Optimization of emerging extended FTTH WDM/TDM PONs and financial overall assessment

ADVERTIMENT . La consulta d’aquesta tesi queda condicionada a l’acceptació de les següents condicions d'ús: La difusió d’aquesta tesi per mitjà del servei TDX (www.tesisenxarxa.net) ha estat autoritzada pels titulars dels drets de propietat intel·lectual únicament per a usos privats emmarcats en activitats d’investigació i docència. No s’autoritza la seva reproducció amb finalitats de lucre ni la seva difusió i posada a disposició des d’un lloc aliè al servei TDX. No s’autoritza la presentació del seu contingut en una finestra o marc aliè a TDX (framing). Aquesta reserva de drets afecta tant al resum de presentació de la tesi com als seus continguts. En la utilització o cita de parts de la tesi és obligat indicar el nom de la persona autora. ADVERTENCIA. La consulta de esta tesis queda condicionada a la aceptación de las siguientes condiciones de uso: La difusión de esta tesis por medio del servicio TDR (www.tesisenred.net) ha sido autorizada por los titulares de los derechos de propiedad intelectual únicamente para usos privados enmarcados en actividades de investigación y docencia. No se autoriza su reproducción con finalidades de lucro ni su difusión y puesta a disposición desde un sitio ajeno al servicio TDR. No se autoriza la presentación de su contenido en una ventana o marco ajeno a TDR (framing). Esta reserva de derechos afecta tanto al resumen de presentación de la tesis como a sus contenidos. En la utilización o cita de partes de la tesis es obligado indicar el nombre de la persona autora. WARNING. On having consulted this thesis you’re accepting the following use conditions: Spreading this thesis by the TDX (www.tesisenxarxa.net) service has been authorized by the titular of the intellectual property rights only for private uses placed in investigation and teaching activities. Reproduction with lucrative aims is not authorized neither its spreading and availability from a site foreign to the TDX service. Introducing its content in a window or frame foreign to the TDX service is not authorized (framing). This rights affect to the presentation summary of the thesis as well as to its contents. In the using or citation of parts of the thesis it’s obliged to indicate the name of the author Philosophy Doctor Thesis Optical Communications Group Signal Theory and Communications Department OPTIMIZATION OF EMERGING EXTENDED FTTH WDM/TDM PONS AND FINANCIAL OVERALL ASSESSMENT Author Sotiria Chatzi Advisors José A. Lázaro Villa Ioannis Tomkos Thesis presented in fulfilment of doctorate program of the Signal Theory & Communications department Technical University of Catalonia July 2013 The work described in this thesis was performed at the Signal Theory and Communications department of the Universitat Politècnica de Catalunya / BarcelonaTech. It was supported by the European Comission through the FP7 Projects SARDANA, EURO-FOS and BONE. Sotiria Chatzi Optimization of Emerging Extended FTTH WDM/TDM PONs and Financial Overall Assessment Subject Headings: Optical fiber communication Copyright © 2013 by Sotiria Chatzi All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means without the prior written consent of the author. Printed in Barcelona, Spain. ISBN: Reg: B. To my grandfather, who taught me how to read, write and play cards, before I was 4 Abstract Fiber-To-The-Home (FTTH) has experienced a boost in the last years which came as a consequence to the continuous technological evolution in hardware as well as in software, which in their turn have drove the evolution of internet applications. Numerous optical fiber network implementations have been realized worldwide. Several architectures and technologies have been tested in real life scenarios and many more have been designed in university labs and telecommunication companies. The point has now been reached, where one refers to nextgeneration networks. At the end of the day the most critical factor in optical networks, is their cost, their effectiveness and the counterbalance among the two elements. A thorough inquiry has been performed on both aspects of Fiber-To-The-x (FTTx) and optical networks in general. Topics on their performance in combination with their cost and financial assessment have been covered for many different infrastructures and technologies, focusing mainly on the Passive Optical Networks (PONs). Hybrid Wavelength Division Multiplexing / Time Division Multiplexing PONs, have been studied scientifically so as to expand their limits. This has been realized by applying elaborate design, in–line remote amplification, use of alternative as well as extended band for the signal transmission, dispersion compensation and signal equalization and with the help of numerous studies. v Table of Contents Chapter I – Introduction 1.1 Motivation.......................................................................................................13 1.2 Objectives........................................................................................................16 1.3 Thesis overview...............................................................................................17 Chapter II – State of the art 2.1 Optical networks worldwide............................................................................19 2.2 Techno -economics..........................................................................................20 2.3 WDM/TDM PON............................................................................................24 Chapter III – Techno-economics 3.1 Drivers and requirements behind FTTH research...........................................29 3.2 Motivation of this study and tools...................................................................31 3.3 FTTx architectures and technologies used in FTTH.......................................32 3.3.1 Time division Multiplexing (TDM) PONs..............................................33 3.3.2 GPON......................................................................................................34 3.3.3 Hybrid Wavelength Division Multiplexing / Time Division Multiplexing (TDM) PON.............................................................................................34 3.3.4 Ethernet Point to Point (EP2P)................................................................35 3.3.5 Active Star...............................................................................................35 3.4 FTTH infrastructure cost considerations – components..................................36 3.4.1 Outside plant............................................................................................36 3.4.2 Methodology of calculations (OSP)........................................................38 3.4.3 Methodology of calculations (Active equipment)...................................42 3.3.4 Results.....................................................................................................43 ix Table of Contents x Chapter IV – SARDANA and challenges of long reach and enhanced PONs 4.1 SARDANA......................................................................................................47 4.2 Main limitations in SARDANA and long reach-enhanced PONs...................50 4.3 Solutions on the main limitattions of long reach-enhanced PONs..................51 4.3.1 Amplification...........................................................................................51 4.3.2 Compensation techniques........................................................................52 4.3.2.1 Dispersion compensation fibers..............................................................52 4.3.2.2 Equalization.............................................................................................52 4.3.2.3 Reduced OLT power...............................................................................57 Chapter V – In-line amplification 5.1 Erbium Doped Fiber Amplifiers basics...........................................................58 5.1.1 Basics.......................................................................................................58 5.1.2 Gain wavelength dependency..................................................................64 5.2 Theoretical approach of in-line amplification.................................................68 5.2.1 General considerations............................................................................68 5.2.2 Definitions and parameters used for the analysis....................................70 5.2.2.1 Input test signal........................................................................................70 5.2.2.2 Signal gain...............................................................................................70 5.2.2.3 Giles parameters......................................................................................70 5.2.2.4 Excess noise power..................................................................................71 5.2.2.5 Optical signal to noise ratio.....................................................................71 5.2.2.6 Noise figure.............................................................................................72 5.2.2.7 Type of EDF............................................................................................73 5.2.3 Analysis of the expected OSNR degradation..........................................74 5.2.3.1 Initial OSNR degradation after the 1st amplification stage.....................74 5.2.3.1.1 Downstream OSNR degradation after 1st amplification stage................75 5.2.3.1.2 Upstream OSNR degradation after 1st amplification stage....................75 5.2.3.2 OSNR degradation in a cascade of in-line amplifiers.............................76 5.2.3.2.1 Downstream OSNR degradation through cascade of in-line amplifiers. 76 5.2.3.2.2 Upstream OSNR degradation through a cascade of in-line amplifiers...77 5.3 Remote amplification approaches and design characteristics.........................77 5.3.1 Network design approach........................................................................78 5.3.2 RN design approach................................................................................81 5.3.3 Design emulation set up and parameters.................................................82 5.3.3.1 Downstream scenario and set up.............................................................84 5.3.3.2 Upstream scenario and set up..................................................................86 5.4 Evaluation of the design solution....................................................................88 5.4.1 Conclusion...............................................................................................91 xi Table of Contents Chapter VI – Alternative band operation of a WDM/TDM PON 6.1 L-band EDFA..................................................................................................94 6.2 L-band fibers...................................................................................................95 6.3 L-band amplification technologies..................................................................96 6.4 In-line and L-band operation of a WDM/TDM PON......................................99 6.4.1 Parameters of the network affecting its operation...................................99 6.4.2 Set-up.....................................................................................................102 6.4.3 Results...................................................................................................103 6.5 Conclusions...................................................................................................104 Chapter VII – Extended band operation of a WDM/TDM PON 7.1 Raman amplification.....................................................................................106 7.2 C + L band gain equalization-Hybrid Raman&in-line amplification............106 7.2.1 System description.................................................................................107 7.2.2 Method used – Results...........................................................................108 7.3 Dual waveband RN for extended reach full duplex 10Gb/s C+L PON........109 7.3.1 Remote node design..............................................................................110 7.3.2 Remote amplification in an extended reach PON.................................114 7.3.3 Conclusion ............................................................................................117 Chapter VIII – Dispersion compensation in optical networks 8.1 Dispersion, source induced chirp and fundamentals of equalization............118 8.2 Dispersion compensating fiber......................................................................119 8.3 Electronic equalizer.......................................................................................120 8.4 Experimental study........................................................................................122 8.4.1 Experimental set-up...............................................................................124 8.4.2 Results...................................................................................................125 8.4.3 Conclusions...........................................................................................130 Table of Contents xii Chapter IX – Combination of in-line amplification / extended band use / equalization use for WDM/TDM PONs 9.1 Technical description.....................................................................................132 9.2 Experimental set-up.......................................................................................134 9.3 Results...........................................................................................................136 9.4 Conclusions...................................................................................................140 Chapter X – Conclusions 10.1 Conclusions...................................................................................................141 10.2 Experimental set-up.......................................................................................144 A. List of Acronyms.........................................................................................I B. Research Publications..........................................................................V C. Bibliography..............................................................................................VIII Biography...........................................................................................................XVI 19 Chapter II State of the art This chapter covers the present situation regarding the subjects dealt within this thesis. As the main research subject of the study is FTTH, initially, the penetration of this technology in Europe and worldwide is presented. Following, aspects of the technoeconomic studies made on the FTTx infrastructures and the corresponding technologies are exposed. Next, an overview of the current state of research on WDM/TDM PONs is given. 2.1 Optical networks worldwide FTTx is spreading very swiftly in all continents, in several forms, with the use of different topologies and different technologies. According to the latest registration (2011) of FTTH Council, more than 75 million users worldwide are connected with fiber optic connections to the home [2]. From those connections, 60 million are used in Asia, 10 million in United States of America (USA), 5.5 million in Europe and 0.5 million in United Arabic Emirates (UAE). Data released from the FTTH Council, rank the countries with the highest penetration worldwide as can be seen in Fig.2.1. In terms of infrastructure, as far as Europe is concerned, there are approximately 28 million homes passed and an average take up rate of 18.4% [2]. Russia though, according to the latest reports [3], has added 2.2 million new FTTH subscribers in the second half of 2012 to reach a grand total of 7.5 million fiber-connected homes. Moreover, French government has announced that it will invest 20 billion Euros into ultra-fast broadband infrastructure, with a special focus on future-proof fiber, in the coming years [4]. Chapter II. State of the art 20 2.2 Techno-economics During recent years, an increasing number of research papers, besides the consultancies’ reports, have been composed within national and international collaborative projects, aiming to contribute to this broadband debate, regarding the cost and the efficiency of each FTTx solution, which has started as early as in the late 80’s. Models have been developed, for the estimation of the several costs involved in the implementation of an optical network. In fact, there are published guidelines and instructions for creating a technoeconomic model [5]. European telecom operator combined these guidelines with a thorough and detailed financial evaluation of several scenarios of deployment, based though, only on current technologies, not covering future ones, such as WDM/TDM PONs [5]. Furthermore, automated network planning tools have been implemented, which helped in techno-economic research [6]. But, these tools have been used for the comparison of FTTx and FTTH architectures that both use commercially available technologies [6]. Since the subject of the infrastructure cost has been a matter of great concern for most of the local authorities in technologically evolved countries, studies have been performed, that concern each country/area locally and individually. Therefore, most of the investigations are focused on specific regional needs. For example, there are deployment studies that focus on the geographical characteristics of a particular area, Fig. 2.1 Market Penetration of FTTH - Ranked by Country (Press conference FTTH Council Europe 15/2/2012) 21 Chapter II. State of the art while there are others that focus on the preferred technological choices of the country’s strongest telecom market players. Therefore, there are studies comparing FTTH deployments with Fiber-To-TheCurb/Very-high-bit rate digital subscriber line (FTTC/VDSL) as in [6]. The network design (geometric) models used to calculate the necessary equipment to deploy the several different infrastructures are based on mathematic norms, and algorithms. These models, that use in addition correction factors [7], give an approximation of the required optical fiber length and consequently of the total material needed for the implementation of the network. Three examples of these models appear in Fig 2.2 and Fig 2.3. a b a b Fig. 2.2 Examples of geometric models used for techno-economic calculations of: a) GPON, b) P2P, fiber distribution [6] Chapter II. State of the art 22 A subject that concerned researchers, is the cost of transition from an FTTx infrastructure to FTTH, or to another version of FTTx [8], but, once more, only commercialized technologies such as GPON and Active Optical Networks (AONs) have been examined. Likewise, disputes have arisen regarding the prevalence of PONs and AONs [8]. The study in [8] focuses, as can be seen in Fig.2.4, on the two FTTH-AON architectures, namely Active star and Home run. As far as PONs are concerned WDM and TDM technologies were investigated. Once more, the hybrid solution of WDM/TDM PON was not included in the study. Fig. 2.3 Geometric model for techno-economic evaluation of FTTC/VDSL and FTTH Roll-Out [7] 23 Chapter II. State of the art The number of the techno-economic evaluations seen in publications is as high as the number of possible combinations of the FTTx infrastructures with all the technologies used for an optical access network. There is big literature regarding the assessment of GPON in contrast with Ethernet Point to Point (EP2P) as well as Active star deployments [9]. One of these comparative studies weighs WDM PONs against TDM PONs [10], giving some interesting results regarding the required infrastructure and the cost difference between Greenfield and Brownfield scenarios, yet again without presenting any actual network technology, but on the contrary, basing the results on a theoretical abstract design. As technology progresses, research focuses in new optical network implementations. For example, we have seen the comparison of two similar hybrid WDM/TDM PONs appearing in [11]. The study has taken into consideration two flavours of WDM/TDM PONs, whose main difference is localized in the ONU. The results of this techno-economic comparison have been extracted for two scenarios, one of an urban and one of a rural area correspondingly and can be seen on Fig 2.5. The comparison has been performed for two similar PONs, not taking into consideration older technologies and how the WDM/TDM PON can offer a financially distinctive choice when compared with existing and deployed technologies. Fig. 2.4 a)AON homerun, b)AON active star, c) TDM PON, d) WDM PON. Comparison of Active star and home run AONs with TDM and WDM PONs [8] Chapter II. State of the art 24 As can be observed, scientific research has been focused on the composition of different elements of this multidimensional study, aiming to derive with the most effective FTTx/FTTH architecture and technology. In this thesis, several technoeconomic comparisons among various existing, emerging and future technologies and infrastructures are presented, in an effort to combine all of these elements. The reason for this was the need to explore the entire cost of different optical access networks and estimate whether the new technologies provide indeed a significant cost reduction. The hypothesis was proven correct, therefore, the estimation of the cost difference among the several infrastructures, followed. Hybrid WDM/TDM PON was found to be the most cost effective solution, nevertheless a complete study on current – commercialized architectures and technologies has been realised as well, in order to have a definite and integral result. This realisation on the cost-effectiveness of the WDM/TDM PON led us to study the ways this technology could be expanded compared to the existing implementations. 2.3 WDM/TDM PON Quite a few studies on WDM/TDM PONs have been performed and are at the moment taking place by researchers worldwide. Research projects currently proceeding or recently completed are presented below. The project PIEMAN (Photonic Integrated Extended Metro and Access Network) [12] is a 6th Framework Program (FP) Specific Targeted Research Project (STReP) that focussed on the development of colourless electro-optical devices of the ONT for WDM/TDM PONs, with bandwidth up to 10 Gbit/s downstream and upstream. This technology can thus be available also at this level. Since this project is mostly a network element - development initiative, it does not focus on network design, therefore does not contribute to network design and optimization. Fig. 2.5 Cost breakdown for the cost per subscriber for different hybrid WDM/TDM PON technologies [11] 25 Chapter II. State of the art MUSE (Multi Service Access Everywhere) [13] is a 6FP Information Society Technologies project whose scope is multi-layered on several wired access techniques. Optical access technologies under study here, are a coarse wavelength division multiplexing (CWDM) resilient ring system and a hybrid fibre radio solution, suited to feed Worldwide Interoperability for Microwave Access (WiMAX) base stations. The subject of Dense Wavelength Division Multiplexing (DWDM) has not been addressed, while the focus is mostly on the end-to-end inter-domain compatibility using Internet Protocol (IP) Ethernet technologies and the Media Access Control (MAC) protocol that was developed as an evolution of the Full Service Access Network (FSAN) GPON system. The project HARMONICS (Hybrid Access Re-configurable Multi-wavelength Optical Networks for IP-based Communication Services) [14] is a hybrid access network, combining an optical feeder network with multiple access technologies. It is a WDM PON and the studies performed in the framework of this project were mainly metro and MAC oriented. HARMONICS used electric switching and fixed lasers and did not reach to the end-user, therefore it consists of a service-metropolitan area network (MAN) solution. It integrated different traffics and proposed the use of a combined WDM/TDM solution, implemented in the MAN ring, with FTTC possible extension. Therefore, the point of FTTH WDM/TDM PON that is the main research subject of this work, has not been raised. Project SUCCESS-HPON (Stanford University aCCESS Hybrid WDM/TDM Passive Optical Network - A Next Generation Optical Access Architecture for Smooth Migration from TDM-PON to WDM-PON) (Stanford University) [15], in Fig 2.6 ,proposes complex ONUs with tuneable sources and active RNs. It is a general architecture of next generation hybrid WDM/TDM optical access architecture. SUCCESS-HPON is well developed on the MAC domain but can present limited physical layer experiments. Chapter II. State of the art 26 Other projects carried out in the past, like PRISMA, TOBASCO, PLANET, SUPERPON, etc, also used hybrid transmission media and/or active devices (like switches, O/E converters, optical amplifiers) along the PON. The main research line followed by these projects is not in accordance to the study represented in this thesis, as the network examined herein is a fully passive solution. ePhoton/One IST Network of Excellence [16], included specific joint activities in topics like next-generation PONs, impairment monitoring and electronic compensation. It has focused of CWDM in contradiction with the studies represented here, who have been focusing on DWDM technology. As far as single research institutes, or collaborations among them or among industry and academic institutions is concerned, there is a number of efforts taking or have been taken place worldwide. Most of them are presented below. Shangai University has focused on the development of a hybrid WDM/TDM PON. Their efforts are concentrated towards the development of a new WDM/TDM PON design based on a dual fiber ring with access-tree topology, which utilizes a reconfigurable optical add-drop multiplexer based on wavelength blocker technology. Furthermore, they have focused on the design of a self seeding Fabry-Perot (FP) fiber laser at the ONU [17]. Huazhong University of science and technology in China has focused on the development of a long reach WDM/TDM PON. This network is based on the use of RSOAs as well as the use of optical add drop multiplexers (OADMs) [18]. Korea Telecom in co-operation with Chungnam National University have focused their Fig. 2.6. SUCCESS-HPON: a) overall architecture, b) OLT block diagram, and c) WDM-PON ONU block diagram [15] 27 Chapter II. State of the art efforts towards the development of a new WDM/TDM PON design based on a dual fiber ring with access-tree topology, which utilizes a reconfigurable optical add drop multiplexer. The design has used RSOAs and remotely pumped EDFAs in combination with 32 WDM channels and 16 TDM channels per wavelength [19, 20]. KTH in co-operation with Ericsson and Zhejiang University researched the transition from TDM to WDM/TDM PON [21]. They have suggested a protection scheme that would serve as the intermediate step for the smooth migration from TDM PON to WDM/TDM PON. The feasibility of such a project in terms of cost has been investigated as well [22]. Korea Advanced Institute of Science and Technology (KAIST) with Samsung electronics have researched a WDM/TDM PON that can serve up to 128 users. This is achieved by cascading 1X16 arrayed waveguide grating (AWG) and 1X8 splitters. The use of a single FP-Laser Diode (LD) has been considered, with an amplified spontaneous emission (ASE) injection. Multipath transmission mechanism from optical line terminal (OLT) in optical wireless converged network architecture has also been investigated [23]. Korea and Bangladesh have introduced a similar design in the form of a self restored tree type WDM/TDM PON. The network is implemented with the use -in this case alsoof a 1X16 AWG in the OLT and 1X4 Band Splitting Wavelength Division Multiplexing (BSWDM) filters both in the OLT as well as in the first stage RNs. In the second stage RN, a pair of 1:16 passive splitter is used so that TDM-PON de-multiplexes and multiplexes signals at channel level respectively [24]. COBRA institute, Eidhoven and Genexis have created a WDM/TDM PON network entitled: “Dynamically reconfigurable Broadband Photonics (BBPhotonics) network”. The configuration is served by an OLT that transmits 8 data-and-continuous wavelengths (CW) wavelength pairs. The CW are transmitted by the OLT for use in the upstream modulation. The CO, where the OLT lies, is connected through a Standard Single Mode Fiber (SSMF) ring with 4RNs. Each RN with the use of a reconfigurable OADM serves 16 reflective optical network units (ONUs). In this framework experiments over 27km were performed, demonstrating simultaneous up and downstream traffic at 10Gb/s per channel. [25]. NTT Access Network Service Systems laboratories in Japan have performed a study on the existing WDM/TDM PON technologies and suggest two different solutions for the way the future research could evolve. NTT’s interest in present and future results of academic research in the WDM/TDM PON technology proclaim an interest on behalf of the operator to use such a technology for future network implementations [26]. They suggest a WDM/TDM system based on their developed technologies. Furthermore, studies have been performed for the use of a dynamic wavelength and bandwidth allocation (DBWA) algorithm so as to use, as efficiently as possible, the available bandwidth resources and to tune them in correspondence with the users’ requests. The DWBA operation has been tested with 40Gb/s tunable laser. ETRI in Korea along with National ICT of Australia and the University of Melbourne suggest a multiwavelength PON (MW) as a useful antecedent of Chapter II. State of the art 28 WDM/TDM PON which will allow a smooth migration from the existing systems to the aforementioned future one [27] The MW PON consists of a remote AWG used as a channel multiplex, a remote splitter and a group of circulators in the RN and small splitters near the ONUs. The seed light is composed of multiple wavelengths rather than a single one. A RSOA is used as a colorless upstream transmitter and a low-cost channel filter is integrated into the RSOA-based transmitter the receiver of the MW PON. ETRI in co-operation with Korea University of science and technology have suggested a protection structure for WDM/TDM PONs. Moreover they have suggested a WDM/TDM PON that operates with tunable laser diodes and considers RSOAs for uplink transmitters [28]. Finally ETRI in cooperation with Korea Telecom have deployed the first commercial WDM//TDM PON in which one feeder fiber can support 512users. The system uses RSOAs in the ONUs and 16X1.25 Gb/s colorless gigabit WDM/PON and a legacy TDM/PON with optical-electrical-optical converters [29]. KDDI R&D Laboratories in Japan propose a ring based WDM/TDM PON implemented with the use of an injection-locked Fabry-Perot LD and a wavelength converter [30]. The institute for Infocomm research in Singapore has proposed a WDM/TDM PON with dynamic virtual PON capabilities, 10Gb/s downstream and 1.25Gb/s upstream. Furthermore, they have suggested the use of DPSK in the downlink and OOK in the uplink operating in the same bandwidth as before [31]. NUST, National University of Science & Technology in Pakistan proposes a WDM/TDM PON architecture based on EPON technology [32]. Corning has performed an experimental study on a 11.1Gb/s WDM/TDM system that supports 8 channels in 100km and 1:128 wavelength split ratio, with the use of EDFA in the experimental set up [33]. As far as WDM/TDM PON research on the MAC layer is concerned, National Taipei University of Technology in Taiwan has studied a delay sensitive multicast mechanism for downstream traffic in WDM/TDM PON [34]. University of California has developed an algorithm for efficient bandwidth usage, using a behaviour-aware user-assignment approach, based on the daily usage profiles [35]. Obviously, there is a large interest worldwide about this new technology and the support it is receiving, is an index that it will be the new technology to be used in FTTx and more specifically FTTH deployments. This, as well as the cost reduction we have estimated that the new technology will bring along, was the driver for the research on the expansion of the limits of a WDM/TDM PON as mentioned already. 35 Chapter III. Techno-economics The WDM/TDM PON investigated in this study is based on the Scalable Advanced Ring-based Dense Architecture (SARDANA) [51]. SARDANA is engineered to incorporate functionalities of metropolitan networks and furthermore its size and limits make it an access-metro convergence solution. It is an architecture that aims at enhancing the performance of PONs, by implementing a scalable access WDM/TDM PON. It is based on a WDM ring for the transport of downstream and upstream information and TDM trees. In the CO a stack of lasers is used to serve with different λs the different tree network segments on a TDM basis. Placed on the ring, passive RN perform optical add and drop functions. With the use of filters, two wavelengths drop at each RN. Each of the two dropped wavelengths is driven to independent TDM PONs. It is a flexible network, whose characteristics in terms of maximum distance, number of wavelengths and splitting ratio can change according to the needs of the area that has to be served. Nevertheless its design characteristics and attributes are thoroughly described in the next chapter of the thesis. In this chapter, since we examine an urban deployment scenario, we consider 32 wavelengths, 1:64 splitting ratio, 20km ring length, feeder length –starting at the RN and ending at the splitterup to 2.9 km and distribution along with drop length equal to 0.1km. The number of end users per ring is therefore set to 2048. 3.3.4 Ethernet Point to Point (EP2P) In a point to point network each user is connected by a dedicated fiber to the CO. A router acts as central element, while the access switches transport the IP packets to the end user. The customer premises equipment (CPE) translates the IP traffic into application signals like Internet or telephony. As the network is scaled to more users, the routing function is distributed over an array of interconnected routers. Eventually, the routing function can be partitioned into two levels: a lower level with distribution routers and a higher level with interconnection routers. The distribution routers distribute the IP traffic to the various access switches, while the interconnection routers enable the interconnection to the service providers [52].The distribution of IP traffic via the Ethernet/IP network is based on standards. In this study the Fast Ethernet standard is considered. Fast Ethernet defines a version of Ethernet with a nominal data rate of 100 Mb/s. 100 Megabit Ethernet100BASE-BX [53] is a version of Fast Ethernet over a single strand of optical fiber. Single-mode fiber is used, along with a special multiplexer which splits the signal into transmit and receive wavelengths. Distances from the switch can be 10, 20 or 40 km. In this work we have considered 20km. 3.3.5 Active Star A Star architecture (also known as a Double Star) is an attempt to reduce the total Chapter III. Techno-economics 36 amount of fiber deployed and hence lower costs by introducing feeder fiber sharing. In a star architecture, a remote node is deployed between the CO and the subscriber’s premises. Each OLT port and the feeder fiber between the CO and the remote node is shared by anywhere from four to a thousand homes (the split ratio) via dedicated distribution links from the remote node. When the remote node contains active devices such as a multiplexer (or Ethernet switch), the architecture is referred to as an Active Star as the remote node needs to be powered. The Remote Node in the Active Star network has a multiplexer / demultiplexer. The remote node switches the signal in the electrical domain (to the intended recipient) and hence OEO conversions are necessary at the remote node. This central device, routes all messages between devices. Since the feeder bandwidth is shared among multiple end points, the maximum sustained capacity available to each home – both upstream and downstream – is less with an active star architecture than with Home Run fiber [38]. 3.4 FTTH infrastructure cost considerations – components 3.4.1 Outside plant One of the most important cost factors in the deployment of an optical network is the outside plant equipment. For the calculation of capital expenditure (CAPEX) per user, one must consider the actual deployment methods of a fiber network and the components needed to create the infrastructure. To describe an FTTH infrastructure network, it is essential to describe some of the basic equipment [54]. An FTTH network normally forms part of an existing access network, connecting a large number of end users back to a central point, the CO, or access node. Each CO contains the required active transmission equipment used to provide the applications and services over optical fiber to the subscriber. Explaining outwards from the CO towards the subscriber, the key FTTH Infrastructure Elements needed are: Central Office or Access Node Feeder Cabling Primary Fiber Concentration points (PCP) Distribution Cabling Secondary Fiber Concentration points (SCP) Drop Cabling Internal Cabling (in-house cabling) The most conventional method to install underground cable, involves the creation of a duct network to enable subsequent installation of cables by pulling, blowing or using floatation techniques. This network comprises of a combination of large main ducts that contain smaller subducts for individual cable installation Fig 3.2 b), (or in other cases large main ducts into which, cables are progressively pulled one 37 Chapter III. Techno-economics over the other as the network grows Fig 3.2 a)) and furthermore small sub-ducts for the installation of a single cable for the drop part of the network. The drop cables will contain only 1 or 2 fibers for the connecting circuitry and possibly additional fibers for backup. The drop cables are installed in special pavement trenches. Summarizing the installation method, for underground networks the drop cabling may be deployed within small ducts, within micro-ducts or by direct burial to achieve a single dig and installation solution. In this study we consider the use of microducts inside trenches. Additionally, we consider the blowing technique for installation of cables in the combined network comprising of ducts/subducts and we do not take into account in CAPEX calculations the inhouse cabling for it can take many forms or be absent whatsoever. Below we mention the necessary main elements of the network taken into account for the estimation of the total CAPEX. Necessary main elements: Trenches Ducts and subducts Fiber cables Branching boxes Splitters Handholes – manholes Y-branches Feeder cables are large size optical cables, with a wide number of optical fibers. The number of feeder fibers in point to point active deployments equals the number of end users (total number of households in FTTH). In PONs it equals the total number of users divided by the splitting ratio. Thus the use of passive fiber splitting devices Fig. 3.2. a) section of duct with fiber cables, b) section of duct with subducts and fiber cables (www.emtelle.com) Chapter III. Techno-economics 38 positioned as close as possible to the end user, enables the use of smaller fiber count cables for the feeder part of the network, providing reduction of the costs. The distribution fiber cables are used between the splitter or the branching boxes and the Y branches. These are usually cables consisting of a low number of fibers. This number depends on the deployment topology (usually based on the topology and density of the premises deployment in an area, for instance the distribution of streets in a rural location). Distribution cables are smaller in number of fibers, therefore in size, than the feeder cables in standardized PONs, AONs and P2P deployment. Drop cables lay between the drop terminals (Y-branches in this case) and ONUs. These are protected single or two fibers armored cables, one for each ONU. Each cable is terminated with a special protected connector that withstands environmental conditions. In Fig 3.3 feeder, distribution as well as drop cables are depicted. 3.4.2 Methodology of calculations (OSP) For the calculations we have used parameters for typical deployment areas with different densities. The number of buildings, namely multi-dwelling units (MDUs) and single-dwelling units (SDUs), as well as the number of HH units are given or calculated. Then the mean distance among buildings is estimated which is referred to as Inter Building Spacing (IBS) [55]. IBS obviously depends on the HH/buildings density. A Ybranch is being used to separate the individual drop cables from the distribution ones. The splitters of PONs as well as the branching boxes of EP2P are placed inside handholes/manholes every several buildings. The switches, are placed inside cabinets in chosen crossroads. Those are the points where we have the transition from distribution to feeder fibers as seen in Fig 3.4 and Fig 3.5. Fig. 3.3. a) feeder cables, b) distribution cables, c) drop cables (www.emtelle.com) 39 Chapter III. Techno-economics The feeder fibers are launched from the OLTs in the COs which are housed in the points of presence (POPs). The CO is placed in the POP of the network which is a building that houses all active transmission equipment. There, all fiber terminations are managed and the interconnection between the optical fibers and the active equipment is facilitated. The main network cables entering the node terminate and run to the active equipment. The feeder cables also connect to the active equipment and run out of the building and into the FTTH network area. Inside the CO, separate cabinets and termination shelves may be considered for equipment and individual fiber management to simplify fiber circuit maintenance as well as avoid accidental interference to sensitive fiber circuits. In the cases where the metro network is also considered (results, Fig 3.8.-3.11.) we consider that the metro network is deployed in special trenches with low fiber count cables. For the cases of GPON and WDM/TDM PON we consider that each CO/POP will serve up to 100.000 users, while for the case of EP2P this number is appointed to 20.000 due to fiber handling constraints at the CO. By dividing the total number of the users with 100.000 we reach to the final number of COs needed [56, 57]. For the calculation of OSP CAPEX per user, we consider: The total length of the different types of trenches needed The total length of the different types of ducts and sub-ducts needed The total length of the different types of fiber cables needed The total number of hand-holes / manholes & branch off closures needed The total number of cabinets needed The total number of y-branches needed Fig. 3.4. POPs, feeder cabling, distribution cabling, drop cabling and ducts with subducts and fiber cables (www.emtelle.com) Chapter III. Techno-economics 40 Table 3.1. OUTSIDE PLANT MATERIAL & INSTALLATION COST Description Unit $ HDPE duct – (24 micro tubes) m 2.25 HDPE duct – (7 micro tubes) m 2.2 HDPE duct – (2 micro tubes) m 0.75 Manhole each 500 Handhole each 400 96 - fiber cable m 2.8 72 - fiber cable m 2.1 12 - fiber cable m 1.1 8 - fiber cable m 1.1 Microcable1f m 0.3 Y-Branch unit each 35 Trench I m 25 Trench II m 20 Microtrench I m 16 Microtrench II m 14 Pavement trench m 35 HDPE duct in trench m 0.55 Cable in subduct m 0.45 Splicing each 5 The total number of the splitters needed The total cost all the aforementioned components The total cost for the installation of the several components To evaluate the actual cost benefits, we performed a detailed techno-economic calculation. For the estimation of the OSP CAPEX per user, for each FTTx network architecture, we have considered the actual deployment methods of a fiber network and the components needed to create the infrastructure as described in Outside Plant Equipment section. To estimate the cost of the FTTN, FTTC and FTTB we assume that the FTTx network forms part of an existing access network, connecting a large number of end users to the CO. On the other hand for calculating the full CAPEX cost of the three technologies (EP2P, GPON, and WDM/TDM PON) we consider metro-access networks where EP2P and GPON form the access part, while on the WDM/TDM PON the metro with the access parts are converged. As far as the FTTH infrastructure is concerned, the cost of the OSP elements considered are namely: the metro and feeder cabling, the PCP, the distribution cabling, the SCP and the drop cabling [54]. The handholes or manholes that house the splitters, the branching boxes and the cabinets that contain the switches are considered as the PCP and the SCP are the Y-branches that help to disjoin the drop cables. Regarding the FTTx infrastructure, only the cost of the fiber network deployment is considered along of course with the necessary OSP additional material. Finally the cables installation, for the needs of the techno-economic study, has been considered to take place with the blowing technique which is one of the most cost effective ones. The cost estimations were based on individual components and 41 Chapter III. Techno-economics civil work costs, as shown in Table 3.1., and were provided by a construction and a telecom company. The potential deployment area considered is a regular rectangular shaped typical residential area with a mixture of MDUs and SDUs and with a varying density of HH per km2. The deployment considered, is based on the geographic model which can be seen in Fig. 3.5 and which has been used in several other studies [55-60]. The feeder part of the network is deployed in big trenches (trench I) with a capacity of 8 subducts, each containing 7 high fiber count cables, therefore 56 high-fiber-count cables, inside sub-ducts for easy installation and replacement. The trench II is a trench with capacity of 6 subducts each containing 24 low-fiber-count cables, which have not been used in this specific deployment. The use of this type of trench would be useful in scarcely populated areas and the model can be very easily modified and adapted to the use of trench II. Nevertheless, the use of this trench is not suggested in densely populated areas, where a larger number of users (subsequently – fibers) should be served by each trench. The distribution part of the network is parceled according to the density of the area into smaller ducts (microtrench I, II) varying from 96 low-fiber-count cables (microtrench I ) capacity to 48 low-fiber-count cables (microtrench II). These are used between the splitter and the Y-branches. The drop part is implemented within a pavement trench with higher cost due to the high restoration cost, by microducts containing 1, 2, 8 or 12 fiber cables. For the estimation of CAPEX per user and as far as the OSP is concerned, we consider the total length of the different types of trenches, ducts, subducts and fiber cables, the total number of handholes/manholes, splitters and Y-branches needed, their total cost and the cost of their installation. drop cable distribution cable feeder cable drop cable distribution cable feeder cable drop cable distribution cable feeder cable Fig. 3.5. POPs, feeder cabling, distribution cabling, drop cabling , y-branch closures and cabinets (www.emtelle.com) Chapter III. Techno-economics 42 A B IBS a PCP SCP CO A B IBSIBS a SCP CO A B IBSIBS a PCP SCP CO A B IBSIBS a SCP CO Fig. 3.6. Geographic model used for the calculation of the outside plant cost [60] The IBS, thus the mean distance among buildings, appearing in Fig 3.6 is calculated by considering the area under investigation rectangular with a length α. || 2^ buildings a IBS  Eq. 3.1 The PCPs are placed in crossroads. The choice of the crossroads depends on the building density and it is made on such a way that every street is served by a single duct. What appears as a double line in each street are the fibers serving the building, nevertheless all the cables are placed in a duct placed on one side of the street. Vertical lines are considered to serve the buildings on the opposite side. The distance among the COs (A+B) is set to two times the maximum reach of each technology. max2reach Eq. 3.2 3.4.3 Methodology of calculations (Active equipment) The active equipment cost calculation includes the cost of the OLTs and their individual parts, the cost of the ONUs and their parts, as well as the cost of the routers and switches placed in the field for EP2P. For the case of WDM/TDM PON we include also the cost of the RNs –even though they don’t include any active elementswhich are new components introduced in this novel architecture. Every RN cost is shared by two PON trees, since two wavelengths are dropped/added in each one of them. The use of a single wavelength per tree is exploited by the reflective semiconductor optical amplifier in the 43 Chapter III. Techno-economics ONU, which reflects, amplifies and remodulates the signal, offering a low cost solution. Since the prices for the new components are not available, estimation was made, based on existing prices for the individual parts and the overall costs were extrapolated in time. 3.4.4 Results The application of the techno-economic model/methodology for the various cases of FTTx architectures revealed the results presented in Fig 3.7 in terms of the OSP costs. In the case of FTTx, we do not consider that the same bandwidth is delivered to each end user. Further more, as mentioned already, only the estimation of the fiber infrastructure cost is taking place. The copper cables are considered to be already deployed (brownfield deployment); therefore their cost is not included in the calculations. It is shown that there is a significant cost difference among the different FTTx architectures which is due to the difference on the distance among the end point of the infrastructure and the customer. FTTN appears to be the cheapest solution, while FTTB the most costly one. This explains the fact that the incumbent operators are mainly deploying xDSL since they own the already deployed copper part of the network. Of course in FTTx cases examined, the cost savings are translated in degradation of the services offered due to the narrow bandwidth and high crosstalks. Comparative Infrastructure Costs 0 200 400 600 800 1000 1200 1400 1600 01000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Households' density (HH/km2) CAPEX/user ($) FTTB FTTC FTTN Fig. 3.7. Comparative OSP infrastructure costs for FTTx architectures, where X stands for Building, Curb and Node [60] Chapter III. Techno-economics 44 When EP2P, GPON and WDM/TDM PON are compared in terms of cost, the cost for both the OSP and the active equipment is estimated for each of the three architectures and the results are shown in Fig 3.8. and 3.9. respectively. The overall cost comparison is shown in Fig.3.10. As one can observe, there is a reduction in the range of 20% in CAPEX/user when the metro & GPON solution is compared to the metro & EP2P network, while there is a reduction in the range of 40% when WDM/TDM PON is compared to the last one. When WDM/TDM PON is compared to GPON the decrease is approximately 20%. All these results are shown in Fig.3.11. As can be observed the OSP cost is higher for EP2P which is expected due to its intrinsic characteristic of a dedicated fiber per customer. This leads to a higher number of cables, ducts and subducts as well as larger and more costly trenches. Furthermore it requires the use of more and larger COs which nevertheless are not included in the CAPEX estimation. The GPON CAPEX/user is larger than in the case of WDM/TDM PON, since a smaller maximum reach as well as a smaller splitting ratio is required in order to achieve the same bandwidth. WDM/TDM PONWDM/TDM PON Fig. 3.8. OSP equipment CAPEX/user vs. HH density [59] 51 Chapter IV Challenges 4.3 Solutions on the main limitations of long reach/enhanced PONs The suggested solutions for overcoming limitations in a fully passive optical network infrastructure are the results of a combination of two requirements. The first is to fight the challenges presenting themselves in optical networks due to natural phenomena, with the effective use of other phenomena, without the use of electro-optical conversion and the second to avoid the use of any active equipment whatsoever in the field. Nevertheless, electronic devices have been suggested for use in the ONU. Below we present four of these suggested solutions which will be further investigated in the next chapters of the thesis. 4.3.1 Amplification In order to preserve the passiveness of an optical network, two solutions have been suggested, that are based purely on qualities of the material used in optical networks and to their interaction with light. The first is amplification with the use of Erbium Doped Fibers. Er 3 3+ + as will be explained on the next chapter are risen with the help of optical pumping, in other words with the help of photons, from their ground state into excited states. This requires three energy levels. The top one to which the electron is elevated should lie energetically above the desired lasing level. After reaching its excited state, the electron must release some of its energy and drop to the desired lasing level. From this level, a signal photon can then trigger it into stimulated emission, whereby it releases its remaining energy in the form of a new photon with a wavelength identical to that of the signal photon. Since the pump photon must have a higher energy than the signal photon, the pump wavelength is shorter than the signal wavelength [64]. This attribute is successfully used in passive optical networks for amplification that avoids electro-optic conversion. Raman amplification is based on a non-linear effect, the so-called stimulated Raman scattering, that in some cases, as for example WDM transmission, could be harmful for the quality of the signals, but could also be beneficial when used for providing amplification. In this case study SRS has been used as an amplification mechanism. The phenomena behind Raman amplification is that if two or more signals at different wavelengths are injected into a fiber, SRS causes power to be transferred from the lower-wavelength channels to the higher-wavelength channels. The energy of a photon at a wavelength λ is given by hc/ λ, where h is Planck’s constant. Thus a photon of lower λ has a higher energy. So the transfer of energy from a signal of lower wavelength to a signal of higher energy corresponds to emission of photons of lower energy caused by photons of higher energy [65]. Therefore in this case, with the use of an auxiliary wavelength (used also as pump signal for EDFAs) we use SRS to our benefit. Chapter IV. Challenges 52 Since we want to preserve the full passiveness of the design, remote pumping has been introduced. Nevertheless, attenuation and dispersion degrade the efficacy of optical pump and of course, cost efficiency, safety regulations and non-linear effects do not allow for unlimited pump power use. 4.3.2 Compensation techniques 4.3.2.1 Dispersion compensation fibers Chromatic dispersion can be effectively compensated by using optical techniques, which is only natural to expect, given that chromatic dispersion originates in the optical domain. Special chromatic dispersion compensating fibers (DCFs) have been developed that provide negative chromatic dispersion in the 1550nm wavelength range. Some of the typical values where DCFs can provide total chromatic dispersion are between -340 and -1360ps/nm and these modules are commercially available. DCFs can be used as pre-compensators when located after an optical amplifier and before the transmission fiber. They are characterised as post-compensators when they are placed after the transmission fiber and before the optical amplifier. Their characteristics are concentrated in plots called dispersion maps. More specifically in a dispersion map one can find the accumulated dispersion and the power level as functions of the distance along the fiber. Their use in the OLT, allows for a simultaneous dispersion precompensation of all the WDM signals. 4.3.2.2 Equalization An optical system, such as a laser, fiber and photo detector constitute a channel over which the signal is being transmitted, if non linearities are ignored, the main distortion caused by this channel is the dispersion –induced broadening of the pulse. Dispersion is a linear effect, and hence the effect of the channel on the pulse, due to dispersion, can be modelled by the response of a filter with transfer function HD(f). Therefore, in principle, by using the inverse of this filter, say H-1D(f)as the equalization filter, this effect can be cancelled completely at the receiver. This is what an equalization filter is trying to accomplish. On the other hand, for conventional direct detection RXs, the linear distortion that is induced by CD in the optical domain is transformed into a nonlinear distortion in the electrical signal, which explains why only limited performance improvements can be achieved by using a linear baseband equalizer with only one baseband received signal. This also explains why nonlinear techniques, such as Non – linear Decision Feedback Equalization (DFE) and Maximum Likelihood Sequence Estimation (MLSD), are more effective in combating CD in direct detection RXs, [6668]. A commonly used filter structure for equalization is shown in Fig. 4.3. This filter structure is called a transversal filter. It is essentially a tapped delay line: the signal is delayed by various amounts and added together with individual weights. The choice of the weights, together with the delays, determines the transfer function of the 53 Chapter IV Challenges equalization filter. The weights of the tapped delay line have to be adjusted to provide the best possible cancellation of the dispersion-induced pulse broadening. Electronic equalization involves a significant amount of processing that is difficult to do at higher bit rates, such as 10 Gb/s. Nevertheless, we have used it quite successfully. Following we see some models of optical electronic equalizers in more detail. There are three main architectures for electrical channel equalizer (ECE). The most basic scheme is the feed-forward equalizer (FFE) (Fig. 4.3). In this set-up, a finiteimpulse-response (FIR) filter is added to the transmission line after the optical-toelectrical conversion. The filter has several stages, each consisting of a delay element, a multiplier and an adder. After every delay element, an image of the non-delayed input is multiplied with a coefficient and added to the signal. The number of filter stages used and the coefficients chosen are crucial for effective dispersion cancellation. Automatic control of the filter coefficients is essential. The filters can be designed based on prior information and kept fixed in the RX (the RX filter used after detection to suppress the out of band noise is a typical example), or they can be adaptive, and designed each time a connection is established, and are called adaptive regardless of the computation mechanism used for determining their coefficients, i.e., adaptively or not. When the distortion is time varying the coefficients need to be recomputed at given time intervals or continuously adapted to track the variations in the channel. The most common linear filter structure is the feed-forward (tapped-delay line) structure and can be implemented either in continuous or discrete time (Fig 4.4). It produces, as mentioned already, an output equal to the sum of weighted combination of different delayed signals. The coefficients of the filter are computed by optimizing a suitably chosen metric. The minimum mean squared error (MMSE) and the least squares (LS) are the most commonly used two criteria. MMSE filter coefficients are given by the Wiener-Hopf equations and can be adaptively estimated using gradient optimization, and the LS weights, using recursive estimates of the input correlation matrix based on the zero-forcing (ZF) algorithm. Least Mean Squares (LMS) algorithm uses an instantaneous estimate of the statistics required for the gradient updates of the Wiener-Hopf solution, and is a simple and effective solution that has been successfully used in most adaptive filtering applications to date. Fig. 4.3. Optical electronic equalizer (Feedback Forward Equalizer) Chapter IV. Challenges 54 A second approach is the decision-feedback equalizer (DFE) (Fig. 4.4, 4.5). This structure is an FFE with a second FIR filter added to form a feedback loop. Again, the coefficients of both filters require active control. Today's integration levels permit either FFE or DFE to be built into a clock-data recovery or demultiplexer chip with an extra power requirement of about 500 mW. The tap coefficients of the filter are calculated and are always adjusted in an adaptive operation according to an algorithm that runs in parallel with the purpose to minimize the error. The type of the algorithm and more significantly the way that this algorithm is optimized are particularly important in order to minimize the error and enhance the transmission properties of the system. The general operating modes of the adaptive equalizer include training and decision mode. During the training mode the algorithm adjusts to the channel characteristics and calculates the filter taps to compensate for the introduced impairments. In the decision mode, small variations in the taps allows for the compensation of the time varying effects of the channel. The most common algorithm that is used in order to calculate the taps is the LMS. The goal of that algorithm is to minimize the MSE between the desired equalizer output and the actual equalizer output. It is controlled by the error signal which is derived by the output of the equalizer with some other signal which is the replica of transmitted signal. The DFE version of the equalizer is a nonlinear process that uses the same algorithm but it subtracts the interference by the already detected data offering advanced performance characteristics. DFEs have long been used in digital communication systems. In channels with high ISI they outperform linear equalizers such as the ZF or MMSE equalizers. A DFE is a nonlinear equalizer that employs previous decisions to eliminate the ISI caused by previously detected symbols on the current symbol to be detected. A simple block diagram for a DFE is shown in Fig. 4.5. Fig. 4.4. Block diagram of the Decisionfeedback equalizer (DFE) structure 55 Chapter IV Challenges The DFE consists of two filters. The first filter is a feed-forward filter, identical in form to the linear equalizer described above. It is generally a fractionally spaced FIR filter, meaning that the tap spacing must be smaller than the symbol interval, with adjustable tap coefficients. Its input is the received filtered signal sampled at some rate that is a multiple of the symbol rate. The second filter is a feedback filter. It is implemented as an FIR filter with symbol-spaced taps having adjustable coefficients. Its input is the set of previously detected symbols. The output of the feedback filter is subtracted from the output of the feed-forward filter to form the input to the detector. The detector determines which of the possible transmitted symbols is closest in distance to the input signal. What makes the DFE nonlinear is the nonlinear characteristic of the detector that provides the input to the feedback filter. The tap coefficients of the feed-forward and feedback filters are selected to optimize some desired performance measure. For mathematical simplicity, the MSE criterion is usually applied, and a stochastic gradient algorithm is commonly used to implement an adaptive DFE. The most sophisticated structure is the maximum-likelihood (sequence) detector (MLD/MLSD) (Fig. 4.6). This is effectively a digital signal processor that performs the necessary mathematical operations on the incoming data stream to reconstruct the transmitted signal. More specifically, an MLSE receiver compares a long section of the noisy received signal with all the possible waveforms of the same length that could be received and chooses the one that is “closer” to the received. The drawback of this method is that it needs the signal to be in digitized form at the input, which requires an analogue-to-digital converter running at full line rate and at a high resolution. MLSE estimates the channel and decides the most likely sequence sent at the transmitter based on the received signal. MLSE is a Viterbi decoder. It has two parts: channel estimation (ISI estimation) and decoding. Decoding complexity of MLSE is the same as the Viterbi decoders. Channel estimation can be thought as the encoder that encodes the original signal with the weighted neighboring bit values. Due to encoders' nature, the channel is assumed to be linear. If the channel is time varying, then the estimation should be able to track the changes. MLSE also requires soft decisions for decoding to maximize the gain. This is analogous to hard and soft decision Viterbi decoding performance difference. The effectiveness of MLSE in the compensation of the chromatic dispersion may extend into several hundred kilometre range, but at such long range, the complexity of the MLSE processor is very large. It is important to examine and optimize the effectiveness of MLSE equalization and provide techniques to reduce complexity without impacting performance. As indicated already, digital equalizers based on MLSE are regarded as most efficient electronic equalization leading to lowest penalties (Fig 4.6). In difference to simple RXs, where each bit is decided immediately, in a MLSE the decision is Fig. 4.5. Block diagram of the Decision Feedback Equalizer (DFE). Chapter IV. Challenges 56 performed by calculating the transmitted bitstream with the highest probability. That means, that the probability for all bit combinations for the whole sequence is considered and the combination with highest probability is assumed as the transmitted bit combination. The overall probability of a transmitted bit combination is given by the product of all single bit probabilities. Instead of probability its logarithmic value is used and all separated logarithmic probabilities can be added, which is easier for hardware realization. The signal distortion within an optical channel can be assumed as coding of signal, which is very similar to convolutional codes. The Viterbi algorithm was originally devised for decoding convolutional codes, and thus can be implemented as a realization of the MLSE. The reason for the limited performance of electronic dispersion compensation (EDC) on CD mitigation in optical communication is due to the nonlinear characteristics of the absolute square operation of DD. In [69] a nonlinear electrical equalizer based on Volterra theory to mitigate this kind of nonlinear ISI is proposed. For illustration, the advanced equalization scheme or Nonlinear DF Equalizer setup is shown in Fig 4.7. The NL-DFE can be considered as the extension of a normal DFE including the nonlinear ISI mitigation. NL-DFE outperforms normal DFE and VE. Fig. 4.7. Setup of NL-DFE: nonlinear part of setup marked with dashed boxes. Fig. 4.6. Viterbi Equalizer based on MLSE 57 Chapter IV Challenges However, in this thesis the equalizers that will be used are FFE and DFE. 4.3.2.3 Reduced OLT power The non linear effects cannot be faced at their root, so the best suggested solution is the decrease of the output power at the OLT. In the case of the aforementioned design, the decrease of the power also applies for the optical pump which in this case creates a constraint on the amplification possible in each RN, proportional to the distance of the RN from the OLT. Moreover, safety regulations do not allow for use of power above a certain threshold. 58 Chapter V In line amplification In this chapter we present a design of a WDM/TDM PON with remotely pumped EDFs to be used for amplification. The theoretical background on Erbium Doped fibers is described, as well as their use in signal amplification. Next, we examine the wavelength dependency of the gain in erbium doped amplifiers, which proves to be a very important factor in the design of the aforementioned WDM/TDM PON. Subsequently a theoretical approach of the in-line amplification takes place. Continuing, we expose the effect of noise figure in Optical Signal to Noise Ratio (OSNR) when a number of cascaded RNs placed in-line with the transmission path of the signal and the one subsequent to the other is used, as amplification approach to a WDM/TDM PON. Some design issues on in-line amplification are exposed with more important the wavelength allocation of the network; and finally, the simulation results of a special WDM/TDM PON operating with the help of in-line amplification are presented commented and evaluated. 5.1 Erbium Doped Fiber Amplifiers 5.1.1 Basics The most usual way to model the physics of an Erbium Doped Fiber Amplifier is to use a three-level atomic system. Therefore we consider a three level system as depicted in Fig. 5.1. As one can see the ground state is depicted as 1, the intermediate state into which energy is pumped is labelled 3 and state 2 is the basic excited state. If state 2 has 3 1 2 Γ21 φsσs Γ32 φpσp 3 1 2 Γ21 φsσs Γ32 φpσp Fig. 5.1. The three level system used for the amplifier model [71] 59 Chapter V. In-line amplification a long lifetime (which as will be proven consequently is a characteristic of a good amplifier) it is referred to as metastable level. State 2 is the upper level of the amplifying transition and state 1 is the lower level. The populations of levels 1, 2 and 3 are named N1, N2 and N3 correspondingly. In order to have amplification we need to have population inversion among states 1 and 2 and of course since state 1 is the ground state the number of the excited photons (belonging to state 2) need to be at least half in number of the total population of erbium ions. That means we need a rather high threshold pump power for achieving amplification. Let us consider that the transmission of light in a fiber is a one dimensional problem and that the pump and signal intensities as well as the erbium distribution are constant in the transverse dimensions over an effective cross-sectional area of the fiber. The incident light intensity flux at the frequency corresponding to the 1 to 3 transition is denoted by υp and corresponds to the pump. The incident flux at the frequency corresponding to the 1 to 2 transition is denoted by υs and corresponds to the signal field. There are numerous reasons that cause change in population for each level, such as absorption of photons from the incident light field and spontaneous and stimulated emission. We denote as Γ32 the transition probability from level 3 to level 2, which is mostly non-radiative and Γ21 the transition probability from level 2 to level 1 which in the case of Er 3 3+ + is mostly due to radiative transitions. We denote the absorption cross section for the 1 to 3 transition by σp and the emission cross section for the 2 to 1 transition by σs. The rate equations for the population changes are written as:   PP dt dN  31332 3 Eq. 5.1   SS dt dN  12332221 2 Eq. 5.2     121 2 1 3 2 1P P S S dN dt              Eq. 5.3 In a steady state situation, the time derivatives will all be zero 0 321  dt dN dt dN dt dN Eq. 5.4 and the total population N is given by 321  Eq. 5.5 Using equation 5.1 we can write the population of level 3 as Chapter V. In-line amplification 60 31 32 1 1pp       Eq. 5.6 When Γ32 is large compared to the effective pump rate into level 3, υpσp, N3 is very close to zero, so that the population is mostly in levels 1 and 2. Substituting N3 in equation 2 with the use of equation 5.6 we obtain:   2 32 1 21 pp ss ss          Eq. 5.7 We then use equation 5.5 to derive the populations N1 and N2 and the population inversion N2 – N1: 21 21 21 22 pp s s p p             Eq. 5.8 The condition for population inversion, and thus for gain on the 2 to 1 transition, is that N2 ≥N1. The threshold corresponds to N1 = N2 and results in the following expression for the pump flux required: 21 2 1 th pp      Eq. 5.9 In a situation where the signal intensity is very small and the decay rate is Γ32 is large compared to the transition rate induced by the pump field, υpσp , we can thus write the population inversion as: 21 '1 '1 p p        Eq. 5.10 Where th p p    ' Eq. 5.11 The plot of equation 10 can be shown in Fig. 5.2. Below the pump threshold the inversion is negative; above the pump threshold it is positive. When the inversion is negative, there are more absorptive transitions then emissive transitions at the signal 67 Chapter V. In-line amplification The Fig.5.5 represents the signal gain variation, for a pump power of -7dBm and an EDF HE980. One can see the gain of the EDF as a function of the length and the signal wavelength. The Fig. 5.6 demonstrates the saturation effect of the amplifier. For these cases, the efficiency of the amplification decreases for high input signal power, due to the saturation of the amplifier. In general, until the EDF reaches the saturation point, the lower the input signal power provided to the EDF is, the higher the signal gain. An exponential increase in the noise figure is achieved, when the amplification goes further into saturation. For an efficient amplification, one should not get further into the saturation mode as this causes lowering of the gain, higher pump consumption and higher degradation of the signal [79] As we will explain later on, the wavelength dependency of the EDFA gain can have a serious impact on the function of WDM/TDM PON using cascaded in-line EDFAs. As an effect by itself is neutral, can be used though to the benefit of the network, when the proper wavelength allocation is foreseen. Fig. 5.6. a) Output signal power in function of the input signal power to various supplied pump power; b) Gain in function of the input signal power for different supplied pump power; c) Noise figure in function of the input signal power to various supplied pump power; d) Gain in function of the pump power to different values of signal input power [79] Chapter V. In-line amplification 68 5.2 Theoretical approach of in-line amplification 5.2.1 General considerations In-line EDF amplification refers to an alternative scenario of the original WDM/TDM PON (SARDANA), suggested in chapter 4, where the EDF is placed in the ring fibre rather than at the beginning of the distribution tree. Similar to the drop amplification case, in-line amplification can be constructed as part of the remote node by attaching the appropriate length of EDF before or after the add/drop filters. The pump distribution, carried over the upstream fibre is split accordingly, in order to provide the appropriate pump power level that in combination with the EDF length, offers the required gain. The network set-up as well as the appropriate RN design are shown in the figure below. The advantage of in-line amplification is that a small gain is only required in order to compensate the transmission and passive element losses of the intermediate distance between amplification stages. Therefore, in theory, all the signals can be kept on the appropriate level of signal power per channel with minimum requirements in terms of EDF length and portion of required pump power. Upstream data Pump x X 1-X Er+ X 1-X λ1λ2 Drop Add xX 1-X X 1-X Er+ Downstream data OLT L … … ONUONU … RN-1 … … … … ONUONU … RN- {n/2-1} … … ONUONU … RN-n … … … … ONUONU … RN-{n/2+1} downSTR upSTR upSTR downSTR Fig. 5.7. Network architecture and Remote Node comprising in line remote amplification 69 Chapter V. In-line amplification However, in practice, the fact that a group of wavelength channels require amplification, in combination with the wavelength dependent gain characteristics of EDF amplification in WDM channels, sets significant constrains in the design of the system. More specifically, the following limitations are observed:  All the studied effects (gain, noise figure and OSNR) apply on WDM channels being amplified simultaneously. The wavelength dependent nature of EDF causes power variations due to unequal gain per wavelength.  Any effects present in a system with in-line amplifiers must be studied for a cascade of amplification stages rather than single amplification per channel in the case of drop amplification. This reflects also in the design of such a system.  The number of accumulated channels is not constant as channels are dropped or added in the system. This affects also the amount of gain received per channel. These constraints complicate significantly both the design and the study of the network with in-line amplification. In general for the design of the downstream propagated signals, in-line amplification must provide a relative small gain in order only to compensate for the transmission losses and the losses of the passive elements. As the number of channels is reduced through a cascade of RNs, an even smaller total gain is required. Additionally, since the input power in the in-line amplifiers is strong and also the transmission losses are small, the added and accumulated noise level for the case of down stream propagation is expected to be limited, allowing good OSNR values at the ONU receivers. On the other hand, the high power levels per channels denote that the amplification operation will be performed in the saturation region of the EDF, where the obtained gain with respect to EDF length and pump is small. The design of the upstream signals is totally different than that of downstream propagation. The main difference here is that the power level of the added channels is very small due to the losses in the distribution tree. This is expected to give a significant initial OSNR degradation. Moreover, the power level of the accumulated signals must be kept at relatively low levels so that we don’t have huge variation between the already added and amplified signals and those newly added at each RN. In this case, the pump and EDF length requirements are relaxed, since a good gain value can be achieved for low power signals that are away from the saturation region of the EDFA. For the analysis of in-line amplification, the wavelength dependent nature of EDF is important to be considered. This depends on:  the type of the EDF used  the pump power  the EDF length and  the input wavelength channels and their power In the following sub-sections first the simulation tool is described, providing its calculation capabilities as well as the common setting and parameters used. Based on this tool, an analysis on the wavelength and power dependent issues of amplification is given, with emphasis on the unequal gain per channel and the OSNR degradation effects. Chapter V. In-line amplification 70 The VPI simulation tool was used in order to estimate the effects of in-line amplification in a WDM signal. The analysis was based on the use of the test-amplifier module that among others it calculates the optical frequency/wavelength dependent gain, noise figure and OSNR per WDM channel. The analysis is performed by examining the input and output signals of each amplification stage. 5.2.2 Definitions and parameters used for the analysis 5.2.2.1 Input test signal Since the main purpose of this study is to examine the gain and OSNR degradation of the signal and not other impairments, a simple WDM source was considered based on the generation of a comp of wavelength with certain wavelength position according to the ITU grid and power. In this case, each wavelength channel is considered to have all the power concentrated in a single frequency. The WDM comp source has been combined with a noise generation source that results in an initial OSNR of 40dB which is fixed for all cases. 5.2.2.2 Signal gain The gain of the amplifier is calculated by the difference of the output signal power and the input signal power at each amplification stage. stageionamplificatidBmPdBmPdBGiiniouti )()()( ,, Eq. 5.38 Note: Each amplification stage includes the WDM coupler and splitter before and after the EDF fibre. 5.2.2.3 Giles parameters A derivation of the Extended Black Box Model can be used, not only for providing an accurate description of the performance of EDFA, but as will be shown, for determining the cross sections in the form of Giles parameters of the studied EDFs. For the described Black Box Model, the Giles parameters depending on the emission and absorption cross sections, are given -as mentioned alreadyby the expressions found in [74]:  ()=a()NEr  ()=e()NEr 71 Chapter V. In-line amplification where NEr=drtnEr(rt)Ψ(rt)2 represents the effective erbium concentration over the cross section of the fiber (rt) overlapping with the signal propagation mode Ψ(rt) It can be demonstrated (publication pending) that the absorption Giles parameters, (), can be determined by using the expression:           ,, ,, 1( , ) ln , , , ln , , , [ ] ref ref in ref p ref ref ref ref ref in ref p ref ref T G P P L L L G P P L               Eq. 5.39 Once the absorption Giles parameters of the EDF are obtained, the emission Giles parameters can be determined as well, by using:     ( , ) ref ref ref T               Eq. 5.40 5.2.2.4 Excess noise power The amplifier noise is the ultimate limiting factor for system applications [Agrawal, 2002]. The ASE noise, can be expressed by the following equation:    ddhGP ASEASE  1 0 Eq. 5.41 where ρASE represents the ASE spectral density, that propagates in the same direction of the signal. The impact of ASE is quantified through the noise figure Fn given by Fn =2nsp. The added noise power at the amplification stage is given by: vGhvnP spASE  2 Eq. 5.42 where nsp is the spontaneous emission factor that depends on the relative populations N1 and N2 of the ground and excited states, hv the photon energy, G the amplifier gain and Δv the bandwidth over which the noise is calculated. 5.2.2.5 Optical signal to Noise ratio OSNR is calculated at the output of the amplification stage by separately estimating the amplified channel power and the noise power. This is done by splitting the output signal in two components, i) the pure (parameterized) wavelength channel and ii) the remaining noise level within 100GHz around the signal integrated at 12.5GHz bandwidth. The ratio of these two values defines the OSNR parameter. It is mentioned again that an initial OSNR value of 40dB is considered in all the examined cases. Chapter V. In-line amplification 72 5.2.2.6 Noise figure The noise figure (NF) can be defined as the relation between the input SNR and the output SNR of the EDF [80]. Considering as the main limitation of the signal detection, the signal spontaneous noise power (NS-SP), the noise figure can be calculated by the forward equation: vGhv P NF ASE   Eq. 5.43 where PASE is the ASE power, hv is the energy of photon, Δv is the resolution bandwidth of the Optical Spectrum Analyzer (OSA) (0.1nm in the referenced case) and G is the gain of the EDF. Considering also the shot noise (Nshot) and the spontaneousspontaneous noise (NSP-SP) powers, the NF is calculated respectively with the equations: GvGhv P NF ASE 1    Eq. 5.44     in eASEASE PGvhv hvBBP GvGhv P NF 2 2 0 2 4 2 1      Eq. 5.45 where B0 is the optical bandwidth of the filter after the EDF, Be the bandwidth of the electrical filter in the receiving circuit and Pin is the optical power at the input of the amplifier. The analysis of noise figure is based on the electronic definition according to which the NF is related with the ratio of the input SNRin over the output SNRout. For this calculation with respect to optical noise and PASE the receiver characteristics must be considered and how this electronic circuit is affected by the noise factor. Assuming that the dominant contribution to the receiver noise comes from the beating of spontaneous emission with the signal and neglecting the contribution shot noise then the noise figure term can be simplified to the equation spsp n G G nNF 2 1 2   Eq. 5.46 Hence from the previous equation the following expression is derived:                 vhv P G NF ASE 1 log10 Eq. 5.47 This expression is according to the IEC 61280 definition on obtaining measurements for WDM systems spaced within the ITU grid and particularly applicable 73 Chapter V. In-line amplification in cases of cascaded EDFAs. The measurement bandwidth is defined again at 0.1nm (or 12.5GHz). The frequency  refers to the operating wavelength or the central frequency of WDM channels. 5.2.2.7 Type of EDF In all measurements the EDF type of fibre used was the HE980® fibre by Lucent-OFS. In Fig. 5.8 the gain spectrum and absorption is given. It is the most important feature of an EDFA as it determines the amplification of individual channels when a WDM signal is amplified. The extracted Giles parameters (appearing in Table 5.1) were inserted in the EDF model of the VPI capable to emulate bidirectional signal propagation and amplification in doped fibers. The analytical model used, considers both the geometry of the fibre waveguide and the physical properties of the dopant ions determined by the Giles parameters. Since the gain and absorption properties in the model used for the studies presented here, is described using Giles parameters, very few additional parameters are required. This is because the measured gain and absorption per unit length in an actual fiber already include the effect of parameters such as the excited state lifetimes, and the details of how the optical field overlaps and interacts with the dopant ions. Both Giles parameters and additional fibre geometry related parameters have been provided by the manufacturer. Fig.5.8. The parameters of the EDF type (Lucent Ofs HE980 ®) that was used in the simulation tool Chapter V. In-line amplification 74 5.2.3 Analysis of the expected OSNR degradation 5.2.3.1 Initial OSNR degradation after the 1st amplification stage The first amplification stage in a cascade of amplifiers in a system is the one that determines the OSNR of the transmitted signal. According to the system design presented in the next chapter there two main cases that can be distinguished:  In downstream signal propagation the initial WDM channels have sufficient power levels (0dBm per channel) at the output of the OLT. In this case, the signal attenuation between OLT and the first amplification stage as well as between the next amplification stages is relatively small. Additionally, the gain provided is also small since we are mainly interested in compensating for the losses in between amplification stages. Therefore, the PASE contribution in the non-degraded (in terms of signal and noise power) input signal to the in-line amplifier is expected to be very small. Moreover since PASE depends on the achieved gain, its value will be also small. TABLE 5.1 Giles parameters for (Gain/Attenuation) for the EDF HE980 [OFS] Wavelength (nm) Gain (dB/m) Attenuation (dB/m) 1400 0,00 0,00 1455 0,09 0,69 1480 0,61 1,88 1484 0,69 2,08 1510 1,56 2,43 1519 2,43 3,30 1528 4,00 4,34 1530 4,08 4,08 1534 3,26 3,21 1544 3,08 2,56 1559 2,78 1,48 1585 1,17 0,35 1614 0,52 0,13 1629 0,00 0,00 1656 0,00 0,00 75 Chapter V. In-line amplification  In upstream signal propagation the tributary wavelength channels initiated from ONUs are facing a strong attenuation within the distribution trees before being amplified by the in-line amplification stages in the ring. Therefore, In this case the PASE contribution on the weak signal will be evident, causing an initial strong degradation through the first amplification stage. Next an example is provided for both cases that proves this issue theoretically. The example is based on the actual values used in performance evaluation of the system and the target gain values required. In all cases an initial signal OSNR of 40dB is considered. Therefore, any comparison in terms of OSNR degradation must be made according to this initial value. Assuming an initial OSNRin = Sin/N being amplified by the first in-line amplification stage with gain G, the output OSNR will be: vGhvnGN GS OSNR PGN GS N S OSNR sp in out ASE in out out out     2 Eq. 5.48 5.2.3.1.1 Downstream OSNR degradation after 1st amplification stage The power per channel of the signals being transmitted from the OLT is 0dB (1mW). With an initial OSNR of 40dB defined at 12.5GHz bandwidth, the noise power within the same bandwidth is N=1e-7 W. According to the results that are presented in the next chapter, the first in-line amplification stage provides a gain of 7dB with a noise figure of 6dB. Assuming that the term 2nsp is defined by the NF, and calculating the output OSNR from the last equation (for a channel frequency of 193.1THz, h=6.6261e-34 the Planck's constant, and BW=12.5GHz) we obtain: OSNRout=39.7dB. Therefore it is evident that in the case of downstream propagated signals, the OSNR degradation with respect to the initial transmitted signal OSNR (40dB) is only 0.3dB. If a pure initial signal is assumed with infinite OSNR (i.e. N=0) then the output OSNR of the first in-line amplification stage is OSNRout = 51.9dB. This value defines the absolute OSNR degradation due to the ASE noise in the amplifier. As the initial OSNR value of the transmitted channels is closer to this value then PASE becomes a more dominant factor and the OSNR degradation will increase. 5.2.3.1.2 Upstream OSNR degradation after 1st amplification stage The power per channel of the signals being transmitted from the ONUs is 0dB (1mW) but due to the tree losses these signals are entering the in-line amplification stage with - 20dBm (10uW). With an initial OSNR of 40dB defined at 12.5GHz bandwidth, the noise power within the same bandwidth is N=1e-9 W. According to the design parameters for upstream signal a small gain is only provided at each RN in order to compensate for losses. An average value of 3dB gain Chapter V. In-line amplification 76 can be considered in this case. The noise figure increases compared to the downstream case and can be up to 7dB. Again the term 2nsp is assumed to be defined by the NF; actually for low gain values a more accurate formula to use is NF=2nsp(G-1)/G from which 2nsp is calculated. Finally, the output OSNR from the last equation (for a channel frequency of 193.1THz, h=6.6261e-34 the Planck's constant, and Δv=12.5GHz) is: OSNRout=27.7dB. Therefore, in the case of downstream propagated signals, the OSNR degradation with respect to the initial transmitted signal OSNR (40dB) increases significantly by 12.3dB. If a pure initial signal is assumed with infinite OSNR (i.e. N=0) then the output OSNR of the first in-line amplification stage is OSNRout = 27.8dB. This shows that the generated ASE noise is a dominant factor and affects significantly the output OSNR almost independently of the quality of the initial signal 5.2.3.2 OSNR degradation in a cascade of in-line amplifiers For this case we assume that any additional in-line amplification systems do not provide extra gain in the signal but simply compensate for the losses between the amplification stages. On contrary ASE noise is accumulated. Therefore the signal power S and PASE after the ith amplifier and compared to the (i-1)th amplifier can be expressed as: vhvGnPP PPP ispiASEiASE SiSiS     2 1,, 1,, Eq. 5.49 Therefore the OSNR at the end of the ith section is: S isp ii P vhvGn OSNROSNR    2 11 1 Eq. 5.50 And after a cascade through N amplification stages this becomes: S N iisp NP vhvGn OSNROSNR     1 0 2 11 Eq. 5.51 where OSNR0 is the initial OSNR of the transmitted signal. Extending the example presented above for the case of a cascade of in-line amplifiers we can distinguish again the two cases of downand up-stream propagation. 5.2.3.2.1 Downstream OSNR degradation through a cascade of in-line amplifiers In this case the degradation term of (Ps/2nsp∙hv∙G∙Δv), has a very small contribution as it was calculated for the initial OSNR degradation. Additionally, Ps is assumed constant 83 Chapter V. In-line amplification between the OLT and an end user is 19km. The OLT transmits 32 wavelengths in the C band. As far as the downstream transmission in concerned the OLT output is considered equal to 10dBm per channel. This power level could cause non linear effects during propagation, but the fact that the distances between the RNs as well as the tree fiber length are relatively small prevents the appearance of non-linearities. In the case of upstream transmission the ONU output is equal to 0dBm while the OLT input level is considered -28dBm. In RSOA based ONUs the input power required to achieve adequate performance and output power at 0dBm is considered -20dBm. The pump power is set to 31dBm. The 32 aforementioned wavelength channels have 100GHz spacing and range within the C-band from 192THz-195.1THz. In order to achieve the same gain through the amplification stages for all wavelengths, an optimum design per RN and per in-line amplifier is required. The main purpose in an optimum design is to consider the worst case of resiliency and reserve those wavelengths that can have more gain for the distant nodes. The gain of the EDF is not flat for all frequencies. On the contrary it exhibits a gain tilt which favors some wavelengths in terms of gain received as can be seen in Fig. 5.8. As the pump power fades out over longer distances, these wavelength channels that have received more gain in the previous stages, as stated in the network design section, will be able to reach longer distances. Since this effect is evident in low frequencies, these are allocated to the more distant RNs in the resiliency case, which are RN1 and RN16, if it is considered that the RN1 to OLT or RN16 to OLT connections respectively are down. The wavelength allocation used in the simulation set up, is the one appearing in Table 5.3. In-line amplification stages were placed in the network according to the normal operation, meaning that the channels with the lower frequencies were handled by RN1 & RN16, RN2 &RN15 handle the sequential, RN3 & RN14 the next ones and so on. Therefore RN1-RN8 present a symmetry with nodes RN16-RN9 when referring to the downstream fibre ring in terms of amplification stages, EDF length, dedicated percentage of pump power and gain tilt. The upstream fibre ring as already mentioned is equipped with the same length of EDF in every RN which depends on its OSNR performance and is in this case 2.2m. As aforementioned the tree fiber has a 3km length. Considering the total passive losses of the network it is evident that these depend mainly on the splitting ratio at the distribution trees and less on the fibre losses. Therefore, in order to comply with the given values for transmitter-receiver power levels, the splitting ratio was set in 1:32 for each tree, which results in 1024 customers served. Table 5.3 Frequency allocation RN 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 f1 192 192.4 192.8 193.2 193.6 194 194.4 194.8 195 194.6 194.2 193.8 193.4 193 192.6 192.2 f2 192.1 192.5 192.9 193.3 193.7 194.1 194.5 194.9 195.1 194.7 194.3 193.9 193.5 193.1 192.7 192.3 Chapter V. In-line amplification 84 The total losses considered at the RN are the losses of the filter, the WDM coupler, the 50/50 coupler, and the circulator. Their sum is 4.6dB. 5.3.3.1 Downstream scenario and set up Each of the shaded areas represents a different RN starting from the first one on the upper left corner and ending on RN16 on the lower right corner. The schematic shown here actually represents the worst case scenario of full resiliency, in which all the 32 wavelengths with 100GHz channel spacing are propagating towards the same direction and the last pair drops at RN16. In normal operation the same exactly set-up was used, up to RN8 though and using the 16 of the 32 wavelengths with channel spacing of 200GHz. The dashed circles in the schematic represent the positions of the in-line amplification stages. As it is observed, in-line amplification is not performed at every stage (Only RN1,4,7,10,13,16 have amplification). This is done because when optimizing the system for the worst case scenario it was shown that the signal level can be increased significantly allowing its transmission over a cascade of additional two nodes before amplification is again required; more details are given in the next subsection. It is important to mention that the elimination of in-line amplification stages is allowed because the transmission losses are low between RNs (i.e. 1km of fibre + passive component losses). Alternatively, we have tried to have in-line amplification at every node using a very small percentage of the distributed pump and/or short EDF length. However, in this case, it was observed that due to the slightly unequal gain per channel obtained at each amplification stage, the cascade of in-line amplifiers was Fig. 5.13 Schematic of the network design for the case of downstream propagation under resilient mode. For normal operation the first 8 RNs are only used. 85 Chapter V. In-line amplification resulting in a strong gain spectrum tilt in the more distant nodes. More specifically, this gives rise in the power of low frequency channels which prevents the large frequency channels to be amplified and reach the appropriate power level for correct reception at the ONU. At each RN shown in the figure above, the upper link has an in-line amplifier (if this is required) followed by two drop filters centred at the wavelength of the drop channels. The lower link represents the pump distribution and contains the WDM splitters/couplers for the pump and the tap couplers that distribute the pump to both the down-stream and upstream directions. The design was built following the rule about network symmetry. Therefore, for example the pairs of [RN1, RN16], [RN2, RN15], [RN3, RN14], … have exactly the same characteristics (splitting ratios and EDF length). In this case and under normal operation, the results obtained for the propagation through RN1 to RN8 will be the same as that from RN16 to RN9; actually the only difference is that the wavelength channels in the last 8 RNs are shifted by 100GHz with respect to the first 8 RNs. Additionally, the internal design of the last 8 RNs, is symmetrical with respect to the first 8 RNs and therefore, the in-line amplification stage appears after the drop filters. This is true for the resiliency case, as after RN8 the signals are propagating in the opposite logical direction compared to the propagation in the first 8 RNs. Table 5.4. contains the optimum values of coupling factor and EDF length used in this set-up. The third column shows also the corresponding value of the pump power that is injected in the in-line amplification stages. #RN Table 5.4. Design parameters for the case of downstream scenario Coupling factor Optimal EDF Length [m] Pump Power[dBm] 1 0.9 6 28.05 2 0 0 0 3 0 0 0 4 0.9 5 23.37 5 0 0 0 6 0 0 0 7 0.5 6 25.679 8 0 0 0 9 0 0 0 10 0.5 6 17.989 11 0 0 0 12 0 0 0 13 0.9 5 3.766 14 0 0 0 15 0 0 0 16 0.9 6 -3.925 Chapter V. In-line amplification 86 5.3.3.2 Upstream scenario and set up In the upstream propagation through the SARDANA ring, the added wavelength channels from each RN are propagating towards the OLT in the same fibre as the pump signal, but in a counter propagation mode. The critical difference between upstream and downstream propagation in terms of the amplification stage design, is that the added upstream channels are facing large losses through the distribution tree and when they are added in the ring fibre, their power level is less that -20dBm. Since this power level is away from the saturation region of the in-line amplifier, adequate gain can be provided in the expense of OSNR degradation. Therefore, the choice on the EDF length and the portion of the injected pump is critical in order to keep noise at moderate levels. The following figure provides the schematic used to emulate the performance of the upstream signal propagation. The schematic refers to the worst case of operation, under resilient mode, when all added channels propagate through the same direction upwards to the OLT. The position of the RNs are in reverse order with respect to the downstream scenario and therefore RN1 is the one on the lower right corner of Fig.5.14 (followed by the OLT receiver) and RN16 is the one on the upper left corner. The general design rule about symmetry is also implemented in this set-up and is observed between the first 8 and the last 8 RNs. At each RN, the upper link represents the signal propagation and the lower links the pump distribution system. Although pump and signal are propagated over the same fibre, within each RN these two are split and handled differently. Therefore, in terms of total passive losses, the upstream signals -when compared to the passive losses of downstream signalsare facing the additional loss from two WDM couplers, used to separate and recombine the counter propagated pump at each RN. Above each RN, is a system that distributes the appropriate wavelength channels pairs in the RNs. 87 Chapter V. In-line amplification In this figure, it is observed that in-line amplification stages have been added at each RN. From the system performance point of view, this approach is mandatory, since we should not allow the already weak added channels to loose extra power. This would affect their OSNR performance. On the other hand, the provided gain must be kept low, by using small coupling ratios for the pump and small EDF lengths, so that all channels have almost the same power level after there are added and cascaded in the ring. If this is not guarantied, then any added signal with low power will be buried inside the noise level of the already added and cascaded channels with large powers. In the case of upstream propagation under normal (non-resilient) mode, the last 8 RNs (i.e. RN8 to RN1) in the schematic are only used. Table 5.5 contains the optimum values of coupling factor and EDF length used in this set-up. It is observed, that coupling factors and EDF length parameters are much lower that what was used in the downstream signals. Fig. 5.14 Schematic of the network design for the case of upstream propagation under resilient mode. For normal operation the last 8 RNs are only used. Chapter V. In-line amplification 88 #RN Table 5.5. Design parameters for the case of upstream scenario Coupling factor Optimal EDF Length [m] Pump Power[dBm] 1 0.95 1.2 24.582 2 0.95 1.2 23.405 3 0.95 1.2 22.228 4 0.95 1.2 20.598 5 0.95 1.2 19.42 6 0.9 1.2 21.254 7 0.95 1.2 13.825 8 0.95 1.2 12.648 9 0.95 1.2 11.726 10 0.95 1.2 10.548 11 0.9 1.2 9.376 12 0.95 1.2 4.954 13 0.95 1.2 3.776 14 0.95 1.2 2.146 15 0.95 1.2 0.969 16 0.95 1.2 -0.208 5.4 Evaluation of the design solution The study performed, shows that the described architecture can serve 1024 end users with several hundreds of symmetric data rate in a maximum distance of 19km. More specifically, for downstream transmission, we observe that when operating in normal mode, all the ONUs are reached with a signal power higher than -20dBm. That means -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 191.5 192 192.5 193 193.5 194 194.5 195 195.5 Channels' frequency allocation ONUs' Input Power [dBm] RN1 RN3 RN14 RN16 RN2 RN15 RN4 RN13 RN5 RN12 RN6 RN11 RN7 RN10 RN8 RN9 Fig. 5.15 Optical power of downstream signals when reaching the ONU receiver 89 Chapter V. In-line amplification that all ONUs used, could be based on RSOAs. The ONUs of the trees served by RN1, RN2, RN3, RN4, RN5, RN6, RN7, RN8, RN12, RN13, RN14, RN15 , RN16 are reached with a signal higher than -20dBm, therefore RSOAs would be efficient in these cases in every operation mode (normal or resilient mode). In case of resiliency operation though, the signal can be received by standard GPON APD receivers in the ONUs served by RN9, RN11 and one of the trees of RN10. The upstream operation of the network is ensured in normal as well as in every case of resiliency operation. The reason is that the signal power level reaching the OLT (all frequencies considered) is above -28dBm which is the OLT receiver’s sensitivity. The results for downstream and upstream signals are depicted in Fig. 5.15 and Fig. 5.16. respectively. As can be observed the trees connected to RN9, RN11 and one of the trees served by RN10 in the worst case of a fiber cut (thus a fiber cut between either the OLT and RN1 or the OLT and RN16) are reached by signals with power levels lower than the required for the operative use of an RSOA as a receiver/transmitter in the end part of the network. The lower power level for the upstream signals that reaches the OLT’s receiver is -21.64dBm that comes from the RN with the smaller distance from the OLT. The reason for that is that this signal doesn’t acquire any gain as it doesn’t pass through any amplification stage. A better performance of the network could be achieved with a smaller splitting ratio (e.g. 1:16 which corresponds to 512 users) or a larger value of pump power (e.g. 39dBm). In both the aforementioned cases the RSOA could be used in all ONU’s and in any case of operation. -24 -22 -20 -18 -16 -14 -12 -10 -8 -6 -4 191.5 192 192.5 193 193.5 194 194.5 195 195.5 Channels' frequency allocation OLT's input power RN1 RN3 RN14 RN16 RN2 RN15 RN4 RN13 RN5 RN12 RN6 RN11 RN7 RN10 RN8 RN9 Fig. 5.16. Optical power of upstream signals when reaching the OLT receiver Chapter V. In-line amplification 90 The results for all the cases presented before are provided, in more detail in Fig.5.17Spectrum after RN1 Spectrum after RN4 Spectrum after RN7 Spectrum after RN11 Spectrum after RN14 Fig. 5.17 Optical spectra with the through and dropped wavelength channels at RNs after each amplification stage for the case of downstream scenario in resiliency mode (Resolution Bandwidth = 0.1 nm) Channels added from RN16 to RN9 All channels added from RN16 to RN1 (Spectrum at OLT) OSNR (for channels added RN16 –RN9) OSNR (for all channels, measured at OLT) Fig. 5.18. Optical spectra with the through and dropped wavelength channels at RNs after each amplification stage for the case of downstream scenario in resiliency mode (Resolution Bandwidth = 0.1 nm) 91 Chapter V. In-line amplification 5.20. More specifically the optical spectra after each amplification stage is shown. For the downstream signals the spectra show both the through channels and the dropped channels at the corresponding ONU. On the other hand, for the upstream signals the spectra show the newly added signals together with the already added signals from previous RNs for the middle and the last RN. Moreover in this case the OSNR per channel is provided. This investigation focused in an urban deployment. The same design can, with different parameters be optimized with response to the needs and the targets of the network. For example more users and stronger pump, longer distances and less users. 5.4.1 Conclusion In this study we have shown that a network architecture based on SARDANA, using inline amplification, can reach 19km while serving 1024 users with several hundreds of Mbps. Such a design can be applied in both, normal and resilient operation. If normal operation is considered, then all costumers can be serviced with the use of RSOA’s Spectrum after RN1 Spectrum after RN4 Spectrum after RN7 Fig. 5.19. Optical spectra with the through and dropped wavelength channels at RNs after each amplification stage for the case of downstream urban scenario in normal mode (Resolution Bandwidth = 0.1 nm) All channels added from RN8 to RN1 (Spectrum at OLT) OSNR (for all channels, measured at OLT) Fig. 5.20. Optical spectra with the added wavelength channels and OSNR per channel at OLT for the case of upstream urban scenario in normal mode (Resolution Bandwidth = 0.1 nm) Chapter V. In-line amplification 92 based ONUs. Nevertheless, in the worst case of a fiber cut, thus a fiber cut between the OLT and the first RN, the 160 users that are served by the RNs most affected can be served by standardized GPON APD receivers. The proposed remote amplification solution consists of a fully passive architecture which can expand the limits of already deployed passive infrastructures, using remote amplification. The low pump power required, along with the extended use of RSOAs suggested, make this design very efficient in terms of OPEX. Furthermore, the overall cost of the RSOAs and the different components of the network, consist of a profitable solution in terms of CAPEX. 99 Chapter VI. Alternative band 6.4 In-line and L-band operation of a WDM/TDM PON Our approach, focusing on amplification and more specifically on in-line amplification using for transmission the L-band, is based and tested in a double ring fiber WDM/TDM PON that has been described in chapter 5, but with a few alterations [88]. The goal is to achieve, with given transmission and pump power, the longest reach possible from the OLT to the furthest user. At the same time, the aim is to serve more than 1000 users and retain the signal in a high level. In the scenario examined, the signal level as well as the OSNR obtained, allow the use at the customer’s end of a RSOA. The scalability of the network permits the implementation of several deployments, whose planning depends on the population density and the area considered. These characteristics determine the final number of the RNs as well as the distances among them. However, in every approach the need of remote amplification is necessary in order to have the appropriate input level in the ONU and to maximize the users while counterbalancing for the power losses throughout the network. On this study the use of in-line remote amplification has been considered. Moreover we have investigated the use of an alternative band of wavelengths, namely the L-band. 1024 users have been reached in a distance of 20km with a signal appropriate to be used with an RSOA. 6.4.1 Parameters of the network affecting its operation The parameters taken under consideration are the number of RNs, the signal power, the pump power, the total length of the fiber between the OLT and the final user -which could be considered as the sum of the fiber distances from the OLT to the RN and the tree fiber length-, the losses induced by the RN, the splitting ratio of the tree, the length of the EDF, the percentage of the total pump power used in each RN to pump the corresponding EDF , the sensitivity of the receivers and the wavelength allocation in the sense of appointing which frequencies will be dropped in each RN regarding their distance from the OLT. Continuing, the parameters are analysed in more detail. Number of RNs The number of RNs depends on the number of end users we want to cover. In an urban area a number of 16RNs per fiber ring is considered. Increase of the number of nodes corresponds to increase of the users served and decrease of the power budget. Signal power value The downstream signal power’s value has been set in 10dBm. The reason for that is that with the power link budget performed, the design of in line amplification requires 10dBm initial power in order to reach the users of the first tree, since no amplification is Chapter VI. Alternative band 100 performed for this frequency. Moreover the use of a stronger signal power was avoided, so as to prevent emergence of non linear effects at furthest propagation as well as for restraining OPEX. The upstream signals’ value is set to 0dBm in order to comply with the capabilities of an RSOA. Pump power value The operation of in line amplification design was tested for two values of pump power, 31 and 39dBm respectively. Increase of the pump power makes the network more flexible in terms of users served. The reason is that it allows a more effective amplification in terms of greater gain, thus an increase in the power budget is achieved. Nevertheless, we examined the use of 31dBm, because a smaller pump power complies better with safety regulations as far as the CO is concerned. It should be marked however, in regard of safety regulations, that since the pump circulates only in the upstream fiber ring, there is no danger that it could come to the reach of the end user. The pump is counter propagating with the upstream signals as aforementioned, but in this investigation the loss due to pump depletion is considered negligible, because of the low signal power. Losses induced at the RN Each RN includes filters that perform Add & Drop operation. In the cases of RNs that include EDFs for in line amplification, the use of splitters and WDM couplers is additionally necessary. The filters used along with the splitters and the WDM couplers introduce losses at the RN. Losses are also induced by the WDM couplers and the attenuation of the fiber along the path to the end user. Their sum is 4.6dB. All these losses have to be counterbalanced by the gain of the EDF. Splitting ratio As aforementioned, the splitting ratio of the tree can vary from 1:32, 1:64 and 1:128 for an urban deployment. Since each node serves two trees each with a different λ, these splitting ratios are interpreted into 64, 128 and 256 users per RN respectively. It is obvious that increasing the splitting ratio the number of end users is increasing accordingly. One should take into account though, that this happens in expense of power losses induced to the network. In the scenario presented, a splitting ratio of 1:32 has been used. EDF length The length of the remotely pumped EDF affects the amount of gain achieved, the portion of the pump power consumed, as well as the OSNR. Therefore a careful choice has been made on the lengths of the EDFs used in each RN in respect of its position in the network (distance from the OLT) in order to counterbalance for all the 101 Chapter VI. Alternative band aforementioned variables. Different fiber lengths have been used in the downstream ring, symmetrically placed though, as described also in chapter 5 (i.e. same EDF lengths used in the first and last RN, so as to preserve a symmetrical design) in order to adapt in the case of resiliency operation (for example a fiber cut between the OLT and the first or the last RN). Furthermore, due to the elaborate design some RNs could be excluded (in this case as well as in the design of chapter 5) since the signal had the necessary power to serve the corresponding trees and no EDFs were placed in their interior. In the upstream propagation ring, all RNs were equipped with an EDF of the same length, but carefully optimized in terms of low pump power consumption and OSNR. Distribution of pump power Pump power is distributed in each RN to be used in amplification of both the downstream and the upstream propagating signals. The distribution is implemented by using different percentages in each RN for the amplification of the downstream signals, while the portion used for the amplification of the upstream signals is invariable. As far as the downstream propagation is concerned, the power required, was determined by the need of each RN for amplification. In some distances the signal had the power required to reach the customers’ ONU without amplification, so, some RNs were excluded. A conservative approach was considered on the use of the pump so that it would be adequately distributed to all amplification stages. Receiver’s sensitivity We consider the use of an RSOA. A power value of -20dBm is considered necessary in order to both obtain the downstream data and at the same time use the optical carrier for modulation with the upstream signal and achieve an output of 0dBm. Transmission with low input powers have been already demonstrated [84]. Finally, the receiver at the OLT is considered to have a sensitivity of -28dBm. Frequency allocation 32 channels have been considered, with 100GHz spacing that range within the L-band from 187.4THz-190.5THz. In order to achieve the same gain through the amplification stages for all wavelengths, an optimum design per RN and per in-line amplifier is required. The main purpose in an optimum design is to consider the worst case of Table 6.3 FREQUENCY ALLOCATION FOR L-BAND WDM/TDM PON WITH IN-LINE AMPLIFICATION RN 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 f1 187.6 188 188.4 188.8 189.2 189.6 190 190.5 190.3 189.9 189.5 189.1 188.7 188.3 187.9 187.5 f2 187.7 188.1 188.5 188.9 189.3 189.7 190.1 190.4 190.2 189.8 189.4 189 188.6 188.2 187.8 187.4 Chapter VI. Alternative band 102 resiliency and reserve the frequencies that can have more gain for the distant nodes. The gain of the EDF is not flat for all wavelengths, but exhibits a gain tilt which favors some of them in terms of gain received. As the pump power fades out over longer distances, the channels that have received more gain in the previous stages will be able to reach longer distances. This effect is evident in low frequencies therefore these were allocated to the more distant RNs in the resiliency case, which are RN16 and RN1, if it is considered that the RN16 to OLT or RN1 to OLT connections respectively are down. The wavelength allocation used is shown in Table 6.3. 6.4.2 Set up The set up is considered to be applied in an urban area, therefore it consists of 16 RNs, with a distance of 1km among them, and an OLT connected with two fiber rings. One of them is used for downstream transmission, while in the other; the upstream signals are counter propagating with the pump signal. The tree fiber has a 3km length, therefore the maximum distance from the OLT to the end user is 19km. Considering the total passive losses of the network it is evident that these depend mainly on the splitting ratio at the distribution trees and less on the fibre losses. Therefore, in order to comply with the given values for transmitter-receiver power levels, the splitting ratio was set in 1:32 for each tree, which results in 1024 customers served. The in line amplification is realized in the RN. With the use of fixed filters and splitters two wavelengths are dropped in each RN. The EDF is placed among the two drop filters and before a 50/50 splitter used for resiliency. Each wavelength is then driven to a tree. The placement of the EDF among the drop filters creates the need to transmit in the downstream direction a signal with adequate power to reach the first tree (served by RN1 in normal operation or RN16 when operating in resiliency mode in the case of a fiber cut between the OLT and RN1) as the drop signal of the first tree in this design will not be amplified. An overview of the network, as well as the RN design, is presented in Fig.6.7. 103 Chapter VI. Alternative band 6.4.3 Results The study performed shows that the described architecture along with the use of L-band signals can serve 1024 end users with several hundreds of symmetric data rate in a maximum distance of 19km. More specifically for downstream as can be seen in Fig. 6.8, all the ONUs are reached with a signal power higher than -20dBm which means, they could be based on RSOAs . The upstream operation of the network is also ensured by the fact that the signal power level reaching the OLT is above -28dBm, as can be seen in Fig.6.9 which is the OLT receiver’s sensitivity. RN13 RN14 RN15 RN16 RN2 RN1 RN3 RN4 RN… CO OLT Upstream signals Pump power Downstream RN… 50/50 WDM WDM WDM WDM 50/50 1:32 1:32 λi1&2 200GHZ Pump RN i Downstream signals Downstream signals Upstream signals Fig. 6.7.Network architecture and Remote Node comprising in line remote amplification -24 -22 -20 -18 -16 -14 -12 -10 -8 187 187.5 188 188.5 189 189.5 190 190.5 191 Channels' frequency allocation ONUs' Input power [dBm] RN16 RN14 RN3 RN1 RN15 RN2 RN13 RN4 RN12 RN5 RN11 RN6 RN10 RN7 RN9 RN8 Fig. 6.8.Downstream signals Chapter VI. Alternative band 104 The gain of the special doped EDF has been experimentally investigated for different pumping power values and the results are depicted in Fig.6.10. 6.5 Conclusions In this chapter we reviewed the idea of using an alternative band of frequencies for the implementation of a WDM/TDM PON. After exposing some technologies investigated in the past for L-band amplification and their performance we have reviewed some types of fiber for L-band amplification. Continuing we report the investigation we performed, on the L-band operation of a WDM/TDM passive optical network. More specifically, we have shown that a WDM/TDM passive optical network architecture using L-band in line amplification can reach 19km while serving 1024 users with several hundreds of Mbps while all costumers are serviced with RSOA’s based ONUs. This solution consists of a fully passive architecture which can enlarge the limits of already deployed passive infrastructures, using remote amplification. The operational expenditures are expected to be decreased due to the low pump power required along with the extended use of RSOAs. In addition the overall capital expenditure can be reduced as the RSOAs consist of a profitable solution. -24 -22 -20 -18 -16 -14 -12 -10 -8 187 187.5 188 188.5 189 189.5 190 190.5 191 Channels' frequency allocation OLT's Input power [dBm] RN16 RN14 RN3 RN1 RN15 RN2 RN13 RN4 RN12 RN5 RN11 RN6 RN10 RN7 RN9 RN8 Fig. 6.9.Upstream signals EDF IXF-FGL 0 5 10 15 20 1570 1580 1590 1600 1610 Wavelength [nm] Gain [dB] Pump = 13 dBm Pump = 14 dBm Pump = 15 dBm Pump = 16 dBm Pump = 17 dBm Pump = 18 dBm Pump = 19 dBm Fig.6.10.Characterization of the L-band amplifier in the RN 106 Chapter VII Extended band operation of a WDM/TDM PON In this chapter, we explore the possibilities of the expansion of a WDM/TDM PON with advancing and enhancing its band operation. Such a study can be considered very useful for infrastructure proprietors wishing to lease one fiber to more than one operator. The suggested designs can turn the network much more competitive, hence lowering the final services’ prices and giving a financial boost to the whole technology. A trial network design, due to operate in both C + L band, has been implemented. Further, the combination of Raman amplification with remotely pumped EDFs placed as in the previous chapters, in-line with the transmission path, thus distributed in the WDM ring is shown. Finally, we have investigated the limits and the performance of a WDM/TDM network operating in C + L band, which uses remotely pumped amplification in the drop part of the network, with the help of an especially designed RN. Chapter 7 has been organised in three subsections. In the first we briefly explain the theoretical base of Raman amplification. On the second, the results of applying Raman amplification in a WDM/TDM PON are presented. Raman amplification mechanism is combined with in-line EDF amplification. On the third a RN design is presented, one that is predestined to operate in C+L band but using amplification on the drop part of the network. In other words, designed to enforce two wavelengths, the ones destined to the TDM trees connected to the specific RN. 7.1 Raman amplification Stimulated Raman Scattering (SRS) is an interaction between light waves and the vibrational modes of silica molecules. If a photon with energy hv1 is incident on a molecule having a vibrational frequency Vn, the molecule can absorb some energy from the photon. In this interaction, the photon is scattered, thereby attaining a lower frequency v2 and corresponding lower energy hv2. The modified photon is called a Stokes photon. Because the optical signal wave that is injected into a fiber is the source of the interacting photons, it is often called the pump wave, since it supplies power for the generated wave. 107 Chapter VII. Extended band This process generates scattered light at a wavelength longer than that of the incident light. If another signal is present at the longer wavelength, the SRS light will amplify it and the pump-wavelength signal will decrease in power. Consequently, SRS can severely limit the performance of a multichannel optical communication system by transferring energy from short-wavelength channels to neighbouring higher wavelength channels. This is a broadband effect that can occur in both directions. Powers in WDM channels separated by up to 16 THz (125nm) can be coupled through the SRS effect, in terms of the Raman gain coefficient gR as a function of the channel separation ∆νs .This shows that owing to SRS, the power transferred from a lower wavelength channel to a higher-wavelength channel increases approximately linearly with channel spacing up to a maximum of about ∆νc = 16THz (or = 125nm in the 1550-nm window), and then drops off sharply for larger spacing [89]. 7.2 C + L band gain equalization – Hybrid Raman & in-line EDF amplification mechanism The feasibility of gain enlargement and equalization on extended reach WDM-ring PON by means of hybrid Raman/EDFA amplification has been investigated with the use of simulation. This study has been performed in the framework of a collaborative effort with Berta Neto [90]. 7.2.1 System description The network examined, consists of a 80km WDM ring with 8 nodes in each one of which, 2 channels are added/dropped. It is therefore operated with 16 channels, 8 of which cover the C band while the other 8 cover the L one. In the CO, a bidirectional laser emitting at 1480nm is used to pump the spaced by 400 GHz 16 channels, that are distributed between 192.5 THz and 186.3 THz (see Table 7.1.). The channels leave the central office with 0dBm of optical power. The bypass losses are considered 0.53dB for channels and 1.01 dB for the pump, while the attenuation of channels and pump are set to 0.20dB/km and 0.25dB/km respectively. The channels drop losses are 3dB. In this framework two situations are analyzed, one with Raman amplification and the other with a hybrid amplification scheme composed by in line EDF with Raman. Therefore, an optimized span of EDF is inserted in the mid length of each link according to the dropping channels powers. This procedure was performed with total pump power of 1W (500mW in each direction). The results have demonstrated gain equalization with a ripple of 2.54dB over a bandwidth of 50nm by using spans of EDF with a total length of 22km. Chapter VII. Extended band 108 7.2.2 Method used – Results The 16 analyzed channels, with the exception of the last two, rely on the maximal bandwidth of Raman gain efficiency. In the following picture, the Raman gain efficiency is plotted for a standard single mode fiber, pumped at 1480nm, being the C and L bands represented by black and red arrows, respectively. The optimization strategy followed is the following: (i) drop the channels with the higher gain first in order to settle a maximal gain level and decrease the effect of pump depletion, (ii) whenever the channel power reduction surpasses a predefined value, try several span of EDF fiber in the mid-link to minimize the gain ripple. The obtained optimized results were also compared with simple Raman amplification. The simulation is based on the implementation of Raman propagation equations [91] and Saleh [92] model for EDFA using the semi-analytical average power analysis method [93, 94]. This method leads to quick and accurate solutions. The optimal dropping order is listed in Table 7.1. The idea is that the channels with the maximal Raman gain efficiency are dropped first and then move outwardly, as assigned in Fig. 7.1. The results for total pumping at 1W (500mW for each direction) are depicted in Fig. 7.2. The top graph displays the power after dropping spectra for optimized hybrid Raman/in line EDFA and simple Raman, while the middle one represents the optimized EDF span Fig. 7.1.Raman gain efficiency for pumping at 1480nm. The channels are represented by arrows (black –C band and red L-band). The curve was obtained by interpolation of experimental data. [95] Table. 7.1 Channel dropping order along the ring (2 channels per node) RN # f1 (THz) f2 (THz) C band 1 190.1 190.5 2 190.9 191.3 3 191.7 192.1 4 192.5 192.9 L band 5 189.3 189.7 6 188.5 188.9 7 187.7 188.1 8 186.9 187.3 109 Chapter VII. Extended band and their position in the ring (in terms of relative to the CO position ). The bottom graph displays the available pump power for the EDF spans. On looking at the simple Raman results, we verify that just by choosing a dropping order compliant with the maximal Raman gain efficiency, the power is considerably flat in the 1567-1590nm range and then decreases as the channels’ wavelengths move away from the maximal gain efficiency. On those links the insertion of an EDF span can increase the power above the 0dBm threshold. It should be noted that in order to provide gain in L band, longer spans of EDF are used. Looking at the pump power results, we notice that due to this methodology, the available pump power in distant links is still high enough to pump the EDF spans. Hence, in the hybrid approach a total span of EDF equal to 22m is used (2m for C band amplification and 20m for L band amplification) to attain a ripple of 2.54 dB over a bandwidth of 50nm. 7.3 Dual waveband remote node for extended reach full duplex 10Gb/s C + L band PON The requirements of an extended WDM/TDM PON in terms of optical budget are increased when compared to simple PONs. Moreover, the requirements for extended reach and higher splitting ratios in their TDM segments increase the demands on the system, therefore additional amplification between the OLT and the ONU becomes mandatory. Although wideband C + L band Erbium Doped Fiber (EDF) already exist [96] the indispensable passiveness demands this kind of amplification to be performed remotely. This technique of remote pumping has been already demonstrated for long haul systems in both wavebands [97] and has been introduced also to PONs for the Cband, but with the use of EDFs in the drop part of the RN [98]. For this latter case, a pump is transmitted along its fiber plant towards the RNs. For the case of a hybrid PON from ring + tree architecture, these RNs, situated at the interconnection between the WDM ring and the TDM trees, also perform the task of signal drop and insertion between ring and tree segments. In this experiment a remote node design for C+L band amplification is characterized and shown to cover the advanced optical power budget for a 55 km reach, 1:32 split hybrid passive optical network with symmetrical full-duplex 10 Gb/s transmission. Different types of Erbium-doped fibers, designed for the Cand L-band are evaluated in a bidirectional amplification stage at the remote nodes of a ring + tree network, providing 5-10 dB of power margin for the reception of downand upstream despite degradation of the optical signal-to-noise ratio. The evolution of signal power and the noise accumulation along the light path is discussed for both wavebands. This work has been performed in a collaborative experimental study with Bernhard Schrenk. Chapter VII. Extended band 116 amplification while the EDF for the L-band wavelength does not due to the higher drop losses from the ring. The upstream gain in the RN benefits from the limited net gain in the ONU, which causes low input powers for the EDFs. Slightly higher gain values can be therefore obtained even for a lower pump. However, this unsaturated gain conditions are gathered at the cost of another 3.2 dB of OSNR degradation, which could be avoided with increased ONU net gain. Consequently, most problematic in the PON is the high loss in the splitter, as there is a high OSNR degradation in the consecutive amplifying stage due to its low input power. Especially for the downstream this is critical as the SOA has a high noise figure of 6.4 (6.6) dB in addition. Keeping several input power levels moderate at the input of the amplification stages along the way from the modulator at the OLT transmitter towards the ONU, the ONSR stays at a high level but is also vulnerable to noise that is accumulated at the following amplification stages. The preamplification in the SOA of the ONU degrades the input OSNR therefore by 4.1 (5.3) dB. However, the OSNR for the downstream detection is high enough to have acceptable sensitivities for the downstream detection, as can be seen in the BER measurements shown in Fig. 7.13. The sensitivity for a BER of 10-10 in the back-to-back case, where the ring+tree fibers were replaced by their equivalent attenuation, is -24 and -24.3 dBm for the Cand L-band wavelength, respectively. When fibers and DCFs are added, an additional penalty of 3.6 and 3 dB is obtained. Therefore, a power margin of 5.4 and 5.9 dB is obtained for a BER of 10-10, large enough to cope for additional splices and other unforeseen losses that may arise over time in a PON. The upstream sensitivities for the Cand L-band are -30.2 and -30.3 dBm for the back-to-back case and a penalty of 4.6 and 2.8 dB is caused when fiber and DCFs are used. Nevertheless, for a BER of 10-10 power margins of 10.9 and 7.5 dB for the upstream are provided even without additional error correction techniques. 5.4 dB 10.9 dB -12 -11 -10 -9 -8 -7 -6 -5 -4 -3 -2 -38 -36 -34 -32 -30 -28 -26 -24 -22 -20 ONU input power [dBm] , OLT input power [dBm] log(Bit Error Ratio) L US C US C DS L DS 3.6 dB 3 dB 3 dB 3.2 dB DS FEC gain 2.8 dB 4.6 dB 5.9 dB 7.5 dB 5.4 dB 10.9 dB Fig. 7.13.BER curves for full-duplex 10 Gb/s downstream (DS, hollow markers) and 10 Gb/s upstream (US, filled markers) transmission. Single ended arrows indicate the reception margins for a reach of 56 km and a split of 1:32 in the tree. Dashed lines correspond to the back-to-back case, while solid lines indicate the presence of ring+tree fiber. ▲,● markers show measurements in the Cand L-band, respectively. 117 Chapter VII. Extended band The obtained margins could be used to increase split or distance further. The sensitivity of the downstream which is limiting this extension in split or reach, would allow to increase the splitting ratio in the tree to 1:64 or an increase in reach of about 25 km. If Forward Error Correction (FEC) with an acceptable BER of 10-4 is taken into consideration for the ONU receiver, the margin would increase to 8.4 and 9.1 dB for the Cand the L-band wavelength (Fig. 7.13). This in turn would then allow increasing the splitting in the tree up to 1:128 or the distance by 40 km. 7.3.3 Conclusion An RN design for dual waveband operation, which satisfies the requirements of an extended reach and high split WDM/TDM-PON, has been characterized. Depending on the waveband, a net gain of 8-12 dB was obtained next to incorporating dual wavelength drop functionality and resiliency for the ring-based PON. Despite the reduced OSNR of 32 dB at the OLT receiver, error-free operation can be obtained with a margin of 5-10 dB for Cand L-band, whereas the downstream transmission has been shown to be more critical With the covered wavelength range from 1530-1563 nm in the Cand 15701600 nm in the L-band, in which sufficient gain can be provided for the given power margins, a PON with data transmission on 32 wavelengths in each waveband can be considered, serving with a splitting ratio of 1:32 in the tree altogether 1024 users with a full-duplex data transmission of up to 10 Gb/s. Together with the nominal reach of 56 km for the PON, a capacity-length product of 17.9 Tb/s km is provided. When the margins are eroded for the benefit of increasing the splitting ratio or the reach, a split of 1:128 or additional 40 km of distance can be provided when error correcting codes are utilized at the downstream receiver. 118 Chapter VIII Dispersion compensation in optical networks In this chapter, we present the study of the dispersion imposed restrictions on a WDM/TDM PON. In the first part, we examine the operation of the transmitters and how the laws of physics applied both on the light sources and propagation mean explain the dispersion on the network. Next, we present two possible tools to help combating these limitations, namely, dispersion compensation fiber and electronic equalization. Then, an experimental study performed with the use of those two dispersion compensating mechanisms on a WDM/TDM PON is presented, along with the relevant results, some comments and conclusions on them. 8.1 Dispersion, source induced chirp and fundamentals of equalisation An optical signal becomes increasingly distorted as it travels along a fiber. This distortion is a consequence of intramodal dispersion and intermodal delay effects. These distortion effects can be explained by examining the behavior of the group velocities of the guided modes, where the group velocity is the speed at which energy in a particular mode travels the fiber. Intermodal dispersion or chromatic dispersion is pulse spreading that occurs within a single mode. The spreading arises from the finite spectral emission width of an optical source. This spectral width is the band of wavelengths over which the source emits light. It is normally characterized by the root-mean-square spectral width αλ. Laser diode optical sources have a spectral width of typical value of 10-4 [100]. In this study we have tested the behaviour of two different transmitters in a WDM/TDM PON. Namely, a distributed feedback (DFB) laser source, externally modulated by a Mach-Zehnder modulator and an integrated DFB - Electroabsorption Modulator (EAM), have been examined. One of the characteristics strongly affecting the performance of the transmitters is the laser frequency chirp. The fundamentals of the modulators operation are described below. The MZM is configured as an interferometer, thus it is referred to as MachZehnder Interferometer (MZI). It uses two separate beam splitters to split and 119 Chapter VIII. Dispersion compensation recombine the beams, and has two outputs. The optical path lengths in the two arms may be nearly identical or may be different, e.g. with an extra delay line. In one state, the signals in the two arms of the MZI are in phase and interfere constructively and appear at the output. In the other state, applying a voltage causes a π phase shift between the arms of the MZI, leading to destructive interference and no output signal. The chirp can be controlled very precisely in such devices. The EAM can be fabricated with the same material and techniques used to fabricate semiconductor lasers. This allows the EAM to be integrated along with a DFB laser in the same package, which results in a compact and low cost solution. The EAM uses a material such that under normal conditions, its band gap is higher then the photon energy of the incident light signal. This allows the light signal to propagate through. Applying an electric field to the modulator, results in shrinking the band gap of the material, causing the incident photons to be absorbed by the material. This effect is called Franz-Keldysh effect or Stark effect [101]. The chirping parameter can be estimated by the following expression [102]:      R chirp n4 Eq. 8.1 Where, ∆nR is the change of the real part of the refractive index, which happens with the change of absorption coefficient, when electric field is applied. ∆α is this change of the absorption coefficient. Two effects cause the phase response or chirp of the device. The modulator section presents some transient chirp due to the fact that the refractive index of the medium where the light is propagating changes when it is modulated. If the modulator were isolated from the laser section, there will not be other chirp effects. But due to reflections in the laser modulator interface there is some optical feedback that affects the laser behavior although it is being operated in CW mode. In particular, optical feedback causes a change in the threshold carrier density of the laser that originates a modulation in the effective refractive index of the laser section and a change of the lasing wavelength. This leads to some extra transient chirp and adiabatic chirp. Transient chirp is the responsible of frequency variations when the output optical signal changes and adiabatic chirp represents a change in the optical emitted frequency for different power levels (marks and spaces in a digital signal).As mentioned already, this is the main cause of dispersion in an optical fiber network. The dispersion management is a way to ameliorate the performance of an optical fiber network, thus expand its dimensions. This objective can be reached with several techniques. Two of the ways to reduce the impact of chromatic dispersion, are, using external modulation in conjunction with DFB lasers and chromatic dispersion compensation. In this experimental study and in this thesis we have tried to combine those two factors. We have combined an externally modulated source, as well as an integrated one, with dispersion compensating fiber and electronic equalization. The principles of electronic equalizers and more specifically of the model used are presented subsequently. Chapter VIII. Dispersion compensation 120 8.2 Dispersion compensating fiber The dispersion of a transmission fiber can be compensated with sections of fibers with different designs or with other optical elements. Dispersion compensation modules (DCMs) can contain, e.g., long pieces of dispersion-shifted fibers. Dispersion-shifted fibers have modified waveguide dispersion so as to shift the zero dispersion wavelength into the 1.5-μm region. This is achieved by modifying the refractive index profile of the core. Common index profiles of dispersion-shifted fibers have a triangular, trapezoidal or Gaussian shape. There are also dispersion-flattened fibers with relatively constant group delay dispersion over some wavelength range, i.e., low higher-order dispersion. They can, for example, exhibit near zero dispersion in the telecom C band. Such fibers are important for data transmission with wavelength division multiplexing and for adiabatic soliton compression. They often have a Wshaped profile of the refractive index, although profiles with a graded index and multiple steps have also been developed. In this study we have used, 2.39km of a DCF with a total dispersion compensation value of -680ps/nm, corresponding to around 40km of SSMF. 8.3 Electronic equalizer Electronic dispersion compensation can be used to counterbalance the accumulated dispersion increasing the transmission length of the system. The scheme can be adaptive to the total dispersion depending on the transmission distance. As mentioned already in chapter four, there are three main architectures for ECE and the most basic scheme is the FFE (Fig. 8.1). In this structure, after the optical-to-electrical conversion a FIR filter is added to the transmission line. The filter has several stages. Every stage, consists of a delay element, a multiplier and an adder. After every delay element, an image of the non-delayed input is multiplied with a coefficient and added to the signal. The number of filter stages used and the coefficients chosen are essential for effective dispersion cancellation. Automatic control of the filter coefficients is crucial. Fig. 8.1.The structure of an FFE 121 Chapter VIII. Dispersion compensation A second approach is the DFE appearing in Fig. 8.2. This structure is an FFE with a second FIR filter added to form a feedback loop. Again, the coefficients of both filters require active control. Today's integration levels, permit either FFE or DFE to be built into a clock-data recovery or demultiplexer chip, with an extra power requirement of about 500 mW. The tap coefficients of the filter are calculated and adjusted in an adaptive operation according to an algorithm that runs in parallel. The purpose of this is the minimization of the error. The type of the algorithm and more significantly the way that this algorithm is optimized, are particularly important in order to minimize the error and enhance the transmission properties of the system. Training and decision mode are the operating modes of the adaptive equalizer. During the training mode the algorithm adjusts to the channel characteristics and calculates the filter taps to compensate for the introduced impairments. In the decision mode, small variations in the taps allow for the compensation of the time varying effects of the channel. The LMS algorithm is the most common algorithm used, in order to calculate the taps. The goal of that algorithm is to minimize the MSE between the desired equalizer output and the actual equalizer output. It is controlled by the error signal which is derived by the output of the equalizer with some other signal which is the replica of transmitted signal. The DFE version of the equalizer is a nonlinear process that uses the same algorithm but subtracts the interference by the already detected data offering advanced performance characteristics. In this study, the examined EDC, is a FFE-DFE integrated circuit. clockclock Fig. 8.2 The structure of DFE. {CDR: Clock and data recovery module, ISI: Intersymbol Interference} Chapter VIII. Dispersion compensation 122 8.4 Experimental study Our suggested WDM/TDM PON architecture considers the use of reflective ONUs at the end users, in order to achieve colorless operation. This in general can be achieved either with the use of a dedicated CW signal fed by the optical line terminal (OLT) [103] – [105] or by re-modulating the down-stream propagated signal. In the former case the maximum achievable performance of the reflected ONUs is exploited but bandwidth utilization is reduced to half since each TDM PON tree requires two wavelengths. Furthermore the CAPEX of the network increases as the number of the required optical sources is higher. . In the latter case, the upstream signal is modulated on top of the downstream signal, which in turn must have an increased DC power level (i.e. low extinction ratio (ER)), in order to assure an adequate power level at the ONU for re-modulation [105] and to avoid residual crosstalk on the up-stream data. The increased DC bias corresponds to an increased optical power offset being applied to the downstream signals. Despite the optimized bandwidth utilization, due to the use of a single wavelength for upstream and downstream transmissions, the transmission performance of the downstream signal must be intentionally reduced, in order to allow the re-modulated upstream signal to be transmitted efficiently. Therefore, despite the optimized bandwidth utilization, due to the use of a single wavelength for upstream and downstream transmission, the transmission performance of the downstream signal is significantly reduced. Evidently, and as it has been studied in [105], there is a compromise between the downstream ER and the overall transmission distance that can be achieved. When the ER of the downstream signal is reduced, the performance of the re-modulated upstream signal is increased. At the same time the performance of the downstream signal is reduced. By applying certain performance improvement schemes, either in the optical [106] or electronic domain [107], or both combined, the performance of the downstream signal with reduced ER can be significantly increased. In this case, an upstream signal with increased performance can be generated after re-modulation. It is noted that latest advances in coherent modulation systems are applied in PONs [108], however, due to the extensive use of electronics that is required, such schemes are still too expensive to be complemented at the ONU. The general purpose of the work performed, is to study experimentally, different transmission performance improvement methods applicable in 10Gb/s optical signals, Fig. 8.3.Digital implementation of MLSE. (ADC: Analog – to digital converter) 123 Chapter VIII. Dispersion compensation with typical and reduced ER values, generated by two different types of transmitters, with different chirp characteristics. More specifically, a MZM and an EAM type of transmitter have been considered and operated at two different sets of driving voltages, resulting in ER values of 3dB and 9dB. The signal improvement methods considered were based on: a) the use of fixed dispersion compensation fiber able to compensate for half the maximum transmission distance, b) the use of electronic equalization and c) the combination of the aforementioned techniques. The performance was evaluated for typical, moderate and long reach PON applications up to 100km, for both types of transmitters and ER values and for all the signal improvement methods. The two values of ER that have been considered in this study (3dB and 9dB), are proven to be appropriate for use in WDM/TDM PON in combination with colourless ONUs. First, a signal modulated with a high value of ER, namely 9dB, was tested in the testbed created. This signal could be used in combination with a CW dedicated to upstream remodulation. The value of ER =9dB has been shown to be appropriate for all optical cancellation in the case of colourless ONUs, as in [109]. Second, the case of low ER, namely 3dB, is shown to be ideal for efficient signal remodulation [106], but also for the more advanced feed forward cancellation method presented in [110]. In the case of the 3dB ER, the reduced receiver sensitivity, due to the poor downstream signal quality, becomes a major issue in long reach PONs, mainly because of the chromatic dispersion penalty. This penalty can be mitigated with the use of either DCF, or electronic dispersion compensators (EDCs). In long reach PONs, due to the specific network design considerations and in order not to increase the signal power losses, it is possible to use the DCF as pre-compensator at the OLT side of the PON, providing partial compensation around half of the maximum transmitted distance. The wide operating bandwidth of DCFs allows the sharing of the DCF module among all users in a WDM/TDM PON, reducing significantly the additional cost per user. On the other hand, EDC can be used only as post-compensator, at the receiver side; therefore it must be included at each ONU increasing its cost. However, EDC offers adaptive dispersion compensation with respect to the transmission distance and optimizes the transmission performance particularly for long distances. The extra cost of the EDC can thus be counterbalanced by the consequent reach extension of the PON. It should be noted that currently, WDM/TDM PONs (e.g. Next generation PONs-NGPON2) use tunable filters in ONUs as an alternative way to achieve colourless operation, we have taken under consideration though, that a possible increase in the number of wavelengths used, could render the use of RSOA, more cost effective. In the case of the 9dB ER signal, the behaviour of the network is intensely affected by the existence of chirp. When no chirp is present, it is possible to reach the maximum distance using no means of signal improvement. Overall, it has been observed that with both transmitters – regardless of the chirpthe application of such methods evidently offers a much more uniform behaviour in terms of OSNR vs. BER for several lengths of the network. Therefore it is apparent that, from a practical point of view, the design of a long reach PON with improved performance can become Chapter VIII. Dispersion compensation 124 problematic when transmitters with different chirp characteristics are considered. In this work, the performance of 10Gb/s MZMand EAM-based sources, operated at two different low-driving voltages (resulting in 3 and 9dB of ER correspondingly), considering additionally, both, the aforementioned dispersion penalty and the improvement methods of partial pre-compensation with DCF and adaptive postcompensation with EDC, is practically evaluated. 8.4.1 Experimental set-up The complete experimental setup for all the examined cases is shown in Fig.8.4. The two types of low-driving voltage sources examined are: a) An externally modulated DFB -by the use of a MZMsource, operating at 1541.35nm with peak-to-peak voltage (VP-P) of 1.2V and amplifying bias voltage (Vbias) of 1.3V, being applied at the MZM for the 3dB ER and 2V of VP-P with 1.3V of Vbias, to achieve 9dB ER. b) An integrated DFB-EAM source, emitting at 1539.46nm and being driven at the EAM section via an external bias-T at 530mVP-P and Vbias of - 2V for 3dB ER and at 2VP-P with Vbias of -1.7V for 9dB ER. A 9.952Gb/s signal with 231-1 long PRBS pattern was used and was launched directly by the pattern generator to the modulators, without the use of electrical amplifiers. The VP-P and Vbias values in each case resulted in optical signals of 3 and 9dB ER. The signals are then launched on variable lengths of standard single mode fiber (SSMF), ranging from 0km (back-to-back (b2b)) to 100km. The launch power was fixed at 6dBm in all cases. Before the receiver, an OSNR emulator is used, consisting of a variable optical attenuator (VOA) and an EDFA, in order to alter the level of OSNR for the measurements and emulate the extender box or remote amplification process in long-reach PONs. At the receiver end, the signal was filtered with a 0.8nm Gaussian filter and received by a p-i-n photodiode (PIN) followed by a trans-impendence amplifier (TIA). The receiver input power at the PIN was always adjusted via an optical attenuator at -10dBm. The two compensating schemes considered are indicated with the grey shaded areas in Fig.1. For the case of partial 10Gb/s PPG PRBS: 231-1 ER monitor ER = 3dB OR DFB MZM DFB - EAM L C The two types of Low driving voltage Tx (<2VP-P) V OLT Tx SSMF 0 …100 km ONU Rx OSNR monitor OSNR Emulator PIN+TIA 10% 10Gb/s BERT Clk Rcvr Data Clock Fixed DCF -680ps/nm FFE DFE Tap Controller Equalizer FFE DFE Tap Controller Equalizer 10Gb/s PPG PRBS: 231-1 ER monitor ER = 3dB OR DFB MZM DFB - EAM L C The two types of Low driving voltage Tx (<2VP-P) V OLT Tx SSMF 0 …100 km ONU Rx OSNR monitor OSNR Emulator PIN+TIA 10% 10Gb/s BERT Clk Rcvr Data Clock Fixed DCF -680ps/nm FFE DFE Tap Controller Equalizer FFE DFE Tap Controller Equalizer Fig. 8.4. Experimental set up for the performance evaluation of low-driving voltage DFB-MZM and DFB-EAM transmitters, over 0 to 100km links and for two dispersion compensating schemes DCF partial pre-compensation and FFE-DFE post compensation 125 Chapter VIII. Dispersion compensation optical pre-compensation, a fixed length of DCF was used before the booster EDFA at the output of the OLT, with a total dispersion compensation value of -680ps/nm, corresponding to around 40km of SSMF. For the case of post-compensation with EDC, an integrated 5-tap FFE and a 2-tap decision-feedback equalization DFE circuit was used and controlled by properly adjusting the taps for optimum BER. A BER tester was connected, either directly on the PIN-TIA output, or the EDC output for the cases of DCFand EDC-based compensation, respectively. The reference clock for the measurements is provided either via an external clock recovery circuit, or directly by the EDC circuit, depending on the compensation method. Finally, it should be noted that in the results, presented next, the performance has been studied in terms of BER versus OSNR, since in long-reach access-metro networks, the use of amplification (either with remote amplification schemes [103], or with optical amplifiers [104]) is considered in the network. In these cases, the power budget issues are resolved, but the input signal level in the amplification sections, as well as the type of amplification, affect the signal’s OSNR. 8.4.2 Results For nearly zero-chirped, externally modulated sources (DFB-MZM), the increase of ER, highly decreases the required OSNR (ROSNR), for an error-free transmission without FEC (BER=10-9) [111], as can be seen in Fig.8.5. Even though in commercialized 10Gb/s PONs the use of FEC is nowadays mandatory, in our effort to give a complete study of the behaviour of the system, performance for both BER=10-9 and BER=10-3 has been investigated. Moreover, in Fig.8.5 we observe that when no compensation or equalization is applied, error-free performance is possible, with an ER of 9dB, for lengths that it was impossible with an ER of 3dB. With the use Fig. 8.5. BER vs. ROSNR for BER=10-9 for zero chirped DFB-MZM transmitter in the cases of: no dispersion compensation, DCF only, EDC only, DCF and EDC when signals of 3 and 9dB ER are transmitted Chapter IX. Combination 132 The signals were dropped with the use of three-port thin-film filters before entering the corresponding in-line remotely pumped EDFA. When this RN design (Fig. 9.1) is used for the implementation of the WDM/TDM PON, the signal serving each tree needs to have the adequate power to reach the furthest ONU of the tree, without receiving extra amplification in the RN in which it is being dropped. In other words: if S1 is the signal being dropped in the first TDM tree and RN1 the RN serving this tree, then the signal S1 will be dropped with the use of a filter, before entering the in-line EDFA situated in this RN. This becomes clearer in Fig. 9.1. OLT RN1 RN2 RN3 25km 25km 25km OLT RN1 RN2 RN3 25km 25km 25km 25/50km OLT RN1 RN2 RN3 25km 25km 25km OLT RN1 RN2 RN3 25km 25km 25km 25/50km Fig. 9.2. WDM/TDM PON design operating in C + L band with the use of remote amplification of EDFAs placed in-line with the propagating signals US +Pump US DS A/D A/D A/D A/D DS 1480 / C+L Remote EDFA Pump 50/50 50/50 50/50 RN to / from Ring to / from Ring to/from Tree x/100-x US +Pump US DS A/DA/D A/DA/D A/DA/D A/DA/D DS 1480 / C+L Remote EDFA Pump 50/50 50/50 50/50 RN to / from Ring to / from Ring to/from Tree x/100-x Fig. 9.1.Remote node design operating in C + L band with the use of remote amplification of EDFAs placed in-line with the propagating signals 133 Chapter IX. Combination On the contrary, the rest of the signals, namely S2, S3, …SN that are dedicated to the TDM trees T2, T3, … TN correspondingly, will pass through the EDFA placed between the two thin-film filters in the RN1, all of them being amplified simultaneously. In this experiment we consider a double-fiber WDM ring with 3 RNs positioned in a distance of 25km from the CO, 50km and 75km correspondingly, as they appear in Fig. 9.2. This design is transformed, in the worst case of a fiber cut (e.g. between the OLT and RN1 or RN3), into a 75km double-fiber WDM trunk. Furthermore, in an attempt to extend the limits, a RN has been placed in a 100km distance from the CO for the last set of measurements. The TDM trees have 6km reach and several splitting ratios have been examined. The ONU receivers are implemented with the use of APDs and in some cases the use of an integrated 5-tap feed-forward equalization (FFE) and a 2-tap decisionfeedback equalization (DFE) circuit is considered for the equalization of the signals. The ONU transmitter is implemented with a continuous wave (CW) transmitter, a REAM and an RSOA. 9.2 Experimental set-up The system investigated consists of seven laser sources emitting on the wavelengths: 1541.1, 1549, 1557, 1571, 1586.2, 1598 and 1600nm, correspondingly. Three RNs, especially designed for in-line amplification in both C+L band (Fig. 9.1.), where placed in distances of 25, 50 and 75km, respectively, as can be seen in Figures 9.2 and 9.3. As a maximum reach evaluation, a 100km WDM ring was tested. That was implemented by increasing the distance between RN2 and RN3 by another 25km. The downstream signal is modulated at the CO on a CW signal with the use of a MachVOA (Splitter) CO PPG λD 1 λD 1/99 2-7 C omb S ignal OSA B oos ter MZM C/L C/L VOA Pump Pump P ump AWG 1480 / C+L 25km DS Ring US Ring 25km 25km Dual-fiber WDM R ing 1557 1586 1549 Dual-feeder T ree 6km 6km 1/99 OSA ONU BERT APD 50/50 1 CW PPG ONU P reamplifier 1/99 OSA VOA F ilter BERT PIN VOA 3 express channels E qualiz er VOA (Splitter) VOA (Splitter) CO PPG λD λD 1 λD λD 1/99 2-72-7 C omb S ignal OSA B oos ter MZM C/L C/L VOA Pump Pump C/L C/L VOA Pump Pump C/L C/L VOA C/L C/L C/L C/L VOA Pump Pump P ump AWG 1480 / C+L 25km DS Ring US Ring 25km 25km Dual-fiber WDM R ing 1557 1586 1549 Dual-feeder T ree 6km 6km 1/99 OSA 1/99 OSA ONU BERTBERT APD 50/50 1 CW PPG ONU P reamplifier 1/99 OSA 1/99 OSA VOA F ilterF ilter BERT PINPIN VOA 3 express channels E qualiz er Fig. 9.3. Experimental set up used to test a WDM/TDM PON operating in C + L band with the use of remote amplification of EDFAs placed in-line with the propagating signals Chapter IX. Combination 134 Zehnder modulator (MZM) at 10Gb/s, using a PRBS 231-1. The downstream has been implemented as continuous data stream. The modulated channels under study were coupled together with the remaining wavelengths forming the dummy comb, using a 40channel AWG with 100GHz grid. The modulated channels were namely: one emitting at 1549nm with 2.4dBm, one at 1557nm with 3.6dBm and one emitting at 1586.2nm with 3dBm. The dummy comb consists of wavelengths of both C and L band. More specifically, we have used the following wavelengths: 1541.1nm emitting with a power of -0.6dBm, 1571nm emitting with 2dBm, 1598nm emitting with 4dBm, 1600nm emitting with -4dBm. These signals have not been modulated, just multiplexed with the 3 modulated signals that were dropped in the 3RNs of the network. After the AWG a C/L combiner splits both bands for the booster stage, where a dedicated booster for each band is used. After the booster stage a C/L combiner is used to combine again both bands and launch the comb in the downstream ring. A variable optical amplifier (VOA) controls the total output power. The launch power was 4dBm per channel. No means of dispersion compensation were used at the OLT. 135 Chapter IX. Combination EDF length = 5m 0 5 10 15 20 25 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-10dBm Pp=10dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm EDF length = 3 m 0 5 10 15 20 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-10dBm Pp=10dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm EDF length = 5m 0 5 10 15 20 25 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-10dBm Pp=10dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm EDF length = 3 m 0 5 10 15 20 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-10dBm Pp=10dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm Fig. 9.4. ) Characterization of the EDF used in the amplification stage of the RN, for a) 3m, b) 5m of length and 0dBm, 5dBm,10dBm of input power and various pump power values of 10dBm, 13dBm, 16dBm and 19dBm respectively The dual fiber 75km WDM ring is made of 6 SMF spools of 25kms, 3 for downstream and 3 for upstream. This set-up corresponds to a 75km trunk with 3RNs in the worstcase resiliency scenario (i.e. fiber cut between OLT and RN1 or RN3). The RN drop channels are distributed as follows: 1549nm is dropped at 25km, 1586.2nm at 50km and 1557 at 75km. The TDM tree is composed by 6km of dual-fiber feeder to avoid strong Rayleigh backscattering at the bidirectional tree. A VOA is used to emulate the splitter stage between RN and ONUs. Chapter IX. Combination 136 An avalanche photodiode (APD) is used as the ONU receiver. The signals are equalised with the use of an FFE-DFE equaliser in half of the measurements. As ONU transmitter, a CW seed light was used in combination with a REAM and a RSOA. The upstream was modulated at 10 Gb/s in order to demonstrate symmetric data transmission. The upstream path is symmetrical to the downstream one. At the OLT input, a WDM waveband combiner is used to inject the counter propagating remote pump at 1480nm. The signal is then split in the two bands, C and L, with a C+L splitter. A band-dedicated optical preamplifier is then used to enhance the reception sensitivity. Finally, the channel under study is selected with the help of a tuneable optical bandpass filter and detected by a PIN diode. The RN is a key element in order to increase the reach of the proposed topology as well as a key element for enhancing the resilience performance. The suggested design is based on the SARDANA RN [1]. The add/drop functionality of the tree wavelengths from the ring into the sub-RNs was performed with 3-ports thin-film add/drop filters, while resiliency is provided by a 50/50 coupler (CR) that allows to receive and transmit the data stream to both directions of the downand upstream ring for the case of a fiber EDF length = 8m 0 5 10 15 20 25 30 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-10dBm Pp=10dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm EDF length = 15m 0 5 10 15 20 25 30 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm EDF length = 8m 0 5 10 15 20 25 30 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-10dBm Pp=10dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm EDF length = 15m 0 5 10 15 20 25 30 1530 1540 1550 1560 1570 1580 1590 1600 1610 Wavelength (nm) Gain (dB) Ps=-10dBm Pp=19dBm Ps=-10dBm Pp=16dBm Ps=-10dBm Pp=13dBm Ps=-5dBm Pp=19dBm Ps=-5dBm Pp=16dBm Ps=-5dBm Pp=13dBm Ps=-5dBm Pp=10dBm Ps=0dBm Pp=19dBm Ps=0dBm Pp=16dBm Ps=0dBm Pp=13dBm Ps=0dBm Pp=10dBm Fig. 9.5. ) Characterization of the EDF used in the amplification stage of the RN, for a) 8m, b) 15m of length and 0dBm, 5dBm,10dBm of input power and various pump power values of 10dBm, 13dBm, 16dBm and 19dBm respectively 137 Chapter IX. Combination cut. Since a dual-feeder is used at the tree, no circulator is required at the RN to separate downstream and upstream. This circulator can be rather found at the VOA emulating the tree splitter. For the proposed RN design an in-line amplification approach has been used in order to amplify the upstream or downstream signals. The EDF is placed between the two thin-film filters, connecting both express ports, amplifying in this way all the channels, except the one that is being dropped in the RN (Fig. 9.2). When working with Cand L-band at the same configuration, the usual approach is to split both bands using a C+L coupler and amplifying each band independently, with a specific EDF for each one, as in [113]. This design has some drawbacks, such as an increase in both the complexity and the cost of the designed RN by including C+L splitters, as well as doubling the number of amplifiers requiring extra pump power. In this experiment, this drawback has been overcome by the use of an EDF appropriate for the amplification of both Cand L-band signals and the appropriate wavelength allocation in the network. The wavelength allocation was performed by means of appointing the right wavelength drop, according to the gain each frequency band receives, while taking into account the distinctiveness of this RN design. This distinctiveness lays on the fact that the wavelength dropped has received amplification in the previous amplification stages, but needs to have adequate power to reach the ONU of the RN in which it is being dropped, without receiving any extra gain in this specific RN. In Figures 9.4 and 9.5 one can see the EDF characterization as a function of the signal wavelength and the pump power at 1480nm, for four different EDF lengths and four different signal power levels. As can be observed, even with low pump power values such as 10 or 13 dBm, the EDF can provide gain in the whole spectrum of C+L bands. From Figures 9.4 and 9.5 it is observed that the L-band signals receive less gain than the C band ones, therefore we have dropped the one belonging in L band (1586.2nm) in RN2, placed in the middle of the network, as in this case, the signal is being amplified when going through RN1 and on the same time does not need to travel the whole distance of the network but a moderate one. The two signals belonging to Cband are dropped as following: the one in RN1 (1549nm) without going through any amplification stage and the second in RN3 (1557nm) going through two amplification stages so as to ensure adequate power to reach the last ONU (at 100+6km). The required gain is around 14dB in both EDFs (upstream and downstream) in RN1 and RN2 while the launch pump power is 30dB and when reaching the two RNs it is 24dB and 17dB correspondingly. Therefore, from Fig. 9.4, we have chosen a 3m EDF for RN1 and 5m EDF for RN2. It is also observed, that with low pump power values the gain tilt is rather flat, in contradiction with the case of high pump power such as 16-19dBm. Since we are using a remote pump, this small tilt towards the C-band can be compensated by the extra Raman gain expected for the L-band, as the maximum Raman gain will be located at 1580nm when using a pump at 1480nm [114]. With the use of this elaborate network design, theoretically, the number of the users can be doubled, by the use of both C and L bands, allocating the signals belonging to the second one, to the RNs placed in the middle of the WDM ring. On the contrary, the Chapter IX. Combination 138 signals belonging to C-band, that apparently receive more gain when going through the several amplification stages, which is in fact accumulated, are allocated in the RNs placed near the CO, which transform into the more distant RNs in the worst case of a fiber-cut, as mentioned before. 9.3 Results The MZM was operated at two different sets of driving voltages, resulting in ER values of 3 and 10dB respectively. A value of ER =9dB has been shown to be appropriate for all optical cancellation in the case of colourless ONUs in [109]. Second, the case of low ER, namely 3dB, is shown to be ideal for efficient signal remodulation [115], but also for the more advanced feed forward cancellation method presented in [116]. Therefore, and based on these previous results, we have used two similar values for the signal modulation. In the results, presented in Fig. 9.6, the performance has been studied in terms of BER versus received power. Measurements were taken in the ONU part for the following cases. First we consider RN1, which is situated 25km away from the OLT, where the wavelength of 1549nm is being dropped and the results shown in Fig.9.6 correspond to an ONU situated 6km away from the OLT, therefore in a total distance of 31km from the OLT. Next, an ONU of RN2 at 50+6km with drop wavelength of 1586nm was examined and subsequently we have examined RN3 (1557nm drop wavelength) and its 1557nm_100km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 3NoEDC 3EDC 1557nm_75km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 10EDC 3NoEDC 3EDC 1586nm_50km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 10EDC 3NoEDC 3EDC 1549nm_25km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 10EDC 3NoEDC 3EDC 1557nm_100km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 3NoEDC 3EDC 1557nm_75km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 10EDC 3NoEDC 3EDC 1586nm_50km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 10EDC 3NoEDC 3EDC 1549nm_25km -12 -10 -8 -6 -4 -2 -30 -25 -20 -15 -10 Received power (dBm) log (BER) 10NoEDC 10EDC 3NoEDC 3EDC Fig. 9.6. BER vs. received power for a)RN1 (λ=1548nm, 25km), b)RN2 (λ=1586.2nm, 50km), c) RN3 (λ=1557nm, 75km), d) RN3 (λ=1557nm, 100km), in the cases of i)10dB ER and no EDC use, ii) 10dB ER and EDC use, iii) 3dB ER and no EDC use, iv) 3dB ER and no EDC use 139 Chapter IX. Combination ONU for two different distances. In this framework, the RN3 is situated 75km away from the OLT (therefore its ONU is situated at 75+6km) and then with the use of 25km of SMF, we place the RN3 at a distance of 100km from the OLT (subsequently the ONU is placed in a distance of 100+6km from the OLT). Observing the results for the downstream propagation, it is obvious that the MZM modulated signal has a better performance when modulated on a high ER [112], in this case 10dB, than in the case of a low 3dB ER. Furthermore, the EDC equalizer offers a decrease in the range of 3-5dB of required received power in order to have an error free (BER=10-10) performance. One can notice in the results the existence of chirp even if we have used a MZM modulator. As it has been remarked in [112] and as one can observe from the results, there is a trade-off between a high ER and dispersion control. This is evident in the case of 10dB ER and long distance (RN3-100km) where the dispersion effect in combination with the small but existent chirp is so high that does not allow for acquiring measurements while using EDC. On the contrary the 3dB ER signal arrives at the 106km with less dispersion as the chirp counteracts with the dispersion caused by the propagation of the signal in the fiber. The behaviour of the 3dB ER signal, is again anticipated as in the case of 25km and C-band signal, the EDC offers a small improvement, but when moving in the 50km, without any dispersion compensation and with a larger chirp parameter since the signal belongs in the L-band [117] we observe that a) the results have a great difference from the 10dB ER ones, b) especially in the case of no EDC the performance of the network is much lower and the EDC offers good results in controlling the dispersion. Nevertheless, as can be observed from the results and shown in [116], the use of EDC is not compulsory in order to achieve results that can be used for proper remodulation while using RSOA in the ONU. 9.4 Conclusions We have combined in an experimental set-up and an experimental study, the ideas that have been examined throughout this thesis achieving to reach by means of in-line amplification and C+L-band utilization, 106km in a WDM/TDM ring PON architecture, adding this way 6km to our previous studies, while at the same time increasing the number of users. The capacity of the network with the use of the C + L band and the proper design in the sense of wavelength allocation can theoretically be doubled. This is a subject for further investigation though. The special RN design has helped the realization of this goal. The use of equalization has been examined and it does increase the performance of the network especially in the case of a larger ER in a scale of 4-6dB. Nevertheless, even the 3dB ER without the use of equalizer can offer an acceptable performance, so the cost of an equalizer at least at the ONU part, can be avoided. Further studies should take place for the examination of the equalizer at the OLT part. The C+L band use can offer both an increase on the number of end customers as well as the possibility in infrastructure proprietors to lease one fiber to more than one operator. Chapter IX. Combination 140 The suggested design can render the network much more competitive, giving this way an impetus to the whole technology. Chapter X Conclusions This chapter summarizes the knowledge acquired throughout this thesis project and presents suggestions for possible future work. 10.1 Conclusions The aim of this study was to shed light on the economic and technical aspects of the emerging technology of WDM/TDM PON FTTH networks. To evaluate, in terms of cost, those networks position in the global telecom market; and to suggest as well as to investigate ways to improve their performance. A study on the economic and technical aspects of the FTTH networks, has shown that the most cost effective solution is the one of WDM/TDM PON implemented in P2MP architecture. More specifically, a model has been created that compares -using OSP, active equipment and implementation costseveral architectures and technologies. The possible implementations were presented along with some technical details on the FTTx infrastructure and a description of the model. To give a more complete and thorough estimation, we have compared the costs of different FTTx infrastructures. When estimating the OSP cost of a FTTH metro network with EP2P on the access part with the metro-access FTTH solution of WDM/TDM PON we conclude that there is a 40% reduction when we choose the second option, while in the case of a FTTH metro network with GPON the decrease is one of 20%. When we compare the overall cost of the three FTTH implementations (OSP and active equipment cost) we end up in the result that the PON is 20% less costly than the active solution while the WDM/TDM PON has a maximum cost of approximately 2000€ in the most scarcely populated area, while the prices decrease for every other population density. In general, it is evident that the most cost saving technology in terms of OSP cost is the WDM/TDM PON. The explanation for that is the competent use of the fiber resources, as the feeder and the distribution part, is shared between a higher number of end-users than in any other technology. In continuance, a presentation of the network design used as a reference for WDM/TDM PONs takes place, along with the challenges such a design presents. These are namely signal attenuation, chromatic dispersion and non linear effects. The suggested solutions for overcoming limitations in a fully passive optical network 141 B. Research Publications B.1 Publications in International, Peer-Reviewed Journals 1) S.Chatzi , J.A.Lazaro, J. Prat and I.Tomkos, “A Techno-economic study on the outside plant cost of current and next generation FTTx deployments”, Fiber and Integrated Optics, vol.32, pp.12-27, Feb. 2013. 2) S.Chatzi, C.P. Tsekrekos, D. Klonidis, J.A.Lazaro, and I.Tomkos,”Experimental evaluation and improvement methods for low cost transmitters in long reach PONs”, Optical Fiber Technology, Elsevier. 3) B. Schrenk, S. Chatzi, F. Bonada, J.A. Lazaro, I. Tomkos, and J. Prat, “Dual Waveband Remote Node for Extended Reach Full-Duplex 10Gb/s Hybrid PONs”, IEEE/OSA J. Lightwave Technol., vol. 28, pp. 1503-1509, May 2010. 4) J. Girão; B. Neto; Rocha, Ana M.; Reis, C.; Dionísio, R.P.; S. Chatzi; F Bonada; J. Lazaro; Teixeira, A.T.; André, P.S; "C+L band extended reach amplified next generation access networks", Microwave and Optical Tech. Letters, Vol. 53, No. 10, pp. 2414 - 2418, October, 2011. 5) B. Schrenk, J.A.Lazaro, D. Klonidis, F. Bonada, F. Saliou, E.Lopez, Q.T. Le, P. Chanchlou, L. Costa, A. Texeira, S. Chatzi, I. Tomkos, G.Tosi Beleffi, D. Leino, S.Spirou, G. de Valicourt, R. Brenot, C. Kazmierski and J. Prat, “Demonstration of a remotely Dual-Pumped Long reach PON for flexible deployment”, J. Lightwave Technol., vol. 30, pp. 953-961, Apr. 2012. B.2 Publications in Scientific Congresses 1) S. Chatzi , D. Klonidis, J. A. Lazaro, J. Prat and I. Tomkos, “Design of inline remote amplification for an extended WDM-PON ring architecture”, in Proc. NOC, 12.4, Valladolid, Spain, Jun. 2009. 2) S. Chatzi , I. Tomkos, J. A. Lazaro and J. Prat, “L-band in-line remote amplification for an extended WDM-PON ring architecture”, in Proc. ICTON, Tu.D5.5, Azores Portugal, Jul. 2009. 3) S. Chatzi , J. A. Lazaro, J. Prat and I. Tomkos, “Techno-economic comparison of current and next generation long reach optical access networks”, in Proc. CTTE, Ghent, Belgium, Jun. 2010. V Appendix B. Research Publications VI 4) S. Chatzi , J. A. Lazaro, J. Prat and I. Tomkos, “A quantitative technoeconomic comparison of current and next generation metro/access converged optical networks”, in Proc. ECOC, We.8.B.2., Torino, Italy, Sep. 2010. 5) S. Chatzi , C. Tsekrekos, D. Klonidis, and I. Tomkos, “Performance evaluation and improvement methods for low-driving voltage transmitters in long reach PONs”, in Proc. OSA / OFC/ NFOEC, L.A., (CA), USA, March 2012. 6) S. Chatzi , I. Tomkos, “Techno-economic study of high–splitting ratio PONs and comparison with conventional FTTH-PONs/FTTH-P2P/ FTTB and FTTC deployments”, in Proc. OSA / OFC/ NFOEC, L.A., (CA), USA, March 2012. 7) J. A. Lazaro, J. Prat, V. Polo, M. Omella, F. Bonada, B. Schrenk, D. Klonidis, S. Chatzi, I. Tomkos, P. Chanclou, “Scalable WDM/TDM extended reach access network architecture”, in Proc. OPTOEL, Malaga, Spain, 2009. 8) B. Schrenk, S. Chatzi, F. Bonada, J.A. Lazaro, D. Klonidis, I. Tomkos, and J. Prat, “C+L Band Remote Node for Amplification in Extended Reach Full-Duplex 10Gb/s WDM/TDM Passive Optical Networks”, in Proc. ECOC, We.P6.19, Vienna, Austria, Sept. 2009. 9) B. Neto, R. P. Dionísio, A. M. Rocha, C. Reis, S. Chatzi, F. Bonada, J. A. Lazaro, A. L. J. Teixeira, and P. S. André, “C+L band extended reach next generation access networks through raman amplification: assessment in rural scenario,” in OptoElectronics and Communications Conference, 2010. OECC 2010. 15th, 5-9 2010, pp. 1 –2. 10)B. Neto, A.M. Rocha, J.P. Girao, R.P. Dionisio, C. Reis, S. Chatzi F. Bonada, J. A. Lazaro, A. L.J. Texeira, P.S.Andre,. “C+L band gain equalization for extended reach WDM-ring PON using hybrid Raman/inline EDFA amplification”, in Proc. ICTON, We.P.18, Munich, Germany, Jun. 2010. 11)J. Bauwelinck, C. Antony, F. Bonada, A. Caballero, S. Chatzi, A.M. Clarke, L.N. Costa, M. Forzati, J.A. Lazaro, A. Maziotis, M. Mestre, I.T. Monroy, P. Ossieur, V. Polo, J. Prat, X.Z. Qiu, P.J. Rigole, B. Schrenk, R. Soila, A. Teixeira, I. Tomkos, P.D. Townsend, X. Yin, and H. Avramopoulos, “Optical Line Terminal and Remote Node Sub-Systems of Next-Generation Access Networks”, in Proc. OSA Advanced Photonics Congress / ANIC, AWA5, Karlsruhe, Germany, Jun. 2010. 12)F.Bonada, B.Schrenk, L.Costa, A.Teixeira, S.Chatzi, D.Klonidis, I.Tomkos, J.Prat and J.A.Lazaro, “Wavelength-tuneable remote node for enhanced resilience and optimization of WDM access networks”, in Proc. ICTON, Tu.B.6.5., Stockholm, Sweden, Jun. 2011. 13)J.Prat, J.A.Lazaro, S.Chatzi and I.Tomkos, “Techno-economics of resilient extended FTTH PONs” (Invited), in Proc. ICTON, Tu.A.6.2, Stockholm, Sweden, Jun. 2011. 14)B. Neto, A. Rocha, J. P. Girão, R. P. Dionísio, C. Reis, S. Chatzi, F. Bonada, J. Lazaro, J. A. and, and A. L. J. Teixeira, and P. S. André, VII Appendix B. Research Publications “Comparative analysis of hybrid in line edfa/raman with simple Raman amplification in WDM ring PON for C+L band,” in Networks and Optical Communications, 2010. NOC ’10. 15th European Conference on, vol. 1, June 2010, pp. 15)B. Schrenk, J.A. Lazaro, D. Klonidis, F. Bonada, F. Saliou, E. Lopez, C. Trung, P. Chanclou, L. Costa, A. Teixeira, S. Chatzi, I. Tomkos, G. Tosi Beleffi, D. Leino, R. Soila, S. Spirou, G. de Valicourt, R. Brenot, C. Kazmierski, and J. Prat, “Demonstration of a Remotely Pumped LongReach WDM/TDM 10 Gb/s PON with Reflective User Terminals”, Proc. ECOC Technical Digest © 2011 OSA, Geneva, Switzerland, Sept. 2011. B.3 Non-peer reviewed publications In international magazines oI. Tomkos, S. Chatzi, “Techno-economic Comparison of Next Generation Access Fiber-to-the-X Network Architectures”, Total Telecom magazine, Dec/Jan 2011. http://headley.co.uk/headturner/TTP1210 pp.11-13 B.4 Workshop presentations oS.Chatzi , I.Tomkos, “Hybrid WDM/TDM scablabe ring based PON with in-line remote amplification”, Workshop on “Activities and perspectives of EURO-FOS concept”, paper P16, 28 March 2011. VIII C. Bibliography [1] J.A. Lazaro et al, “Remotely amplified SARDANA: Single-fibre tree Advanced Ring-based Dense Access Network Architecture”, Proc. ECOC’06, Cannes, France, Sept. 2006. [2] FTTH Council Europe, Press conference, Munich, Germany, 15 February 2012. [3] FTTH Council Europe, “Winners and losers emerge in Europe’s race to a fibre future”, London, UK, 20 February 2013. [4] FTTH Council Europe, “The FTTH Council Europe welcomes French ultra-fast broadband infrastructure investment plan”, 25 February 2013, London, UK. [5] A. Vegara et al, “COSTA, a model to analyze next generation broadband access platform competition”, 14th International Telecommunications Network Strategy and Planning Symposium (NETWORKS), pp. 1-6, Warsaw, Poland, Sep. 2010. [6] T. Rokkas et al., “Techno-economic Evaluation of FTTC/VDSL and FTTH RollOut Scenarios: Discounted Cash Flows and Real Option Valuation”, IEEE/OSA J. Optical Communication Networks, vol. 2, pp. 760-772, Sept. 2010. [7] R. Zhao et al., “Dynamic Migration Planning towards FTTH”, in Proc. NETWORKS’10, Warsaw, Poland, Sept. 2010. [8] C.P. Larsen et al, “Comparison of Active and Passive Optical Access Networks”, Proc. CTTE’10, Ghent, Belgium, Jun. 2010. [9] S. Chatzi et al., “A quantitative techno-economic comparison, of current and next generation metro/access converged optical networks”, Proc. ECOC’10, We.8.B.2, Torino, Italy, Sep. 2010. [10] T. Rokkas et al., “Economics of Time and Wavelength Domain Multiplexed Passive Optical Networks”, IEEE/OSA Journal of Optical Communications and Networking, vol.2, no.12, pp.1042-1051, Dec.2010. [11] B. Lannoo et al., “Techno-economic feasibility study of different WDM/TDM PON architectures”, Proc. ICTON’10, Mo.C4.3, Munich, Germany, Jun. - Jul. 2010. [12] www.ist-pieman.org [13] www.ist-muse.org IX Appendix C. Bibliography [14] A. Geha et al., “HARMONICS, an IP based service network over hybrid fibreaccess network supporting QoS”, Proc ICT’03, Tahiti, France, Feb. - Mar. 2003. [15] F.T. An et al., “SUCCESS-HPON: A next-generation optical access architecture for smooth migration from TDM-PON to WDM-POM”, IEEE Communications Magazine, vol.43, no.11, Nov. 2005. [16] www.e-photon-one.org [17] Y. Cao et al. “A novel architecture of reconfigurable WDM/TDM PON”, Proc. WOCC’10, Shanghai, China, May 2010. [18] Y. Qian et al., “RSOA-based distributed access long reach Hybrid WDM-TDM PON with OADMs”, OSA Chinese Optical letters, vol.8, no.9, pp.899-901, Sep. 2010. [19] S.J. Park et al., “Hybrid WDM/TDM PON Using Remotely Pumped Optical Amplifier”, Proc. ECOC’07, Dresden, Germany, 2007. [20] J. M. Oh et al., “Enhanced system performance of an RSOA based Hybrid WDM/TDM –PON system using a remotely pumped erbium doped fiber amplifier”, Proc OSA/NFOEC, PDP9, Anaheim (CA), USA, Mar.2007 [21] J. Chen et al., “Performance Analysis of Protection Schemes Compatible with Smooth Migration from TDM-PON to Hybrid WDM/TDM-PON”, Proc OSA/NFOEC, JWA85, Anaheim (CA), USA, Mar.2007. [22] J. Chen et al., “Cost vs. Reliability performance study of fiber access network architectures”, IEEE Communications Magazine, vol.48, no.12, Feb. 2010. [23] D. J. Shin et al. “Hybrid WDM/TDM PON With Wavelength Selection Free Transmitters”, IEEE Journal of Lightwave Technology, vol.23, no.1, pp. 187-195. Jan.2005 [24] S. Ahsan et al. “Migration to the next generation optical access networks using hybrid WDM/TDM PON”, IEEE Journal of Networks, vol.6, no.1, pp.18-25, Jan.2011. [25] P.J. Urban et al. “Experimental demonstration of a 10Gbit/s wavelength 27km reach WDM/TDM PON based on reconfigurable OADM and colourless ONU”, Proc ECOC’09, 7.5.2, Vienna, Austria, Sep. 2009. [26] S. Kimura et al. “A 10Gbit/s CMOS burst mode clock and data recovery IC for a WDM/TDM PON access network”, Proc LEOS’04, TuR1, Rio Grande, Puerto Rico, Nov. 2004. [27] C. J. Chae et al,, “Multi-wavelength PON as a Cost-effective and Power efficient alternative to WDM/TDM PON for extended reach-applications”, Proc LEOS’10, Denver, Colorado, 2010. [28] D.M. Seol et al,”Passive Protection in a Long Reach WDM /TDM PON”, Proc COIN’10, The Shila Jeju, Korea, Jul. 2010. Appendix C. Bibliography X [29] J. H. Lee et al,”First commercial deployment of a colorless Gigabit WDM/TDM Hybrid PON system using remote protocol terminator”, IEEE Journal of Lightwave Technology, vol.28, no.4, pp. 344-351. Feb. 2010. [30] R. Inohara et al, ”Reconfigurable WDM/TDM –PON ring architecture by using alloptical wavelength converter and injection-locked FP-LD”, Proc OSA/OFC/NFOEC’10, OWG5, San Diego (CA), USA Mar. 2010. [31] X. Cheng et al, ”Hybrid WDM/TDM PON with dynamic virtual PON (VPON) capability” Proc ECOC’10, P6.07, Torino, Italy, Sep. 2010. [32] F. Khan et al, ”NUST hybrid (WDM/TDM) EPON based Access network with triple play support”, Proc HONET’07, Dubai, UAE, Nov. 2007. [33] J. D. Downie et al, ”An 11.1 Gb/s WDM/TDM PON system with 100km reach using ultra low loss fiber and duobinary downstream signals”, Proc LEOS’08, WEE3, Newport Beach (CA), USA, Nov. 2008. [34] C.H. Chen et al, ”A Delay sensitive Multicast Mechanism for differentiated services in WDM/TDM PON”, Proc COIN’08, Tokyo, Japan, Oct. 2008. [35] L. Shi et al, ”Behavior-aware user-assignment in Hybrid PON planning”, Proc OSA/OFC/NFOEC’09, JThA72, San Diego (CA), USA, Mar. 2009. [36] M. Kassir, “Current and future broadband bandwidth demand, promises and Challenges”, Published Capstone, University of Denver, 2006. [37] S. Kulkarni, et al, “FTTH-Based Broadband Access Technologies: Key Parameters for Cost Optimized Network Planning”, Bell Labs Technical Journal, vol.14, pp.297309, 2010. [38] A. Banerjee, M. Sirbu, “Towards technologically and competitively neutral fiber to the home (FTTH) infrastructure,” in Broadband services: business models and technologies for community networks, John Wiley & Sons, 2005. [39] T. Koonen, “Fiber to the Home/Fiber to the Premises: What, Where and When?”, Proc. of the IEEE, vol. 94 no. 5, pp. 547–588, Jun. 2006. [40] D. Nesset et al, “Economic Study Comparing Raman Extended GPON and Midspan GPON Reach Extenders’, Proc .OSA/OFC/NFOEC’10, pp.1-3, March 2010. [41] R.P. Davey et al, “The future of optical transmission in access and metro networksan operator’s view”, Proc ECOC’05, vol. 5, pp. 53-56, Sept. 2005. [42] B.T. Olsen et al, “Models for forecasting cost evolution of components and technologies”, in Telectronik, vol. 100, no. 4, 2004. XI Appendix C. Bibliography [43] D. Rokkas, et al, “Techno-economic Evaluation of FTTC/VDSL and FTTH RollOut Scenarios: Discounted Cash Flows and Real Option Valuation”, IEEE/OSA Journal of Optical Communications and Networking, vol. 2 no. 9, pp.760-772, Nov. 2010. [44] http://en.wikipedia.org/wiki/Fiber_to_the_x [45] ITU-T G.983.2 [46] ITU-T G.984.1 [47] IEEE 802.3 [48] IEEE P802.3av 2009 (9/2009) [49] ITU-T G.987 [50] L. Wosinska et al, “How much to pay for protection in fiber access networks: Cost and reliability tradeoff”, Proc. ANTS’09, pp. 1-3, Dec 2009. [51] J.A.Lazaro et al, “Scalable Extended Reach PON”, Proc .OSA/OFC/NFOEC’08, pp. 1-3, Feb. 2008. [52] J. Aweya, “IP Router Architecture”, in Journal of Systems Architecture 46 (2000) pp.483-511, 1999. [53] IEEE 802.3-2008, “Part 3: Carrier sense multiple access with collision detection (CSMA/CD) access method and physical layer specifications” section 5, clause 58. [54] FTTH Council, Network committee, “FTTH Infrastructure Components and Deployment Methods”, (2007). [55] S. Azodolmolky et al, “A techno-economic study for active Ethernet FTTH deployments”, in Journal of Telecommunications Management, vol.1, no.3, 294-310, (2008). [56] P. Chanclou et al, "Overview of the Optical Broadband Access Evolution: A Joint Article by Operators in the IST Network of Excellence e-Photon/One”, in Communications Magazine, IEEE, vol. 44, no.8, 29-35, Aug. 2006. [57] D. Breuer et al, “Architectural options and challenges for next generation optical access”, in Proc .of 36th European Conference on Optical Communication, 2010 (ECOC 2010), pp. 1-5, Sept. 2010. [58] S. Chatzi et al, “Techno-economic comparison of current and next generation long reach optical access networks”, Proc .of 9th Conference on Telecommunications Internet and Media Techno-economics(CTTE) 2010, pp. 1-6, May 2010. [59] S. Chatzi et al, “A Quantitative Techno-economic Comparison of Current and Next Generation Metro/Access Converged Optical Networks”, Proc .of 36th European Conference on Optical Communication, 2010 (ECOC 2010), pp. 1-3, Sept. 2010. Appendix C. Bibliography XII [60] S. Chatzi et al, “Techno-economic study of high splitting ratio PONs and comparison with conventional FTTH-PONs/FTTH –P2P/FTTB and FTTC deployments”, Optical Fiber Communication (OFC), collocated National Fiber Optic Engineers Conference 2011 Conference on (OFC/NFOEC), JWA15. [61] Optical fiber Communications Gerd Keiser / Mc Graw Hill International editions 2000, pp.92 [62] Optical Networks A practical perspective, Rajiv Ramaswami, Kumar N. Sivarajam, Morgan Kaufman publishers 2002 pp.68 [63] Optical Networks A practical perspective, Rajiv Ramaswami, Kumar N. Sivarajam, Morgan Kaufman publishers 2002 pp. 76 [64] Optical fiber Communications Gerd Keiser / Mc Graw Hill International editions 2000, pp.431 [65] Optical Networks A practical perspective, Rajiv Ramaswami, Kumar N. Sivarajam, Morgan Kaufman publishers 2002 [66] F. Buchali et al, "Reduction of the Chromatic Dispersion Penalty at 10Gbls by integrated Electronic Equalisers ", OFC 2000, ThSl-1, vol.3, pp.268-270 [67] S. Otte et al, "A decision feedback equalizer for dispersion compensation in high speed optical transmission systems", ICTON 1999, We.B.2, pp.19-22 [68] C.R.S Fludger et al, "Electronic Equalisation for Low Cost 10 Gbit/s Directly Modulated Systems", OFC 2004, WM7, vol.1, pp.234-236 [70] C. Xia et al, "Performance enhancement for duobinary modulation through nonlinear electrical equalization", ECOC 2005, Tu.4.2.3, vol.2, pp.257-258 [71] P.C. Becker et al, “Erbium-Doped Fiber Amplifiers”, 1st ed. Ed.USA, Academic Press, 1999, pp. 131-138. [72] J.A. Lazaro et al, “Extended Black-Box Model for Fiber Length Variation of Erbium-Doped Fiber Amplifiers”, IEEE Photon. Technol. Lett,. vol.20, pp.2063-2065, 2008 [73] J. Burgmeier et al, “A black box model of EDFAs operating in WDM systems”, J. Lightw. Technology, vol.16, no.7, pp.1271-1275, Jul. 1998 [74] R. Marz, Integrated Optics: Design and Modelling. Boston, MA: Artech House, 1995. [75] A. Texeira et al., “Black box model of erbium doped fiber amplifiers in C and L bands,” Telecommunications and Networking (ICT 2004), Springer Berlin, 2004, pp. 267-271 XIII Appendix C. Bibliography [76] X.Zhang et al., “A simple black box model for erbium doped fiber amplifiers,” IEEE Photon. Technol. Lett,. vol.12, no.1, pp.28-30, Jan. 2000 [77] G. Jacobsen et al, “Pump power dependent black box EDFA model”, J. Optic Commun. vol. 21, pp. 675-681, 2000 [78] G.P. Agrawal, “Fiber Optic Communication Systems”, 2nd Ed. John Wiley and Sons, Inc., 2002 [79] P.S. André, “Optoelectronic components for high speed photonic networks”, Ph.D. thesis, University of Aveiro, Portugal, 2002 [80] V. Polo et al, “Rayleigh scattering reduction by means of optical frequency dithering in passive optical networks with remotely seeded ONUs”, IEEE Photon. Technol. Lett,. vol.19, no.2, 2007 [81] R. Ramaswami et al., “Optical Networks”, 2nd ed., Ed. USA: Morgan Kauffmann publishers, 2002, pp.158-159. [82] S. Padwal et al., “Modelling of gain in EDFA and its behavior in C and L band”, Intern. Journal of Advanced electrical and electronic engineering, vol.1, pp. 25-29, 2012 [83] Y. Sugaya et al., “1.58 μm band Er3+ doped fiber amplifiactaion with a 1.55 μmband light injection,” in OECC’98 Tech. Dig. 1998, pp. 498-499, paper 16C2-4. [84] J. F. Massicott et al., “Low noise operation of Er3+ doped silica fiber amplifier around 1.6μm”, Electron. Lett., vol.28, no.20, pp.1924-1925, 1992 [85] Y. Sun et al., “80nm ultra-wideband erbium doped silica fiber amplifier,” Electron. Lett., vol. 34, no15, pp.1509-1510, 1998 [86] J. Lee et al., “Enhancement of Power conversion efficiency for L-band EDFA with a secondary Pumping effect in the unpumped EDF section”, IEEE Photon. Technol. Lett,. vol.11, no.1, pp.42-44, 1999 [87] B. H. Choi et al., “New pump wavelength of 1540nm band for long wavelength band erbium doped fiber amplifier (L band EDFA)”, IEEE Journal of Quantum electronics, vol.39, no.10, 2003 [88] S. Chatzi et al., “L-Band in-line remote amplification for an extended WDM/PON ring architecture” in ICTON 2009, Tu.D5.5, pp.1-4, 2009 [89] G. Keiser, “Optical fiber Communications”, 3rd ed. , Ed. Singapore: McGraw –Hill, 2000, pp.491-492. [90] B. Neto et al, “C+L Band Gain Equalization for Extended Reach WDM-Ring PON Using Hybrid Raman / in Line EDFA Amplification”, in Proc. ICTON’10, We.P.18, Munich, Germany, Jun.-Jul. 2010. Appendix C. Bibliography XIV [91] X. Liu , “Powerful solution for simulating nonlinear coupled equations describing bidirectionally pumped broadband Raman amplifiers”, Optics express, vol.12, no.4, pp.545-550, Feb. 2004. [92] A. A. M. Saleh et al, “Modelling of gain in erbium doped fiber amplifiers”, IEEE Photon. Technol. Lett. vol.2, no10, pp. 714-717, Oct. 1990. [93] B. Min et al, “Efficient formulation of Raman amplifier propagation equation with average power analysis”, IEEE Photon. Technol. Lett. vol.11, no10, pp. 1486-1488, Nov. 2000. [94] T. G. Hodgkinson “Improved average power analysis technique for erbium-doped fiber amplifiers” IEEE Photon. Technol. Lett. vol.4, no11, pp. 1273-1275, Nov. 1992 [95] M. C. Fugihara et al, “Low cost Raman amplifier for CWDM systems”, Microwave. Opt. Technol. Lett., vol.50, no2, pp.297-301, 2008. [96] Y. Sun et al., “80 nm ultra-wideband erbium-doped silica fibre amplifier,” El. Lett. 33, 1965 (1997). [97] R.E. Neuhauser et al, “New remote pump scheme enabling high-capacity (3.2 Tb/s) unrepeatered C + L band transmission over 220 km,” Proc. OFC’02, TuR2 (2002). [98] J. A. Lazaro et al., “Power Budget Improvement for Passive Outside Plant Long Reach High Density Access Network using High Bit Rate RSOA-ONUs”, Proc. ECOC’07, We6.4.3 (2007). [99] F. Bonada, et al, “Remotely Pumped Erbium Doped Fibre Bidirectional Amplifier for Gain Transient Mitigation”, Proc. ICTON’09, Tu.D5.4 (2009). [100] G. Keiser “Optical fiber Communications”, 3rd ed., Ed. Singapore: McGraw –Hill, 2000, pp.104. [101] R. Ramaswami et al., “Optical Networks”, 2nd ed., Ed. USA: Morgan Kauffmann publishers, 2002, pp.188. [102] Z.B. Hao et al., “Theoretical analysis of lnGaAsP/lnGaAsP multiple quantum wells electroabsorption modulators for the application of high speed low driving voltage integrated light source” Journal of Korean Physical Society, vol.43 April 1999, pp S105-S108. [103] J.A. Lazaro, et al, “Remotely amplified combined ring-tree dense access network architecture using reflective RSOA-based ONU”, IEEE/OSA J. of Optical Networking, vol. 6, no. 6, pp.801-807, June 2007. [104] C. Antony et al., “Demonstration of a carrier distributed, 8192-split hybrid DWDM-TDMA PON over 124km field installed fibers”, in proc. OFC/NFOEC’10, PDPD8, San Diego, USA, March 2010.