Studies on Placement of Semiconductor Optical Amplifiers in Wavelength Division Multiplexed Star and Tree Topology Networks by

Yatindra Nath Singh

submitted in fulfilment of the requirement of degree of Doctor of Philosophy (Ph.D.) to Electrical Engineering Department Indian Institute of Technology, Delhi Hauz Khas, New Delhi 110016 India September 1996

Certificate

This is to certify that the thesis entitled "Studies on Placement of Semiconductor Optical Amplifiers in Wavelength Division Multiplexed Star and Tree Topology Networks" being submitted by Mr. Yatindra Nath Singh to the Department of Electrical Engineering, Indian Institute of Technology, Delhi is the record of the bonafide research work carried out by him. He has worked under our supervision and guidance during the period December 1992 to August 1996. He has fulfilled all the requirements for submission of the thesis which has reached the requisite standard.

The results contained in this thesis have not been submitted either in part or in full to any other university or institute for the award of any degree or diploma.

(Prof. Hari Mohan Gupta)

(Dr. Virander Kumar Jain) Thesis Supervisors Department of Electrical Engineering Indian Institute of Technology, Delhi Hauz Khas, New Delhi India

ii

Acknowledgement

I am highly indebted to my thesis supervisors Prof. Hari Mohan Gupta and Dr. Virander Kumar Jain for giving me the opportunity to work under their supervision. They had encouraged me in period of distress and anxiety and guided me to accomplish this research work.

I must acknowledge the staff of optical communication laboratory, IIT Delhi (Mr. A. P. Thukral, Ms. Neeru Asija and Mr. J. P. Naudiyal) for their support and help during the research work.

My parents had kept me free from my duties at home so that I can devote more time to my research work. I acknowledge them for their blessings and support.

My friends Mr. Brejesh Lall and Mr. Bharat Gupta must also be acknowledged for their patience while listening to me about my research work and for their suggestions and encouragement.

I sincerely thank all the persons who have helped and supported me directly or indirectly in the course of this research work.

(Yatindra Nath Singh)

iii

iv

Abstract

This thesis is mainly concerned with the use of semiconductor optical amplifiers (SOAs) in wavelength division multiplexed (WDM) broadcast topologies viz. star and tree. The aim of investigations is to determine the increase in the number of users when SOAs are placed in the above topologies. Various placement options have been considered in the above investigations.

Error correcting codes can also be used in optical networks to increase the number of users by improving the available power budget. Therefore, an on-off keying (OOK) communication system has been investigated with SOA as preamplifier and an error correcting code. The coding gain in the single channel point-to-point link has been computed. In this study, the baseband filter in the receiver is assumed to be optimum for uncoded system. Hence, the coding gain in the above includes the effect of increased intersymbol interference (ISI). It is compared with the gain obtained due to placement of SOA in the link. The suitability of SOA over coding has been discussed based on the above comparison. It is shown that SOA provides much higher gain than the use of error correcting code.

A detailed study on the use of SOAs in WDM star topology is also carried out. Two schemes have been considered viz. SOAs as postamplifiers and preamplifiers. In the postamplifier scheme, SOAs are placed after the transmitter and in the preamplifier scheme before the receiver. In the postamplifier scheme, three cases have been investigated. First case corresponds to unsaturated SOAs, second to gain saturated SOAs and in third case effect of reflection noise is studied. Similarly, in the preamplifier case unsaturated SOAs, average gain

saturated SOAs with and without gain fluctuations have been considered. Further, reflection noise is considered as in postamplifier scheme. It is observed that for typical values of system parameters, SOAs as preamplifiers perform better than the postamplifiers.

The study on placement of SOAs has been extended to WDM tree topology passive broadcast network i.e. WDM tree-net. The tree-net consists of star as main topology and folded bus as auxiliary topology. Star portion consists of a star coupler. A tree-net with b branches uses bxb star coupler. The SOAs can be placed in the star portion of the tree-net to increase the supportable number of users. However, the number of SOAs can be smaller than b to support the same number of users as in star. In this study unsaturated SOAs and average gain saturated SOAs with and without gain fluctuations have been considered. On comparison with star, it is observed that a tree-net can support more users than a star for a given number of SOAs.

It is concluded that SOAs as preamplifiers perform better in star topology. When SOAs are used in tree topology, number of users supported can be more than in star. Further, the required number of SOAs would be lesser in tree than in star for a given number of users.

vi

Table of Contents
List of Figures List of Tables List of symbols Abbreviations Chapter 1 Introduction 1.1 Need for Optical Communications 1.2 Fiber Optic Networks . . . . . . . . . 1.3 Multiplexing Schemes . . . . . . . . . 1.4 WDM Broadcast Networks . . . . . 1.5 Thesis Outline . . . . . . . . . . . . . . ix xiii xv xix

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . .

1 1 2 4 7 10

Chapter 2 Optical Amplifiers in Broadcast Networks 2.1 Optical Amplifiers . . . . . . . . . . . . . . . . . 2.2 Optical Amplifiers in Broadcast Networks 2.2.1 Bus . . . . . . . . . . . . . . . . . . . . . . 2.2.2 Dual Bus . . . . . . . . . . . . . . . . . . 2.2.3 Ring . . . . . . . . . . . . . . . . . . . . . 2.2.4 Star . . . . . . . . . . . . . . . . . . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

. . . . . .

13 13 20 20 22 23 26

Chapter 3 Semiconductor Optical Amplifier and Coding in OOK 3.1 Uncoded System . . . . . . . . . . . . . . . . . . . . 3.2 Coded System . . . . . . . . . . . . . . . . . . . . . . 3.3 Uncoded System with Optical Amplifier . . 3.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . 3.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . .

System ...... ...... ...... ...... ......

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

. . . . .

37 40 43 45 48 49

Chapter 4 Semiconductor Optical Amplifiers in WDM Star Topology 4.1 System Description . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2 Amplifier Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3 Evaluation of Minimum Required Transmitter Power 4.4 Star without Amplifier . . . . . . . . . . . . . . . . . . . . . . . 4.5 Star with Postamplifier . . . . . . . . . . . . . . . . . . . . . . . 4.5.1 Unsaturated Amplifiers . . . . . . . . . . . . . . . . 4.5.2 Effect of Gain Saturation . . . . . . . . . . . . . . . 4.5.3 Effect of Reflection Noise . . . . . . . . . . . . . . . 4.6 Star with Preamplifier . . . . . . . . . . . . . . . . . . . . . . . 4.6.1 Unsaturated Amplifier . . . . . . . . . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

. . . . . . . . . .

51 52 54 56 56 59 59 62 64 69 70

vii

4.6.2 Effect of Gain Saturation 4.6.3 Effect of Reflection Noise 4.7 Example . . . . . . . . . . . . . . . . . . . 4.8 Conclusions . . . . . . . . . . . . . . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

70 75 79 88

Chapter 5 Semiconductor Optical Amplifiers In WDM Tree-net 5.1 System Description . . . . . . . . . . . . . . . . . 5.2 Analytical Model . . . . . . . . . . . . . . . . . . 5.3 Example . . . . . . . . . . . . . . . . . . . . . . . . . 5.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . Chapter 6 Conclusions References Appendix - I Appendix -II

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

. . . .

91 . . 92 . . 96 . 100 . 107

109 113 123 129

viii

List of Figures
Fig.2-1 Fig.2-2 Fig.2-3 Fig.2-4 Fig.2-5a Fig.2-5b Fig.2-5c Fig.2-6 Fig.2-7 Fig.2-8 Fig.2-9 Fig.2-10 Fig.2-11 Fig.2-12 Fig.2-13 Fig.2-14 Fig.2-15 Fig.2-16 Fig.2-17 Fig.2-18 Fig.3-1a Fig.3-1b Fig.3-2 psds of (i) ASE-signal beat noise, (ii) ASE-ASE beat noise and (iii) ASE-shot noise components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Forward pumped doped fiber amplifier . . . . . . . . . . . . . . . . . . . . . . Optical amplifiers in bus topology. . . . . . . . . . . . . . . . . . . . . . . . . . Dual bus network using fiber as point-to-point link. . . . . . . . . . . . . . Schematic of non-regenerative photonic dual bus (NI is the node interface). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Node interface for head of the bus (node 1). . . . . . . . . . . . . . . . . . . . Node interface for node 2 to N. . . . . . . . . . . . . . . . . . . . . . . . . . . . Ring network with distributed EDFA. . . . . . . . . . . . . . . . . . . . . . . . 4x4 star coupler with single EDFA. . . . . . . . . . . . . . . . . . . . . . . . . . 8x8 coupler based on 4x4 star couplers. . . . . . . . . . . . . . . . . . . . . . . Implementation of distributed Nu x Nu (Nu=m2) reflective star coupler based on m number of m x m couplers. . . . . . . . . . . . . . . . . . . . . . . 16x16 distributed reflective star coupler. . . . . . . . . . . . . . . . . . . . . . Implementation of distributed Nu x Nu (Nu= 2 m2) coupler based on 2m number of mxm couplers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32x32 distributed reflective star coupler. . . . . . . . . . . . . . . . . . . . . . FDM/OFDM distributed expandable amplified star coupler configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Star coupler with amplifiers and without bandpass and bandstop filters. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8x1 coupler based on seven 2x2 couplers. . . . . . . . . . . . . . . . . . . . . Forward pumped EDFA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . mmc x mmc reflective star. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . mc x mc reflective star coupler based on mcx1 tree coupler. . . . . . . . . . The system block diagram for optical communication system with coding scheme. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The system block diagram for optical communication system with SOA as preamplifier. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Worst-case waveform patterns for (a) bit 1 and (b) bit 0 before and after the Gaussian filter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 18 21 23 23 24 24 25 27 27 28 29 30 31 32 33 34 35 35 36 38 38 42

ix

Fig.3-3

Variations of BER vs received power level (dBm) for (i) uncoded system, (ii) coded system without decoding, (iii) coded system with decoding and (iv) uncoded system with SOA as preamplifier. . . . . . . . An 8x8 star coupler consisting of 3 dB 2x2 couplers. . . . . . . . . . . . . Star network without optical amplifier. . . . . . . . . . . . . . . . . . . . . . . . Star network with postamplifiers (A, B, ...F is echo path). . . . . . . . . . Star network with preamplifiers (A, B, ...G is echo path). . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.0 in postamplifier scheme . . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.05 in postamplifier scheme . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.10 in postamplifier scheme . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.15 in postamplifier scheme . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.0 in preamplifier scheme . . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.05 in preamplifier scheme . . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.10 in preamplifier scheme . . . . . . . . . . . Variations of minimum required average transmitter power Pta with number of users Nu for = 0.15 in preamplifier scheme . . . . . . . . . . . Variations of Pta with for different Nu in postamplifier scheme. Solid lines correspond to gain saturated SOAs and dashed lines to gain saturated SOAs with reflection noise. . . . . . . . . . . . . . . . . . . . . . . . . Variations of Pta with for different Nu in preamplifier scheme. Solid lines correspond to gain saturated SOAs and dashed lines to gain saturated SOAs with reflections. . . . . . . . . . . . . . . . . . . . . . . . . . . . Basic tree-net topology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Structure of star coupler in tree-net without amplifiers. . . . . . . . . . . . Structure of branches in a tree-net without amplifiers. . . . . . . . . . . . . A 4x4 star coupler with one amplifier in tree-net. . . . . . . . . . . . . . . . A 4x4 star coupler with two amplifiers in tree-net. . . . . . . . . . . . . . . A 4x4 star coupler with four amplifiers in tree-net. . . . . . . . . . . . . . .

49 53 58 60 77 80 81 82 83 84 85 86 87

Fig.4-1 Fig.4-2 Fig.4-3 Fig.4-4 Fig.4-5 Fig.4-6 Fig.4-7 Fig.4-8 Fig.4-9 Fig.4-10 Fig.4-11 Fig.4-12 Fig.4-13

88

Fig.4-14

89 92 93 94 94 95 95

Fig.5-1 Fig.5-2a Fig.5-2b Fig.5-3a Fig.5-3b Fig.5-3c

Fig.5-4a Fig.5-4b Fig.5-5a Fig.5-5b Fig.5-6a

Variations of Nu with Na (unsaturated SOAs) for n=2. The values in the brackets are corresponding required Pta in dBm. . . . . . . . . . . . . . . . Variations of Nu with Na (unsaturated SOAs) for n=3. The values in the brackets are corresponding required Pta in dBm. . . . . . . . . . . . . . . . Variations of Nu with Na (average gain saturated SOAs) for n=2. The values in the brackets are corresponding required Pta in dBm. . . . . . . Variations of Nu with Na (average gain saturated SOAs) for n=3. The values in the brackets are corresponding required Pta in dBm. . . . . . . Variations of Nu with Na (average gain saturated SOAs with gain fluctuations) for n=2. The values in the brackets are corresponding required Pta in dBm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Variations of Nu with Na (average gain saturated SOAs with gain fluctuations) for n=3. The values in the brackets are corresponding required Pta in dBm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Reflected and transmitted signals in a Fabry-Perot amplifier. . . . . . .

102 102 103 103

104

Fig.5-6b

104 124

Fig.A-1

xi

xii

List of Tables
Table 4-1 Table 4-2 Table 5.1 Table 5.2 Table 5.3 Number of users supported in the postamplifier scheme for =0.10. . . Number of users supported in the preamplifier scheme for =0.10. . . . Number of users and required Pta for the tree-net without SOAs. . . . Maximum number of users and Pta for a given number of SOAs. Numbers in brackets are the corresponding values of n. . . . . . . . . . . Comparison of star network and tree-net in terms of number of SOAs required for a given number of users. . . . . . . . . . . . . . . . . . . . . . . 86 90 105 107 108

xiii

xiv

List of symbols
a0() a1() b bc br B Be Bo Cf Dth e eecho er erfc(x) Worst-case waveform for bit 0 in the presence of instersymbol interference (ISI) Worst-case waveform for bit 1 in the presence of ISI Number of branches in the tree-net Bit in the desired channel Bit whose echo produces the reflection noise Bit rate per user Bandwidth of electrical filter in receiver Optical filter/demultiplexer bandwidth Velocity of light in fiber Decision threshold level Electronic charge Echo signal electric field Signal electric field at the receiver

Ei Eo Er g go G G0 Gav GFP Gp Gp(br) h

Input electric field amplitude at Fabry-Perot cavity Output electric field amplitude at Fabry-Perot cavity Reflected electric field amplitude from Fabry-Perot cavity Gain coefficient of SOA Unsaturated gain coefficient of SOA Amplifier gain Unsaturated amplifier gain Average gain of SOA when multiple channels are being amplified Gain of F-P amplifier Gain of SOA when signal corresponding to bit br is amplified Gain of SOA when signal corresponding to bit br is amplified Plancks constant

xv

h() H(f) iecho-sig Isig k kb L Las1 Lc Lcp Lcv Li Lfi Lg Lsp Lss Lti Ltr mc M n nsp N Na Nu P Pe Pes(ad) Pes(bd) Pin Pint

Impulse response of electrical filter in the receiver Frequency response of electrical filter in the receiver Echo and signal beat noise current Signal current at photodetector Number of data symbols in a coded block Boltzmanns constant Length of fiber from user to star coupler Loss between SOA and splice 1 Coherence length of laser linewidth Amplifier coupling loss Star coupler variability Insertion loss of 3 dB 2x2 coupler Insertion loss of filter or demultiplexer Length of gain medium Fiber splice loss Distance between two consecutive users on a branch in the tree-net Total insertion loss of star coupler Loss between transmitter and receiver Size of central coupler Maximum number of users on a bus segment without SOA Number of users per branch in the tree-net Spontaneous emission factor Number of symbols in a block Number of amplifiers Number of users (stations) Optical power Probability of error Probability of symbol error after decoding Probability of symbol error before decoding Input optical power to SOA Total input power to amplifier

xvi

Pout Pr Pr(0) Pr(1) Pra Psat Pt Pta Pt(0) Pt(1) R R1 R2 Ramp Rf RL Ro Rsp S0 S1 SASE-shot(f) SN(f) Sshot Ssp Sspi Sth t T Tamp Tb Tmax

Output power of SOA Received optical signal power at the photodetector Received optical signal power at the photodetector for bit 0 Received optical signal power at the photodetector for bit 1 Average received optical power at the photodetector Saturation power level of SOA Transmitted optical power Average transmitted optical power Transmitted optical power for bit 0 Transmitted optical power for bit 1 Data rate Reflection coefficient of facet 1 Reflection coefficient of facet 2 Reflectivity of SOA Facet reflectivity of SOA (Rf = R1 = R2) Receiver load resistance Responsivity of photodetector Reflection coefficient of splice Signal level for bit 0 at time t=0 for worst-case bit sequence in presence of ISI Signal level for bit 1 at time t=0 for worst-case bit sequence in presence of ISI psd of shot noise due to ASE psd of additive Gaussian noise psd of shot noise Single sided ASE noise psd due to amplifier psd of noise at the receiver due to ith amplifier in postamplifier scheme in star network psd of thermal noise Number of correctable errors in a coded block Absolute temperature in Kelvin Transmittance of amplifier cavity Bit period Maximum value of Tamp

xvii

W1 L c d p s m 2(1) 2(0) 2(bc) 2ASE-ASE 2ASE-shot 2ASE-sig 2n

Wavelength band 1 in WDM/FDMA system Attenuation coefficient of fiber in dB/km Propagation constant of electric field in a given media Extinction ratio Laser linewidth Longitudinal mode spacing of FP cavity Quantum efficiency of photodetector Operating wavelength Wavelength for clock broadcast in photonic dual bus Wavelength for data broadcast in photonic dual bus Pump wavelength Signal wavelength Optical frequency Resonant frequency of mth peak in FP cavity transmission spectrum Noise variance for bit 1 Noise variance for bit 0 Noise variance for bit bc ASE-ASE beat noise variance Shot noise variance due to ASE noise ASE-signal beat noise variance Noise variance

xviii

Abbreviations
ASE BER BPF BSF CDM CDMA dc DFA DQDB EDFA FCS FDDI FDM FPA FT-FR FT-TR HIPPI ISI LAN MAC MAN OA O/E OOK psd RATO SAN SDH SOA Amplifier Spontaneous Emission Bit Error Rate Band Pass Filter Band Stop Filter Code Division Multiplexing Code Division Multiple Access Direct Current Doped Fiber Amplifier Distributed Queue Dual Bus Erbium Doped Fiber Amplifier Fiber Channel Standard Fiber Distributed Data Interface Frequency Division Multiplexing Fabry-Perot Amplifier Fixed Transmitter-Fixed Receiver Fixed Transmitter-Tunable Receiver HIgh Performance Parallel Interface Inter Symbol Interference Local Area Network Media Access Control Metropolitan Area Network Optical Amplifier Optical-to-Electronic On-Off Keying Power Spectral Density Random Access Time Out Subscriber Access Network Synchronous Digital Hierarchy Semiconductor Optical Amplifier

xix

SONET TDM TT-FR TT-TR TWA WAN WDM

Synchronous Optical NETwork Time Division Multiplexing Tunable Transmitter-Fixed Receiver Tunable Transmitter-Tunable Receiver Travelling Wave Amplifier Wide Area Network Wavelength Division Multiplexing

xx

Chapter 1

Introduction

1.1 Need for Optical Communications

Technical breakthroughs are affecting the human society for ages. It is well known that discovery of fire, agriculture and wheel has made the human beings superior over all the other species on the earth. In the last two centuries, industrial revolution started by the invention of steam engine by James Watt has changed the face of humanity. Nowadays, we are witnessing information revolution which is affecting the humanity.

Information revolution implies that the information can be presented, analyzed, transported in an efficient manner. Computers are facilitating the presentation and analysis of information, while the transportation of information is achieved using communication networks. With the passage of time, communication networks are becoming independent of type of information. Now, the information moving over the networks can be voice, video, computer data or text. The networking for supporting such information is termed as multimedia networking. Multimedia networks are designed to carry a large amount of information bits per second. The high bit rates are especially needed by the upcoming high bandwidth real-time video services [58] (e.g. video conferencing and video on demand). At present, optical fiber is the only transmission medium offering such large bandwidth [27, 28, 39]. Optical fiber also provides the low loss communication links as compared to

radio or electrical cables. In addition to large bandwidth and low loss, there are many other advantages of using optical fiber. Optical fibers are immune to electromagnetic interference; comparatively lighter and cheaper than copper cables required to carry the same amount of information. As the transmission of information is through dielectric media, transmitting and receiving ends are electrically isolated. Optical links are more reliable and can support future application due to inherently large available capacity.

Optical fiber links are best suited for fixed user locations. These links require specialized tools and skills for installation and maintenance. Further, the technology of optical devices and components, especially optical sources and detectors is still evolving to match the system requirements.

Optical fiber is finding increasing use in the communication networks due to above advantages [42]. The telecommunication service providers have laid down the optical fiber links on the land and in the oceans for trunk traffic. Optical fibers are finding increasing use in subscriber loop too [20, 43, 52, 65, 85, 95]. It is also used in computer networking [6, 66].

1.2 Fiber Optic Networks

Optical fiber has found use in switched communication networks by replacing the other media viz. microwave, electrical cable. Nowadays, almost all the trunk lines of existing networks are using optical fiber. This large scale use of optical fiber has led to the development of synchronous digital hierarchy (SDH)/ synchronous optical network (SONET) standards for switched networks [35, 73, 82]. In future, the link between the user and the

switch i.e. subscriber loop will also be based on fiber. Consequently, new services (e.g. videophony, video on demand) can be made available by telecommunication service providers.

The fiber has also replaced the media in the broadcast and select networks which are commonly used as local area networks (LANs). Consequently, ethernet, token bus and token ring have been implemented on optical fiber. As the fiber is used to replace conventional media, these networks should be designed to harness the full fiber capabilities. The broadcast networks can be classified as active or passive. When the signal is optoelectronically regenerated or amplified, the network is active broadcast network otherwise it is passive broadcast network. Ethernet and token bus are the examples of passive broadcast network, while token ring is of active broadcast network. In passive broadcast networks, the transmitted optical power is split for distribution among the receivers. Therefore, increase in number of users implies that each receiver will receive reduced amount of optical power for a fixed transmitted power level. The reduction in received optical power level will worsen the receiver performance. To keep the receiver performance acceptable, the number of users have to be limited. Such a limitation on number of users does not exist in active broadcast networks, but the active broadcast networks suffer with other problems e.g. low reliability, higher cost etc..

New networks e.g. fiber distributed data interface (FDDI) [25, 26, 74] and distributed queue dual bus (DQDB) have also been deployed with optical fiber as media. These networks operate at higher bit rates and cover larger distances. FDDI uses ring topology. It contains two counter propagating rings for high reliability. In case of failure of a node or a link, the fault is bypassed using the other ring. DQDB uses dual bus or dual ring as topology. Each

of the two bus or ring is unidirectional and carries data in opposite directions. In the above networks, optical fiber is used as point-to-point link connecting two consecutive nodes. However, it is possible to use fiber as passive broadcast media by using passive taps in DQDB network. These networks operate at bit rates of the order of 100 Mb/s due to optical media. The network standards with still higher bit rates (of the order of Gb/s) have been developed e.g. fiber channel standard (FCS) and high performance parallel interface (HIPPI). HIPPI uses multiple parallel links, each of which is operating at bit rates of the order of 100 Mb/s. The parallel links can use electrical cable for shorter distances (few meters) or optical fiber for longer distances (few kms). Further increase in capacity of broadcast networks for efficient utilization of optical fiber has been hindered by electronic bottleneck. The transmitter and receiver become extremely costly with the increase in bit rates above Gb/s. At high bit rates, optical multiplexing and demultiplexing techniques are used to attain large capacity networks. Large capacity switched optical networks are also possible [34, 93]. In these networks, many switches are interconnected. The signal remains optical from transmitting to receiving end. Therefore, switching is optical in nature and optical multiplexing and switching techniques will be used for increasing the capacity [2].

1.3 Multiplexing Schemes

The bandwidth requirement is expected to increase with time. Therefore, research in optical fiber networks is directed towards attaining high capacity optical networks. As mentioned above, high capacities in these networks can be achieved by using various multiplexing techniques. There are three possible approaches: (i) wavelength division

multiplexing (WDM), (ii) time division multiplexing (TDM) and (iii) code division multiplexing (CDM). Combinations of the above can also be used.

A WDM system is similar to frequency division multiplexed (FDM) radio broadcasting system. The optical spectrum is sliced in multiple channels centred at different wavelengths [8, 9, 53, 77, 78, 79]. Ideally, the information in various channels do not interfere with each other. However, interchannel interference may occur when the information passes through optical amplifiers (OAs) or any other nonlinear device in the link. Implementation of WDM systems require couplers [69], wavelength multiplexers and demultiplexers, tunable or fixed wavelength sources and detectors capable of operating over the entire wavelength range. Also the various parameters characterizing the OAs, couplers, fibers etc. are desired to be uniform in the operating wavelength range. There are many issues which must be tackled in a WDM system. Two of these are: (i) wavelength stability of sources and (ii) channel spacing. If the transmitter source wavelength is not stable, it may drift towards the neighbouring wavelength channels. Consequently, channels spacing will reduce resulting in crosstalk. Various wavelength stabilizing schemes have been proposed to avoid drift [70]. Most of these schemes compare the transmitting wavelength with a reference and adjust the transmitting source wavelength continuously. Channel spacing is an important issue [81]. Smaller spacing implies more channels in a given spectral range. But the demultiplexers resolution decides the spacing. For smaller channel spacing, narrow bandwidth filters are required. When coherent receivers are used, channel spacing can be very small (2 to 4 times the signal bandwidth) as compared to when noncoherent receivers are used [10, 11]. The WDM systems with a smaller channel spacing are termed as optical FDM systems [33].

In TDM systems, optical pulses from various sources carrying the information are time multiplexed [4, 41]. In these systems, the pulse width reduces with the increase in number of users when bit rate per user is fixed. It must be 1/(NuB) where Nu is number of users and B the bit rate per user. Such a system requires the optical sources which generate very narrow width optical pulses. Generally, mode locked lasers can be used for this purpose. The output of these are then externally modulated depending upon the information to be transmitted. If internal modulation is used on commonly used laser sources, pulse compression has to be applied to generate narrow pulses. Such a compression is generally achieved using nonlinear processing. In optical TDM, the pulses from various sources require very tight delay adjustment so that pulses can be multiplexed together. Further, exact timing information is required at all the receivers for demultiplexing [3] which can be obtained by centralized clock or synchronized distributed clocks. Therefore, implementation of optical TDM system requires tight clock jitter control and clock synchronisation at very high frequencies ( 1 THz).

In CDM systems, each user is assigned a unique waveform code to represent the information bit [1, 21, 51]. The codes are selected in such a manner that code sequence of each user has very small correlation with code sequences of all the other users. Thus codes are equivalent to independent channels. At the receiving end, optical correlators are used for identifying the information destined for it. As the number of waveform codes are equal to number of users, increase in number of users implies more number of pulses in waveform code for a bit. Consequently, for a given bit rate the pulse width will decrease with the increase in number of users. In a code division multiple access (CDMA) broadcast network having more than 100 users, bit rate of 1 Gb/s per user implies pulses of the order of 10 picoseconds. These short pulses can be generated by a mode locked laser. With these

narrow pulses and proper optical delay lines, suitable waveform code can be generated [40]. At the receiver end, delay lines are used to decode the required waveform. In a broadcast situation either encoders or decoders or both can be tunable. Each waveform code is used as an independent channel. In encoders and decoders, optical fiber delay lines which require tight fiber length tolerances are used. To make the delay lines, tunable switching can be used. Fine tuning of delay can be achieved using Piezoelectric tuning of delay lines.

It may be mentioned that for implementing a WDM system, fixed/tunable sources, wavelength multiplexers and demultiplexers and fixed/tunable receivers are required. These devices are available at the current state of the art technology [5, 15, 16, 29, 56]. In comparison, implementation of encoders/decoders and time multiplexers/demultiplexers at Tb/s is very difficult. Therefore, WDM implementation is comparatively easier than that of TDM and CDM [13]. It is expected that WDM switched as well as broadcast networks will dominate the future for quite some time. As WDM systems can provide large bandwidth, complexity of switched network can be avoided by using broadcast network topologies.

1.4 WDM Broadcast Networks

Many issues must be investigated for implementing a WDM broadcast network. Some of the important issues are: (i) topology, (ii) tunability characteristics of transmitter, (iii) tunability characteristics of receiver, (iv) size of the network, (v) number of channels and (vi) media access control protocols. Some issues are inter related also.

Fiber layout to connect the different users constitute the physical topology of the networks [66]. A physical broadcast topology must have some desirable features. A few of these are given below.

(i)

In a WDM system, there must be large number of users in the topology to utilize the available bandwidth. Ideally, each receiver must receive minimum required power for a given bit error rate (BER) to maximize the number of users supported. A topology must also provide each receiver with equal amount of power so that receivers with smaller dynamic range can be used. Sometimes, equal distribution of power can also lead to more number of users (e.g. star topology supports more users than the bus).

(ii)

Fiber used in the network topology must be minimum. This is an important cost consideration.

(iii)

Transmitter to receiver loss within the topology must be minimum. This would mean more power is available to receivers. Therefore, number of users in the network can be increased.

(iv)

Topology must be extendible. It means that number of users can be increased easily. When the above is not met, the topology is designed for more number of users than actually present. The unused ends are latter assigned to the new users. It is also desired that media access control (MAC) protocol is independent of number of users in the network.

(v)

WDM optical network are generally very high speed networks [12]. In such networks, transmission time is much less than the propagation time. Therefore, the average propagation time in the network must be minimized while selecting a topology. This is extremely important for better performance.

(vi)

Minimization of number of couplers, multiplexers and demultiplexers, transmitters and receivers and optical amplifiers in the network topologies is also required. This is an important cost reduction measure in a large network.

In the broadcast network topologies, star, bus, tree or ring can be used. Among these, star is the best option as it can support maximum number of users for a given power budget. The reason for the above is uniform distribution of transmitted power by a star coupler.

In WDM networks, efficient methods (i.e. MAC protocols) for accessing the bandwidth are desired [31, 36, 44, 45, 60, 64, 71, 103]. In these protocols, contention for the wavelength channels to be used and time when the packet is to be transmitted are resolved. There are several possibilities depending on the tunability of transmitters and receivers. One can have fixed transmitter - fixed receiver (FT-FR), fixed transmitter - tunable receiver (FT-TR), tunable transmitter - fixed receiver (TT-FR) and tunable transmitter - tunable receiver (TT-TR). The protocols for the cases when tunable components are used are required to take into consideration the tuning range and tuning time of the tunable components. When the tunable range of components is small, it may not be possible to implement the MAC protocol on all the channels simultaneously. Therefore, the protocol needs to be modified depending upon the number of channels covered by the tunable range.

High capacity WDM networks [97] should be able to support large number of users. At the current level of technology, the broadcast optical networks support very few users ( 64 users). The number of users can be increased by (i) increasing the power budget, (ii) reducing the loss in the network, (iii) using the opto-electronic regenerators and (iv) using optical amplifiers (OAs). The available power budget can be improved by using transmitters with higher power sources and receivers with better sensitivity. Higher transmitter power and better receiver sensitivity need improvement in device technology. The error control coding can be used to provide coding gain and hence the improvement in the receiver sensitivity which in turn increases the available power budget [38]. Coding techniques require the encoder and decoder blocks in the transmitter and receiver respectively. The reduction of loss in the network requires the improvement in splicing techniques and optical components. The optoelectronic regenerators involve optical-to-electronic (O/E) conversion and vice-versa. This reduces the reliability of the network as the regenerator is an active device. Further, the upgradation of multichannel WDM network will require separate regenerator for each channel. In contrast to the above, an optical amplifier can amplify multiple WDM channels.

1.5 Thesis Outline

This thesis is mainly concerned with the use of semiconductor optical amplifiers (SOAs) in WDM broadcast topologies viz. star and tree. The aim of investigations is to determine the increase in the number of users in the above topologies when SOAs are placed. Further, various SOA placement options are studied. In chapter 2, the characteristics and suitability of SOAs in networks has been reviewed. The relevant SOA model and parameters have been presented.

10

In chapter 3, the improvement in an on-off keying (OOK) communication system due to coding and semiconductor optical preamplifier has been investigated. The coding gain in the single channel point-to-point link has been computed. In this study, the baseband filter is assumed to be optimum [37] for uncoded system. Hence, the coding gain in the above includes the effect of increased intersymbol interference (ISI). It is compared with the gain obtained due to placement of SOA in the link. The suitability of SOA over coding has been discussed based on the above comparison [91, 92, 96, 99, 102].

Chapter 4 contains a detailed study on the use of SOAs in WDM star topology. Two placement schemes, postamplifier and preamplifier, have been considered. In the postamplifier scheme, SOAs are placed after the transmitter and in the preamplifier scheme before the receiver. In the postamplifier scheme, three cases have been investigated. First case corresponds to unsaturated SOAs, second to gain saturated SOAs and in third case reflection noise is incorporated. Similarly, in the preamplifier scheme unsaturated SOAs and average gain saturated SOAs with and without gain fluctuations have been considered. Further, the effect of reflection noise is also investigated like in postamplifier scheme [98, 100].

In chapter 5, the study on placement of SOAs is extended to WDM tree topology passive broadcast network i.e. WDM tree-net. The tree-net consists of star as main topology and folded bus as auxiliary topology. Star portion consists of a star coupler. The deployable number of SOAs has been investigated in the tree-net. This number varies from 1 to b, where b is number of branches in the tree-net. In this study, unsaturated SOAs and average gain saturated SOAs with and without gain fluctuations have been considered [101].

11

In chapter 6, conclusions of the thesis are presented. The limitations and shortcomings of approach in the thesis are discussed. Further possibilities of research and extensions have been suggested.

12

Chapter 2

Optical Amplifiers in Broadcast Networks

Optical amplifiers (OAs) can increase the number of users supported by a broadcast network. Many investigations have been carried out to increase the network size using OAs. Some of the schemes are based on semiconductor optical amplifiers (SOAs) and others on doped fiber amplifiers (DFAs). This chapter consists of review of OAs in broadcast networks. Section 2.1 describes operational principles of OAs. Various types of OAs are compared in this section. Specifically, SOAs are modelled for their subsequent analysis in networks. The effectiveness of OAs in various schemes, as reported, has been discussed in section 2.2.

2.1 Optical Amplifiers

Optical amplifier is a device which amplify the input optical signal. This device works on the principle of stimulated emission [7]. There are two types of OAs which are used in communication system; (i) semiconductor optical amplifiers (SOAs) and (ii) doped

The work reported in this chapter has resulted in the following publication
1. Y. N. Singh, V. K. Jain and H. M. Gupta, "WDM Data Network," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994.

13

fiber amplifiers (DFAs). The SOAs are basically semiconductor lasers which operate below lasing threshold [7]. These devices require population inversion so that stimulated emission and hence amplification can take place. The population inversion is achieved by means of electrical energy. A SOA has two facets and reflectivities of these facets decide whether the device will operate as a SOA or a semiconductor laser. When the facet reflectivities are zero, input signal passes through the device only once. Such an amplifier is called travelling wave amplifier (TWA). When reflectivities are non-zero but quite small, the signal passes through the cavity several times. The signal within the cavity reduces with the successive passes and dies out ultimately. This type of SOA is called Fabry-Perot amplifier (FPA).

In SOAs, gain of amplifier reduces as the input signal power increases. This phenomenon is called gain saturation. The gain coefficient g is given by [30]

(2.1)

where g0 is the unsaturated gain coefficient, Psat the saturation power level, P the optical power to the amplifier. The rate of increase of optical power with distance is given by

(2.2)

14

For a TWA, the output power is obtained by using Eqn.2.1 in Eqn.2.2 and integrating Eqn.2.2 from 0 to Lg with initial condition P(0)=Pin, where Lg is the length of gain medium and Pin the input optical power to SOA. The output power is obtained as

(2.3)

where G0 (= exp (g0L)) is unsaturated amplifier gain. The amplifier gain G (=Pout/Pin) from the above equation is given by

(2.4)

A semiconductor laser amplifier alongwith the amplified signal, also produces amplified spontaneous emission (ASE) noise. It is modelled as a white Gaussian noise with single sided power spectral density (psd) [30] (2.5)

where nsp is the spontaneous emission factor, h the Plancks constant and the optical frequency. Typically, the unsaturated gain, G0, of SOAs is in the range 8 - 29 dB and the Psat in the range 3.1 - 15.6 dBm. The noise figure for SOAs ranges from 6 to 8.5 dB which corresponds to nsp of 2 - 3.5 [19, 47, 62, 87]. When the amplified signal and accompanying ASE noise are made to incident at the photodetector, the noise beats with itself and also with the signal. Thus ASE-ASE and ASE-signal beat noise components are produced. The

15

psds of the above noise components and ASE shot noise which arises due to dc component of ASE-ASE beat noise are shown in Fig.2-1 [63]. In this figure, Bo is the bandwidth of optical filter used for limiting the ASE noise of OA, Ro the responsivity of photodetector and Pr the received optical power level at the photodetector.

Fig.2-1

psds of (i) ASE-signal beat noise, (ii) ASE-ASE beat noise and (iii) ASE-shot noise components

In a FPA, the gain is given by [30]

(2.6)

In the above equation, R1 and R2 are the facet reflectivities, G() the gain spectrum when the reflectivities are zero, L the longitudinal mode spacing and m the resonant

16

frequencies for the cavity. When [30].

the FPA can be approximated as TWA

In DFAs, optical fiber core is doped by rare earth elements e.g. erbium [17, 18, 55], praseodymium [48, 88] etc.. These amplifiers like SOAs also use stimulated emission for amplifying the optical signals. In contrast to SOAs, population inversion in these amplifiers is achieved by optical pumping. The basic scheme for optical pumping is as shown in Fig.2-2. A coupler is used for combining the information signal and the pump signal. When these signals travel in a doped fiber, power at the pump wavelength p is absorbed and it creates population inversion. The signal at s, the signal wavelength, amplifies as it passes through the doped fiber core having population inversion. The pump power is generated by semiconductor laser diodes at suitable wavelengths (e.g. 980 nm and 1480 nm for EDFAs). In Fig.2-2, both pump and information signals propagate in the same direction. Such type of pumping scheme is called forward pumping. It is possible to introduce the pump power at the second coupler so that pump and information signal travel in the opposite direction. This pumping scheme is referred as backward pumping. Both types of pumping can also be used simultaneously and referred as bidirectional pumping. As the pump power reduces along the propagation direction, the population inversion also reduces. The signal amplification depends upon the population inversion profile along the length of doped fiber. For higher amplification, more population inversion is desired over the larger length. For a given population inversion profile, there is an optimum length of doped fiber which maximises the overall gain [30].

17

Fig.2-2

Forward pumped doped fiber amplifier The relaxation time of excited ions/atoms is an important parameter in determining

the effect of gain variations on the incoming information signal. In DFAs, it is of the order of few milliseconds. Therefore, signal for both bit 1 and 0 experiences same amplifier gain when very high bit rate signals are amplified. In contrast to this, the relaxation time in SOAs is of the order of few nanoseconds. This means that at high bit rates (109 b/s), gain for bit 1 and 0 would be different. Due to this reason, fluctuations in gain for bit 1 and 0 are more in SOAs than in DFAs. This results in comparatively less cross-saturation effect in DFAs. The cross-saturation effect means the reduction of gain in desired channel due to the presence of other channels when wavelength division multiplexed (WDM) signals are amplified by OAs.

The DFAs can be used as in-line component in fiber optic links because these can be easily spliced with the fiber. Consequently, there is smaller coupling loss. The signal polarisation does not affect the gain of these amplifiers because of their circularly symmetric cross-section. In SOAs, there is a large coupling loss (3.5 dB/facet typically) and gain varies with input signal polarisation. The gain sensitivity to signal polarisation in SOAs can be reduced by using various techniques [30, 89]. SOAs have two distinct

18

advantages over doped fiber amplifiers. These are: (i) SOAs can be easily integrated in integrated optic transmitter/ receiver and (ii) SOAs are available over a wide wavelength range (0.8 m - 1.55 m).

The most significant application of optical amplifiers is for amplifying the optical signals in communication links. This increases the regenerator spacing in power budget limited optical fiber links. In fiber optic broadcast networks, the transmitter power is distributed among all the users. The number of users are restricted by the transmitter power level. Use of OAs will increase the available power budget and hence the number of users supported. This aspect of SOAs has been considered in detail in the following chapters of the thesis.

In a point-to-point link, OA can be used as (i) postamplifier, (ii) preamplifier and (iii) in-line amplifier. Postamplifier implies that an OA is used just after the transmitter. It is also possible that OA is integrated with the source to form a high power optical source. In this application, OA with higher saturation power is desired to reduce the effect of gain saturation. DFAs can be used for this application because of their high saturation power level. However, such a high power optical source would be very bulky. When an optical amplifier is used as preamplifier, it is placed just before the receiver. This can also be integrated with the receiver to form a high sensitivity receiver module. The OAs in this application must produce low noise. Higher value of saturation power level is not required as in postamplifier. In-line amplifiers are used in the optical fiber link itself [23, 24, 49, 80]. DFAs are better suited for this application because these can be easily spliced in the link. Remote pumping can also be used to avoid electrical supply to the amplifier.

19

2.2 Optical Amplifiers in Broadcast Networks

As mentioned above, OAs can be used in passive broadcast networks for increasing the number of users supported. The number of OAs in a network should be minimum for a given number of supportable users. Conversely, number of supportable users must be maximized for a given number of OAs. The optimal utilization of OAs can be achieved by (i) proper placement of OAs in a given network topology and (ii) modifying the existing topologies to utilize the OAs effectively. In view of the above, many studies have been made to investigate the effectiveness of OAs in various topologies e.g. bus, ring, star and some multilevel topologies. In the following, above studies and their results are reviewed.

2.2.1 Bus

Bus topology is very popular in copper based networks. It requires less copper cable as compared to other topologies. Further, standard media access protocols e.g. IEEE 802.3 can be used. When the bus is implemented on optical fiber, there is a limitation on the number of users supported by it. This is because of non-uniform distribution of power among the users, requiring large dynamic range for the receivers. With limited dynamic range, number of users are limited on bus network.

Bus topology supports very few users (typically < 20) [84]. The optical amplifier can be used to keep the distribution of received power level more uniform. This results in increase in supportable number of users. Wagner [84] has considered the use of SOAs in a single channel bus network with uniformly distributed users. The SOAs are assumed to

20

be placed such that there are M users in between as shown in Fig.2-3. The maximum value of M depends upon the dynamic range of received and transmitted power levels. It is equal to number of users which can be supported by the bus without any SOA.

Fig.2-3

Optical amplifiers in bus topology.

There are two factors which limit the increase of number of users when SOAs are used: (i) ASE noise and (ii) gain saturation. Initially, only the effect of ASE noise is studied assuming the SOAs to be unsaturated. When a signal passes through many SOAs, the ASE noise accumulates and degrades the system performance. It limits the number of SOAs and hence number of supportable users. When the gain saturation is considered, accumulated ASE noise saturates the SOAs alongwith the signal. This further reduces the number of supportable users.

Use of heterodyne detection alongwith SOAs has also been studied [84]. In the study, it is considered that in each section having M users and a amplifier, end-to-end signal gain compensates the loss. Therefore, whatever signal power enters in a section also emerges out of the section. To study the degrading effect of gain saturation, gain saturation due to last amplifier is considered and for all other amplifiers the gain equal to the loss is

21

assumed. With this assumption, the study gives an upper bound on number of users. The analysis is limited to single channel bus. It can be modified for a WDM network.

2.2.2 Dual Bus

Dual bus topology is used for implementing distributed queue dual bus (DQDB) metropolitan area networks (MANs). It consists of two buses which carry data in opposite direction. The data is transferred in time slots which are generated by the Head of the Bus (HoB). The two buses use optical fiber for connecting the two consecutive nodes on the bus. The fiber is basically used as point-to-point link (Fig.2-4). This network has a disadvantage of being unreliable as it does not use passive bus over the whole network. It can be made reliable by connecting the nodes to the passive bus using 2x2 couplers. This limits the number of nodes which can be supported by the network. Study on the use of OAs in such a passive dual bus for increasing the number of nodes has been carried out [50]. The above scheme is shown in Fig.2-5. Two wavelengths c and d are used in this scheme (Fig.2.5b and 2.5c). The wavelength c is used to broadcast the clock signal and d to transmit/receive the data. As the network is to operate at 10 Gb/s, broadcast of clock signal on c is essential. This is because at such a high bit rate, clock recovery and synchronisation is very difficult. Use of clock at c and data at d, requires phase delay compensation to compensate for smaller phase delays due to environmental variations (e.g. temperature). The type of OAs although not specified have been modelled for noise and gain saturation as in SOAs. It has been shown that a photonic dual bus using OAs (with unsaturated gain of 12 dB and saturation power level of 4 dBm) with 100 nodes spanning

22

hundreds of kilometres is possible. Therefore, such a network is suitable as wide area network (WAN) or MAN.

Fig.2-4

Dual bus network using fiber as point-to-point link.

Fig.2-5a

Schematic of non-regenerative photonic dual bus (NI is the node interface).

2.2.3 Ring

Optical fibers have been used for connecting consecutive nodes in rings. Fiber distributed data interface (FDDI) is such a ring network designed with fiber as the

23

Fig.2-5b

Node interface for head of the bus (node 1).

Fig.2-5c

Node interface for node 2 to N.

transmission medium [25, 26]. Ring has also been proposed for self healing synchronous optical network (SONET) architectures. In these self healing rings, reliability is increased by creating alternate paths and providing isolation of faulty links. FDDI uses two counter rotating rings. It employs a loopback mechanism for high reliability in case of node failures. Passive optical rings can provide high reliability because failure of a node does not break the ring. But these have power budget problem limiting the number of nodes which can be

24

used on it. This problem can be overcome to a great extent by using optical amplifiers. A passive ring having distributed erbium doped fiber amplifier has been proposed [22]. The scheme is shown in Fig.2-6. In the analysis of this ring structure, g/ is considered to be a constant along the ring. Here, g is gain coefficient and the attenuation coefficient of the fiber in the ring. It was shown that when g is less than , ASE noise psd within the ring attains a constant value which is quite low even when -g << 1. Thus distributed optical amplification can be used in the ring to increase the supportable number of users. It has been shown that a 300 km ring at 2.5 Gb/s cannot support a single user without amplifier, but with the distributed amplification over 450 users can be supported. In the study, quite a few assumptions have been made. Pump power is assumed to cause such a population inversion profile that g/ is constant. Some other factors which restrict the number of users are ignored e.g. (i) dynamic range of receivers, (ii) back reflection at various points and (iii) remnants of recirculating optical signals.

Fig.2-6

Ring network with distributed EDFA.

25

2.2.4 Star

Ideally, star topology distributes optical power equally at output ports. Therefore, it can support maximum number of users without OAs as compared to any other topology. At the present level of technology, number of users supported by star is small (typically 64). Many modified topologies which use star couplers and OAs have been proposed [59, 94, 104]. These can support higher number of users as compared to a passive star topology. The star topology can use either transmissive or reflective stars or both. In reflective star coupler, input and output ports are same. The user transmits the signal in a given port and receives the signal from all other users from the same port [67]. The propagation direction of transmitted and received signals are opposite in the port. Two schemes of distributed reflective star couplers which use the erbium doped fiber amplifiers (EDFAs) have been proposed [94]. In both the schemes, 4x4 couplers having an EDFA as shown in Fig.2-7 is used as basic element. An 8x8 star coupler can be made using 4x4 couplers and 2x2 couplers as shown in Fig.2-8. In the same manner, star coupler of larger size can be made using smaller size couplers. In the first scheme (Fig.2-9), m number of mxm star couplers are used. Implementation of each mxm coupler require m/4 EDFAs and 6m/4 + (m/2) log2(m/4) number of 2x2 couplers [94]. It can be seen from Fig.2-9 that reflective star made of m number of mxm star couplers can support Nu=m2 number of users. One output port of each coupler is terminated with a mirror and other (m-1) ports are connected to remaining (m-1) number of mxm couplers. A 16x16 distributed reflective star coupler based on this scheme is shown in Fig.2-10

26

Fig.2-7

4x4 star coupler with single EDFA.

Fig.2-8

8x8 coupler based on 4x4 star couplers. In the second scheme (Fig.2-11), 2m number of mxm couplers are used to implement

a distributed reflective star coupler which can support Nu=2m2 number of users [94]. The mxm couplers are being grouped to form m pairs. In each pair, one output of each mxm coupler is connected using 2x2 reflective star coupler. Each pair of mxm couplers is connected to every other pair by a 2x2 coupler. A 32x32 reflective star coupler based on this scheme is shown in Fig.2-12. In these star couplers, less fiber is required as compared

27

Fig.2-9

Implementation of distributed Nu x Nu (Nu=m2) reflective star coupler based on m number of m x m couplers.

28

Fig.2-10

16x16 distributed reflective star coupler.

to a centralized star coupler. The EDFAs are shared by all the users. In the above proposition, no mention has been made regarding the utility of this star in a WDM system. The use of above star in a WDM system will require wide bandpass filters in EDFAs. Further, the required gain of EDFAs for a given number of users is not given, which is needed in design of such systems. Also the injection of pump in EDFAs will require wavelength selective coupler.

Another star coupler configuration using EDFAs for WDM/FDMA network (Fig.2-13) has also been proposed [59]. WDM/FDMA network means that the signal spectrum consists of many frequency band at different wavelengths. Each band contains many channels which are wavelength multiplexed with very narrow spacings. In this

29

Fig.2-11

Implementation of distributed Nu x Nu (Nu= 2 m2) coupler based on 2m number of mxm couplers.

scheme, there is one central coupler of dimension mcxmc. All other couplers are auxiliary couplers. There are mc wavelength bands and a unique wavelength band is allocated to the users on auxiliary coupler. Each wavelength band has multiple channels to be used for transmission by the users on the corresponding auxiliary coupler. When a user on auxiliary coupler AC1 transmits the signal on a wavelength channel in the assigned band W1, it is broadcast to all the users on the same coupler. A part of the signal is transmitted to central coupler. At the input, only the wavelength band W1 is allowed to pass through by means of an optical bandpass filter (BPF). This filter allows only the signals from users on AC1

30

Fig.2-12

32x32 distributed reflective star coupler.

to be passed, which are then amplified and broadcast to the other auxiliary star couplers. The output fiber connected to AC1 from central star contains a bandstop filter (BSF) for wavelength band W1. Ideally, the use of BSF prevents the recirculation of the optical signal in wavelength band W1. One of the output ports of mcxmc central coupler can be used as the input for the pump power for all the EDFAs. This scheme has the advantage that existing networks on auxiliary stars are easily interconnected. The network size is easily expandable by using auxiliary stars until all the ports of the central star are filled. Sharing of pump source, amplification of a part of spectrum by EDFAs are the main features of the scheme. The disadvantage is that recirculating signals can produce interference in practical BSFs. The interference can be minimized by suitably matching the BPFs and BSFs. The

31

mxm star coupler must be wideband as the pump signal also traverse through the central coupler.

Fig.2-13

FDM/OFDM distributed expandable amplified star coupler configuration.

A few other configurations using star couplers with OAs have been reported [59, 104]. The star coupler proposed in [59] uses BPFs and BSFs for each wavelength band. This coupler is modified by Yung-Kuang et.al [104] and the modified version is shown in Fig.2-14. This is a mmcxmmc centralized star coupler. The 1xmc and mcx1 couplers can be made using (mc-1) number of 2x2 couplers as shown in Fig.2-15. The mcxmc coupler is a transmissive star coupler made using (mc/2)log2(mc) number of 2x2 couplers. Each OA is a forward pumped EDFA as shown in Fig.2-16. This scheme supports more number of users than m(mc-1)xm(mc-1) coupler [59]. The number of required 2x2 couplers is less than

32

that of star coupler [59] for m 32. This coupler does not require dedicated BPF and BSF as in [59].

Fig.2-14

Star coupler with amplifiers and without bandpass and bandstop filters.

As each user must be connected to the centralized star coupler, there is a large fiber requirement. The loss between transmitter and receiver is more than the loss in the coupler with BPFs and BSFs. The fiber requirement can be reduced by making the above star coupler distributed. This can be achieved by putting the 1xm and mx1 couplers with the group of m users. It implies that fiber is required only to connect the 1xm and mx1 couplers to mcxmc star. This saves the required fiber as well as the required conduit length for fiber layout. Further, the number of 1xm couplers can be reduced by using reflective mcxmc star coupler. The resulting mmcxmmc coupler is shown in Fig.2-17. The above modification also reduces the required fiber to half as the transmitted signal and signal to

33

be received share the same fiber in opposite direction. These signals can be separated by using diplexers at the user ends. The mcxmc reflective star coupler can be made by using a mcx1 coupler whose single port is terminated by a mirror (Fig.2-18). There are also two other possible ways to form a mxm reflected star coupler. These are shown in Figs.2-9 and 2-11.

Fig.2-15

8x1 coupler based on seven 2x2 couplers.

In all the above studies, degrading effects due to OAs have not been considered. Therefore, these give an upper bound on the supportable number of users. In some of the studies, new schemes have been proposed which use the OAs more efficiently. But their practical implementation require some special devices e.g. mirrors for terminating the ports and 2x2 reflective star couplers.

34

Fig.2-16

Forward pumped EDFA.

Fig.2-17

mmc x mmc reflective star.

An accurate analysis is needed to compute the supportable number of users which is expected to be lower than the upper bound. Therefore, the previous performance bounds cannot be used to compare the various schemes. Only a qualitative comparison can be made. All recent studies are mostly based on EDFA due to its popularity. It is remarked that the doped fiber can only be used in distributed star coupler or in bus and ring

35

Fig.2-18

mc x mc reflective star coupler based on mcx1 tree coupler.

networks. In centralized star couplers, SOAs can be easily integrated with integrated optic implementation. Further, SOAs are also better suited as post or preamplifier since these can be integrated with either the transmitter or the receiver chips. Therefore, SOAs have been considered in the studies reported in the following chapters.

36

Chapter 3

Semiconductor Optical Amplifier and Coding in OOK System

Fiber optic broadcast networks can support only a limited number of users because of constraint on the transmitter power. It is desirable to have large number of users to utilize the capacity of network. The number of users can be increased by increasing the power budget, reducing the loss in the network, using electronic repeater and optical amplifiers. The power budget can be improved by using coding techniques. The use of coding and semiconductor optical amplifier (SOA) to increase the network size have been considered in this chapter. For

The work reported in this chapter has resulted in the following publications.
1. V. K. Jain, Y. N. Singh and H. M. Gupta, "Power Penalty due to Optical Amplifier Induced Crosstalk in Non-Coherent OOK Transmission Systems," Journal of Optical Communications, Vol.16, No.5, Oct.1995, pp.194-196. Y. N. Singh, V. K. Jain and H. M. Gupta, "Effect of Reed-Solomon Code on Laser Linewidth Requirements of BPSK Homodyne Optical Communication Systems," Journal of Optical Communications, Vol.16, No.6, Dec.1995, pp.207-210. Y. N. Singh, V. K. Jain and H. M. Gupta, "Reed Solomon Code and Semiconductor Optical Preamplifier in OOK Communication System: A Comparative Study," Journal of Optical Communications (accepted for publication). Y. N. Singh, V. K. Jain and H. M. Gupta, "Effect of Error Correcting Codes on Laser Linewidth Requirements in Optical Binary Phase Shift Keying Communication Systems," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. V. K. Jain, Y. N. Singh and H. M. Gupta, "Effect of Optical Amplifier Induced Crosstalk in Two Channel Non-Coherent OOK Transmission System," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994.

2.

3.

4.

5.

37

their comparative study, a point-to-point on-off keying (OOK) communication link has been considered. The receiver structure considered is shown in Fig.3-1. It consists of a PIN photodetector followed by a Gaussian filter with the frequency response

(3.1)

Here, f is the frequency and Be the electrical bandwidth of the filter. This filter is followed by sampling, threshold and decision circuits.

Fig.3-1a

The system block diagram for optical communication system with coding scheme.

Fig.3-1b

The system block diagram for optical communication system with SOA as preamplifier.

38

The electrical signal at the receiver output is corrupted by additive shot and thermal noise of photodetector. Both types of noise are considered to be white Gaussian with power spectral densities (psds) [30] (3.2a)

and

(3.2b)

respectively. In the above equations, Ro is the responsivity of the photodetector, Pr the received optical power level which becomes Pr(1) for bit 1 and Pr(0) for bit 0. These are given by

(3.3a)

and

(3.3b)

where is the extinction ratio and Pra (= [Pr(1) + Pr(0)]/2) the average received optical power level. In the above equations, e is the electron charge, kb the Boltzmanns constant, T the temperature of receiver in Kelvin and RL the receiver load resistance.

39

The receiver performance in terms of required recieved optical power level has been computed for (i) uncoded system, (ii) coded system with decoding and (iii) uncoded system with optical preamplifier. The Gaussian filter in the receiving system is optimized for uncoded transmission. The coding in the system increases the bit rate and hence intersymbol interference (ISI). In order to evaluate the degradation due to ISI, performance of coded system without decoding has been evaluated. Any coding gain should be measured with respect to the above performance. Section 3.1 presents the analysis of uncoded system. In section 3.2, coded system without decoding and with decoding has been considered. Uncoded system with SOA has been analyzed in section 3.3. A comparative study of above systems based on a practical example is given in section 3.4. Section 3.5 contains the observations and conclusions of this study.

3.1 Uncoded System

The Gaussian filter in the receiver reduces noise by removing noise outside the frequency band of interest. But it also causes ISI. Increase in Be results in decrease in ISI and increase in noise after the filter and vice-versa. Hence, there exists an optimal filter bandwidth.

Worst-case degradation due to ISI can be modelled by considering worst-case waveforms for bit 1 and 0. These waveforms decide the eye opening. For bit 1 and 0, the worst-case waveforms correspond to bit sequences ...0001000... and ...1110111... respectively. The waveforms before and after the filter are shown in Fig.3-2. When the sampling is done

40

in the middle of bit, signal levels S1 and S0 corresponding to worst-case sequences for bit 1 and 0 are given by

(3.4a)

and

(3.4b)

where

(3.5a)

and

(3.5b)

In the above equations, h() is the impulse response of the Gaussian filter, Tb the bit duration. The a1() and a0() are the worst-case waveforms for bit 1 and 0 respectively. The eye opening will be S1-S0.

41

Fig.3-2

Worst-case waveform patterns for (a) bit 1 and (b) bit 0 before and after the Gaussian filter. For an additive Gaussian noise with psd SN(f), variance after the Gaussian filter is

given by

(3.6a)

If the noise is white i.e. SN(f)=SN. Using Eqn.3.1 in the above equation, n2 is given by

42

(3.6b)

In an uncoded system, SN corresponding to bit 1 and 0 is given by sum of Sshot and Sth. Noise variances 2(1) and 2(0) corresponding to S1 and S0 are determined from Eqn.3.6b. The bit error rate (BER) which equalises the error rates for S1 and S0 is given by

(3.7)

The above analysis is quite general and the system performance can be evaluated for any bit rate R (= 1/ Tb). The filter can be optimized to minimize the effect of ISI at a given bit rate. When bit rate is increased (as in coding), the above optimization is not valid and the system performance degrades due to increased ISI. The degradation in performance is evaluated using the above general model (Eqns.3.4 to 3.7).

3.2 Coded System

Coding is introduced in the system by encoding the input signal to the transmitter and decoding the receiver output signal (Fig.3-1a). As mentioned earlier, the bit rate increases in the system due to coding. The increase depends on the amount of redundancy introduced by coding scheme. When the data rate is R b/s and (N,k,t) block code is used, the coded bit rate becomes R(N/k) b/s. Here, N is the number of symbols in the block, k the data symbols in the block and t the number of symbols which can be corrected by the error correcting code. The

43

increased bit rate leads to more ISI as the receiver is same as in the uncoded system. It results in increase in BER. Some errors are corrected in the decoding process leading to decrease in BER. If the decrease in BER is more than the increase due to ISI, there will be a coding gain.

For a (N,k,t) block code, the symbol error rate after decoding Pes(ad) is given by [72, 99, 102]

(3.8a)

where Pes(bd) is the symbol error rate before decoding. It is related to bit error rate Pe by

(3.8b)

In the above, n is the number of bits in a symbol. For example, in systematic Reed-Solomon (R-S) (31,27,2) code, there are 31 five bit symbols, 27 data symbols and 4 parity symbols. Therefore, in Eqn.3.8, values of N, k, t and n are 31, 27, 2 and 5 respectively.

In the coded system, the bit rate becomes higher and the BER can be determined using the same approach as in the uncoded system. This BER is used to determine the symbols error rate before decoding Pes(bd) from Eqn.3.8b. The R-S code will reduce Pes(bd). The reduced symbol error rate after decoding Pes(ad) can be determined using Eqn.3.8a. To compare the

44

performance of coded system with the uncoded system, BER for coded system is required to be known. This can be determined from Eqn.3.8b by replacing Pes(bd) with Pes(ad) and then determining Pe using a suitable search technique. The BER also includes the errors in the parity symbols which are discarded after decoding. Hence, the BER after discarding parity symbols is Pe(k/N). The system performance for R-S (31,27,2) code is evaluated using the above approach and results are shown in Fig.3-3.

3.3 Uncoded System with Optical Amplifier

In this analysis, an SOA has been considered as preamplifier. It is modelled as travelling wave amplifier (TWA). The SOA alongwith the amplified signal also produces ASE noise which beats with signal and itself at the photodetector. The psd of various noise components at the photodetector is shown in Fig.2-1.

There is a dc component due to ASE-ASE beat noise which gives rise to shot noise with psd (3.9)

The above is modelled as white Gaussian noise [63]. It adds to shot noise produced by the signal. In Eqn.3.9, Ssp is the ASE noise psd and Bo the bandwidth of optical filter. The Ssp is given by

45

(3.10)

where nsp is the spontaneous emission factor, G the gain of OA, h the Plancks constant and the optical frequency.

The noise variance at the filter output due to ASE beat noise components is determined by using Eqn.3.6a and noise psds are shown in Fig.2-1. These are given by (i) ASE-ASE beat noise

(3.11a)

(ii) ASE-signal beat noise

(3.11b)

46

(iii) ASE shot noise

(3.11c)

In the above, Pin is the optical power at the input of optical preamplifier. The total noise variance at filter output is given by

(3.12)

In order to compute the BER, received power levels for bit 1 and 0 are determined for a given average received power level and extinction ratio from Eqn.3.3. Corresponding amplifier gain G(1) and G(0) for bit 1 and 0 respectively are determined using these received power levels and the amplifier gain saturation formula given in Eqn.2.4 [30]. The optical power levels falling on photodetector for bit 1 and 0 are Pin(1)G(1) and Pin(0)G(0) respectively. Here, Pin(1) and Pin(0) are input optical power levels to SOA for bit 1 and 0 respectively. The signal current after the Gaussian filter is determined for bit 1 and 0 following the same approach as in section 3.1. There is an optical bandpass filter of bandwidth Bo between SOA and photodetector (refer to Fig.3-1b). It is used to limit the ASE noise. The noise variances for bit 1 and 0 are determined from Eqns.3.11 and 3.12 using the ASE noise psd, optical signal power levels for bit 1 and 0 and optical filter bandwidth Bo. These are used to determine BER from Eqn.3.7.

47

3.4 Example

The results for a sample system are computed and analyzed in this section. Following are the system parameters:

Bit rate (uncoded system), R Optimum BW of Gaussian filter, Be Operating wavelength, Quantum efficiency of photodetector, Coding scheme Bit rate after coding Unsaturated gain of SOA, G0 Saturation power level, Psat BW of optical filter, Bo Extinction ratio,

1 Gb/s 1.5 GHz 1.55 m 0.95 R-S (31,27,2) 1.14 (=32/27) Gb/s 29 dB 10 dBm 10 GHz 0

The BER is determined as a function of average received signal power level for (i) uncoded system, (ii) coded system without decoding, (iii) coded system with decoding and (iv) uncoded system with optical preamplifier. Variations of log10(BER) with received signal power level Pra (in dBm) are shown in Fig.3-3.

It is seen from the above figure that a coded system without decoding performs worse than the uncoded system. This is expected as coding increases bit rate which in turn increases the ISI and therefore the performance becomes worse. Once the received signal is decoded, the performance improves because of error corrections. R-S(31,27,2) code for a BER of 10-9

48

Fig.3-3

Variations of BER vs received power level (dBm) for (i) uncoded system, (ii) coded system without decoding, (iii) coded system with decoding and (iv) uncoded system with SOA as preamplifier.

provides coding gain of 1.7 dB. When optical preamplifier is used instead of coding, improvement in receiver sensitivity is about 21.2 dB. It implies that use of optical amplifier improves the system performance much more than the use of coding.

3.5 Conclusions

It is concluded that in an OOK communication system, use of SOA is advantageous as it provides more improvement ( > 20 dB) as compared to coding. Although the above conclusion is based on R-S code, it will remain valid for all other codes too. The level of

49

improvement depends upon the received optical power level and coding scheme used. Use of coding cannot provide coding gains in the vicinity of 20 dB.

Another advantage of SOA over coding is that it can amplify signals at high bit rates (> 1 Gb/s). The practical implementation of encoder and decoder at such a high bit rate is difficult. Further, a single SOA can handle multiple WDM channels. Therefore, in chapters 4 and 5 of the thesis only the use of SOAs in network has been considered.

50

Chapter 4

Semiconductor Optical Amplifiers in WDM Star Topology

In a star topology network, the transmitted signal from all the users are combined in a star coupler and then distributed to all the receivers. The star coupler can be centralized or distributed. Generally, the star coupler distribute the transmitted signal power equally among the users. Consequently, star topology supports maximum number of users as compared to other passive broadcast topologies viz. bus, ring and tree.

The semiconductor optical amplifiers (SOAs) can be placed in the star network to amplify the optical signal. These will increase the supportable number of users. The SOAs can be placed in a star either after the transmitters as postamplifiers or before the receivers as preamplifiers. In case of WDM star networks, postamplifiers amplify single channel signal only, while preamplifiers amplify signals in multiple channels. In the latter case, there may be interchannel crosstalk.

The work reported in this chapter has appeared in the following publications.
1. Y. N. Singh, V. K. Jain and H. M. Gupta, "Semiconductor Optical Amplifiers in WDM Star Networks," IEE-Proceedings Optoelectronics, Vol.143, No.2, April 1996, pp.144-152. Y. N. Singh, V. K. Jain and H. M. Gupta, "On Placement of Semiconductor Laser Amplifiers in WDM Star Networks," Proc. National Conference on Optical Communications, J.K.Institute of Applied Physics & Technology, University of Allahabad, Allahabad, India, Feb.22-24, 1995, pp.66-78.

2.

51

In this chapter, the above SOA placement schemes are analyzed. Section 4.1 presents the description of the star network. Section 4.2 discusses the reflection due to amplifier. Noise variances due to ASE and signal beat noises are determined when the receiver uses an ideal filter. In section 4.3, approach used in analyses is described. The analyses of star without SOA, with postamplifier and preamplifier are presented in sections 4.4, 4.5 and 4.6 respectively. In section 4.7, an example with practical system parameters has been considered. Finally, conclusions are given in section 4.8.

4.1 System Description

In a NuxNu star network, every user is provided with a transmitter and receiver pair. In the simplest arrangement [32, 61], each user is assigned a dedicated transmit wavelength in the 1.5 m waveband. The modulation scheme considered is the binary intensity modulation. When extinction ratio is zero, it becomes an on-off keying system. The transmitted signal power is equally distributed to all the receivers by the star coupler. Each user receives the signals on all the wavelengths. One of the wavelengths is selected using an optical filter and then detected using a photodetector. The above mentioned star coupler is made of 3 dB 2x2 couplers. An 8x8 star coupler consisting of 2x2 couplers is shown in Fig.4-1. In general, a NuxNu star coupler has log2Nu stages and requires (Nu/2)log2Nu number of 3 dB 2x2 couplers. The total insertion loss Lti of a NuxNu star coupler is given by (4.1)

52

where Li is the insertion loss of an individual 3 dB 2x2 coupler. The power split due to star coupler may not be uniform due to practical limitations. This nonuniformity has been considered in the analysis in terms of coupler variability, Lcv, which is the worst-case reduction of power in the output port. Each user is connected to star coupler using a singlemode dispersion flattened fiber. Consequently, the input and output ports of NuxNu coupler are spliced to optical fiber. Each of these splices introduces a loss of Lsp dB.

Fig.4-1

An 8x8 star coupler consisting of 3 dB 2x2 couplers.

At receivers, ideal optical filters are used. Therefore, there is no interchannel crosstalk due to these filters. These filters introduce a insertion loss of Lfi dB. The filters can be tunable or fixed wavelength type depending on the media access control (MAC) protocol. When receiver arrays are used to receive more than one channel at a time, demultiplexers can be employed instead of filters. Optical filters are followed by PIN photodetectors in each of the receivers to detect the optical signal.

53

The use of SOAs is expected to increase the number of users or decrease the required transmitter power in the star network. However, the presence of amplifier noise and gain saturation effect is expected to partially reduce the above advantages. When SOAs amplify multiple channels, total optical power at the SOA input varies randomly since light is independently modulated in each channel. This results in interchannel crosstalk. The signal power at the amplifier output in a channel varies according to the gain fluctuations induced by modulation in other channels even when the input power in the channel is constant. This is the crosstalk induced by gain saturation in the amplifier and is often referred as cross-saturation. The cross-saturation effect increases with the increase in the number of wavelength channels. Therefore, it is more severe when larger number of wavelengths are used in the system. In the above system, maximum cross-saturation occurs as Nu wavelengths are used. Hence this will provide upper bound on the degradation due to cross-saturation for the WDM star network.

4.2 Amplifier Model

The gain of SOAs reduces with the increase in input power level. This reduction arises due to gain saturation. The saturated gain of SOAs is given by Eqn.2.4. The amplified spontaneous emission (ASE) noise is also present in the amplifier output alongwith the amplified signal. The psd of ASE noise is given by Eqn.2.5. At the photodetector, ASE noise beats with itself, signal and echo signal if present. The psds of ASE-ASE and ASE-signal beat noise components are shown in Fig.2-1. The noise variances of the various noise components, assuming an ideal electrical filter of bandwidth Be in the receiver, are given by

54

(i)

ASE-ASE beat noise [63, 90] (4.2a)

(ii)

ASE-signal beat noise [63, 90] (4.2b)

When an echo signal is present, ASE-echo beat noise variance is given by Eqn.4.2b by replacing Pr by the echo power Pecho. ASE-ASE beating also produces a dc component resulting in shot noise. The variance of this shot noise is given by (4.3)

In addition to signal amplification, the amplifiers also reflect signals. The reflection coefficient of the amplifier is given by (Appendix-I).

(4.4)

In the above, R1 and R2 are the facet reflectivities and G the single pass gain of SOA.

55

4.3 Evaluation of Minimum Required Transmitter Power

Once the relationship between bit error rate (BER) and transmitter power, Pta, is known for a given number of users, Nu, minimum required Pta can be determined using this relationship. Therefore, models to determine BER as a function of Pta and Nu are formulated in the following sections. These models are used to evaluate the performance of star network without amplifier, with postamplifier and with preamplifier. In the analysis, reflection noise due to filters and coupler have been neglected. Optical filters are assumed to be ideal so that these do not introduce any crosstalk. Further, responsivity of photodetectors is assumed to be independent of wavelength. The effect of finite extinction ratio of transmitted pulse has been considered in the analysis.

4.4 Star without Amplifier

The WDM star topology is shown in Fig.4-2. Each transmitter is assigned an unique operating wavelength and the coupler distributes signals of all wavelengths to all the receivers. The power received at the photodetector for bit bc (bc is either 1 or 0) is given by

(4.5)

where

56

(4.6a)

(4.6b)

In the above equations, Pta represents the average transmitter power and the extinction ratio. The loss Ltr between transmitter and receiver excluding the split loss of star coupler is given by

(4.7a)

or equivalently (4.7b)

The above does not include the split loss of star coupler which has been included by the factor 1/Nu in Eqn.4.5. In the above equation, is the attenuation coefficient of fiber in dB/km, L the length of fiber from user to star coupler in km, Lfi the filter insertion loss, Lcv the loss due to nonuniformity in the power splitting by the star coupler and Lsp the splice loss in dB.

57

Fig.4-2

Star network without optical amplifier. The signal current and noise variance for bit bc (bc = 0 or 1) at photodetector output

are given by (4.8a)

and

(4.8b)

where kb is the Boltzmanns constant, T the temperature in Kelvin, R0 the responsivity and RL the load resistance of the photodetector. The first term on right hand side of Eqn.4.8b represents the shot noise and second term the thermal noise. The average probability of error with the threshold level which equalises the BER for bit 1 and 0 is given by

58

(4.9)

It is obvious that Pe is determined from Isig(1) and Isig(0), (1) and (0) which in turn are determined from Pr(bc) and hence from Pt(bc). The Pt(bc) is determined from Pta. Thus Pta determines Pe. Alternatively, for a given Pe, Pta can be computed by a suitable search technique.

4.5 Star with Postamplifier

In order to reduce the required Pta, SOAs can be used in the star topology. In this section, SOAs are placed immediately after the transmitters (postamplifiers). The performance of star network with unsaturated postamplifiers has been analyzed in this section. Subsequently, the effects of amplifier gain saturation and reflection have been included in the analysis.

4.5.1 Unsaturated Amplifiers

As the gain of unsaturated postamplifier is G0, the received signal power Pr(bc) is given by

59

Fig.4-3

Star network with postamplifiers (A, B, ...F is echo path).

(4.10)

where Lta is the loss between transmitter and amplifier and Lar the loss between amplifier and receiver. These are given by (4.11a)

and (4.11b)

In Eqn.4.11b, Lcp is the coupling loss of an amplifier facet. The psd of ASE noise at the receiver will be

60

(4.12)

This psd is not reduced by the factor 1/Nu due to the following reason. Let Sspi be noise psd at the receiver due to ith amplifier. It is given by

(4.13)

The total noise psd at the receiver should be sum of all the psds due to Nu amplifiers. Hence Ssp will be

(4.14)

This is same as given in Eqn.4.12.

The signal current and noise variance for bit bc using Eqns.4.2 and 4.3 are given by (4.15a)

and

61

(4.15b)

The first term in Eqn.4.15b corresponds to shot noise due to signal and ASE, second term to ASE-signal beat noise, third term to ASE-ASE beat noise and fourth term to thermal noise. The signal currents Isig(bc) and noise variance 2(bc) from Eqn.4.15 are used in Eqn.4.9 to determine average probability of error Pe.

4.5.2 Effect of Gain Saturation

In general, gain of SOAs is not constant and it is input power Pin dependent. The gain reduces with the increase in Pin. This will degrade the receiver performance which has been investigated below. Let the amplifier gain corresponding to bit 1 and 0 be G(1) and G(0) respectively. Therefore, the received signal power level for bit bc will be

(4.16)

The received ASE noise psd is different for bit 1 and 0. When the desired channel has bit 1, one half of the remaining (Nu-1) channels are expected to have bit 0 and the other half bit 1. The ASE noise from each amplifier gets distributed equally among the

62

outputs. Therefore, ASE noise psd from each amplifier will get reduced by a factor of 1/Nu at the receiver input. The total ASE noise psd at the receiver for bit bc is

(4.17)

The first term in the above equation corresponds to the desired channel, second term to channels having bit 0 and the third term to channels having bit 1.

The signal current and noise variance for bit bc are given by (4.18a)

and

(4.18b)

The first term in Eqn.4.18b correspond to shot noise due to signal and ASE, second term to ASE-signal beat noise, third term to ASE-ASE beat noise and last term to thermal noise.

The average Pe in this case can be determined using Eqn.4.18 alongwith Eqn.4.9.

63

4.5.3 Effect of Reflection Noise

Splices in the network produce back reflections resulting in echoes at the receivers. A signal which has undergone 2i reflections will generate i-pass echo. Generally, the reflection coefficient of splices is below -20 dB, therefore the echoes with two and more passes can be neglected. When the delays suffered by echoes are greater than the coherence time of optical signal, coherence between signal and echoes need not be considered. In the network under consideration, distance between the reflection points is much greater than coherence length. Hence, incoherent addition of signal and echo power has been considered.

In the network shown in Fig.4-3, reflection can occur at amplifier or at splice 1 or at splice 2. Consequently, there would be three possible echo signals. Out of these, the echo signals due to SOA and splice 1 will be strongest. This is so because loss between splice 1 and splice 2, and SOA and splice 2 (insertion loss of star coupler + splice loss + fiber loss) is much higher than loss between SOA and splice 1. Hence, echo signal arising due to splice 2 can be neglected.

Worst-case occurs when the signal bit and echo signal bit arrive synchronously at the receiver. This gives upper bound on degradation due to echo. The probability of error in this worst-case is obtained as follows.

Let the echo signal of bit br interfere with the signal corresponding to bit bc. The echo signal power at various points in the echo path A, B, ..., F (Fig.4-3) are given by

64

(4.19a)

(4.19b)

(4.19c)

(4.19d)

(4.19e)

(4.19f)

The received echo power, Pecho, is same as echo power at point F (Eqn.4.19f). In the above equation, Las1 is the fiber loss between amplifier and splice 1, Rsp the splice reflection coefficient. The parameter Gp(br) represents the gain of amplifier when signal corresponding to bit br passed through the amplifier. It is obtained by replacing Pin with Pin(br, brr) in Eqn.2.4. The Pin(br, brr) is given by

65

(4.20)

where Rf is the facet reflectivity of amplifier (Rf=R1=R2). In the above equation, it is presumed that echo of bit brr interfered with the signal corresponding to bit br. The parameter Gp is the gain when signal corresponding to bit brr was amplified. In the worstcase situation, it is assumed that no echo was present when signal corresponding to brr was amplified. Further, brr is assumed to be zero for maximum echo signal. In the above equation, signal corresponding to br and echo of brr are added together and used as Pin in Eqn.2.4. As such, the two optical signal powers cannot be added as these are entering the amplifier from opposite ends. But the above addition corresponds to the worst-case scenario.

The amplifier gain G for the signal corresponding to bit bc under the same assumption is given by

(4.21a)

where

(4.21b)

66

It is seen from Eqn.4.4 and the above equation that Ramp is function of G which in turn depends upon bc and br.

The received signal power at the photodetector input will be

(4.22a)

The ASE noise psd at the photodetector is given by

(4.22b)

In the above equation, first term corresponds to bit bc in the desired channel, second term to bit br in the desired channel, third term to signals from the remaining Nu-1 amplifiers and last term to echo signals from the remaining Nu-1 amplifiers. Further, bits bci, bri and brri are bits in the ith channel corresponding to the bits bc, br and brr respectively in the desired channel. In the worst-case situation, bits brr and brri can be taken as 0. The bci and

67

bri can be either 1 or 0 with equal probability. When Pr, Pecho and Ssp are known, the signal current and noise variance are given by (4.23a)

and

(4.23b)

In the Eqn.4.23b, additional fourth term represents the signal-echo beat noise.

It is evident from Eqn.4.23a that there are two signal levels corresponding to bit 1 as well as bit 0 depending on bit br. The threshold is obtained using the highest level corresponding to bit 0, the lowest level corresponding to bit 1 and the corresponding noise variances. The threshold which equalizes the BERs for these levels is given by

(4.24)

With this threshold, the probabilities of error for bit 1 and 0 are given by

68

(4.25a)

and

(4.25b)

For equiprobable bit 1 and 0, average probability of error is given by

(4.25c)

4.6 Star with Preamplifier

In this scheme, all the wavelengths are amplified by each preamplifier and one of these is selected by the filter. As in the postamplifier scheme, performance of star network with ideal preamplifiers has been analyzed first. This analysis is extended to include the effect of amplifier gain saturation and reflection noise.

69

4.6.1 Unsaturated Amplifier

The received power for bit bc is given by

(4.26)

where

(4.27)

is the loss between transmitter and amplifier without the split loss of coupler. The Lar represents the loss between amplifier and receiver which consists of only the insertion loss of filter, Lfi, and amplifier coupling loss, Lcp.

The ASE noise psd, Ssp, at the photodetector is determined from Eqn.4.12. It will be same for bit 1 and bit 0. Substituting Pr(bc) and Ssp in Eqn.4.15, the signal currents (Isig(1) and Isig(0)) and noise variances (2(1) and 2(0)) for bit 1 and 0 are obtained and average Pe is computed using Eqn.4.9.

4.6.2 Effect of Gain Saturation

The cross-saturation leads to decrease in the average gain because of increase in the number of channels and gain fluctuations due to randomness in the number of channels

70

having bit 1. This degrades the performance [46, 75, 76]. In this section, degradation due to average gain reduction and gain fluctuations has been analyzed.

(i) Average Gain Reduction The input power to the amplifier corresponding to bit bc is

(4.28)

Let N1 channels out of total Nu channels are having bit 1. Therefore, the total input power to the amplifier is N1Pin(1) + (Nu-N1)Pin(0). The corresponding saturated gain of the amplifier is given by

(4.29)

In order to determine the average gain, probability distribution of number of channels having bit 1 is considered to be binomial. The probability of N1 channels having bit 1 will be

(4.30)

Therefore, the average gain is given by

71

(4.31)

The signal power received for bit bc will be (4.32)

As Gav is independent of signal bit, the ASE noise psd for both bit 1 and 0 will be same. It is given by (4.33)

Eqns.4.32 and 4.33 are used to determine signal currents (Isig(1) and Isig(0)) and noise variances (2(1) and 2(0)) in conjunction with Eqn.4.15. Then Pe is computed using Eqn.4.9.

(ii) Gain Fluctuations Let N1 channels out of Nu-1 interfering channels are having bit 1. The corresponding total amplifier input power for the bit bc in the desired channels is

Pin(bc) + N1Pin(1) + (Nu-1-N1)Pin(0). Therefore, the saturated gain of the amplifier will be

72

(4.34)

The probability that N1 channels out of Nu-1 channels are having bit 1 is given by

(4.35)

For bit bc, the signal power and ASE noise psd at the photodetector is given by (4.36a)

and (4.36b)

The above equation is used to determine the signal current and noise variance corresponding to bit bc. These are given by (4.37a)

and

73

(4.37b)

Both Isig(1) and Isig(0) will have Nu levels depending upon N1. The threshold corresponding to highest level of Isig(0) and lowest level of Isig(1) will be

(4.38)

The probability of error under the condition that N1 interfering channels are having bit 1 is given by

(4.39a)

The average Pe will be

(4.39b)

74

4.6.3 Effect of Reflection Noise

In order to analyze the effect of reflection noise, the echo path A, B, ..., G as shown in Fig.4-4 has been considered. Reflection between SOA and splice 2 has only been considered because the loss between splice 1 and splice 2 and splice 1 and SOA is much higher. The amplifier is saturated by the signal and echo signal power in the desired channel and also the signal and echo signal power in the interfering channels. Let the echo of bit br affect the signal corresponding to bit bc. Further, let the M1 interfering channels, with probability PM 1, were having bit 1 when the signal corresponding to br was reflected by SOA to form the echo signal. The average echo signal power at the amplifier input in the desired channel is given by

(4.40)

The average echo signal power in the remaining interfering channels will be

(4.41)

The PM 1 is obtained from Eqn.4.35 by replacing N1 by M1. The Ls2a represents the loss between splice 2 and amplifier (fiber and amplifier coupling loss). The Ramp depends on the gain G(br,M1) which is given by

75

(4.42a)

where (4.42b)

Let N1 channels out of Nu-1 interfering channels have bit 1 when signal corresponding to bit bc is amplified. The gain of amplifier for this signal is

(4.43a)

where (4.43b)

In Eqn.4.43b, first term is the signal power in the desired channel, second term the echo signal power in the desired channel, third term (in the square bracket) the signal power in the interfering channels and last term the echo signal power in the interfering channels.

The signal power, echo signal power and ASE noise psd at the receiver are given by

76

Fig.4-4

Star network with preamplifiers (A, B, ...G is echo path).

(4.44a)

(4.44b)

and (4.44c)

The signal current and noise variance are determined using the above equations, which are given by (4.45a)

and

77

(4.45b)

As before there are many signal levels for both bit 1 and 0 and the threshold is computed which equalizes the probability of error for the highest level for bit 0 and the lowest level for bit 0. The probability of error Pe(br,N1) for a given br and N1 will be

(4.46)

The average Pe will be

(4.47)

where PN 1 is given by Eqn.4.35.

78

4.7 Example

In this example, a star network with Nu users (Nu=2, 4, 8, 16, ...) has been considered. Data rate for each channel has been assumed to be 1 Gb/s. Typical values of different parameters in the network are as follows [19, 47, 62, 87].

Quantum efficiency of photodetector, Length of fiber from user to star coupler, L Attenuation coefficient of fiber, Unsaturated amplifier gain, G0 Saturation power level of amplifier, Psat Optical filter bandwidth, Bo Insertion loss of each 2x2 coupler, Li Insertion loss of splice, Lsp Output power variability of NuxNu coupler, Lcv Amplifier facet reflectivities, Rf Amplifier coupling loss, Lcp Receiver temperature, T Load resistance, RL Spontaneous emission factor, nsp Electrical bandwidth of receiver, Be Operating wavelength, Filter insertion loss, Lfi

0.95 1 km 0.2 dB/km 29 dB 10 dBm 10 GHz 0.5 dB 0.5 dB 0.5 dB 10-5 (-50 dB) 3 dB 3000 K 100 Ohms 3.0 1 GHz 1.55 m 0.5 dB

79

For this network, numerical computations have been made for various values of Nu and a specified BER of 10-9. The required transmitter power is determined (i) without SOA, (ii) with optical postamplifier and (iii) optical preamplifier. The above is computed for different values of extinction ratio i.e. =0.0, 0.05, 0.10 and 0.15. Numerical computations are also made to include the effect of gain saturation and reflection noise. These results are shown in Figs.4-5 to 4-14. In the following, results and inferences based on the above figures are discussed.

Fig.4-5

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.0 in postamplifier scheme

i.

Use of an ideal amplifier reduces the required minimum transmitter power level in

both postamplifier and preamplifier schemes. The reduction is more in the postamplifier

80

Fig.4-6

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.05 in postamplifier scheme

than in the preamplifier because ASE noise undergoes higher attenuation in the former scheme.

ii.

Gain saturation in the amplifier results in an increase in the required transmitter

power Pta. The difference in power level is referred to as power penalty. In the postamplifier scheme, the penalty becomes more severe with increasing Nu. It can be explained as follows. As Nu increases, split loss of the star coupler also increases. Therefore, more transmitter power is required to compensate for this loss. It results in decrease in gain of the amplifier owing to an increase in gain saturation. The penalty becomes so severe after Nu=512 that it overcomes the amplifier gain resulting in degradation of system performance. The reason for gain saturation penalty exceeding

81

Fig.4-7

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.10 in postamplifier scheme

unsaturated amplifier gain is that for large Nu, gain for bit 1 is almost unity due to saturation effect and hence ASE noise psd is negligible. However for bit 0, gain is much more than unity and this results in substantial ASE noise psd. Therefore, on an average the noise level is more than the noise level without amplifier. It degrades the system performance for large Nu. The same trend is observed for all values of extinction ratio, . The gain saturation effect increases with the increase in .

In the preamplifier scheme, the effect of average gain saturation is small for low values of Nu. It increases with an increase in Nu. It is observed that degradation due to gain saturation effect is much less in this scheme than in the postamplifier scheme because of much lower average power at the amplifier input.

82

Fig.4-8

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.15 in postamplifier scheme The effect of gain fluctuations in the preamplifier first increases with Nu and after

attaining a maximum value decreases with the increase in Nu (Figs.4-9 to 4-12). Since the effect of gain fluctuations depends on the variance of probability density function of gain which increases with Nu, for low value of Nu. However, for large Nu average gain shifts towards the lower extreme of the range (1 to G0) and the levels are cluttered together. This results in a decrease in variance with Nu at large Nu. For some intermediate value of Nu, the variance will be maximum. This explains the variations of effect of gain fluctuations with Nu.

iii.

The effect of reflection noise reduces with the increase in Nu for both post and

preamplifier schemes. As Nu increases, split loss of the star coupler also increases which

83

Fig.4-9

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.0 in preamplifier scheme

results in increased required transmitter power. In the postamplifier scheme, it will give rise to gain saturation and hence a reduction in amplifier gain, while in the preamplifier scheme, the total power at the amplifier input increases with the increase in number of channels. Therefore, amplifier gain reduces with an increase in Nu in the preamplifier scheme also. A reduction in amplifier gain means a decrease in amplifier reflectance which reduces the degrading effect of reflection. It is also observed that preamplifier scheme is more severely affected by reflection noise than the postamplifier scheme. This can be attributed to more gain saturation in the postamplifier.

84

Fig.4-10

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.05 in preamplifier scheme

iv.

Figs.4-13 and 4-14 show the variations of average required transmitter power, Pta,

with . It is observed that Pta increases with an increase in as expected.

Figs.4-7 and 4-11 are used to present the data in tabular form in Tables 4-1 and 4-2. The following observations are made from these tables.

v.

With the use of unsaturated SOAs in the network, number of users is increased for

both the post and preamplifier schemes. Gain saturation and reflection noise reduce the above advantages. For example, in the postamplifier scheme, Nu increases to 4096 from 64 for transmitter power level of 0 dBm. When gain saturation and reflection noise are

85

Fig.4-11

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.10 in preamplifier scheme

Number of users Pta (dBm) Without SOAs Unsaturated SOAs Gain saturated SOAs 256 128 32 4 Gain saturated SOAs with reflection (Rsp=-20 dB) 256 64 0 0

0 -10 -20 -30

64 8 0 0

4096 512 32 4

Table 4-1

Number of users supported in the postamplifier scheme for =0.10.

considered, Nu is 256. In general, the number of users will increase as the reflection

86

Fig.4-12

Variations of minimum required average transmitter power Pta with number of users Nu for = 0.15 in preamplifier scheme

coefficient of splice decreases for a fixed transmitter power level.

vi.

In the preamplifier scheme, greater number of users are supported except when Pta

is very low and Rsp is very high e.g. Pta = -10 dBm and Rsp = -20 dB. It is observed that the supportable number of users reduces with the decrease in Pta and the increase in Rsp. However, networks are usually not designed for very low transmitter powers and reflection coefficients due to splices is much lower than -20 dB.

vii.

For a typical Pta of 0 dBm, the preamplifier scheme performs better than the

postamplifier scheme in terms of an increase in the number of users. When the transmitter power level is low (less than -10 dBm) and reflection coefficient is high, the postamplifier

87

Fig.4-13

Variations of Pta with for different Nu in postamplifier scheme. Solid lines correspond to gain saturated SOAs and dashed lines to gain saturated SOAs with reflection noise.

scheme is better. However, networks are not implemented for such low transmitter power levels and high reflection coefficients.

4.8 Conclusions

In this chapter, placement of optical amplifiers in a WDM star network has been investigated. The number of users supported by the star topology for post and preamplifier schemes has been determined. The effect of gain saturation and reflection noise in both these schemes has been evaluated in terms of number of users supported or the required transmitter power level for a given BER. The analyses also consider the effect of extinction

88

Fig.4-14

Variations of Pta with for different Nu in preamplifier scheme. Solid lines correspond to gain saturated SOAs and dashed lines to gain saturated SOAs with reflections.

ratio. When there is no gain saturation and reflection noise in the network, postamplifier scheme performs better. In presence of gain saturation, postamplifier scheme performs better for low transmitter power levels. However, such power levels are not used in practice. When number of users and consequently transmitter power increases, performance in postamplifier scheme degrades, while in the preamplifier scheme it improves. The power penalty due to reflection noise reduces in both the schemes with the increase in the number of users.

Overall, the preamplifier is a better scheme than the postamplifier in WDM star topology networks. The formalism developed in this chapter has been used to evaluate the

89

Number of users Pta (dBm) Without SOAs Unsaturated SOAs Average gain saturated SOAs Average gain saturated SOAs with gain fluctuation 512 128 16 2 Average gain saturated SOAs with gain fluctuations and reflection (Rsp=-20 dB) 512 0 0 0

0 -10 -20 -30

64 8 0 0

1024 128 16 2

1024 128 16 2

Table 4-2

Number of users supported in the preamplifier scheme for =0.10.

performance of tree-net topology in chapter 5. A comparative study of star network and tree-net has been made in the same chapter.

90

Chapter 5

Semiconductor Optical Amplifiers In WDM Tree-net

In past few years, tree-net topology has received the attention of network designers. It provides large geographical area coverage with less amount of fiber and easy expandability. The protocols can be designed for these networks to achieve low latency and bounded delay [57]. Such networks can be configured in broadcast mode using passive components or active regenerators. However, use of passive components is preferred for the reasons of reliability and lower cost. The capacity of these networks can be tremendously increased by making use of wavelength division multiplexing (WDM). The tree-net is a two level topology (Fig.5.1) consisting of star as main topology and folded bus as auxiliary topology. In a purely passive tree-net, the number of users supported is limited by split and distribution losses in star couplers. The number of users supported can be increased by using semiconductor optical amplifiers (SOAs).

Like in a star network, several SOA placement schemes are also possible in a treenet. As shown in the previous chapter, preamplifier scheme (SOAs as frontend of receivers) is a better arrangement in the star topology [100]. With this in view, placement of SOAs

The work reported in this chapter has resulted in the following communications
1 Y. N. Singh, H. M. Gupta and V. K. Jain, "Semiconductor Optical Amplifiers in WDM Tree-net," IEEE/OSA Journal of Lightwave Technology (under review).

91

after the root (Fig.5-1) has been considered in the tree-net. In this chapter, tree-net has been analyzed with and without SOAs. The results have been computed in terms of increase in the number of users or reduction in the required transmitter power level with the placement of SOAs. Subsequently, a comparative study of tree-net and star network has been made.

Fig.5-1

Basic tree-net topology.

5.1 System Description

In a tree-net with b (= 2i, i is a natural number) branches and n number of users per branch, total number of supported users are Nu (= bn). A tree-net topology with b=4 and n=4 is shown in Fig.5-1 [57]. In this topology, the users termed 1 are first order users, 2 are second order users and so on. Therefore, order of users varies from 1 to n.

92

The power loss in the above network can be reduced by realising the star portion as multistage star coupler (Fig.5-2) [66].

Fig.5-2a

Structure of star coupler in tree-net without amplifiers.

In the tree-net topology, Na (=2k, k=0, 1, 2, ...) number of SOAs can be used. There are three cases for incorporating SOAs in tree-net viz. (i) Na is one (20), (ii) Na ia less than b and (iii) Na is equal to b. These configurations are shown in Fig.5-3a, 5-3b and 5-3c respectively for 4x4 star portion of a tree-net. In these configurations, 3 dB couplers are used to make the star coupler of dimension NaxNa (Na > 1). The amplifiers are attached at the output of this star coupler. Further, the dimension of the above is increased by using the 3 dB 2x2 couplers as combiners at inputs (A, B, ...) and splitters at outputs (A, B, ...). The 3 dB 2x2 couplers are also used to connect the users to the folded bus.

93

Fig.5-2b

Structure of branches in a tree-net without amplifiers.

Fig.5-3a

A 4x4 star coupler with one amplifier in tree-net.

In the following analysis, number of wavelength channels in the tree-net are assumed to be b. In a simple media access control (MAC) protocol, users on each branch transmit on an unique wavelength channel. The contention for transmission among the users on same branch can be resolved using variety of single channel protocols such as random access time out (RATO) [14], ALOHA etc.. Each user can receive simultaneously on all the

94

Fig.5-3b

A 4x4 star coupler with two amplifiers in tree-net.

Fig.5-3c

A 4x4 star coupler with four amplifiers in tree-net.

wavelengths using an array of receivers. In contrast to WDM star [100], more than one user can share a wavelength channel in tree-net. This reduces the required size of receiver arrays at all the receivers. The available bandwidth in a wavelength channel is shared among the users in a branch. Like in star network, the acknowledgement of packets is not required in the tree-net also as the transmitting users listen to their own packet. If the transmitted packet is received back correctly, it is expected that the packet has also been successfully transmitted to its destination. It is also assumed that each transmitter in the tree-net adjusts its transmitter power such that the power received from its own transmitter is same as that

95

received from highest order user (4th order user in Fig.5-2b). Thus a farthest (highest order) user from the root on the bus transmits maximum power. This implies that the optical power level at the amplifier input in each channel is same.

5.2 Analytical Model

In order to analyze the effectiveness of SOAs in tree-net, following four cases are considered i.e. tree-net (i) without SOAs, (ii) with unsaturated SOAs, (iii) with SOAs having average gain saturation and (iv) with SOAs having average gain saturation and gain fluctuations. The optical amplifier model used in the analysis is same as used in previous chapters.

For the worst-case analysis, the highest order transmitter has been considered. It is remarked that the choice of receiver for worst-case analysis is not obvious. This is because signal as well as the ASE noise is minimum at the highest order receiver. For the lowest order receiver, both the signal as well as the ASE noise are at the maximum. Therefore, BER for different receivers is computed and the receiver giving maximum BER or worstcase receiver is identified. The computed results show that highest order receiver always gives the worst performance. The BER for all the cases mentioned above are computed following the same approach as in chapter 4 [100]. In the tree-net without SOAs, receiver of highest order has been considered for the worst-case analysis as the received signal is minimum for it. The loss Ltr between transmitter and receiver depends upon number of users, number of branches, attenuation in fiber, splice and insertion losses. It is given by

96

(5.1a)

where Xtr is the above loss in dB and is given by

(5.1b)

Here, Lss is the fiber loss between two consecutive users on a bus, Li the insertion loss of a 3 dB 2x2 coupler, the attenuation coefficient of fiber in dB/km, L the length of fiber between order one user and star portion of the network in km, Lsp the splice loss and Lfi the insertion loss of wavelength demultiplexer. The received optical power is determined using the given transmitter power level and loss Ltr. The noise variances at the receiver (sum of shot noise and thermal noise) are determined for received optical power levels for bit 1 and 0. The BER which equalizes the probability of error for bit 1 and 0 is computed using the signal currents and noise variances.

When SOAs without gain saturation are used in the tree-net, the optical power level in a channel at the amplifier input is determined using given transmitter power level and loss between transmitter and amplifier Lta. In addition to parameters which affect Ltr, the Lta also depends on number of SOAs. It is given by

97

(5.2a)

where Xta is the above loss in dB and is given by

(5.2b)

The received signal and ASE noise at the receiver of order j (1 j n) are determined using the unsaturated gain G0 of the amplifier and loss, Lar, between amplifier and the receiver of order j. The loss Lar is given by

(5.3a)

where

98

(5.3b)

At the receiver of order j, noise variance due to ASE-signal beat noise, ASE-ASE beat noise, shot noise (due to signal and dc component of ASE-ASE beat noise) and thermal noise are determined for bit 1 and 0. The BER which equalises the probability of error for bit 1 and 0 is determined using corresponding signal levels and noise variances.

In presence of average gain saturation, signal power at the input of SOA is determined as mentioned above. It is presumed that bits in all the channels at the input of SOA are synchronised which produce maximum degradation [75, 76]. For a given number of channels having bit 1, total input power to SOA is determined and used for finding the saturated gain. The average gain is determined by averaging this gain over the binomial distribution of number of channels having bit 1. The received optical signal and ASE noise at the input of the receiver are determined using this average gain. These are used to obtain the signal levels and noise variances at the receiver output which are used to determine the BER.

99

In the last case i.e. tree-net with SOAs having average gain saturation and gain fluctuations, the BER is determined for a given bit pattern in the interfering channels. It is then averaged using the binomial distribution of interfering channels having bit 1 to determine overall BER. For a given channel, the power level at the input of SOA is determined as in case of tree-net with unsaturated SOAs. The saturated gain of SOA for bit 1 and 0 in desired channel is determined for a given number of interfering channels having bit 1. Then received optical signal power and ASE noise psd are determined at the input of receiver of order j. These are used to determine signal levels and noise variances at the receiver output for bit 1 and 0 in the desired channel. As there are (b-1) interfering channels, b signal levels for both bit 1 and 0 will exist depending upon number of interfering channels having bit 1. The highest signal level for bit 0 is always less than the lowest level for bit 1 [75]. Using these two levels, threshold is determined which equalises the probability of error for these levels. The probability of error for each signal level is determined using the corresponding noise variance and the computed threshold. This probability of error is averaged presuming equiprobable bit 1 and 0 in the desired channel and the binomial distribution of number of interfering channels.

5.3 Example

The effectiveness of optical amplifiers on a sample tree-net is studied using the above model. The tree-net has the following parameters:

Maximum allowed transmitter power, Pta Desired BER

0 dBm 10-9

100

Insertion loss in 2x2 coupler, Li Unsaturated gain of SOAs, G0 Saturation power level in SOAs, Psat Distance between the root and first order user on a branch, L Distance between consecutive nodes on a branch, Lss Attenuation coefficient of fiber, Load resistance of receiver, RL Operating wavelength, Optical BW of a passband in wavelength demultiplexer, Bo Electrical bandwidth of receiver, Be Temperature of receiver, T Spontaneous emission factor, nsp Insertion loss of splice, Lsp Insertion loss of wavelength demultiplexer, Lfi Extinction ratio,

0.5 dB 29 dB 10 dBm 1 km 100 m 0.2 dB/km 100 1.55 m 10 GHz 1 GHz 300 oK 3.0 0.5 dB 0.5 dB 0.10

Numerical results for the sample tree-net are computed using the above parameter values. Table 5.1 shows the number of users supported, Nu, and minimum required transmitter power level, Pta, for various values of n without SOAs. Variations of Nu with Na for three cases viz. tree-net with (i) unsaturated SOAs, (ii) average gain saturated SOAs and (iii) average gain saturated SOAs with gain fluctuations are shown in Figs.5-4, 5-5 and 5-6 respectively. The numbers shown in brackets are the minimum required transmitter power levels (dBm) for the corresponding values of Nu and Na. For example, in Fig.5-4a when n=2 and Na=32, Pta is -5.4 dBm and corresponding Nu is 512. It is remarked that plot

101

Fig.5-4a

Variations of Nu with Na (unsaturated SOAs) for n=2. The values in the brackets are corresponding required Pta in dBm.

Fig.5-4b

Variations of Nu with Na (unsaturated SOAs) for n=3. The values in the brackets are corresponding required Pta in dBm.

for n=1 (not shown here) is similar to the ones for n=2. It is observed that for a given value of n, increase in Na may result in increased Nu or reduced PT or both. Increase in Na

102

Fig.5-5a

Variations of Nu with Na (average gain saturated SOAs) for n=2. The values in the brackets are corresponding required Pta in dBm.

Fig.5-5b

Variations of Nu with Na (average gain saturated SOAs) for n=3. The values in the brackets are corresponding required Pta in dBm.

does not necessarily increase Nu. Since in tree-net, Nu varies in discrete steps, the next

103

Fig.5-6a

Variations of Nu with Na (average gain saturated SOAs with gain fluctuations) for n=2. The values in the brackets are corresponding required Pta in dBm.

Fig.5-6b

Variations of Nu with Na (average gain saturated SOAs with gain fluctuations) for n=3. The values in the brackets are corresponding required Pta in dBm.

higher value may not be supported if the required Pta is above 0 dBm. For example, Nu is

104

192 for n=3 and Na=4 (Fig.5-4b). When the Na is increased to 8, Nu is still 192 because for the next admissible number Nu=384, required Pta is more than 0 dBm. But the next higher value of Na i.e. 16 results in increase of Nu to 384. It is noticed that increase in Na from 4 to 8 results in reduction in Pta from -2.6 dBm to -5.7 dBm and no change in Nu. Therefore, Na=8 is not useful as the available Pta is 0 dBm and there is no increase in Nu.

Number of users per branch, n

Tree-net without SOAs Nu=bn Pta (dBm) -2.2 -2.2 -1.8

1 2 3

64 32 12

Table 5.1

Number of users and required Pta for the tree-net without SOAs.

When average gain saturated SOAs are considered, a decrease in Nu is observed compared to above. For example, when n=2, Na=16 supports 512 users for unsaturated SOAs and 256 users for average gain saturated SOAs. This behaviour is true for all values of n. Further, the gain fluctuations in SOAs either increase Pta or reduces Nu. For example, when n=2, Na=16 still supports same number of users as in average gain saturation case, but Pta has increased to -2.6 dBm from -3.8 dBm showing a power penalty of 1.2 dB. An example of reduction in Nu is n=3, Na=8 when Nu reduces from 192 to 96 (see Figs.5-5b and 5-6b).

105

In tree-net topology, n=3 provides the values of Nu which are not admissible in star topology. This is advantageous in certain cases as seen in Table 5.2. The table shows the values of Nu corresponding to each Na for all the three cases. For each case, minimum required Pta has also been computed and shown. It is observed that maximum number of users are supported by n=2 in most of the cases. The worst-case losses in n=1 and 2 are almost same, but the number of wavelengths used in n=2 are halved. This results in reduced effect of average gain saturation and fluctuations. In some cases, n=3 supports more number of users than n=2 (e.g. Na=32 supports a maximum of 384 users for n=3 in average gain saturation case). This is due to the peculiarity of number of users admissible in tree-net for n=3.

A comparison of star network with tree-net is given in Table 5.3. It is observed from this table that both the networks can support 64 users without SOAs. Both star network and tree-net with SOAs can support 128 users. However, the former needs 128 SOAs and the latter 4 SOAs. Similarly, star network can support 256 and 512 users with 256 and 512 SOAs respectively, while the tree-net requires 16 and 64 SOAs to support the same number of users respectively. It is also observed that 1024 users are supported by the tree-net only. The above is possible as the amplifiers are integrated within the star coupler and the amplifier coupling loss is negligible. Therefore, the tree-net requires lesser number of SOAs for a given number of supportable users. Both star network and tree-net cannot support more than 1024 users irrespective of number of SOAs used.

106

Number of Unsaturated SOAs Average gain saturated Average gain saturated amplifiers, SOAs SOAs with gain fluctuations Na Nu (n) Pta (dBm) Nu (n) Pta (dBm) Nu (n) Pta (dBm) 1 2 4 8 16 32 64 128 256 512 1024 2048 128 (2) 128 (2) 256 (2) 256 (2) 512 (2) 512 (2) 1024 (2) 1024 (2) 2048 (2) 2048 (2) 2048 (2) 2048 (1) -3.0 -6.1 -2.6 -5.7 -2.2 -5.4 -1.8 -3.8 -0.3 -1.5 -1.8 -1.8 64 (2) 64 (2) 128 (2) 192 (3) 256 (2) 384 (3) 512 (2) 512 (2) 1024 (2) 1024 (2) 2048 (2) 2048 (1) -3.4 -10.0 -3.8 -0.3 -3.8 -0.3 -3.4 -6.9 -2.2 -4.6 -0.3 -1.1 64 (2) 64 (2) 128 (2) 128 (2) 256 (2) 256 (2) 512 (2) 512 (2) 1024 (2) 1024 (2) 1024 (2) -0.7 -8.9 -1.8 -8.9 -2.6 -8.5 -3.0 -6.5 -1.8 -3.4 -3.0 -

Table 5.2

Maximum number of users and Pta for a given number of SOAs. Numbers in brackets are the corresponding values of n.

5.4 Conclusions

It has been observed that the gain saturation in SOAs affects the performance of tree-net. It reduces number of users supported as compared to tree-net with unsaturated SOAs. The gain fluctuations in SOAs further deteriorate the performance. Since the star topology supports maximum number of users without SOAs, star network and tree-net are

107

Number of users, Nu 64 128 256 512 1024

Number of SOAs required Star Tree 0 0 128 256 512 4 16 64 256

Table 5.3

Comparison of star network and tree-net in terms of number of SOAs required for a given number of users.

compared. It is shown that the tree-net require lesser number of SOAs as compared to star for a fixed number of supportable users. This is because of topological advantage of treenet in terms of wider choices of users per SOA. It is remarked that in comparison to star network, tree-net offers distinct advantages in terms of more geographical coverage, less usage of fiber and transmitter sharing between more than one user [83, 86].

108

Chapter 6

Conclusions
In the research work reported in this thesis, usage of semiconductor optical amplifiers (SOAs) in wavelength division multiplexed (WDM) star network and tree-net has been studied. The SOAs are used to alleviate the limitation on power budget and hence number of users supported. Error correcting codes can also be used for this purpose. The use of SOA and an error correcting code has been investigated in a point-to-point OOK link. It has been observed that the improvement in power budget due to SOA exceeds the improvement due to coding by approximately 20 dB. Therefore, the use of SOA is a much better option than the use of an error correcting code. In view of this, only the SOAs are considered for the placement in WDM star and tree networks.

Two SOA placement schemes (i.e. postamplifier and preamplifier) in WDM star network have been investigated. The number of SOAs in these schemes are always equal to number of users. Each user is assigned an unique wavelength for transmission. The study shows that when unsaturated SOAs are used, the postamplifier scheme is better than the preamplifier scheme. The reason for the above is that the amplified spontaneous emission (ASE) noise is attenuated more in the postamplifier scheme than in the preamplifier scheme. Consequently, the effect of ASE noise is less in the postamplifier scheme. When gain saturated SOAs are considered, the supportable number of users is reduced in both the postamplifier and the preamplifier schemes. The degradation increases with an increase in the number of users in both the schemes. This is quite expected. However, the degrading effect of gain saturation is relatively more in the postamplifier scheme. The SOAs in the

109

preamplifier scheme are affected by both average gain saturation and gain fluctuations. The average gain saturation is dependent on number of WDM channels and hence number of active users. The gain fluctuations depend on bit patterns in active channels. The analysis in chapter 4 shows that the degrading effect of average gain saturation increases with an increase in the number of users. However, the degradation due to gain fluctuations first increases to a maximum value and then decreases. When reflection noise due to splices is also considered, the performance of both the schemes degrades. The above degradation is more severe in the preamplifier scheme than in the postamplifier scheme. The effect of nonzero extinction ratio has also been studied. As expected, it has been observed that the degradation increases with an increase in extinction ratio. In general, the preamplifier scheme performs better than the postamplifier scheme and hence supports more number of users.

The tree-net topology is a more general topology of which star is a special case. The tree-net can support 2in (i=1,2,..., n=1,2,...) users while star supports only 2i users. In the treenet, there exists the possibility of sharing transmitter sources using passive access node (PAN) interface [86]. In this thesis, number of wavelengths in the tree-net is chosen to be equal to number of branches. The SOAs are placed in the star coupler part of the tree-net. In contrast to the star network, number of SOAs in the tree-net can be less than the number of branches. The number of users supported by tree-net with unsaturated SOAs and saturated SOAs is determined. As expected, the supported number of users reduces when SOAs are gain saturated. When tree-net is compared with the star, it is observed that the tree-net can support more number of users than star for a given maximum transmitter power level and number of amplifiers. This advantage accrues due to the topology of the tree-net which provides wider possible assignment of number of users for a given number of SOAs. In brief, it is concluded

110

that use of SOAs is a better choice to increase the supportable number of users in both star network and tree-net. Further, the tree-net utilizes the SOAs more effectively as compared to the star network.

In the work presented in this thesis, following assumptions and simplifications have been made. The noise of the transmitting sources has not been considered. The dispersion in fiber is neglected as distances involved are small for an appreciable dispersion. Modal noise due to fiber [68] has been assumed to be absent. The couplers used are considered to be wavelength independent. Further, the demultiplexers and filters are taken to be ideal. Therefore, no crosstalk occurs due to these components. The responsivity of photodetectors has been assumed to be wavelength independent over the range of operation.

In the foregoing analyses, SOAs are considered as travelling wave amplifiers (TWAs) and the gain profile is assumed to be uniform in the operating wavelength range. Further, the crosstalk in the SOAs are considered to arise only due to the gain saturation.

In this thesis, use of SOAs in only two of the many topologies i.e. WDM star and WDM tree-net has been studied. Therefore, OA placement options can also be explored in other broadcast topologies including multilevel topologies. It is remarked that multilevel topologies may be advantageous in terms of number of OAs required for a given number of users.

As mentioned in chapter 1, switched optical networks will be attractive in the future. Therefore, the placement of OAs in these networks requires investigations. These switched

111

networks include wavelength routed networks in which wavelengths are directed to those parts of the network where these are needed the most. The proper routing of wavelengths allows wavelengths reuse in different parts of network [54].

In this thesis, SOAs have been considered because these can be easily integrated in star coupler, transmitter or receiver. However, doped fiber amplifiers (DFAs) are finding applications in wide area networks as in-line amplifiers. Therefore, analysis presented in this thesis can be extended to the networks using DFAs.

Another important area where optical amplifier would be very useful is subscriber access networks (SANs). In these, both WDM techniques and OAs can be used for connecting large number of users to a central office. The topologies used in SANs have different requirement as compared to LANs and WANs. The OA placement studies can be extended to SANs considering these requirements.

In the above, use of OAs as signal amplifier for the purpose of increasing the size of network has been suggested. The OAs can also function as switches and wavelength convertors. Therefore, studies of optical networks which include these additional functionalities of OAs are suggested.

It is concluded from the above that OAs will form an important part of the future high capacity optical networks.

112

References

[1]

A. A. Hassan, J. E. Hershey and N. A. Reja, "Spatial Optical CDMA," IEEE Journal on Selected Areas in Communications, Vol.13, No.3, April 1995, pp.609-613. Photonic Switching and Interconnects, Eds. Abdellatif Marrakchi, Marcel Dekker, 1994. A. D. Ellis, K. Smith and D. M. Patrick, "All Optical Clock Recovery at Bit Rates upto 40 Gbit/sec," Electronics Letters, Vol.29., No.15, 22nd July 1993, pp.1323-1324. A. D. Ellis, D. M. Patrick, D. Flannery, R. J. Manning, D. A. O. Davies and D. M. Sprit, "Ultra High-Speed OTDM Networks using Semiconductor Amplifier Based Processing Nodes," IEEE/OSA Journal of Lightwave Technology, Vol.13, No.5, May 1995, pp.761-770. A. R. Salvatore, T. Schraus and A. Yariv, "Wavelength Tunable Sources of Subpicosecond Pulses from CW Passively Mode Locked Two-Section Multiple Quantum Well Laser," IEEE Photonics Technology Letters, Vol.5, No.7, July 1993, pp.756-758. A. S. Tanenbaum, Computer Networks, 2nd Ed., Prentice Hall of India, 1992. A. Yariv, Optical Electronics, 4th Ed., Saunders College Publishing, 1991. B. Mukherjee, "WDM-Based Local Lightwave Networks Part I: Single-Hop Systems," IEEE Network, May 1992, pp.12-27. B. Mukherjee, "WDM-Based Local Lightwave Networks Part II: Multihop Systems," IEEE Network, July 1992, pp.20-32. B. S. Glance, J. Stone, K. J. Pollock, P. J. Fitzgerald, C. A. Burrus, Jr., B. L. Kasper and L. W. Stulz, "Densely Spaced FDM Coherent Star Network with Optical Signals confined to Equally Spaced Frequencies," IEEE/OSA Journal of Lightwave Technology, Vol.6, No.11, Nov.1988, pp.1770-1781. B. S. Glance and O. Scaramucci, "High-Performance Dense FDM Coherent Optical Network," IEEE Journal on Selected Areas in Communications, Vol.8, No.6, August 1990, pp.1043-1047. B. W. Abeysundra and A. E. Kamal, "High Speed Local Area Networks and Their Performance: A Survey," ACM Computing Surveys, Vol.23, No.2, June 1991, pp.227-264.

[2]

[3]

[4]

[5]

[6] [7] [8]

[9]

[10]

[11]

[12]

113

[13]

C. A. Brackett, "Dense Wavelength Division Multiplexing Networks: Principles and Applications," IEEE Journal on Selected Areas in Communications, Vol.8, No.6, August 1990, pp.948-964. Cavour Yeh, Maurice Lin, Mario Gerla and Paulo Rodrigues, "RATONET: A Random Access Protocol for Unidirectional Ultra-High-Speed Optical Fiber Network," IEEE/OSA Journal of Lightwave Technology, Vol.8, No.1, January 1990, pp.78-89. C. Cremer, M. Schier, G. Baumeister, G. Ebbinghaus, H. Huber, W. Kunkel, J. G. Bauer, G. Kristen, J. Rieger, R. Schimpe and R. Strzoda, "WDM Receiver Chip with High Sensitivity," Electronics Letters, Vol.30, No.19, 15th Sept.1994, pp.16251626. C. E. Zah, J. Gamelin, B. Pathak, F. Favire, P. S. D. Lin, N. Andreadakis, R. Bhat, C. Caneau, L. Curtis, D. D. Mohoney, W. C. Young and T. P. Lee, "Multiwavelength Light Sources with Integrated DFB Laser Array and Star Coupler for WDM Lightwave Communication," International Journal of High Speed Electronics Systems, Vol.5, No.1, March 1994, pp.91-109. Christian Lester et.al, "Reflection Insensitive Erbium-Doped Fiber Amplifiers," IEEE Photonics Technology Letters, Vol.7, No.3, March 1995, pp.293-296. D. A. Chapman, "Erbium-Doped Fiber Amplifiers: The Latest Revolution in Optical Communications," IEE Electronics and Communication Engineering Journal, Vol.6, No.2, April 1994, pp.59-67. D. J. Blumenthal and N. C. Kothari, "Coherent Crosstalk in Multichannel FSK/DD Lightwave Systems Due to Four-Wave Mixing in Semiconductor Optical Amplifiers," IEEE Photonics Technology Letters, Vol.8, No.1, January 1996, pp.133-135. D. W. Faulkner, D. B. Payne, J. R. Stern and J. W. Ballance, "Optical Networks for Local Loop Applications," IEEE/OSA Journal of Lightwave Technology, Vol.7, No.11, Nov.1989, pp.1741-1751. D. Zaccarin and M. Kavehrad, "An Optical CDMA System Based on Spectral Encoding of LED," IEEE Photonics Technology Letters, Vol.4, No.4, April 1993. E. Goldstein, "Optical Ring Networks with Distributed Amplification," IEEE Photonics Technology Letters, Vol.3, No.4, April 1991, pp.390-393. E. L. Goldstien, V. da Silva, L. Eskildsen, M. Andrejco and Y. Silberberg, "Inhomogeneously Broadened Fiber-Amplifier Cascade for Wavelength-Multiplexed Systems," IEEE Photonics Technology Letters, Vol.5, No.5, May 1993, pp.543-545. E. Patzak, "Calculation of Bit Error Rate in DPSK System with Interfering Echoes from Optical Chain," Electronics Letters, Vol.25, No.14, 6th July 1989, pp.902-904.

[14]

[15]

[16]

[17]

[18]

[19]

[20]

[21]

[22]

[23]

[24]

114

[25]

F. E. Ross, "FDDI: A Tutorial," IEEE Communication Magazine, Vol.24, No.5, May 1986, pp.10-17. F. E. Ross, "An Overview of FDDI: The Fiber Distributed Interface," IEEE Journal on Selected Areas in Communications, Vol.7, No.7, Sept.1989, pp.1043-1051. G. Grau and W. Freude, Optische Nachrichtentechnik Eine Einfhrung, SprigerVerlag, 1991 (Berlin). G. Kieser, Optical Fiber Communications, 2nd Ed., McGraw Hill Inc., 1991. G. M. Foster, J. R. Rowsthorne, J. P. Hall, M. Q. Kearley and P. J. Williams, "OEIC WDM Transreceivers Modules for Local Access Networks," Electronics Letters, Vol.31, No.4, 19th Jan.1995, pp.132-133. G. P. Agrawal, Optical Communication Systems, John Wiley, 1992. G. R. Pieris and G. H. Sasaki, "Scheduling Transmission in WDM Broadcast-AndSelect Networks," IEEE/ACM Transactions on Networking, Vol.2, No.2, April 1994, pp.105-110. H. Kobrinski, R. M. Bulley, M. S. Goodman, M. P. Vecchi and C. A. Brackett, "Demonstration of High Capacity in the Lambdanet Architecture: A Multiwavelength Optical Network," Electronics Letters, Vol.23, No.18, 30 July 1987, pp.824-826. H. Toba, K. Oda, K. Nosu and N. Takato, "Factors Affecting the Design of Optical FDM Information Distribution Systems," IEEE Journal on Selected Areas in Communications, Vol.8, No.6, Aug.1990, pp.965-972. I. Chlamatac, A. Ganz and G. Karmi, "Lightpath Communication: An Approach to High Bandwidth Optical WANs," IEEE Transactions on Communications, Vol.40, No.7, July 1992, pp.1171-1182. "Introduction to SONET/SDH," Editorial, IEEE Communication Magazine, Vol.31, No.9, Sept.1993. J. C. Lu and L. Kleinrock, "A Wavelength Division Multiple Access Protocol for High-Speed Local Area Networks with a Passive Star Topology," Performance Evaluation, Vol.16, 1992, pp.223-239. J. Franz, C. Rapp and G. Sder, "Influence of Baseband Filtering on Laser Phase Noise in Coherent Optical Transmission Systems," Journal of Optical Communications, Vol.7, No.1, 1986, pp.15-20.

[26]

[27]

[28] [29]

[30] [31]

[32]

[33]

[34]

[35]

[36]

[37]

115

[38]

J. Franz, C. Correll, F. Dolainsky, R. Schweikert and B. Wandernoth, "Error Correcting Coding in Optical Transmission Systems with Direct Detection and Heterodyne Detection," Journal of Optical Communications, Vol.14, No.5, 1994, pp.194-199. J. Franz and V. K. Jain, Optical Communication Systems, Narosa Publishing House, Delhi, 1996. J. -G. Zhang and G. Picchi, "Tunable Prime-Code Encoder/Decoder for All Optical CDMA Applications," Electronics Letters, Vol.29, No.13, 24th June 1993, pp.1211-1212. J. -G. Zhang and G. Picchi, "Self Synchronized All-Optical Time-Division MultipleAccess Broadcast Network," Electronics Letters, Vo.29, No.21, 14th Oct.1993, pp.1871-1872. J. P. Raina, "Futuristic Trends in Ultra High Speed Network," Proc. National Conference on Optical Communications, J.K.Institute of Applied Physics and Technology, University of Allahabad, Allahabad, Feb.22-24, 1995, pp.1. K. Asatani, "Lightwave Subscriber Loop Systems Towards Broadband ISDN," IEEE/OSA Journal of Lightwave Technology, Vol.7, No.11, Nov.1989, pp.1705-1714. K. A. Williams, T. Q. Dam and D. H-C. Du, "A Media-Access Protocol for Time- and Wavelength-Division Multiplexed Passive Star Networks," IEEE Journal on Selected Areas in Communications, Vol.11, No.4, May 1993, pp.560-567. K. Bogineni, K. M. Sivalingam and P. W. Dowd, "Low-Complexity Multiple Access Protocols for Wavelength-Division Multiplexed Photonic Networks," IEEE Journal on Selected Areas in Communications, Vol.11, No.4, May 1993, pp.590-604. K. Inoue, "Crosstalk and Power Penalty in Multichannel Transmission due to Gain Saturation in a Semiconductor Laser Amplifier," IEEE/OSA Journal of Lightwave Technology, Vol.7, No.7, July 1989, pp.1118-1124. Kyo Inoue and Manabo Yoshino, "Gain Dynamics of a Saturated Semiconductor Laser Amplifier with 1.47 m LD Pumping," IEEE Photonics Technology Letters, Vol.8, No.4, April 1996, pp.506-508. K. Kikushima, T. Whitley, R. Cooke, K. Stalley, M. Fabe and E. Lawrence, "Multichannel FM-TV Transmission using and Engineered 1.3 m PraseodymiumDoped Fluoride Fibre Amplifier," Electronics Letters, Vol.30, No.17, 18th August 1994, pp.1431-1432. K. Rottwitt, J. H. Povlsen and A. Bjarklev, "Long Distance Transmission Through Distributed Erbium-Doped Fibers," IEEE/OSA Journal of Lightwave Technology, Vol.11, No.12, Dec.1993, pp.2105-2115.

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

116

[50]

K. T. Koai and R. Olshansky, "Nonregenerative Photonic Dual Bus with Optical Amplifiers," IEEE Photonics Technology Letters, Vol.4, No.4, April 1993, pp.482-485. L. A. Rusch and H. V. Poor, "Effect of Laser Phase Drift on Coherent Optical CDMA," IEEE Journal on Selected Areas in Communications, Vol.13, No.3, April 1993, pp.577-591. L. Bickers, M. H. Reeve, P. A. Rosher, S. C. Fenning, A. J. Cooper, S. G. Muthley and S. Hornung, "The Analog Local Loop: A Growing Revolution in Optical Transmission," IEEE/OSA Journal of Lightwave Technology, Vol.7, No.11, Nov.1989, pp.1819-1823. L. G. Kazovsky, Charles Berry, M. Hickey, C. A. Naronha, Jr. and P. Poggiolini, "WDM Local Area Networks," IEEE LTS Magazine, May 1992, pp.8-15. M. A. Marson, A. Bianco, E. Leonardi and F. Neri, "Topologies for Wavelength Routing All-Optical Networks," IEEE/ACM Transaction on Networking, Vol.1, No.5, Oct.1995, pp.534-546. M. Artiglia, "Optical Fiber Amplifiers: Physical Model and Design Issues," Optical and Quantum Electronics, Vol.26, No.6, June 1996, pp.585-608. M. Fallahi et. al, "Grating Demultiplexer Integrated with MSM Detector Array in InGaAs/AlGaAs/GaAs for WDM," IEEE Photonics Technology Letters, Vol.5, No.7, July 1993, pp.794-797. M. Gerla and L. Fratta, "Tree Structured Fiber Optic MANs," IEEE J. Selected Areas in Communications, Vol.6, No.6, July 1988, pp.934-943. M. H. Lyons, K. D. Jensen and I. Hawker, "Traffic Scenarios for the 21st Century," British Telecom Technology Journal, Vol.11, No.4, Oct.1993, pp.73-84. M. I. Irshid and M. Kavehrad, "Expansion of FDM/WDM Star Networks Using HighPower Erbium-Doped Fiber Amplifiers," Journal of Optical Communications, Vol.14, No.3, 1993, pp.114-119. M. S. Chen, N. R. Dono and R. Ramaswami, "A Media Access Protocol for PacketSwitched Wavelength Division Multiaccess Metropolitan Area Networks," IEEE Journal on Selected Areas in Communications, Vol.8, No.6, August 1990, pp.1048-1057. M. S. Goodman, H. Kobrinski, M. P. Vecchi, R. M. Bulley and J. L. Gimlett, "The LAMBDANET Multiwavelength Network: Architecture, Applications, and Demonstrations," IEEE Journal on Selected Areas in Communications, Vol.8, No.6., August 1990, pp.995-1004.

[51]

[52]

[53]

[54]

[55]

[56]

[57]

[58]

[59]

[60]

[61]

117

[62]

Manabu Yoshino and Kyo Inoue, "Improvement of Saturation Output Power in a Semiconductor Laser Amplifier through Pumping Light Injection," IEEE Photonics Technology Letters, Vol.8, No.1, January 1996, pp.58-59. N. A. Olsson, "Lightwave Systems with Optical Amplifiers," IEEE/OSA Journal of Lightwave Technology, Vol. LT-4, July 1986, pp.864-869. N. -F. Huang and C. -S. Wu, "An Efficient Transmission Scheduling Algorithm for A Wavelength-Reusable Local Lightwave Network," IEEE/OSA Journal of Lightwave Technology, Vo.12, No.7, July 1994, pp.1278-1290. N. Kashima and K. Kikushima, "New Optical Star Bus Network for Subscriber," Journal of Optical Communications, Vol.11, No.2, 1990, pp.42-49. P. E. Green, Fiber Optic Networks, Prentice Hall, 1993. P. J. Duthie, M. J. Wale and I. Bennion, "A New Architecture for Large Integrated Optic Switch Arrays," Proceedings of First Topical Meeting on Photonic Switching, Incline Village, Nevada, March 18-20, 1987, Photonic Switching, Springer-Verlag. P. Mohana Shankar, "Bit Error Rate Degradation due to Modal Noise in Single-Mode Fiberoptic Communication Systems," Journal of Optical Communications, Vol.10, No.1, March 1989, pp.19-23. P. Mohana Shankar, Lloyd C.Bobb and Howard D.Krumboltz, "Coupling of Modes in Bent Biconically Tapered Single-Mode Fibers, " IEEE/OSA Journal of Lightwave Technology, Vol.9, No.7, July 1991, pp.832-837. Q. Jiang and M. Kavehrad, "A Frequency Calibration Method for Multilocation Optical FDM Networks," IEEE Photonics Technology Letters, Vol.5, No.10, Oct.1993, pp.1253-1256. R. Chipalkatti, Z. Zhang and A. S. Acampora, "Protocols for Optical Star-Coupler Network using WDM: Performance and Complexity Study," IEEE Journal on Selected Areas in Communications, Vol.11, No.4, May 1993, pp.579-589. R. E. Blahut, Theory and Practice of Error Control Codes, Addison Wesley, 1983. R. J. Boehm, "SONET: An International Standard," Telecommunications, March 1988. R. Radha Krishna Pillai, Anand Mardi and A. Selvarajan, "Improving Fault Detection and Reliability in FDDI Networks," Computer Communications, Vo.15, No.9, Nov.1992, pp.586-592. R. Ramaswami, Issues in Multi-Wavelength Optical Network Design, IBM Research Report RC15829, 31 May 1990.

[63]

[64]

[65]

[66] [67]

[68]

[69]

[70]

[71]

[72] [73] [74]

[75]

118

[76]

R. Ramaswami and P. A. Humblet, "Amplifier Induced Crosstalk in Multichannel Optical Networks," IEEE/OSA Journal of Lightwave Technology, Vol.8, No.12, Dec.90, pp.1882-1896. R. Ramaswami, "Multiwavelength Lightwave Networks for Computer Communications," IEEE Communication Magazine, Feb.1993, pp.78-88. R. Ramaswami, "Fiber Optic Networks: Future Trends," Journal of IETE, Vol.39, No.2, March-April 1993, pp.59-68. S. C. Liew and K. -W. Cheung, "A Broad-Band Optical Network Based on Hierarchical Multiplexing of Wavelengths and RF Subcarriers," IEEE/OSA Journal of Lightwave Technology, Vol.7, No.11, Nov.1989, pp.1825-1838. S. M. Huang and A. E. Willner, "Guidelines for Optimizing System Performance for 20 WDM Channels Propagating Through a Cascade of EDFAs," IEEE Photonics Technology Letters, Vol.5, No.10, Oct.1993, pp.1190-1193. S. Huang, S. Xie, B. Zhou and Y. Gao, "Crosstalk of WDM Optical Communication Systems using Fabry-Perot Demodulators," Journal of Optical Communications, Vol.15, No.3, 1994, pp.101-103. S. P. Ferguson, "Implications of SONET and SDH," Electronics and Communication Engineering Journal, June 1994, pp.133-142. S. Rao, M. Potts and R. Beeler, "Metropolitan Area Network with Passive Node Structure and Fast Bit Flip Access Mechanism," Proceedings of International Conference On Computer Communication, ICCC90, New Delhi, India, Eds. S. Ramani, H. Shrikumar, S. V. Raghavan, Narosa Publishing House, Delhi, pp.264-268. S. S. Wagner, "Optical Amplifier Applications in Fiber Optic Local Networks," IEEE Transactions on Communications, Vol.35, No.4, April 1987, pp.419-426. S. S. Wagner and H. L. Lemberg, "Technology and System Issues for a WDM-based Fiber Loop Architecture, " IEEE/OSA Journal of Lightwave Technology, Vol.7, No.11, Nov.1989, pp.1759-1768. Th. Martinson, "Oscar and the Very High Speed Optical Loop," F&T Digest, No.2, 1990, pp.9-12. Thomas Ducellier and Marie Bndicte Bibey, "Study of Optical Phase Conjugation in Bulk Travelling Wave Semiconductor Optical Amplifiers," IEEE Photonics Technology Letters, Vol.8, No.4, April 1996, pp.530-532.

[77]

[78]

[79]

[80]

[81]

[82]

[83]

[84]

[85]

[86]

[87]

119

[88]

T. J. Whitley, R. Wyatt, D. Szebesta and S. T. Davey, "Towards a Practical 1.3 micron Optical Fiber Amplifier," British Telecom Technology Journal, Vol.11, No.2, 1993, pp.115-127. T. Toyonaka, S. Tsuji, M. Honeda and Y. Ono, "Polarisation-independent Semiconductor Optical Amplifier Module Using Twin Grin Road Lenses," Electronics Letters, Vol.27, No.11, 23rd May 1991, pp.941-942. V. K. Jain," Effect of Background Noise in Space Optical Communication Systems," International Journal of Electronics and Communication (AE), Vol.47, No.2, 1993, pp.98-107. V. K. Jain, Y. N. Singh and H. M. Gupta, "Effect of Optical Amplifier Induced Crosstalk in Two Channel Non-Coherent OOK Transmission System," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. V. K. Jain, Y. N. Singh and H. M. Gupta, "Power Penalty due to Optical Amplifier Induced Crosstalk in Non-Coherent OOK Transmission Systems," Journal of Optical Communications, Vol.16, No.5, Oct.1995, pp.194-196. Y. Cai, M. Fortenberry and R. S. Tucker, "Demonstration of Photonic Packet-Switched Ring Network with Optically Transparent Nodes," IEEE Photonics Technology Letters, Vol.6, No.9, Sept.1994, pp.1139-1145. Y. K. Chen, "Amplified Distributed Reflective Optical Star Couplers," IEEE Photonics Technology Letters, Vol.4, No.6, June 1992, pp.570-573. Y. M. Lin and D. R. Spears, "Passive Optical Subscriber Loops with Multiaccess," IEEE/OSA Journal of Lightwave Technology, Vol.7, No.11, Nov.1989, pp.1759-1968. Y. N. Singh, V. K. Jain and H. M. Gupta, "Effect of Error Correcting Codes on Laser Linewidth Requirements in Optical Binary Phase Shift Keying Communication Systems," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. Y. N. Singh, V. K. Jain and H. M. Gupta, "WDM Data Network," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. Y. N. Singh, V. K. Jain and H. M. Gupta, "On Placement of Semiconductor Laser Amplifiers in WDM Star Networks," Proc. National Conference on Optical Communications, J.K.Institute of Applied Physics & Technology, University of Allahabad, Allahabad, India, Feb.22-24, 1995, pp.66-78.

[89]

[90]

[91]

[92]

[93]

[94]

[95]

[96]

[97]

[98]

120

[99]

Y. N. Singh, V. K. Jain and H. M. Gupta,"Effect of Reed-Solomon Code on Laser Linewidth Requirements of BPSK Homodyne Optical Communication Systems," J. Optical Communications, Vol.16, No.6, Dec.1995, pp.207-210.

[100] Y. N. Singh, V. K. Jain and H. M. Gupta, "Semiconductor Optical Amplifiers in WDM Star Networks," IEE proceedings - Optoelectronics, Vol.143, No.2, April 1996, pp.144-152. [101]. Y. N. Singh, H. M. Gupta and V. K. Jain, "Semiconductor Optical Amplifiers in WDM Tree-net," IEEE/OSA Journal of Lightwave Technology (under review). [102] Y. N. Singh, V. K. Jain and H. M. Gupta, "Reed-Solomon Code and Semiconductor Optical Preamplifier in OOK Communication System: A Comparative Study," Journal of Optical Communications (accepted for publication). [103] Y. Ofek and M. Sidi, "Design and Analysis of a Hybrid Access Control to An Optical Star using WDM," Journal of Parallel and Distributed Computing, Vol.17, 1993, pp.259-265. [104] Yung-Kuang Chen, S. Chi and Wann-Yih Guo, "High-flexibility Configurations of Amplified Star Coupler for Optical Networks," Journal of Optical Communications, Vol.15, No.2, 1994, pp.56-62.

121

122

Appendix - I
Derivation of Ramp(G)

Let the reflectivities of two facets of SOA be R1 and R2 as shown in Fig.A-1. The Ei is the amplitude of the input electric field and Eo1, Eo2, ...Eop are the output electric field amplitudes after 0, 2, ...2(p-1) reflections. Here, p is an integer. Let Er1, Er2, ...Erp be the reflected electric field amplitudes after 1, 3, ...(2p-1) reflections (see Fig.A-1). The amplitudes of the output electric fields are given by (A-1a)

(A-1b)

and (A-1c)

where is the propagation constant in the gain media, G the single pass gain in cavity and Lg the length of gain media. The total output electric field is given by

(A-2)

123

Fig.A-1

Reflected and transmitted signals in a Fabry-Perot amplifier.

In the above equation, p has been taken as infinity. Normally, it must be an integer part of Lc/(2xLg). The Lc is the coherence length of laser light and is related to laser linewidth as [7]

(A-3)

where Cf is the speed of light in the fiber and the laser linewidth. The coherence length of laser output for a linewidth of 1 MHz and Cf=2x108 m/s in glass fiber will be 63.64 m [7]. Therefore, m would be 31,820 assuming the OA cavity length to be 1 mm. Further, the amplitude of the output field decreases at each reflection. It means higher order reflections will have negligibly small contribution in the output electric field. Therefore, the approximation of extending m to infinity is valid. Equation A-2 can be written as

124

(A-4)

When

G R1R2 <1, the above series converges to

(A-5)

Substitution of Eo1 from Eqn.A-1a gives

(A-6)

The transmittance Tamp is given by

(A-7)

It is seen from the above equation that Tamp will be maximum when 2Lg = 2i where i is an integer. The maximum value of Tamp is referred as Tmax and is given by

125

(A-8)

Like the transmitted electric fields, reflected electric fields are given by (A-9a)

(A-9b)

or (A-9c)

and (A-9d)

Following the same approach as used in the evaluation of the total output electric field, the total reflected electric field is given by

126

(A-10)

This equation is used to determine the reflectance Ramp of the amplifier. As mentioned earlier, Tamp will be maximum when 2L=2i. Under the same condition, reflectance of the amplifier Ramp is given by

(A-11)

127

128

Appendix -II
Statistics of Signal and Echo Beat Noise

In presence of reflection, the information signal and echo signal beat at the photodetector to produce the signal-echo beat noise. The electric fields due to echo and the information signals are given by (B-1a)

and (B-1b)

At the photodetector, eecho and er beat and produce the beat noise current (B-2a)

where (B-2b)

Assuming to be uniformly distributed random variable between - and , its pdf is given by

129

(B-3)

Therefore, the mean and variance of beat noise current, iecho-sig are given by (B-4a)

(B-4b)

130

Biography of Author

The author was born in Delhi in 1969. He obtained B.Tech. in Electrical Engineering with honours from Regional Engineering College, Hamirpur, Himachal Pradesh in July 1991 and M.Tech. in Optoelectronics & Optical Communications from Indian Institute of Technology, Delhi in December 1992. Since then he is pursuing his Ph.D in Electrical Engineering Department, Indian Institute of Technology, Delhi. He is an associate member of Institution of Electronics and Telecommunication Engineers (IETE), India. His academic interests include Optical Computing, Optical Networks, Photonic Switching and Optical Communications. He is also interested in MAC Protocols and Philosophy of Science.

131

132

List of Publications Resulted from the Thesis Research

A. Journals 1. V. K. Jain, Y. N. Singh and H. M. Gupta, "Power Penalty due to Optical Amplifier Induced Crosstalk in Non-Coherent OOK Transmission Systems," Journal of Optical Communications, Vol.16, No.5, Oct.1995, pp.194-196. Y. N. Singh, V. K. Jain and H. M. Gupta,"Effect of Reed-Solomon Code on Laser Linewidth Requirements of BPSK Homodyne Optical Communication Systems," Journal of Optical Communications, Vol.16, No.6, Dec.1995, pp.207-210. Y. N. Singh, V. K. Jain and H. M. Gupta, "Semiconductor Optical Amplifiers in WDM Star Networks," IEE-Proceedings Optoelectronics, Vol.143, No.2, April 1996, pp.144-152. Y. N. Singh, V. K. Jain and H. M. Gupta, "Reed-Solomon Code and Semiconductor Optical Preamplifier in OOK Communication System: A Comparative Study," Journal of Optical Communications (accepted for publication). Y. N. Singh, H. M. Gupta and V. K. Jain, "Semiconductor Optical Amplifiers in WDM Tree-net," IEEE/OSA Journal of Lightwave Technology (under review).

2.

3.

4.

5.

B. Conferences 6. Y. N. Singh, V. K. Jain and H. M. Gupta, "Effect of Error Correcting Codes on Laser Linewidth Requirements in Optical Binary Phase Shift Keying Communication Systems," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. V. K. Jain, Y. N. Singh and H. M. Gupta, "Effect of Optical Amplifier Induced Crosstalk in Two Channel Non-Coherent OOK Transmission System," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. Y. N. Singh, V. K. Jain and H. M. Gupta, "WDM Data Network," Presented in IXth National Convention of Electronics and Telecommunication Engineers, University of Roorkee, Roorkee, India, March 30-31, 1994. Y. N. Singh, V. K. Jain and H. M. Gupta, "On Placement of Semiconductor Laser Amplifiers in WDM Star Networks," Proc. National Conference on Optical Communications, J.K.Institute of Applied Physics & Technology, University of Allahabad, Allahabad, India, Feb.22-24, 1995, pp.66-78.

7.

8.

9.

133