as follows. In section 4.1, the system parameters which are common

to all analyses are detailed. In section 4.2, the two allocation schemes used in this work are

described. In section 4.3, the computation and analysis of the MZM’s optimal modulation index

is presented. In the following sections, the performance evaluation of the ASA-OFDMA-PON

is analysed following three categories; in section 4.4, the performance of the proposed scheme

at channel estimation, is presented; in section 4.5, the evaluation of the OFDMA-based OAN

performance, when all the ONUs are at the same distance from the OLT, is presented. In

section 4.6, the evaluation of the OFDMA-based OAN performance when the ONUs are at

different distances from the OLT is presented.

42 PERFORMANCE ANALYSIS OF THE ASA-OFDMA-PON SCHEME

Number of ONUs Ls [dB]1 016 15.2432 19.0564 22.86

Table 4.1: Splitter insertion losses.

MZM 1:N Splitter Optical Circulators SSMF attenuation0 dB Ls dB 2 × 0.7 dB 0.2 dB/km

Table 4.2: Insertion losses considered in the downstream transmission.

4.1 System parameters

The configuration of the studied ASA-OFDMA-PON scheme, whose performance is evaluated

using numerical simulation, is illustrated in Fig. 3.1 in Chapter 3.

In all the studies accomplished, a stream of 128 OFDM symbols is considered for the trans-

mitted OFDM signal. When signal equalization is used at the receiver, 10 training symbols are

used for channel estimation and the remaining 118 are information symbols.

In order to evaluate the influence of the subcarrier modulation format on the OFDMA-based

OAN performance, three different modulation formats are used. These are 4, 16 and 32-QAM.

The constellation mapping and de-mapping processes are further detailed in Appendix A.1.1.

PON configurations with 1, 16, 32 and 64 ONUs are analysed. When 1, 16 and 32 ONUs

are considered, OFDM symbols with 128 subcarriers are transmitted. When the OLT serves 64

ONUs, OFDM symbols with 256 subcarriers are considered, so that the number of subcarriers

assigned to each ONU is closer to the previous cases.

Two subcarrier allocation schemes, detailed in section 4.2, are analysed. These schemes

describe the assignment of the OFDM signal subcarriers by the users. The distribution of

subcarriers follows the requirements detailed in Chapter 3.

The insertion losses considered for this analysis comprehend the splitter losses, the circulator

losses and the fiber attenuation. The splitter losses, presented in Tab. 4.1, are calculated as

described in the appendix A.2.2 and they differ according to the number of ONUs. The insertion

losses, considered for the downstream transmission in the configuration analysed in this work,

are presented in Tab. 4.2.

The optical signal is modulated using a laser source, with null phase and constant amplitude,

and a MZM biased in a quadrature bias point. These are described in Appendix A.2.1. The

values considered for the laser output power, Pout, are 4 mW (6 dBm) and 2 mW (3 dBm) and

the MZM output power is 2 mW (3 dBm) and 1 mW (0 dBm), respectively. According to [17],

an average power at the fiber input less than 4 dBm is sufficient to neglect non-linearities in the

Subcarrier allocation schemes 43

SSMF. Despite the fact that no insertion losses are considered in the MZM, an additional 3 dB

attenuation is introduced since the MZM is biased in quadrature.

Different noise power spectral density (PSD) values are considered to evaluate the inves-

tigated ASA-OFDMA-PON resilience to the electrical noise of the receiver of the ONU. The

values considered for the electrical noise PSD are Sc = 1× 10−24 A2/Hz, Sc = 1× 10−23 A2/Hz,

Sc = 1× 10−22 A2/Hz.

4.2 Subcarrier allocation schemes

Two subcarrier allocation schemes are used to access the system performance. Since one of the

most promising features of the studied ASA-OFDMA-PON is the ability to dynamically allocate

the bandwidth according to the user demand, it is interesting to study a configuration where

the system capacity is equality distributed among the users, opposed to the configuration where

a certain capacity is attributed to each user. These schemes are created in accordance with the

conditions presented in Eqs. 3.8 and 3.10 from the section 3.4, and are presented in Appendix

C.

Tabs. C.1, C.2 and C.3 illustrate the subcarrier distribution by the ONUs where the same

number of subcarriers, NscN , (where N is the number of ONUs of the PON) is assigned to

the 16, 32 and 64 ONUs, respectively. The importance of studying the situation presented

by this scheme, where all users are assigned the same capacity, is to analyse the impact of

other parameters (i.e. the ONUs network reach, constellation format, electrical noise PSD) in

the proposed ASA-OFDMA-PON performance. This scheme will be referred to as subcarrier

allocation scheme 1.

Tabs. C.4, C.5 and C.6 illustrate the situation where the system capacity is not uniformly

distributed among the ONUs. As presented in Tab. C.4, when 16 ONUs are considered, a group

of N4 ONUs is assigned Nsc

23 subcarriers, N4 ONUs are assigned Nsc

24 subcarriers and the remaining

N2 ONUs are assigned Nsc

25 subcarriers.

When the subcarriers are to be assigned to 32 ONUs (Tab. C.5), a group of N4 ONUs is

assigned Nsc

24 subcarriers, N4 ONUs are assigned Nsc

25 subcarriers and the remaining N2 ONUs are

assigned Nsc

26 subcarriers. The same distribution holds for 64 ONUs (Tab C.6), since 256 OFDM

subcarriers are available. This scheme will be referred to as subcarrier allocation scheme 2.


4.3 Optimal modulation index

In this work, a MZM is used to modulate the optical signal. The MZM modulation index is a

parameter which describes the measure of the amplitude variation surrounding an unmodulated

carrier.

The MZM modulation index is expressed as:

mi =VRMS

Vπ(4.1)

where VRMS represents the root mean square (RMS) voltage of the OFDM signal applied to the

MZM.

The performance of the system varies according to the MZM’s modulation index values. It

is important to find the optimal modulation index for each PON configuration considered, for

which the system performance is the best.

The value of the optimal modulation index varies with the optical modulator bias point, the

fiber length and the O/E converter considered (such as PIN, APD.)

In this work, a study of different modulation indexes was performed for different configura-

tions which have the parameters presented in Table 4.3.

Table 4.3: Parameters used to obtain the optimal modulation indexes for the different ASA-OFDMA-PON setups studied in this work.

Setup SI SII SIII SIVBit rate, Rb 10 Gb/s

Number of ONUs, N 1 16 32 64Number of OFDM symbols, NOFDM 128

Number of subcarriers per ONU, NscONU 128 12816

12832

25664

Guard interval, ∆G 0 sModulation format, M -QAM 16-QAM

PSD of electric circuit noise, Sc(f) 1× 10−23 A2/HzRF carrier frequency, fRF 4 GHz

Laser output power, Pout3 dBm6 dBm

Distance, d 20 kmNumber of training symbols, NTS 10 symbols

Fig. # 4.1 4.2 - -

Figs. 4.1 and 4.2 present the results of the EVM as function of the modulation index for the

setups SI and SII related to the situations where the system has 1 and 16 ONUs, respectively.

The results of the optimal modulation indexes for these and the remaining setups, (SIII and

SIV), presented in Tab. 4.3, are summarized in Tab. 4.4.

In order to establish a compromise between the electrical noise and the distortion impacts

Optimal modulation index 45

5 10 15 20 25 30 35ï50

ï40

ï30

ï20

ï10

Modulation index [%]

EVM

[dB

]

Pout=0 dBmPout=3 dBm

36

Figure 4.1: EVM as a function of the modulation index when transmitting a 16-QAM OFDMsignal to 1 ONU when d=20 km and Sc = 10−23 A2/Hz. Results for setup SI.

Pout=3dBm

Pout=0dBm

1 6 11 16 21 26 31 36 41 46 51ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5


EVM

[dB]

3 dBm

6 dBm

Figure 4.2: EVM as a function of the modulation index when transmitting a 16-QAM OFDMsignal to 16-ONUs when d=20 km and Sc = 10−23 A2/Hz. Results for setup SII.

Table 4.4: Optimal modulation indexes obtained for the different setups of the studied ASA-OFDMA-PON.

SetupOptimal modulation index [%]Pout=3 dBm Pout=6 dBm

SI 7 8SII 26 21SIII 31 26SIV 44 33

on the signal degradation, the electrical noise PSD considered to obtain the optimal modulation

index is Sc = 10−23 A2/Hz. From the results obtained for the PONs with 32 and 64 ONUs, it

was verified that when Pout=3 dBm, the system performance is below the EVM performance


limit level for 16-QAM. This limit is of -16.5 dB. The EVM limits are obtained for a BER of

10−3 as explained in Appendix B. Therefore, for the remaining performance evaluations, only

the laser output power of 6 dBm, is considered.

For lower modulation index, electrical noise is responsible for the EVM degradation (since,

for lower values of VRMS , the MZM still operates linearly). For higher modulation index values,

the modulator and PIN nonlinearities are responsible for the performance degradation observed

(since the MZM+PIN are no longer operating in the linear zone due to high values of VRMS).

Comparing the optimal modulation indexes obtained for the situation where the PON has

only one ONU (Fig.4.1) with the situation where the PON has 16 ONUs (Fig.4.2) and observing

the results presented in Tab. 4.4 it is possible to conclude that, for the setup SII the optimal

modulation index takes much higher values than for setup SI. Due to the splitter losses the signal

is much more attenuated in setup SII. Therefore, for lower modulation index values, where the

electrical noise is responsible for the EVM degradation, the EVM is much more degraded in the

setup with more attenuation. In this case, it is necessary more power-higher optimal modulation

values-to reach the region of balance between the noise contributions to the signal degradation

and the contribution of the system components non-linearities.

2 4 6 8 10 12 14 16 18 20ï55

ï50

ï45

ï40

ï35

ï30

ï25


EVM

[dB]

0 km20 km40 km

Figure 4.3: EVM as a function of the modulation index when a 16 QAM OFDM signal istransmitted to 1 ONU, considering an electrical noise PSD of Sc = 10−23 A2/Hz, for differentnetwork reaches.

In order to take conclusions on the influence of the network reach on the optimal modulation

indexes values, the EVM as a function of the modulation index when considering different

network reaches, were obtained. These results are presented in Fig. 4.3.

Comparing the curves in Fig. 4.3 it can be seen that, for longer reaches, the minimum EVM

corresponds to slightly greater modulation indexes. It is also noticeable that the minimum EVM

Performance of the preprocessing matrix as channel characteristic estimator 47

value is much less defined for longer network reaches. The minimum corresponds to the case

where the contribution of the electrical noise and the distortion effects in the system reach an

balance. When the network reach increases, the system is more prone to the distortion effects

and fiber attenuation. Therefore, its performance will never improve as much when increasing

VRMS , because the signal is more degraded. It should be taken into account that choosing a

value in this region of balance implies that, if the system configuration characteristics are slightly

altered from the ones considered in the optimal modulation index analysis, the MZM may start

to operate in the non-linear region.

4.4 Performance of the preprocessing matrix as channel char-

acteristic estimator

As explained in chapter 3, the PPM, used to preprocess the subcarriers so that each ONU is able

to recover the information carried by its assigned subcarriers, also performs signal equalization.

For the situation where there is only one ONU, and all the subcarriers carry data assigned

to the same ONU, the performance of the PPM relative to its channel estimation and signal

equalization properties, is evaluated. Figure 4.4 presents the EVM results when a 16-QAM

OFDM signal is transmitted to one ONU. For this analysis, a situation where the electrical noise

effect can be neglected in comparison with the distortion effects, is considered. The distortion

effects can be estimated and compensated by the equalizer. It can be perceived from Fig. 4.4

that, when there is no channel compensation, the network reach is severely limited (around 20

km). The optimal modulation index obtained for the setup SI, mi = 8%, was used.

From the curves presented in Fig. 4.4, it is possible to conclude that, while the PPM presents

some channel estimation capability, it is not so efficient as using an equalizer at the receiver. This

can be explained by the difference between the training symbols used to obtain the equalizer,

which are similar to the ones transmitted as information symbols, and the training symbols

used to obtain the PPM, which are not modulated according to an M -QAM scheme and were

chosen specifically to study the frequency response of the transmission system at each subcarrier

frequency, fk.

In order to obtain the maximum reach for the proposed ASA-OFDMA-PON, signal equal-

ization is used at the receiver in all the remainder analyses.


0 10 20 30 40 50 60 70 80 90ï50ï45ï40ï35ï30ï25ï20ï15ï10ï5

Distance [km]

EVM

[dB

]

With PPM and with EQWith PPM and no EQWithout PPM and with EQWithout PPM and no EQ

Figure 4.4: EVM as a function of the distance when transmitting a 16 QAM OFDM signal whenPout = 6dBm and Sc=10−24A2/Hz.

4.5 ASA-OFDMA-PON performance with all ONUs at the same

distance from the OLT

In this section, the performance of the ASA-OFDMA-PON scheme, when all the ONUs are

at the same distance from the OLT, is analysed. In section 4.5.1, the maximum reach of the

network, when all the ONUs are assigned the same number of subcarriers, is presented. The

impact of the mapping format used to modulate the OFDM subcarriers on the network reach,

is also analysed in this section.

In section 4.5.2, the performance of the studied ASA-OFDMA-PON, when different ONUs

are assigned a different number of subcarriers, is analysed. The consequences of assigning a

different number of subcarriers to each user on the system performance, when opposed to having

the system capacity equally distributed among the users, is assessed.

4.5.1 Performance results when the system capacity is uniformely distributed

by ONUs

Figs. 4.5(a), 4.5(b), 4.5(c) present the EVM results as a function of the network reach when

a 16-QAM OFDM signal is transmitted in a ASA-OFDMA-PON serving 16, 32 and 64 clients,

respectively. All the ONUs are assigned the same number of subcarriers, according to the

subcarrier allocation scheme presented in section C.1 of Appendix C. The dashed horizontal line

shown in all three figures is set to −16.5 dB, corresponding to the EVM limit for a 16-QAM

OFDM signal when a BER threshold of 10−3 is defined (as explained in section B.2 of Appendix

B).

ASA-OFDMA-PON performance with all ONUs at the same distance from the OLT 49

Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

0 10 20 30 40 50 60ï30

ï25

ï20

ï15

ï10

ï5

Distance [km]

EVM

[dB]

(a) 16-ONUs

Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

0 10 20 30 40 50 60ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

(b) 32-ONUs

Sc=10ï22 A2/HzSc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

0 10 20 30 40 50 60ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

(c) 64-ONUs

Figure 4.5: EVM as a function of the distance when transmitting 16-QAM OFDM signal sub-carriers over (a) 16-ONUs, (b) 32-ONUs and (c) 64-ONUs according to the subcarrier allocationscheme 1 when all the ONUs are at the same distrance from the OLT.

The maximum network length achieved by the studied ASA-OFDMA-PON in the described

conditions is 60 km. This is achieved for the lowest electrical noise PSD value considered,


0 10 20 30 40 50 60ï30

ï25

ï20

ï15

ï10

ï5

Distance [km]

EVM

[dB]

4ïQAM16ïQAM32ïQAM

(a) 16-ONUs

0 10 20 30 40 50 60ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB

]


(b) 32-ONUs

0 10 20 30 40 50ï21

ï19

ï17

ï15

ï13

ï11

ï9

ï7

ï5

Distance [km]

EVM

[dB]


(c) 64-ONUs

Figure 4.6: EVM as a function of the distance for different M -QAM OFDM signals transmittedto (a) 16-ONUs, (b) 32-ONUs and (c) 64-ONUs ASA-OFDMA-PON, according to the subcarrierallocation scheme 1 when all the ONUs are at the same distance from the OLT.

Sc = 10−24 A2/Hz, and for the PON serving less clients. In Tab. 4.5, the values of the

maximum network reach seen in Fig. 4.5 are summarised.

When the PON serves 16 ONUs, the EVM performance is under the accepted EVM limits


Table 4.5: Maximum network reach of the studied ASA-OFDMA-PON when all the ONUs areat the same distance from the OLT and the subcarrier allocation scheme 1 is used.

Sc [A2/Hz ] Number of ONUs Network reach [km]

10−2416 6032 5064 35

10−2316 4232 2864 10

10−2216 2032 564 -

for all electric noise levels considered for up to 20 km. For PONs with 32 and 64 ONUs, this

does not occur. For a 32 ONUs PON, only 5 km are achievable when Sc = 10−22 A2/Hz. For

a 64 ONUs PON, the signal is not received under the acceptable EVM limits when the highest

electrical noise PSD value is considered.

As it can be seen in Fig. 4.5(a), the maximum reach of the 16-ONU PON decreases around 20

km when the electrical noise PSD value increases by one order of magnitude. This degradation

is more accentuated for PONs with more ONUs, where the reduction, per electrical noise PSD

value order of magnitude increase, is about 25 km. The maximum network length decreases

between 10 and 15 km, for PONs serving twice the number of ONUs. The link attenuation

increases considerably with the number of ONUs, as presented in Table 4.1. The difference of

performance between PONs with more and less ONUs is greater when higher electrical noise

PSD values are considered. When the noise levels are very high, the signal gain amplification,

performed in the receiver to retrieve the severely attenuated signal, amplifies not only the signal

amplitude level, but also the noise amplitude level, causing the signal to become unrecoverable.

This difference is also attributed to the increase of the system complexity with the number of

ONUs, resulting in additional numerical errors when processing and inverting the PPM.

The impact of the chosen OFDM subcarriers modulation format on the network reach is

analysed in Figs. 4.6(a), 4.6(b) and 4.6(c). These curves present the average of the EVM results

of all ONUs of the PON when the different modulation formats are used to modulate the OFDM

signal subcarriers.

Despite the number of ONUs, the behaviour of the curves is quite similar for all the subcarrier

modulation format. While there is not a great variation in the performance of the system when

different constellation modulation formats are used, the acceptance levels of EVM for each one

vary greatly. This results in network reaches much higher for lowerM -QAMmodulation formats.


Table 4.6: Maximum network reach of the studied ASA-OFDMA-PON when all the ONUs areat the same distance from the OLT and the subcarrier allocation scheme 1 is used.

M -QAM Number of ONUs Network reach [km]

4-QAM16 5832 4564 32

16-QAM16 4232 2864 10

32-QAM16 <3232 <2064 -

The maximum network reaches for each modulation format for the presented ASA-OFDMA-

PON configurations are presented in Tab. 4.6.

The subcarrier modulation format which allows longer network reaches, regardless the num-

ber of ONUs in the system, is 4-QAM. However, it has to be taken into account that choosing

this modulation format compromises the signal spectral efficiency (as explained in Chapter 2).

4.5.2 Performance results when the capacity is non-uniformely distributed

by ONUs

This section presents the analysis of the performance of the studied ASA-OFDMA-PON when

a configuration where the ONUs are at the same distance from the OLT and the subcarrier

allocation scheme 2 is adopted. It is important to perform this analysis, since it allows to draw

conclusions on the impact of the number of subcarriers assigned to each user on each ONU

performance.

The EVM results as a function of the distance of the ONU to the OLT, for a 16-ONU

ASA-OFDMA-PON when a 16-QAM OFDM signal is transmitted, are depicted in Fig.4.7. For

different network lengths, each mark represents one ONU which is assigned either 16, 8 or 4

subcarriers. The results obtained for the 32 and 64-ONUs ASA-OFDMA-PONs are depicted in

Figs. E.1 and E.2 in Appendix E, respectively.

In Fig. 4.7 a), it can be seen, for lower network reaches (0-20 km), a difference of 1-2 dBs in

the EVM between the ONUs which are assigned more and less subcarriers. For short distances

and low PSD electrical noise values, the distortion is the the greatest contributor for the signal

degradation. This difference becomes less noticeable either when the network reaches increase,

increasing the attenuation as well, or when the electrical noise PSD level increases (Fig. 4.7 b)

and c)).


This difference of 1-2 dB on the EVM can be explained by the fact that the same modula-

tion index is used to modulate the signal at the OLT, which is transmitted to all the ONUs.

This means that the non-linearities of the MZM, resultant of being working with the optimal

modulation indexed, may have an impact over some subcarriers. The ONUs with more subcar-

riers should show a bigger impact of the MZM non-linearities in their performance since more

subcarriers are exposed to these.

Table 4.7: Maximum network reach of the studied ASA-OFDMA-PON when all the ONUs areat the same distance from the OLT and the subcarrier allocation scheme 2 is used for PONswith 16, 32 and 64 ONUs.

Sc [A2/Hz ] Number of subcarriers per ONUNetwork reach [km]

16 ONUs 32 ONUs 64 ONUs

10−2416

60 50 3584

10−2316

42 29 1084

10−2216

22 8 -84

Tab. 4.7 summarizes the maximum network reaches observed for the ASA-OFDMA-PON,

when an EVM limit of -16.5 dB is considered. For the 16-ONU ASA-OFDMA-PON, network

reaches between 20 and 60 km are achieved, depending on the electrical noise PSD level.

For the 32-ONU ASA-OFDMA-PON, maximum reaches of 5 and 50 km are achievable for

the highest and the lowest electrical noise PSD values, respectively. A disparity of 1-2 dB on

the EVM results of ONUs assigned more and less subcarriers, can also be seen in Fig. E.1, for

short distances and low electrical noise values.

For the 64-ONU ASA-OFDMA-PON, maximum reaches between 10 and 35 km are achievable

when Sc = 10−23 A2/Hz and Sc = 10−24 A2/Hz, respectively.

When comparing the results presented in Tab. 4.7, with the ones presented in Tab. 4.6,

related to the situation when all the ONUs are assigned the same number of subcarriers, it can

be concluded that no difference is seen in the maximum reach of the studied ASA-OFDMA-PON

when the users are assigned a different number of subcarriers. This means that the scheme of

allocation of the subcarriers to different ONUs does not compromise the system performance.

The impact of the electrical noise on the signal degradation is analysed using the constella-

tions presented in Fig. 4.8. In these figures, the impact of different different levels of electrical

noise PSD on the constellations of a 16-QAM OFDM signal received by one ONU, which is


0 10 20 30 40 50 60ï30

ï28

ï26

ï24

ï22

ï20

ï18

ï16

Distance [km]

EVM

[dB

]

NscONU=16NscONU=8NscONU=4

(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï30

ï28

ï26

ï24

ï22

ï20

ï18

ï16

ï14

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï26

ï24

ï22

ï20

ï18

ï16

ï14

ï12

ï10

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz

Figure 4.7: EVM as a function of the distance when transmitting a 16-QAM OFDM signalsubcarriers following the subcarrier allocations scheme 2, for a network with 16-ONUs

assigned 16 subcarriers and located at 20 km from the OLT, is presented.

As it can be seen in the constellation of Fig. 4.8 b), for low electrical noise PSD levels, the

influence of the electrical noise on the system constellation is not noticeable, since it behaves as

Performance evaluation of the PPM with ONUs at different distances from the OLT 55

ï1 ï0.5 0 0.5 1ï1

ï0.5

0

0.5

1

I

Q

(a) No noise considered

ï1 ï0.5 0 0.5 1ï1

ï0.5

0

0.5

1

I

Q(b) Sc=10−24A2/Hz

ï1 ï0.5 0 0.5 1ï1

ï0.5

0

0.5

1

I

Q

(c) Sc=10−23A2/Hz

ï1 ï0.5 0 0.5 1ï1

ï0.5

0

0.5

1

I

Q

(d) Sc=10−22A2/Hz

Figure 4.8: Constellations of the 16-QAM OFDM signal received in a ONU which is assigned16 subcarriers and is at 20 km from the OLT, for different electrical noise PSD levels.

if there was no electrical noise contribution, which is the case presented in Fig. 4.8 a). In these

situations, distortion is the factor which contributes the most for the signal degradation.

As the electrical noise contribution gets more prominent, by increasing the distance or by

imposing higher electrical noise PSD values, the degradation of the signal quality is more no-

ticeable. This degradation is depicted in the constellation of Fig. 4.8 c) and d), which are more

scattered for higher electrical noise PSD levels.

As depicted in Fig.4.8, when distortion has a greater impact than electrical noise, the network

reach increase does not contribute for the signal degradation. Thus, from 0 to 20 km, the system

performance is similar.

This explains the behaviour of the curves presented in Fig. 4.7 a) which present an almost

constant EVM between 0 and 20 km. As the distance increases, the signal level decreases,

the noise contribution gets more noticeable and the degradation of the signal quality is steeper.

When higher electrical noise PSD values are imposed to the system, Fig 4.7 b) and c), regardless

of the network length, the SNR values are lower, leading to a faster degradation of the signal.

4.6 Performance evaluation of the PPM with ONUs at different

distances from the OLT

In a real multiple access transmission system, the ONUs are not all at the same distance from the

OLT. In this section, the perfomance of the investigated ASA-OFDMA-PON, when its ONUs

are not all at the same distance from the OLT, is evaluated. Different scenarios, representing

different spatial distributions of the PON are defined to perform this analysis. Tables D.1-D.12,

found in Appendix D, present these scenarios.

Scenario D1 is presented in Tabs. D.1, D.6 and D.11, for 16, 32 and 64-ONUs, respectively.

In this case, network length ranges from 10 to 50 km for different ONUs. In D4, (Tabs. D.2,

D.7 and D.12), the same spatial distribution is considered, but for ONUs assigned different


subcarriers. This is so that it could be ascertain that the perfomance of each ONU, at a certain

distance, is independent of the subcarriers assigned it. Scenario D2 is presented in Tabs. D.3,

D.8 and D.13, for 16, 32 and 64-ONUs, respectively. In this case, a scenario where the ONUs

are close to each other and in short reach to the OLT, is considered. In scenario D3, presented

in Tabs. D.4, D.9 and D.14, a situation where the ONUs are close to each other, but farther

from the OLT, is considered.

In section 4.6.1, the ASA-OFDMA-PON performance is evaluated when subcarrier allocation

scheme 1 is used. In section 4.6.2, the results for the ASA-OFDMA-PON performance when the

subcarriers are assigned following the subcarrier allocation scheme 2 are presented.

4.6.1 Performance results when the capacity is uniformely distributed by

ONUs

The results obtained for each one of the ONU distances distribution plans, D1-D4, for a 16-ONU

ASA-OFDMA-PON when transmitting a 16-QAM OFDM signal are depicted in Figs. 4.9 and

4.10. For each electrical noise PSD level considered, the ONUs located at the distances presented

in the referred figures, represent the ONUs of the ASA-OFDMA-PON.

For the ASA-OFDMA-PON with 32 ONUs, the corresponding results are shown in Figs. E.3

and E.4. When the number of users is 64, the corresponding results are depicted in Figs. E.5

and E.6. These figures can be found in section E.3.1 of Appendix E.

Figs. 4.9(b) and 4.9(c) show the results for a 16-ONU ASA-OFDMA-PON when distribu-

tions D2 and D3 are considered, respectively. In 4.9(b), the curves present an almost constant

behaviour for the lowest electrical noise PSD levels, and the degradation becomes evident when

the electrical noise PSD values increase. In Fig. 4.9(c), where every ONU has a network length

above 30 km, the increase of the EVM results is visible even for Sc = 10−24 A2/Hz.

From the results presented in Figs. 4.9 and 4.10, the EVM degradation is visible when the

noise has an influence over the system greater than the distortion effects. This can be seen,

either for higher levels of imposed electrical noise PSD or for greater network reaches.

In Fig. 4.9(a), the performance degradation is observable for the ONUs located at more than

20 km (for the lowest electrical noise PSD level considered) from the OLT. This conclusion is in

agreement with what had been seen in section 4.5.1, for the case where all the ONUs were at

the same distance from the OLT.

Similar maximum network reaches are achieved for the ASA-OFDMA-PON when its ONUs

are all at the same distance from the OLT, (results depicted in Fig. 4.5(a) in section 4.5.1) and

when its ONUs are at different distances from the OLT (results depicted in Fig. 4.9(a)). In



Sc=10ï24 A2/Hz10 20 30 40 50

ï29ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB

]

(a) Scenario D1

Sc=10ï22 A2/Hz


10 12 14 16 18 20ï29

ï27

ï25

ï23

ï21

ï19

ï17

ï15

Distance [km]

EVM

[dB]

(b) Scenario D2

Sc=10ï22 A2/Hz Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

30 32 34 36 38 40ï28ï26ï24ï22ï20ï18ï16ï14ï12ï10ï8

Distance [km]

EVM

[dB]

(c) Scenario D3

Figure 4.9: EVM results as a function of the distance when transmitting a 16-QAM OFDMsignal to the 16-ONUs of the ASA-OFDMA-PON which are distributed according to D1 (a), D2(b) and D3 (c).

conclusion, similar performance will be obtained for an ONU located at a certain distance from

the OLT, regardless of the network length of the other ONUs in the PON. From inspection of the


Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

10 20 30 40 50ï29ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

Figure 4.10: EVM results as a function of the distance when transmitting a 16-QAM OFDMsignal to the 16-ONUs of the ASA-OFDMA-PON which are distributed according to D4.

Figs. E.3 and E.5, it is noticeable that the same conclusion holds for the ASA-OFDMA-PON

with 32 and 64 ONUs.

It was also concluded that the behaviour of each ONU, in the proposed ASA-OFDMA-

PON, is independent of the spectral position of its assigned subcarriers. This conclusion holds

independently of the number of ONUs in the PON.

From the conclusions drawn in section 4.5.2, it is expectable that the performance results

of each ONU are also independent of the number of subcarriers assigned to it. This analysis is

performed in the next section.

4.6.2 Performance results when the capacity is non-uniformely distributed

by ONUs

The results obtained for each ASA-OFDMA-PON spatial distribution plan, when a different

number of subcarriers is assigned to the users, and the OLT serves 16 ONUs are presented in

Figs. 4.11-4.17 in this section. The results obtained when 32 and 64-ONUs are considered, are

presented in Figs.E.7-E.8 and Figs.E.9-E.10, respectively. These results can be found on section

E.2 of Appendix E.

An additional scenario for the network reaches of each ONU in the ASA-OFDMA-PON, is

introduced in this section. In scenarios D1-D4, ONUs assigned the same number of subcarriers,

are at the same distance. In scenario D5, a different situation occurs. Tabs. D.5, D.10 and D.15

shown in Appendix D, present the network reach of each ONU in the configuration D5 when

the network has 16, 32 and 64 ONUs, respectively. The performance of the ASA-OFDMA-PON

obtained for this scenario is presented in Figs. 4.16, 4.17, E.8 and E.10.

Fig. 4.11 presents the results for the achievable network reach of the ASA-OFDMA-PON


0 10 20 30 40 50 60ï28

ï26

ï24

ï22

ï20

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï27

ï25

ï23

ï21

ï19

ï17

ï15

ï13

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz

Figure 4.11: EVM as a function of the distance when transmitting a 4-QAM OFDM signal tothe 16-ONUs of the ASA-OFDMA-PON, which are distributed according to D1.

when 16 ONUs, located at different distances from the OLT, are not all assigned the same

number of subcarriers of the transmitted OFDM signal. Network reaches between 20 and 50 km

are achieved for the highest and the lowest electrical noise PSD levels considered, respectively.


8 10 12 14 16 18 20 22ï30

ï29

ï28

ï27

ï26

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

8 10 12 14 16 18 20 22ï27

ï26

ï25

ï24

Distances [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

8 10 12 14 16 18 20 22ï21

ï20

ï19

ï18

ï17

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz


Comparing the EVM results as a function of the distance, illustrated in Fig. 4.11, with the

values presented in Tab. 4.7, it can be seen that they are in agreement. Therefore, it can be

concluded that, when the subcarrier allocation scheme 2 is considered, the performance of each


28 30 32 34 36 38 40 42ï26

ï25

ï24

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

28 30 32 34 36 38 40 42ï22

ï21

ï20

ï19

ï18

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

28 30 32 34 36 38 40 42ï14

ï13

ï12

ï11

ï10

ï9

ï8

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz


ONU of the ASA-OFDMA-PONs is independent of the network reach of the remaining ONUs

in the network. A similar conclusion had been drawn with respect to the subcarrier allocation

scheme 1 in the previous section.


0 10 20 30 40 50 60ï30

ï28

ï26

ï24

ï22

ï20

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï28

ï26

ï24

ï22

ï20

ï18

ï16

ï14

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï22ï20ï18ï16ï14ï12ï10ï8ï6ï4

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz


Fig. 4.11 and 4.14 depict the EVM of the ASA-OFDMA-PON as a function of the distance

when configurations D1 and D4 are considered. As no difference can be seen, similar conclusions

regarding the performance of the system being independent of the frequency of the subcarriers


0 10 20 30 40 50 60ï30

ï28

ï26

ï24

ï22

ï20

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï28

ï26

ï24

ï22

ï20

ï18

ï16

ï14

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï22ï20ï18ï16ï14ï12ï10ï8ï6ï4

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz


assigned to each ONU can be drawn when subcarrier allocation scheme 2 is used.

The disparity of 1-2 dB between the ONUs with more or less subcarriers, discussed in section

4.5.2, is seen as well for the present situation. In Fig. 4.15(a), this difference is seen, and it is


0 10 20 30 40 50 60ï32

ï30

ï28

ï26

ï24

ï22

ï20

ï18

ï16

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]


(b) Sc=10−22A2/Hz


attenuated, as expected, either for ONUs located farther from the OLT, or for higher electrical

noise PSD levels. In Fig. 4.12(a), a closer inspection can be done regarding this matter. The

EVM values of the ONUs assigned more capacity, which are closer to the OLT, are higher than

the ones which are farther from the OLT. This is only seen for network reaches up to 20 km,

since the same behaviour can not be seen when inspecting Fig. 4.13(a).

The ASA-OFDMA-PON performance, when different M -QAM OFDM subcarier modulation

formats are considered, is evaluated in Figs.4.16 and 4.17, for 4-QAM and 32-QAM, respectively.

These results are for the situation where the ONUs in the proposed ASA-OFDMA-PON are

distributed according to the scenario D5.

These results are in accordance with the ones observed in Fig. 4.6(a). The EVM values

required for the acceptable performance limit when a BER threshold of 10−3 is considered,

are higher for lower modulation M -QAM formats. Therefore, a much longer network length

is achievable for lower modulation M -QAM formats. However, the EVM values as a function


0 10 20 30 40 50 60ï30ï29ï28ï27ï26ï25ï24ï23ï22ï21ï20

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]


(b) Sc=10−22A2/Hz


of the distance are lower for higher M -QAM modulations. The power fading which affects the

signal with larger bandwidth is the reason for this to happen. The impact of the power fading

on the signals with lower M -QAM formats is analysed in section 2.4.2.2.

The conclusions stated above hold for the PONs with 32 and 64 ONUs. However, and as

described in the previous sections, the network reaches decrease when the number of ONUs

increases. The reasons for this were described in section 4.5.2. The achievable network lengths

for the ASA-OFDMA-PON with 32 and 64 ONUs are in concordance with the ones presented in

Tab. 4.7 of section 4.5.2, as can be observed from inspecting Figs. E.7-E.8 and Figs. E.9-E.10,

for 32 and 64 ONUs, respectively.

Similar achievable PON network reaches to the ones observed before, were obtained for the

case where the ONUs are at different distances from the OLT and the capacity of the system

is distributed to the users according to the subcarrier allocation scheme 2. For the 16-ONU

ASA-OFDMA-PON, the performance was under the acceptable EVM limits for distances from


20 to 50 km, for the highest and lowest electrical noise PSD level considered, respectively.

For the 32-ONU ASA-OFDMA-PON, maximum reaches of 50 km are achievable for the

lowest electrical noise PSD values. Since the spatial distributions considered (D1-D5) didn’t

present distances lower than 10 kms no EVM values for the highest electrical noise PSD values

considered were under the acceptable EVM limit of -16.5 dB.

For the 64-ONU ASA-OFDMA-PON, maximum reaches between 10 and 35 km are achievable

when Sc = 10−23 A2/Hz and Sc = 10−24 A2/Hz, respectively.

4.7 Conclusions

In this chapter, the performance of the studied ASA-OFDMA-PON was assessed. The subcarrier

allocation schemes used to demonstrate the ASA-OFDMA-PON operation were detailed.

A study regarding the MZM optimal modulation indexes for the different configurations

considered to evaluate the ASA-OFDMA-PON performance was presented.

The performance of the PPM as an optical channel estimator was evaluated. It was concluded

that while the PPM performs some channel estimation, it is not as efficient as using an equalizer

at the receiver.

In the studies developed to assess the performance of the proposed ASA-OFDMA-PON for

the NG-OANs when transmitting an OFDM signal at 10 Gb/s, network reaches of 60, 50 and

35 km were achieved for PONs with 16, 32 and 64 subcarriers, respectively. These values were

obtained for the lowest electrical noise PSD level considered, 10−24 A2/Hz. The achievable

network reaches of the PONs with 16, 32 and 64 ONUs, are severely reduced when higher

electrical noise PSD values are considered.

It was determined that the M -QAM format for which greater network lengths are achieved,

is 4-QAM. However, using such a low order signal modulation format compromises the signal

spectral efficiency.

Regarding the operation of the proposed ASA-OFDMA-PON, it was confirmed that the

performance of each ONU is independent, not only of the subcarriers’ assignment, but also of

the characteristics (distance to the OLT, capacity assigned) of the remaining ONUs in the PON.

67

Chapter 5

Conclusions and future work

In this chapter, the final conclusions of the work developed in this dissertation are summarized.

Some proposals for future work are presented.

5.1 Final Conclusions

In this work, a scheme of multiple access based on OFDM signal for the NG-OANs has been

studied.

In chapter 1, a survey of the different OFDMA-based schemes proposed for NG-OANs is

presented. Two categories were identified. OFDM-NG-OAN based on a wavelength-division

multiplexing (WDM) architecture, and OFDM-NG-OAN based on sub-carrier allocations. It

was decided to study the performance of the latter for the NG-OANs due to its advantages in

dynamic bandwidth allocations and low cost on optical devices.

In chapter 2, the OFDM signal fundamentals were explained. The electrical and optical

elements, which constitute the OFDM system architecture used in this work, were described.

An architecture, where the signals are processed in the digital domain using IFFTs and FFTs,

was chosen for the OFDM system studied in this work. Electrical I/Q modulation is performed,

followed by the E/O conversion using a single MZM and a LD. Direct detection with a PIN

was the implementation selected for the O/E conversion. The DSB OFDM signal is transmitted

through a SSMF where training symbols are used to perform the channel estimation.

In chapter 3, the proposed ASA-OFDMA-PON scheme architecture and operation are de-

tailed. This scheme allows for dynamic digitally controlled system capacity allocation and it

is based on a preprocessing matrix which is used to modulate the signal so the users are able

to select and recover their information using lower rate ADCs and smaller sized FFTs. It was

concluded that the preprocessing works as intended, meaning that the ONUs are able to recover

68 CONCLUSIONS AND FUTURE WORK

all the information carried by the subcarriers assigned to them. However, the applicability of

the proposed ASA-OFDMA-PON for the NG-OANs is compromised by the limitations in the

subcarrier allocation scheme imposed by the sampling process, as well as the synchronization

required so that each ADC starts the signal sampling at the exact time instant which allows for

the complete data recovery.

In chapter 4, the results of the performance analysis of the proposed ASA-OFDMA-PON

scheme are presented. Several configurations were considered for the evaluation of the NG-

OAN performance and assessment of the OFDMA-based NG-OAN capacity. Namely, different

M -QAM symbol mappings, number of clients being served by the network and number of sub-

carriers assigned to each client. The results showed that network reaches of 60, 50 and 35 km

were achieved for PONs with 16, 32 and 64 subcarriers, respectively. These values were obtained

for the lowest electrical noise PSD level considered, 10−24 A2/Hz. It was concluded that the

electrical noise PSD levels had a great impact in the system performance, since the network

reaches achieved were considerably lower when higher electrical noise PSD values were consid-

ered. It was also concluded that the performance of each ONU is independent of the subcarriers’

assignment and of characteristics of the remaining ONUs in the PON.

5.2 Future Work

The following items are proposed as future work on the topic of this dissertation:

• Evaluation of the ASA-OFDMA-PON performance when optical amplification is per-

formed and, consequently, optical noise is introduced in the system.

• Assessment of the resilience of the proposed scheme to the de-synchronization at the re-

ceiver side. Evaluation of the performance of the system when the ADC starts the sampling

at neighbouring time instants from the one which allows the recovery of all the information

carried in the subcarrier.

• Proposal of an alternative preprocessing technique which is not so dependent on the ADC

start sampling instant.

• Study of the ASA-OFDMA-PON performance when more subcarriers per OFDM signal

are considered, with alternative subcarrier allocation schemes.

• Study of the ASA-OFDMA-PON performance for system capacities used in the NG-OAN,

as 40, 100 Gb/s.

Future Work 69

• Study of the impact of the laser characteristics in the ASA-OFDMA-PON performance,

in particular of the laser linewidth.

• Experimental implementation of the proposed ASA-OFDMA-PON scheme. Comparison

of the results obtained in the experimental setup for the NG-OAN performance with the

numerical simulation results presented in this dissertation.

70 CONCLUSIONS AND FUTURE WORK

71

References

[1] I. W.Shieh, Orthogonal frequency multiplexing for optical communications. Elsevier, 1 ed.,

2010.

[2] N. Cvijetic, “OFDM for next-generation optical access networks,” Journal of Lightwave

Technology , vol. 30, pp. 384–398, February 2012.

[3] L. Cheng, H. Wen, X. Zheng, H. Zhang, and B. Zhou, “Channel characteristic division

OFDM-PON for next generation optical access,” Optics Express, vol. 19, pp. 19129–19134,

September 2011.

[4] N. Cvijetic, M.-F. Huang, E. Ip, Y. Shao, Y.-K. Huang, M. Cvijetic, and T. Wang, “1.92

Tb/s coherent DWDM-OFDMA-PON with no high-speed ONU-side electronics over 100km

SSMF and 1:64 passive split,” Optics Express, vol. 19, pp. 24540–24545, November 2011.

[5] P. Duhamel and H. Hollmann, “Split radix FFT algorithm,” Electronics Letters , vol. 20,

pp. 14–16, 5 1984.

[6] N. Cvijetic, M. Cvijetic, M.-F. Huang, E. Ip, Y.-K. Huang, and T. Wang, “Terabit optical

access networks based on WDM-OFDMA-PON,” Journal of Lightwave Technology , vol. 30,

pp. 493–503, February 2012.

[7] L. Kazovsky, N. Cheng, W.-T. Shaw, D. Gutierrez, and S.-W. Wong, Broadband Optical

Access Networks. John Wiley & Sons, 2011.

[8] N. Cvijetic, D. Qian, J. Hu, and T. Wang, “Orthogonal frequency division multiple access

PON (OFDMA-PON) for colorless upstream transmission beyond 10 Gb/s,” IEEE Journal

on Selected Areas in Communications, vol. 28, pp. 781–790, August 2010.

[9] D. Qian, J. Hu, J. Yu, P. Ji, L. Xu, T. Wang, M. Cvijetic, and T. Kusano, “Experimental

demonstration of a novel OFDM-A based 10Gb/s PON architecture,” in 33rd European

72 REFERENCES

Conference and Ehxibition of Optical Communication , 2007 , pp. 1–2, September 2007,

5.4.1.

[10] D. Hewitt, “Orthogonal frequency division multiplexing using baseband optical single side-

band for simpler adaptive dispersion compensation,” in Optical Fiber Communication and

the National Fiber Optic Engineers Conference, 2007 , pp. 1–3, March 2007, OME7.

[11] N. Cvijetic, M.-F. Huang, E. Ip, Y.-K. Huang, D. Qian, and T. Wang, “1.2 tb/s symmetric

WDM-OFDMA-PON over 90km straight SSMF and 1:N 32 passive split with digitally-

selective ONUs and coherent receiver OLT,” in Optical Fiber Communication Conference

and Exposition (OFC/NFOEC), 2011 and the National Fiber Optic Engineers Conference,

pp. 1–3, 2011.

[12] D. Qian, J. Hu, J. Yu, P. N. Ji, L. Xu, T. Wang, M. Cvijetic, and T. Kusano, “Experimental

demonstration of a novel OFDM-A based 10Gb/s PON architecture,” in 33rd European

Conference and Ehxibition of Optical Communication (ECOC), 2007 , pp. 1–2, 2007.

[13] D. Qian, N. Cvijetic, J. Hu, and T. Wang, “108 Gb/s OFDMA-PON with polarization

multiplexing and direct-detection,” Journal of Lightwave Technology , vol. 28, pp. 484–493,

February 2010.

[14] W. Lim, P. Kourtessis, M. Milosavljevic, and J. Senior, “Dynamic subcarrier allocation for

100 Gbps, 40 km OFDMA-PONS with SLA and COS,” Journal of Lightwave Technology ,

vol. 31, no. 7, pp. 1055–1062, 2013.

[15] W. Shieh, “OFDM for flexible high-speed optical networks,” Journal of Lightwave Tech-

nology , vol. 29, pp. 1560–1577, May 2011.

[16] T. Alves and A. Cartaxo, “Performance degradation due to OFDM-UWB radio signal

transmission along dispersive single-mode fiber,” Photonics Technology Letters, IEEE ,

vol. 21, no. 3, pp. 158–160, February,.

[17] T. Alves and A. Cartaxo, “Distribution of double-sideband OFDM-UWB radio signals in

dispersion compensated long-reach pons,” Journal of Lightwave Technology , vol. 29, no. 16,

pp. 2467–2474, 2011.

[18] R. Shafik, S. Rahman, and R. Islam, “On the extended relationships among EVM, BER

and SNR as performance metrics,” in International Conference on Electrical and Computer

Engineering, 2006. ICECE ’06., pp. 408–411, 2006.

REFERENCES 73

[19] P. Vitthaladevuni, M.-S. Alouini, and J. Kieffer, “Exact BER computation for cross QAM

constellations,” IEEE Transactions on Wireless Communications, vol. 4, pp. 3039 – 3050,

November 2005.

[20] X. Zhang, H. Yu, and G. Wei, “Exact symbol error probability of cross-QAM in AWGN

and fading channels,” EURASIP Journal on Wireless Communications and Networking ,

vol. 2010, no. 1, p. 917954, 2010.

74 REFERENCES

75

Appendix A

OFDM system details

In this appendix, a description of the proposed ASA-OFDMA-PON architecture components

characteristics and their implementation, is presented. In section A.1, the electrical OFDM

system elements, as the constellation mapping and demapping, the DAC, ADC and LPF, the

I/Q modulation and demodulation, the equalization, the electrical noise and the synchronization

in OFDM systems, are described. In section A.2, the optical elements of the implemented

system, as the MZM, the optical fiber and the optical receiver, are presented. In section A.3,

the computation of a DSB OFDM signal delay spread is presented.

A.1 Electrical OFDM transceiver

In this section, the electrical OFDM system transceiver components (transmitter + receiver)

presented in Fig. 3.1 which were not previously analysed in detail, are described.

A.1.1 Constellation mapping and demapping

A constellation map illustrates the modulated signal which was modulated using a digital mod-

ulation scheme, as PSK or QAM. It displays the symbols in a two-dimensional complex plane.

In OFDM systems, QAM modulation is usually adopted.

Three constellation types, (4, 16 and 32-QAM), were implemented to analyse the system

performance. Each M -QAM modulated symbol consists of log2M mapped bits, bi.

For all M -QAM constellations, Gray coding was used for the symbols generation. This

binary numeral system was chosen since it arranges the data so that the bit patterns of adjacent

constellation points, differ by only one bit. The bit mapping for 4 and 16-QAM used in this

work is presented in the tables A.1, A.2 and for simplicity the bit mapping for 32-QAM is shown

as a constellation in Fig.A.1, where I and Q represent the in-phase and quadrature coordinates

76 OFDM SYSTEM DETAILS

in the complex plane, respectively.

Table A.1: Bit mapping for 4-QAM using Gray coding.b0 b1 I Q0 0 -1 -10 1 -1 +11 1 +1 +11 0 +1 -1

Table A.2: Bit mapping for 16-QAM using Gray coding.b0 b1 I b2 b3 Q0 0 -3 0 0 -30 1 -1 0 1 -11 1 +1 1 0 +11 0 +3 1 1 +3

-6 -4 -2 0 2 4 6-6

-4

-2

0

2

4

6

In-phase (AU)

Qua

drat

ure

(AU

)

Scatter Plot 32-QAM

0 1

3 2

4

5

7

6

12

13

15

14

8

9

11

10

24

25

27

26

28

29

31

30

20

21

23

22

16 17

19 18

Figure A.1: Constellation mapping of 32-QAM using gray coding.

So that each modulated signal has the same average power at the mapper output, it is

necessary to normalize each constellation power to one. This normalization is done dividing

each symbols absolute value by the maximum absolute value of the constellation.

The constellation demapping follows the inverse process of the constellation mapping. De-

cision thresholds were defined for each constellation, and the sequence of bits are retrieved

according to the decision region where the received symbol I and Q coordinates, are.

A.1.2 DAC and ADC

In the numerical computation environment where the proposed system was implemented, it is

not possible to generate an analog signal. The simulation in a digital environment is done using

Electrical OFDM transceiver 77

discrete samples. To recreate an analog signal in these conditions, an infinite number of samples

would be required.

Therefore, to emulate a continuous analog waveform in the digital domain, a number of

discrete samples with zero value, Nc, is used between each sample at the IDFT output. In this

work, Nc = 25 samples were used to this effect, that will be referred as oversampling.

An LPF, consisting of an rectangular ideal filter is used after the oversampling, and a con-

tinuous waveform is obtained at the LPF output.

The ADC has a simple implementation, which consists of selecting the original digital samples

when the signal is received. The introduced samples with zero value are discarded.

A.1.3 Up-converter and down-converter

As described in the section 2.4, the up an down-converter are complex-to-real and real-to-

complex converters. The baseband signal is composed by complex values, however as the signal

transmitted through the optical channel is necessarily real-valued, the up-conversion from a

complex to real signal is necessary. The down-conversion executes the conversion in the opposite

order.

The bandpass real-valued signal is mathematically expressed by the equation Eq.2.15.

The complex baseband signal at the down-converter output is mathematically expressed by:

sr(t) = sOFDM (t) cos(2πfRF t+∆φ)− sOFDM (t) sin(2πfRF t+∆φ) (A.1)

where sOFDM (t) is the received bandpass signal at the channel output. ∆φ is the sine and

cosine phase offset compensation. This parameter is obtained by transmitting the carriers in

the channel and analysing the phase offset which occurs between the up and down-converter.

The phase compensation is done in order to guarantee the same sine and cosine phase at the

transmitter and receiver. Thus, the transmitted and received signals will be synchronized.

In the Figs.A.2 and A.3, the spectra of the signals sOFDM (t) and sr(t) in back-to-back

electrical OFDM system, are presented.

The bandpass signal is transmitted in the carrier frequency fRF = 4 GHz. There are other

configurations for OFDM systems that do not require these converters, since the signal is already

real valued. In these systems, the OFDM signal is modulated with Hermitian symmetry, where

half the subcarriers contain the complex conjugate of the modulated symbols [1].


ï10 ï8 ï6 ï4 ï2 0 2 4 6 8 10ï40

ï30

ï20

ï10

0

10

20

Frequency [GHz]

PSD

[dBW

/Hz]

Figure A.2: Spectrum of the signal sOFDM after the up-conversion is performed.

ï4 ï2 0 2 4

ï20

ï10

0

10

20

30

Frequency [GHz]

PSD

[dBW

/Hz]

Figure A.3: Spectrum of the signal sr(t) after the down-conversion is performed.

A.1.4 Equalizer

An equalizer is used in OFDM systems to reduce the distortion of the OFDM symbols. The

signal equalization comprises of estimating, at the receiver, the channel and other components

distortions which affected the signal during the transmission, and to apply an equalization

function to the received signal according to that estimation. Training symbols are commonly

used to obtain the equalization function.

The distortion estimation is done using the channel frequency response, HE , which relates

the transmitted and the received OFDM symbols. For a sequence of training symbols, TS, the

channel frequency response is expressed as:

HE =sTS

STS(A.2)

where sTS is the received training sequence and STS is the generated training sequence.

This transfer function has all the information in the frequency domain of the transmission

effects and is sufficient to relate all the amplitude and phase differences of each subcarrier. To

Optical components of the transmission system 79

compensate these effects, the equalizer uses H−1E as the transfer function to mitigate all the

distortions present in the transmission channel. The equalizer transfer function, GE is given by:

GE =1

HE=

STS

sTS(A.3)

In this work, NT = 10 training symbols were used to generate the equalization function.

A.1.5 Electrical Noise

The electrical noise is intrinsic in every electrical component of a real system. However, for

implementation simplicity, it is considered that the electrical noise in the system is present in

the OFDM receiver input.

Usually, the electrical noise is characterized by the square root of the PSD of the electrically

generated noise given by:√

Sc(f) =

√

4 · kBTr

Rbias· fn,e (A.4)

where kB is the Boltzmann constant, Tr the room temperature which is usually 290 Kelvin, Rbias

the bias resistance, and fn,e the noise figure representing the influence of the active components.

Typical values for√

Sc(f) are between 1 pA/√Hz and 10 pA/

√Hz [2]. In this work, the values

considered for the sqare root of the PSD were√

Sc(f)=1, 10 and 100 pA/√Hz.

A.2 Optical components of the transmission system

In this section, the optical components of the optical fiber communication system are described.

The optical modulators are detailed in section A.2.1, the optical passive filter in section A.2.2,

the optical fiber in section A.2.3 and the optical receiver in section A.2.4.

A.2.1 Optical modulator

The optical transmitter converts the electrical signal that transports the information to an

optical signal, and launches it into the optical fiber.

In order to generate the optical signal, an optical source based on semiconductors is com-

monly used. The light-generation process occurs in certain semiconductor materials due to

recombination of electrons and holes in positive-negative (PN) junctions, under direct biasing

[2]. Depending on the nature of the recombination process, different semiconductor light sources

can be classified as either light-emitting diodes (LEDs), or laser diodes (LDs).

In this work, the laser model considered is a laser diode (LD) , that generates a carrier with

zero linewidth (a single frequency implies no phase noise) and with a constant power.


An optical modulator is a device which modulates an electrical signal into an optical car-

rier. The modulator used in this work is the Mach-Zehnder modulator. MZMs are external

modulators, which are based on the electro-optic effect (the effect that in certain materials, the

refractive index changes with respect to the voltage applied across electrodes) [1].

The transfer function of an ideal MZM can be described using a periodic cosine function,

illustrated in Fig.A.4, which relates the optical intensity and the optical field against the drive

voltage.

It can be observed in Fig. A.4 that the MZM bias point for the optical intensity modulation

can be set at the quadrature point, where the behaviour of the optical intensity is approximately

linear. As a note, for the optical field modulation, a null bias point is better suited since it is

the point where the behaviour of the optical field curve is approximately linear.

Figure A.4: Transfer functions of the optical intensity and the optical field against the drivevoltage.

The optical signal at the MZM output is given by:

Es(t) = ELD(t) cos

(

π

2·sOFDM (t) + VDC

Vπ

)

· ejωLDt · ejφLD(t) (A.5)

where ELD(t) is the optical field amplitude of the source laser diode, sOFDM(t) is the bandpass

signal, centered in the angular frequency fRF , VDC is the direct current (DC) bias voltage of

the MZM, Vπ is the half-wave switching voltage and ωLD and φLD(t) are the frequency and the

phase of the source laser, respectively.

In this work, a quadrature bias point was used to polarize the MZM. According to the

Fig.A.4, when the MZM has a quadrature bias point, VDC is −Vπ2 . The half-wave switching

voltage value considered was Vπ = 5 V.

A.2.2 Optical power splitter

Optical splitters are passive components used to split optical power into multiple fibers. Being

passive components they do not require any external power to perform their functionalities, thus

Optical components of the transmission system 81

being more reliable and require less maintenance effort [7].

According to the splitting ratio, 1N , where N is the number of branches of the power splitter,

the optical power, Pin, at the splitter input is split into N with equal power Pout = PinN if

assuming a splitter that divides the input power equally with excess losses.

The splitter has considerable losses. The following equation was used in this work to calculate

the splitter losses:

Ls = 10 log10(N) + Le log2(N) (A.6)

where N is the number of ONUs and Le are the excess losses. It was considered a value of 0.8

dB for the splitter excess losses. As it is possible to conclude from the equation A.6, the splitter

losses will increase with the number of ONUs considered.

A.2.3 Optical fiber

Optical fibers are used as the transmission medium in optical communication systems. The

optical fibers can be classified into two main categories, multi-mode fibers (MMFs) and single-

mode fibers (SMFs), according to the number of propagation modes which supported by the

fiber.

Standard single-mode fiber (SSMF), due to its low-cost, is used in most PON deployments.

For this reason, it was the transmission medium chosen to be used in this work.

While the optical fibers are non-linear transmission mediums, it was assumed, for simplifi-

cation purposes, a linear behaviour for the optical fiber.

The model used in this work to represent the optical fiber accounts for the fiber disper-

sion, which describes the effect where different components of the transmitted signal propagate

at different velocities, thus arriving at different times at the receiver. This effect causes the

broadening of the signal impulses by spreading the signal over time originating inter-symbol

interference (ISI). Fiber losses are also considered in the model used in this work to represent

the optical fiber. Fiber losses, usually under 0.5 dB/km, are a characteristic aspect of these

transmission mediums. The losses in the fiber attenuate the signal power.

The optical fiber model used in this work is described by the following transfer function:

Hfib(ν) = e−jβ(2πν)L · e−α2 L (A.7)

where α[Np/m] represents the fiber power losses coeficient, L[m] is the total fiber length, includ-

ing both feeder and distribution fibers, ν[Hz] is the optical frequency and β[rad/m] is the fiber

propagation constant.


The propagation constant, β, depends on the optical frequency, and its exact value cannot

be calculated [1]. An approximate value is obtained expanding a Taylor series around the optical

carrier frequency ν0. The series expansion is:

β(Ω) ≈ β0 + β1 · Ω+β22

· Ω2 +β36

· Ω3 (A.8)

where Ω = 2π(ν − ν0)[Hz] is the equivalent angular baseband frequency. The β0 and β1 val-

ues account for the propagation constant in the carrier frequency and the propagation delay,

respectively. These terms are neglected in this work, since signal synchronization was assumed,

meaning that no temporal broadening is imposed to the signal. The terms β2 and β3 account

for the group velocity dispersion (GVD) and second-order GVD, respectively. They are given

by:

β2 =Dλ · λ2

2πc(A.9)

β3 =Dλ · λ3

2π2c2+

(

λ2

2πc

)

· Sλ0

where Dλ[ps/nm/km] is the fiber dispersion parameter, λ0[m] is the wavelength related to the

optical carrier frequency by λ0 =cν0, c[m/s] is the speed of light in vaccum and Sλ0 [ps/nm

2/km]

is the fiber dispersion slope.

In this work, the values considered for the referred parameters are λ0=1552.52 nm, Dλ=17

ps/nm/km, Sλ0=70 [fs/nm2/km], c=2997.92×108 m/s and α=0.2 dB/km.

Finally, combining the Eqs. A.7 and A.8, the optical fiber model used in this work is

expressed as:

Hfib(ν) = e−j

!

β22 Ω2+

β36 ·Ω3

"

L · e−α2 L (A.10)

A.2.4 Optical receiver

An optical receiver converts the optical signal to an electrical signal. This is accomplished by

a photodiode. The photodiode absorbs photons in the incoming optical signal and converts

them back to the electrical domain in a process which is the inverse of the one taking place

in semiconductor lasers [1]. Common photodiodes are the positive-negative photodiode (PN),

positive-intrinsic-negative photodiode (PIN), avalanche photodiode, and metal-semiconductor-

metal photodetectors. Of all these common photodiodes, the one used in this work is the PIN

photodiode, which consists of an intrinsic region placed between the positive and negative layers.

The photocurrent at the PIN output, Iout, is proportional to the optical power present at

Optical DSB OFDM signals dispersive time delay spread computation 83

the photodetector input,

Iout(t) = Rλ· | Er(t) |2 (A.11)

where Rλ represents the PIN responsivity and | Er(t) |2 represents the instantaneous power of

the received optical field Er(t).

A.3 Optical DSB OFDM signals dispersive time delay spread

computation

The dispersive time delay spread, td, is given by the Eq. 2.5 presented in section 2.2.

The value used in this work for the dispersion parameter of the fiber is 17 ps/nm/km and

since the percentage of CP required in an OFDM signal increases with the fiber length, to

compute the minimum guard interval required it is used the maximum reach considered in this

work, which is 50 km.

The optical DSB (ODSB) signals presented in Figs.2.3 and 2.4, centered in the carrier fre-

quency fRF = 4 GHz, considered in this work, occupy a total bandwidth which is given by:

∆ν =

(

fRF +Bw

2

)

−(

−fRF −Bw

2

)

= 2fRF +Bw (A.12)

where Bw is the baseband OFDM signal bandwidth described in Eq. 2.15.

The spectral width, ∆λ, of the ODSB OFDM signal in the fiber in nanometres is given by:

∆λ =λ20

c∆ν (A.13)

where λ0 is the central optical wavelength. In this work, the signals are transmitted in the third

window, therefore the value considered for λ0 is 1552.52 nm.

Having computed ∆λ, it is now possible to obtain the dispersive delay spread using the

expression presented in Eq. 2.5.

85

Appendix B

Performance analysis

In this work, the error vector magnitude (EVM) as a function of the root-mean-square (RMS)

voltage of the OFDM signal was adopted as the parameter to evaluate the performance of the

proposed architecture. In this appendix, the EVM description and usage, as well as the EVM

performance limits defined from the bit error rate (BER) values, are presented.

B.1 Error vector magnitude

The EVM is an useful parameter to quantify the digital transmission quality. The EVMmeasures

the difference of amplitude and phase between the value of the actual received symbol and the

expected value for the same symbol.

Fig.B.1 illustrates the EVM vector theory. In Fig.B.1, the actual received symbol is rep-

resented by a square, whereas the expected symbol, which has the mapped constellation ideal

value, is represented by a circle. The EVM is the distance between the two markers. In an ideal

transmission system, this distance is zero.

Q

IIkref

Qkref

Qk

Ik

Error Vector

Figure B.1: EVM concept.

86 PERFORMANCE ANALYSIS

In decibel (dB), the EVM is defined as:

EVMdB = 20 log10

⎛

⎜

⎜

⎜

⎝

NOFDM∑

l=1

Nsc∑

k=1

[

(

Il,kref − Il,k

)2+

(

Ql,kref −Ql,k

)2]

NOFDM∑

l=1

Nsc∑

k=1

[

Il,kref 2 +Ql,k

ref 2]

⎞

⎟

⎟

⎟

⎠

(B.1)

where NOFDM stands for the number of OFDM information symbols that contribute to the

performance evaluation, Nsc is the number of subcarriers within one OFDM symbol, Il,kref and

Ql,kref represent the in-phase and quadrature ideal constellation symbols respectively, and Il,k

and Ql,k represent the in-phase and quadrature measured constellation symbols, respectively.

B.2 BER

The EVM performance limits are calculated for a defined BER value. The BER is defined as

the ratio between the number of bit errors and the number of transmitted bits.

Assuming the error of the constellation symbols at the QAM demapper as Gaussian dis-

tributed, the BER of each OFDM subcarrier, k, can be evaluated from the EVM of each sub-

carrier resorting to the expressions used to evaluate the performance of a M -QAM modulation

format when M = 22n, with n as an integer [18]:

BER(k) = 21− 1√

M

log2Merfc

⎛

⎝

√

3 log2√M

M − 1

1

EVM(k) log2M

⎞

⎠ (B.2)

where erfc(x) is the complementary error function, and the EVM of the kth subcarrier is given

by:

EVM(k)dB = 20 log10

⎛

⎜

⎜

⎜

⎝

NOFDM∑

l=1

[

(

Il,kref − Il,k

)2+

(

Ql,kref −Ql,k

)2]

NOFDM∑

l=1

[

Il,kref 2 +Ql,k

ref 2]

⎞

⎟

⎟

⎟

⎠

(B.3)

The Eqs. B.2 and B.3 can only be used for rectangular QAM (M = 22n) constellations, which

is the case for both 4 and 16-QAM. For 32-QAM, which has a non-rectangular constellation,

the previous equations cannot be used to derive the BER value. To derive the expression to

calculate the exact value of the BER for these constellations extensive calculations have to be

made, since the bits for the different decision regions have different contributions for the BER

equation [19].

However, an approximate expression of the symbol-error-rate (SER), which translates the

error between the received and the transmitted modulated symbols of these constellations, it is

BER 87

easier to deduct. According to [20], the SER for a 32-QAM constellation is given by:

SER =1

M·1

8

(

13 · erfc(√

y

20

))

−

(

23

4· erfc

(√

y

5

)2)

+1

2· erfc

(

2 ·√

y

5

)

(B.4)

where erfc() is the complementary error function and y is the SNR in linear units given by:

y =1

10EV M10

(B.5)

The BER depends on the bit to symbol mapping, so with Gray-coded bits and for high

SNRs, the BER is approximately given by:

BER ≈SER

log2M. (B.6)

Since Gray coding was used for everyM -QAM constellation in this work, the equations described

above were used to calculate the BER for the 32-QAM constellation considered.

Fig.B.2 presents a comparison of the BER values as a function of the EVM for the M -QAM

constellations implemented in this work.

ï30 ï25 ï20 ï15 ï10 ï5

10ï1510ï1210ï10

10ï510ï3100

EVM[dB]

BER


Figure B.2: BER as a function of the EVM for different M -QAM constellations.

There is a performance degradation, i.e. the EVM values required to achieve the same BER

values decrease, as the number of bits per symbol increases. The reasons for this degradation

were referred in section 2.3.

Considering 10−3 and 10−12 as the BER limits before forward error correction (FEC) and

after FEC, respectively, the EVM values required from each contellation to achieve this perfor-

88 PERFORMANCE ANALYSIS

mance are presented in the Tab.B.1.

Table B.1: EVM limits when BER is 10−3 and 10−12.EVM [dB]

M -QAM BER=10−3 BER=10−12

4-QAM -9.8 -16.916-QAM -16.5 -23.932-QAM -19.4 -26.9

As can be seen in Tab. B.1, the desired performance for 32-QAM requires an EVM 10 dB

lower than the one required by 4-QAM, and 3 dB lower than the one required by 16-QAM.

89

Appendix C

Subcarrier allocation schemes

In this appendix, the two subcarrier assignment schemes used to evaluate the performance of the

proposed ASA-OFDMA-PON, are detailed. In section C.1, a scheme were the system capacity

is equally distributed among the ONUs, is presented. In section C.2, an alternative scheme,

where the system capacity is not uniformly distributed among the ONUs, thus representing the

situation where the subcarriers are assigned according to the user demand, is presented.

For the cases where 16 and 32 ONUs are served, OFDM symbols with 128 subcarriers are

transmitted. When the PONs serve 64 ONUs, OFDM symbols with 256 subcarriers are used.

The computation complexity required to demonstrate the ASA-OFDMA-PON operation using

OFDM symbols with 256 subcarriers is very high. However, to study the 64 ONU PON in similar

conditions to the 16 and 32-ONUs PONs, 256 subcarriers are necessary, despite the computation

complexity necessary.

The subcarriers assignment schemes are illustrated using tables where for each ONUx, 1 ≤

x ≤ N , the indexes k of the subcarriers assigned to that ONU, are presented.

The distribution of the OFDM subcarriers on the frequency spectrum is illustrated in Fig.C.1.

The table presented in Fig.C.2 details the correspondence between the indexes which are assigned

to the subcarriers and the frequency, fk they occupy in the spectrum.

f [Hz]

......

0−2Δf −Δf Δf 2ΔfNsc

2Δf− Nsc

2−1⎛

⎝⎜⎞⎠⎟ Δf

Figure C.1: Distribution of the OFDM subcarrier frequencies in the frequency spectrum.

90 SUBCARRIER ALLOCATION SCHEMES

fk − Nsc

2−1⎛

⎝⎜⎞⎠⎟ Δf −

Nsc

2− 2⎛

⎝⎜⎞⎠⎟ Δf −2Δf −Δf Δf 2Δf Nsc

2−1⎛

⎝⎜⎞⎠⎟ Δf

Nsc

2Δf

Nsc

2+1 Nsc

2+ 2 Nsc − 2 Nsc −1 Nsc

Nsc

2−1 Nsc

2

... 0 ...

k ... 1 2 ...

Figure C.2: Correspondence between the OFDM subcarriers indexes k and their position in thefrequency spectrum.

C.1 Scheme 1 - System capacity uniformely distributed by the

ONUs

In this scheme all the ONUs are assigned the same number of subcarriers. The subcarrier

allocation schemes are as follows:

Table C.1: Subcarriers assignment scheme 1 for 16 ONUs.ONUx Subcarrier index k

1 1 17 33 49 65 81 97 1132 2 18 34 50 66 82 98 1143 3 19 35 51 67 83 99 1154 4 20 36 52 68 84 100 1165 5 21 37 53 69 85 101 1176 6 22 38 54 70 86 102 1187 7 23 39 55 71 87 103 1198 8 24 40 56 72 88 104 1209 9 25 41 57 73 89 105 12110 10 26 42 58 74 90 106 12211 11 27 43 59 75 91 107 12312 12 28 44 60 76 92 108 12413 13 29 45 61 77 93 109 12514 14 30 46 62 78 94 110 12615 15 31 47 63 79 95 111 12716 16 32 48 64 80 96 112 128

Tables C.1, C.2 and C.3 illustrate the situation where the same number of subcarriers NscNONUs

is assigned to the 16, 32 and 64 PONs, respectively.

In the subcarrier allocation scheme 1, the distribution of a certain subcarrier, k, by each

ONU x, is expressed by:

k = x+NONU · n,

n ∈ N |Nsc

NONU− 1

(C.1)

Scheme 1 - System capacity uniformely distributed by the ONUs 91


1 1 33 65 972 2 34 66 983 3 35 67 994 4 36 68 1005 5 37 69 1016 6 38 70 1027 7 39 71 1038 8 40 72 1049 9 41 73 10510 10 42 74 10611 11 43 75 10712 12 44 76 10813 13 45 77 10914 14 46 78 11015 15 47 79 11116 16 48 80 112

ONUx Subcarrier index k

17 17 49 81 11318 18 50 82 11419 19 51 83 11520 20 52 84 11621 21 53 85 11722 22 54 86 11823 23 55 87 11924 24 56 88 12025 25 57 89 12126 26 58 90 12227 27 59 91 12328 28 60 92 12429 29 61 93 12530 30 62 94 12631 31 63 95 12732 32 64 96 128



1 1 65 129 1932 2 66 130 1943 3 67 131 1954 4 68 132 1965 5 69 133 1976 6 70 134 1987 7 71 135 1998 8 72 136 2009 9 73 137 20110 10 74 138 20211 11 75 139 20312 12 76 140 20413 13 77 141 20514 14 78 142 20615 15 79 143 20716 16 80 144 20817 17 81 145 20918 18 82 146 21019 19 83 147 21120 20 84 148 21221 21 85 149 21322 22 86 150 21423 23 87 151 21524 24 88 152 21625 25 89 153 21726 26 90 154 21827 27 91 155 21928 28 92 156 22029 29 93 157 22130 30 94 158 22231 31 95 159 22332 32 96 160 224


33 33 97 161 22534 34 98 162 22635 35 99 163 22736 36 100 164 22837 37 101 165 22938 38 102 166 23039 39 103 167 23140 40 104 168 23241 41 105 169 23342 42 106 170 23443 43 107 171 23544 44 108 172 23645 45 109 173 23746 46 110 174 23847 47 111 175 23948 48 112 176 24049 49 113 177 24150 50 114 178 24251 51 115 179 24352 52 116 180 24453 53 117 181 24554 54 118 182 24655 55 119 183 24756 56 120 184 24857 57 121 185 24958 58 122 186 25059 59 123 187 25160 60 124 188 25261 61 125 189 25362 62 126 190 25463 63 127 191 25564 64 128 192 256

Scheme 2 - System capacity non uniformely distributed by the ONUs 93

C.2 Scheme 2 - System capacity non uniformely distributed by

the ONUs

In this section, the allocation scheme where the ONUs are not all assigned the same number of

subcarriers, is detailed.

Following the subcarrier allocation requirements of the ASA-OFDMA-PON which were spec-

ified in section 3.4 in chapter 3, the presented assigned schemes were obtained for PONs with

16, 32 and 64 ONUs.


1 1 9 17 25 33 41 49 57 65 73 81 89 97 105 113 1212 3 11 19 27 35 43 51 59 67 75 83 91 99 107 115 1233 5 13 21 29 37 45 53 61 69 77 85 93 101 109 117 1254 7 15 23 31 39 47 55 63 71 79 87 95 103 111 119 1275 2 18 34 50 66 82 98 1146 4 20 36 52 68 84 100 1167 6 22 38 54 70 86 102 1188 8 24 40 56 72 88 104 1209 10 42 74 10610 12 44 76 10811 14 46 78 11012 16 48 80 11213 26 58 90 12214 28 60 92 12415 30 62 94 12616 32 64 96 128

Tables C.4, C.5 and C.6 illustrate the situation where the system capacity is not uniformly

distributed among the ONUs.



1 1 17 33 49 65 81 97 1132 3 19 35 51 67 83 99 1153 5 21 37 53 69 85 101 1174 7 23 39 55 71 87 103 1195 9 25 41 57 73 89 105 1216 11 27 43 59 75 91 107 1237 13 29 45 61 77 93 109 1258 15 31 47 63 79 95 111 1279 2 34 66 9810 4 36 68 10011 6 38 70 10212 8 40 72 10413 10 42 74 10614 12 44 76 10815 14 46 78 11016 16 48 80 11217 18 8218 20 8419 22 8620 24 8821 26 9022 28 9223 30 9424 32 9625 50 11426 52 11627 54 11828 56 12029 58 12230 60 12431 62 12632 64 128

Scheme 2 - System capacity non uniformely distributed by the ONUs 95


1 1 33 65 97 129 161 193 2252 3 35 67 99 131 163 195 2273 5 37 69 101 133 165 197 2294 7 39 71 103 135 167 199 2315 9 41 73 105 137 169 201 2336 11 43 75 107 139 171 203 2357 13 45 77 109 141 173 205 2378 15 47 79 111 143 175 207 2399 17 49 81 113 145 177 209 24110 19 51 83 115 147 179 211 24311 21 53 85 117 149 181 213 24512 23 55 87 119 151 183 215 24713 25 57 89 121 153 185 217 24914 27 59 91 123 155 187 219 25115 29 61 93 125 157 189 221 25316 31 63 95 127 159 191 223 25517 2 66 130 19418 4 68 132 19619 6 70 134 19820 8 72 136 20021 10 74 138 20222 12 76 140 20423 14 78 142 20624 16 80 144 20825 18 82 146 21026 20 84 148 21227 22 86 150 21428 24 88 152 21629 26 90 154 21830 28 92 156 22031 30 94 158 22232 32 96 160 224


33 34 16234 36 16435 38 16636 40 16837 42 17038 44 17239 46 17440 48 17641 50 17842 52 18043 54 18244 56 18445 58 18646 60 18847 62 19048 64 19249 98 22650 100 22851 102 23052 104 23253 106 23454 108 23655 110 23856 112 24057 114 24258 116 24459 118 24660 120 24861 122 25062 124 25263 126 25464 128 256

97

Appendix D

Distributions of the ONU-OLT distance in

the ASA-OFDMA-PON network

In section 4.6 of chapter 4, the performance of the proposed ASA-OFDMA-PON when the ONUs

are at different distances from the OLT, is assessed.

In order to perform this analysis, different ONU spatial distributions were created. These

distributions are described in this appendix.

Each ONU is assigned an index x. These indexes are in accordance with the ones presented

in Appendix C with respect to the subcarriers assignment by the ONUs.

D.1 ASA-OFDMA-PON with 16 ONUs

The following scenarios are considered when the proposed ASA-OFDMA-PON serves 16 clients.

Table D.1: ONUs distances to the OLT. Scenario D1: ONUs with long and short networkreaches.

Distances [km] ONU x index10 1 220 3 4 530 6 7 8 9 1040 11 1250 13 14 15 16

98DISTRIBUTIONS OF THE ONU-OLT DISTANCE IN THE ASA-OFDMA-PON

NETWORK

Table D.2: ONUs distances to the OLT. Scenario D4: Same as D1 but for different ONUs.Distances [km] ONU x index

50 1 240 3 4 530 6 7 8 9 1020 11 1210 13 14 15 16

Table D.3: ONUs distances to the OLT. Scenario D2: Short network reaches for every ONU.Distances [km] ONU x index

10 13 14 15 1612 11 1214 6 7 8 9 1016 3 4 520 1 2

Table D.4: ONUs distances to the OLT. Scenario D3: Long network reaches for every ONU.Distances [km] ONU x index

30 13 14 15 1632 11 1236 6 7 8 9 1038 3 4 540 1 2

Table D.5: ONUs distances to the OLT. Scenario D5: ONUs assigned the same number ofsubcarriers at different distances from the OLT.

Distances [km] ONU x index10 1 5 9 1420 2 6 1030 3 7 1140 4 12 1550 8 13 16




Distances [km] ONU x index10 1 2 3 420 5 6 7 8 9 1030 11 12 13 14 15 16 17 18 19 2040 21 22 23 2450 25 26 27 28 29 30 31 32

ASA-OFDMA-PON with 64 ONUs 99


50 1 2 3 440 5 6 7 8 9 1030 11 12 13 14 15 16 17 18 19 2020 21 22 23 2410 25 26 27 28 29 30 31 32


10 1 2 3 412 5 6 7 8 9 1014 11 12 13 14 15 16 17 18 19 2016 21 22 23 2420 25 26 27 28 29 30 31 32


30 1 2 3 432 5 6 7 8 9 1036 11 12 13 14 15 16 17 18 19 2038 21 22 23 2440 25 26 27 28 29 30 31 32


Distances [km] ONU x index10 1 6 9 14 19 24 2920 2 7 10 15 20 25 3030 3 8 11 16 21 26 3140 4 12 17 22 27 3250 5 13 18 23 28




Distances [km] ONU x index10 1 2 3 4 5 6 7 820 9 10 11 12 13 14 15 16 17 18 19 2030 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 4040 41 42 43 44 45 46 47 4850 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64

100DISTRIBUTIONS OF THE ONU-OLT DISTANCE IN THE ASA-OFDMA-PON

NETWORK


50 1 2 3 4 5 6 7 840 9 10 11 12 13 14 15 16 17 18 19 2030 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 4020 41 42 43 44 45 46 47 4810 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64


10 1 2 3 4 5 6 7 812 9 10 11 12 13 14 15 16 17 18 19 2014 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 4016 41 42 43 44 45 46 47 4820 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64


30 1 2 3 4 5 6 7 832 9 10 11 12 13 14 15 16 17 18 19 2036 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 4038 41 42 43 44 45 46 47 4840 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64


Distances [km] ONU x index30 1 6 11 16 21 26 31 36 41 46 51 56 6132 2 7 12 17 22 27 32 37 42 47 52 57 6236 3 8 13 18 23 28 33 38 43 48 53 58 6338 4 9 14 19 24 29 34 39 44 49 54 59 6440 5 10 15 20 25 30 35 40 45 50 55 60

101

Appendix E

Additional Results

In this appendix, the results obtained using the proposed ASA-OFDMA-PON but not included

in chapter 4, are presented.

E.1 ASA-OFDMA-PON performance when the ONUs have the

same network reach

In this subsection, the results when the subcarrier allocation scheme 2 is used and the ONUs

are all at the same distance from the OLT, are presented. Fig. E.1 presents the results obtained

when the ASA-OFDMA-PON serves 32 ONUs and Fig.E.2 presents the results obtained when

the ASA-OFDMA-PON serves 64 ONUs.

102 ADDITIONAL RESULTS

0 10 20 30 40 50 60ï28

ï26

ï24

ï22

ï20

ï18

ï16

Distance [km]

EVM

[dB

]

NscONU=8

NscONU=4

NscONU=2

(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

NscONU=8

NscONU=4

NscONU=2

(b) Sc=10−23A2/Hz

0 10 20 30 40ï21

ï19

ï17

ï15

ï13

ï11

ï9

ï7

ï5

Distance [km]

EVM

[dB]

NscONU=8

NscONU=4

NscONU=2

(c) Sc=10−22A2/Hz

Figure E.1: EVM as a function of the distance for an 16-QAM OFDM signal being transmittedover 32-ONUs.

ASA-OFDMA-PON performance when the ONUs have the same network reach 103

0 10 20 30 40 50 60ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB

]

NscONU=8

NscONU=4

NscONU=2

(a) Sc=10−24A2/Hz

0 10 20 30 40 50ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

NscONU=8

NscONU=4

NscONU=2

(b) Sc=10−23A2/Hz

0 10 20ï17

ï15

ï13

ï11

ï9

ï7

ï5

Distance [km]

EVM

[dB]

NscONU=8

NscONU=4

NscONU=2

(c) Sc=10−22A2/Hz

Figure E.2: EVM as a function of the distance for an 16-QAM OFDM signal being transmittedover 64-ONUs.


E.2 ASA-OFDMA-PON performance when the ONUs have dif-

ferent network reaches

Figs. E.3 and E.4 present the EVM results as function of the distance, obtained when the 32

ONUs of the PON are distributed according to the scenarios D1-D5, presented in section D.2,

and the subcarrier allocation scheme 1 is used. Figs. E.5 and E.6 present these results when

the PON serves 64 ONUs.

Figs. E.7 and E.8 present the EVM results as a function of the distance when the OFDM

subcarriers are assigned to the ONUs according to subcarrier allocation scheme 2 and the 32

ONUs of the PON are distributed according to the scenarios D1 and D5, respectively. Figs. E.9

and E.10 present these results when the PON serves 64 ONUs.

ASA-OFDMA-PON performance when the ONUs have different network reaches 105

Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

10 20 30 40 50ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

(a) Scenario D1


Sc=10ï24 A2/Hz

10 12 14 16 18 20ï26

ï24

ï22

ï20

ï18

ï16

ï14

ï12

ï10

Distance [km]

EVM

[dB]

(b) Scenario D2

Sc=10ï22 A2/Hz


30 32 34 36 38 40ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

(c) Scenario D3

Figure E.3: EVM as a function of the distance when transmitting a 16-QAM OFDM signal tothe 32-ONUs of the ASA-OFDMA-PON which are distributed according to (a) D1, (b) D2 and(c) D3.


Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

10 20 30 40 50ï27ï25ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

Figure E.4: EVM as a function of the distance when transmitting a 16-QAM OFDM signal tothe 32-ONUs of the ASA-OFDMA-PON which are distributed according to D4.


Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

10 20 30 40 50ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB

]

(a) Scenario D1

Sc=10ï22 A2/Hz

Sc=10ï24 A2/Hz

Sc=10ï23 A2/Hz

10 12 14 16 18 20ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB]

(b) Scenario D2


30 32 34 36 38 40ï21

ï19

ï17

ï15

ï13

ï11

ï9

ï7

ï5

Distance [km]

EVM

[dB]

(c) Scenario D3

Figure E.5: EVM as a function of the distance when transmitting a 16-QAM OFDM signal tothe 64-ONUs of the ASA-OFDMA-PON which are distributed according to (a) D1, (b) D2 and(c) D3.


Sc=10ï22 A2/Hz

Sc=10ï23 A2/Hz

Sc=10ï24 A2/Hz

10 20 30 40 50ï23ï21ï19ï17ï15ï13ï11ï9ï7ï5

Distance [km]

EVM

[dB

]



0 10 20 30 40 50 60ï26

ï24

ï22

ï20

ï18

ï16

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï23

ï21

ï19

ï17

ï15

ï13

ï11

ï9

ï7

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï16

ï14

ï12

ï10

ï8

ï6

ï4

ï2

0

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz



0 10 20 30 40 50 60ï28

ï26

ï24

ï22

ï20

ï18

ï16

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï26ï24ï22ï20ï18ï16ï14ï12ï10ï8ï6

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï16

ï14

ï12

ï10

ï8

ï6

ï4

ï2

0

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz



0 10 20 30 40 50 60ï21

ï19

ï17

ï15

ï13

ï11

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï18ï16ï14ï12ï10ï8ï6ï4ï2

0

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï13

ï11

ï9

ï7

ï5

Distances [km]

EVM

[dB]


(c) Sc=10−22A2/Hz



0 10 20 30 40 50 60ï25

ï23

ï21

ï19

ï17

ï15

ï13

ï11

Distance [km]

EVM

[dB

]


(a) Sc=10−24A2/Hz

0 10 20 30 40 50 60ï20ï18ï16ï14ï12ï10ï8ï6ï4ï2

Distance [km]

EVM

[dB]


(b) Sc=10−23A2/Hz

0 10 20 30 40 50 60ï13

ï11

ï9

ï7

ï5

Distance [km]

EVM

[dB]


(c) Sc=10−22A2/Hz


Study&of&a&Scheme&of&Multiple&Access&in& … · iii Abstract Multiple access based on orthogonal...

Documents

Transcript of Study&of&a&Scheme&of&Multiple&Access&in& … · iii Abstract Multiple access based on orthogonal...