UPPER BOUND ON THE PERFORMANCE OF SUBSCRIBER ACCESS NETWORKS FOR DOWNSTREAM TRAFFIC CONSIDERATIONS FOR BROADBAND APPLICATIONS

Transcription

1 UPPER BOUND ON THE PERFORMANCE OF SUBSCRIBER ACCESS NETWORKS FOR DOWNSTREAM TRAFFIC CONSIDERATIONS FOR BROADBAND APPLICATIONS A Thei Submitted in Partial Fulfillment of the requirement for the Degree of MASTER OF TECHNOLOGY by T. M. Praanna to the DEPARTMENT OF ELECTRICAL ENGINEERING INDIAN INSTITUTE OF TECHNOLOGY, KANPUR May, 005

2 ii CERTIFICATE Thi i to certify that the work contained in thi thei entitled Uer Bound on the Performance of Subcriber Acce Network for Downtream Traffic Conideration for Broadband Alication, ha been carried out by T M Praanna (Y ) under my uerviion and that thi work ha not been ubmitted elewhere for a degree. May 005 Dr. Yatindra Nath Singh Deartment of Electrical Engineering Indian Intitute of Technology Kanur-08016

3 iii Abtract Increaed demand for bandwidth arie from a roliferation of alication uch a voice, video and data traffic a well a by the boottraing effect of increaed conumtion reulting from lower rate and otical fiber enjoy monooly in roviding long ditance communication with a remarkable error rate erformance. Otical communication i driven by WDM technology that emloy Erbium Doed Fiber Amlifier. WDM carve u the huge bandwidth of ingle mode fiber (SMF) into channel whoe bandwidth are comatible with eak electronic roceing eed. The thei deal only with the unidirectional aect of the Dual Bu Architecture for Subcriber Acce Network with aive otical litting being emloyed at the Otical Network Unit (ONU). The aim i to give an uer bound on the number of WDM channel that can be tranmitted and the number of ubcriber that can be accommodated (by maximizing the number of ower lit) for broadcat alication. The receiver enitivity i comared for different detection cheme. Three cae arie a (i) Analog broadcat channel along with unicat tranmiion (alo called witched ervice) (ii) Digital broadcat channel with witched ervice (iii) Hybrid Multichannel cae. The analog broadcat i AM-VSB (Vetigial Side Band) and digital broadcat i M-QAM or QPSK modulated and they are ub-carrier multilexed. The thei alo analye the benefit of AM/OFDM than AM/M-QAM hybrid ervice in term of bit error rate erformance and rooe other cheme like Forward Error Correction coding. The imlified gain model of EDFA a a reamlifier ha been adoted to exemlify it alication in otical communication, and algorithm i given for the deign of Subcriber Acce Network, with an examle deign alo carried out. Thu, the Wavelength Diviion Multilexing (WDM) alo called, a Data in a Rainbow concet will cater to the eventual need of greater caacity and fater acce.

4 iv Acknowledgement I take thi oortunity to acknowledge the hel and uort of my thei uervior Dr.Yatindra Nath Singh in guiding me to accomlih thi reearch work. I mut alo acknowledge Dr.Anjan Kumar Ghoh, Head of our Deartment who taught u the coure on Fiber Otic Sytem that motivated me to do my reearch in thi field. I would alo like to acknowledge Dr.Katuri Vaudevan and Dr. Ajit Kumar Chaturvedi who taught u the coure on digital communication and their alication in wirele technologie that made me comfortable to handle the concet in thi work. Alo, I mut acknowledge Dr.A.R.Harih, Dr.Animeh Biwa, Dr.Utal Da whoe coure enriched my knowledge and alo all the faculty member of our deartment who heled in my learning through eminar and dicuion. I remember the day when our Director Dr. Sanjay Govind Dhande aid The contribution of any intitute i gauged by the amount of oitive tranformation it bring in it tudent and I am highly indebted to my alma mater Indian Intitute of Technology, Kanur for bringing out uch a brilliant tranformation in me by enriching my technical knowledge along with the overall eronality imrovement. I mut alo acknowledge the faculty member of my undergraduation intitute Bhilai Intitute of Technology, Durg who taught me the baic coure and all the mentor who have contributed in my learning in one way or the other. I would alo like to thank the eronnel in my Deartment for roviding me with the required reource for reearch. I would like to thank my friend Navin whoe comany I alway cherihed, and other like Swanil, Bhaumik and Prabodh who made my tay at thi lace a memorable one. I would like to acknowledge Mr. Rajiv Srivatava for ome fruitful dicuion that heled me in thi work. Lat, but certainly not the leat; I would like to thank my family who tood by me in rough time of the life and Almighty for howering her bleing on me that made me to tread on thi ath of knowledge. T M Praanna

5 v Content Chater 1: Introduction 1.1 Why there i a need for more and more bandwidth? 1 1. The Race for Bandwidth continue with WDM but how? 1.3 Fiber Baed Acce Network: An Overview Migration toward Broadband Alication Iue ertaining to Sytem Deign Thei Outline. 9 Chater : Overview of the Sytem 11.1 Subcarrier Multilexed Lightwave Sytem 11. Erbium Doed Fiber Amlifier (EDFA) 1.3 Comariion between different Detection cheme Otical Preamlification with Direct Detection 15 (Not alicable for two level ignal).3. Otical Preamlification with Avalanche Photodetection Digital Coherent Detection Digital Direct Detection (Finite and Infinite Extinction Ratio) 19.4 Conideration of Architecture emloyed in Subcriber Acce Network 0 Chater 3: Performance Analyi of Subcriber Acce Network Salient Feature Regarding the Architecture 3 3. Analyi of All Analog Broadcat along with Switched ervice Deigning of a branch in Subcriber Acce Network Deigning of bu of a Subcriber Acce Network Analyi of All Digital Broadcat (M-QAM/QPSK) with witched ervice Analyi of Hybrid Multichannel AM-VSM/M-QAM video Lightwave 41 Tranmiion Sytem with witched ervice 3.5 Conideration of iue ertaining to ractical deloyment of SAN 56

6 vi Chater 4: Some Inight into Practical Deloyment Conideration Holitic icture of tranmiion imairment Prominent Nonlinear Interaction AM/OFDM i better than AM/M-QAM for hybrid tranmiion Crotalk Analyi Dierion Effect Laer Phae Noie How the imairment were taken care of and inight into future 65 digital ytem imlementation Chater 5: Concluion and Scoe for Future Work 67 Aendix A: Comariion of SCM/WDM with WDM 70 Aendix B: Rate and Proagation Equation in EDFA 71 Aendix C: M-QAM modulation cheme: An Introduction 74 Aendix D: Error Correcting Code: Reed Solomon Code 75 Aendix E: Hermite Polynomial 76 Reference 77 Bibliograhy 80

7 vii 1.1 Broadcat and Select Network 1. Wavelength Routed Network. 1.3 Broadband Service in a nuthell. Lit of Figure 1.4 Overview of a fiber baed otical Network 1.5 Simlified chematic of Broadband Network.1 Schematic of Subcarrier Multilexing.. Comariion between different Detection cheme with Otical Preamlification.3 Detection erformance in Coherent Detection Analyi..4 Performance Analyi in Digital Direct detection when both finite and infinite extinction ration are conidered..5 Architecture emloyed for the deign of Subcriber Acce Network. 3.1 Individual Noie to Carrier Ratio with reect to rm modulation index for AM-VSB channel. 3. Number of ubcriber in one branch w.r.t. Number of AM-VSB channel for All Analog Broadcat. 3.3 Inut ignal ower with reect to gain of an EDFA for FM Broadcat video for different number of ubcriber er branch of Subcriber Acce Network. 3.4 Subcriber Acce Network conidering downtream traffic mode only. 3.5 Number of EDFA and number of ubcriber in SAN w.r.t. number of WDM channel for analog video broadcat along with witched ervice. 3.6 Number of uer uorted on the SAN with the number of EDFA uorted on bu for analog video broadcat along with witched ervice. 3.7 Number of EDFA uorted on the bu with reect to the number of WDM channel. 3.8 Algorithm for otimum deigning of SAN for Analog Video Broadcat ervice 3.9 Number of Digital Receiver with number of digitally modulated channel mounted on one wavelength for All Digital Broadcat with witched ervice Number of Uer in a SAN for all Digital Video Broadcat with Switched Service Number of Subcriber in SAN with Number of WDM channel in all digital video Broadcat along with witched ervice. 3.1 Deigning of a reamlifier. The linearity of the gain of EDFA with the inut ignal ower over the deired window of interet continue to be key aumtion in thi work Analyi of two regime of oeration in Hybrid AM-VSB/M-QAM tranmiion cae in SAN to find the otimum oerating oint.

8 viii 3.14 Performance Analyi of AM-VSB/64-QAM in SAN when 64-QAM ignal i le a comared to AM-VSB ignal ower Uer Bae Size and Channel caacity in AM-VSB/64-QAM in SAN when 64- QAM ignal i le a comared to AM-VSB ignal ower Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i le a comared to AM-VSB ignal ower when number of ubcriber i fixed 3.17 Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i le a comared to AM-VSB ignal ower when QAM channel i fixed Uer Bae Size and Channel caacity in AM-VSB/56-QAM in SAN when 56-QAMignal i le a comared to AM-VSB ignal ower Performance Analyi of AM-VSB/64-QAM in SAN when 64-QAM ignal i comarable a comared to AM-VSB ignal ower. 3.0 Uer Bae Size and Channel caacity in AM-VSB/64-QAM in SAN when 64- QAM ignal i comarable a comared to AM-VSB ignal ower. 3.1 Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i comarable a comared to AM-VSB ignal ower when number of ubcriber i fixed. 3. Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i le a comared to AM-VSB ignal ower when number of QAM channel i fixed.

9 1 Chater 1 Introduction The evolution of communication technologie ha enviaged the better connectivity and better ervice between different lace in the globe. Internet and Cable Televiion (Common Antenna Televiion-CATV) ha become a houehold name now. There i an inatiable hunger for bandwidth. 1.1 Why there i a need for more and more bandwidth? Increaed demand for bandwidth arie from a roliferation of alication uch a voice, video and data a well a by the boottraing effect of increaed conumtion reulting from lower rate. The video tranfer may be either of till image over internet or have moving icture a in cable televiion network. There ha been a henomenal growth in data traffic becaue of myriad of web age available and their continuouly increaing uage. Alo, the tranfer of voice traffic ha increaed to great limit with decreaing cot and introduction of new technologie like Voice over IP (VoIP) [Bib 1]. The emergence of digital trend in communication ha given birth to better quality of audio and video tandard like HDTV. The data traffic i mainly of SONET/SDH rate and trend are of acket tranmiion in connectionle IP network. FTTH (Fiber to the Home) i conidered a a technology for future anwering to the increaing bandwidth requirement. The term Broadband Service i coined for the hybrid tranfer of data, voice and video ervice. Technologie like Video on Demand, Teleconferencing and Virtual Reality that require enormou bandwidth can be now realized. Thi dream i enviaged by otical communication along with Dene Wavelength Diviion Multilexing (DWDM) technology. Otical Communication uing otical fiber ha a virtual monooly when bit rate exceed a few Mb or ditance exceed a few hundred meter. WDM became the tranmiion technology of choice with the commercial availability of Erbium Doed Fiber Amlifier (EDFA). Tranmiion Sytem Engineering ha a come a long way ince then by imrovement in modulation, coding, dierion comenation and combating fiber nonlinearitie technique. Thu, the coming lightwave tranmiion technologie will enhance the coe of broadband (voice, video, WWW) alication.

10 1. The race for Bandwidth continue with WDM but how? With the ublic network at croroad, ince data traffic i uraing voice traffic, the bottleneck at the hyical layer tarted being felt and multilexing cheme were thought of a a viable olution. With Sace Diviion Multilexing (SDM), Time Diviion Multilexing (TDM) and WDM being in the icture, the quetion wa to emloy which multilexing cheme o a to make the bet ue of the vat bandwidth available in the otical fiber, and WDM technology wa the referred choice owing to ome engineering reaon. DWDM i a tranmiion technology that utilize a comoite otical ignal carrying multile information tream, each tranmitted on a ditinct otical wavelength. The concet of ending the whole traffic over one fiber in a multilexed fahion intead of uing multile fiber in arallel i a very cot effective aroach. The traffic carried on an otical wavelength i frequently generated by ome SDH (Synchronou Digital Hierarchy) equiment containing motly voice that i the maintay of conventional telecommunication network. The Time Diviion Multilexing (TDM) functionality of SDH, i.e, combining lower rate tream into higher rate tream i available with the DWDM equiment and thi integration of TDM multilexing functionality with DWDM contribute to the cot aving in the network economy. The ingle DWDM amlifier relace multile SDH regenerator, thu reducing the unit cot of bandwidth. Alo, ince WDM uort many channel of lower bit rate, the ditance limit due to chromatic dierion i much larger for WDM ytem than for equivalent TDM ytem. Similarly, Polarization Mode Dierion (PMD) doe not imoe ignificant ditance limitation at lower bit rate. The WDM aroach reult in an increae in tranmiion caacity in a modular fahion by adding wavelength whenever the caacity increae i required rather than having the whole big infratructure in the tarting itelf a of in TDM aroach. WDM ytem being tranarent to different bit rate and rotocol format (Data Link Layer and Network Layer iue) i an added advantage a it follow the comlete layered aroach enabling the ugrade in the network without many intricacie. TDM i limited by electronic eed, becaue to have the receiver electronic oerating at very high bit rate lace reure on today VLSI technology, although the electronic for lower bit rate are being deloyed uccefully, thu making WDM a ractical otion. WDM carve u the huge bandwidth

11 3 of ingle mode fiber (SMF) into channel whoe bandwidth are comatible with eak electronic roceing eed. Reliability iue give WDM an edge over higher eed TDM ytem a the latter reliability i not that well quantified. Alo, WDM Add Dro Multilexer/ Demultilexer (ADM) can be ued to bya through traffic without terminating them in SONET equiment, thu reducing the number of Synchronou Otical Network (SONET) ADM needed in the network reulting in network cot aving effectively. Needle to ay, today fiber baed acce network ue a combination of thee acce cheme with WDM-SCM combination (Aendix A) already being deloyed on mot of the regional and acce network, thu emloying the benefit of both. With DWDM being adoted on backbone/trunk route; the hyical layer contraint have been reduced to a lot more extent. Thu, thi Data in a Rainbow concet will cater to the eventual need of greater caacity and fater acce and will leafrog to an Ultra Long Haul DWDM technology for tranoceanic ditance in the near future. 1.3 Fiber baed Acce Network- An Overview The lethora of information being available timulated the deloyment of network [1] for haring, thu giving birth to oerator roviding the full ervice encomaing the trile lay- voice, video and data. With ervice being claified a broadcat or witched, with the former one meaning the ditribution of the ame information to all ubcriber a in CATV and the latter one a ubcriber ecific data being tranmitted in the lat leg like in Internet and telehone ervice; aved the way for the claification of network. The final link between the ervice rovider facility to the home or buine i called the Acce Network []. The conventional telehone network and cable televiion network differ in many reect like in bandwidth conideration and witching technique, but with the merging of ervice, the acce network were made to be comatible with all tye of traffic. Acce network can be further claified a feeder network; the one that i between Hub/Head-end and the Remote Node, and ditribution network between the Remote Node (RN) and the Network Interface Unit (NIU). Ditribution network are claified a broadcat and witched network. Broadcat network are well uited for broadcat ervice with all

12 4 NIU being identical and intelligence in them only, accounting for their eay deloyment. On the contrary, witched network being more uitable for witched ervice, with much eaier fault location than the former one and NIU being imler, reulting in the intelligence in the network Claification of traffic according to their nature a burty, like internet traffic or non-burty, a telehony or video, ha driven the claification of feeder network accordingly. Burty ervice do not require the dedicated bandwidth link, a it will be more efficient to hare a total large bandwidth among many NIU, with ome form of media acce control (MAC) [1] rotocol being required to coordinate acce to the hared bandwidth by the NIU. The diadvantage of the aroach i that the receiver electronic ha to oerate at the total bandwidth of the network a ooed to the bandwidth needed by the NIU a in dedicated bandwidth link. Dedicated bandwidth link baed network are able to give each NIU a certain Quality of Service (QoS) guaranteed. The future acce network are aimed toward uorting the convergence of ervice with Integrated Service Digital Network (ISDN) roviding 18 Kb and Aymmetric Digital Subcriber Line (ADSL) [1] roviding higher bandwidth than the former one being already deloyed. But with ISDN uffering from bandwidth limitation and ADSL being limited by ditance in the final leg, it led to the develoment of FTTC/FTTB/FTTH technologie a acce network for the third and fourth generation lightwave ytem. The FTTC/B/H aroach ha a higher initial cot, but rovide bandwidth deeer in the network and rove to be a better long time olution. With nowaday network being claified a Broadcat and Select for broadcat ervice and Wavelength Routed Network [] for witched ervice, the future call for the hybrid of both to uort the hybrid of ervice. The ditribution network need not be a oint-to-oint link ecifically. It can be any network in itelf like tar, bu, meh, ring [3] or hybrid of them. For examle, the ONU can give connection to an Ethernet Local Area Network (LAN) [1] or Fiber Ditributed Data Interface (FDDI) network []. The ONU may be a imle aive device uch a an otical tar couler or a tatic wavelength router. The number of NIU/RN that can be uorted in a network i limited by the litting lo in the tar couler. The deign of network a regional or acce network i different in aroach owing to ome long ditance imairment like chromatic dierion being dominant in the former one. Alo, ower budget lay a ignificant role in the deign.

13 5 Fig 1.1 Broadcat and Select Network. Fig 1. Wavelength Routed Network (Figure 1.1 and 1. taken from Communication of the ACM, Vol 4, No.6, June 1999) A wavelength routed network carrie data from one acce tation to another without any intermediate otical to electronic converion and i referred to a an All Otical Network. The following figure how a rough ketch of acce network a emloyed in thi thei. It i a wavelength-routed network in the ditribution art and a broadcat network in the feeder art thu catering to the hybrid requirement of Broadband communication. Head end/ Central Office SMF λ 1 EDFA λ 1, λ, λ3 Otical Network Unit NIU λ 1, λ, λ3 λ 1, λ, λ3 NIU RN λ λ 3

14 6 1.4 Migration toward Broadband Alication Fig 1.3 Broadband Service in a nuthell. (Figure taken from the webite: htt:// Fig 1.4 Overview of a fiber baed otical Network (Figure taken from the webite: htt://

15 7 Fig 1.5 Simlified chematic of Broadband Network. A broadband network i different from the conventional telehone or CATV network in the ene that it ha to uort the high data rate in addition to low bandwidth requirement with certain QoS. The broadband network are bi-directional with alication being claified a ymmetric or aymmetric one. To achieve the low intalled cot, Sub Carrier Multilexing (SCM) [4], [Bib ]; aive otical litting, FTTC technique are being emloyed. The network i deigned to rovide high quality broadband ervice, while imultaneouly achieving low intalled fiber cot and affording graceful migration to High Definition Televiion (HDTV) and digital video. The ue of aive otical litting at the ONU enable the cot of tranmitter and fiber lant to be hared by variou ubcriber. The utream requirement differ from the downtream one in the bandwidth requirement, a the latter one account for more. The utream tranmiion can be carried out at different wavelength o a to avoid interference a the ame fiber i ued for both, thu enabling full dulex oeration. The aive litter at the ONU hould alo include the roviion for witching other than the tar couler that act a a litter. With CMOS witche being well adoted for thi uroe, the witching requirement i atified for witched alication along with broadcat one. Although the deigning of a whole broadband network involve iue of other layer like network rotocol (Network Layer), ignaling rotocol (Data Link Layer), ue of Tranmiion Control Protocol (TCP)/ Internet Protocol (IP) a a art of tranort layer in network, thi work aim at the toic from the hyical layer oint of view and involvement of other layer in the deign can be taken a an extenion of the ame roviding full fledged ervice. Other oible extenion of the work i in making

16 8 it bidirectional and introduction of queuing analyi in the network to account for latency in the delivery of acket from the ource or among eer-to-eer tranmiion in the network. 1.5 Iue ertaining to Sytem Deign The deign of any network whether it i a long ditance regional network or a hort ditance LAN, involve many iue to be taken care of. Toology, ize of the network, receiver bae ize, tranmitter and receiver tuning characteritic, choice and lacement of amlifier, ower budget driving the number of amlifier/regenerator are ome of the cardinal factor ruling thi deign arena [3]. Deigning ha to very well comlement the otimization of reource; the mot imortant one are maximum tranmitter ower and the otical fiber bandwidth. Many of the factor are interdeendent. The receiver mut receive the minimum required ower for a given Bit Error Rate (BER) o that the number of uer that can be uorted can be made maximum. The ower ditribution to receiver mut be done in uch a fahion that receiver with maller dynamic range can be ued. With amlifier cot being reflected in the unit cot of the bandwidth, the deign hould try to reduce their number without comromiing much on ower budget. Alo, other aive comonent like Multilexer, Demultilexer and couler hould be minimized. Tranmiion ha to be carried out in the roer window. With otical fiber being le loy in 1550 nm window, the tranmiion need to be done in thi range. Thi low lo in the fiber, reulting in more ower at the receiver will traightly hel in the enhancement of the number of uer. Toology mut be deigned in uch a manner that it can be eaily ugraded at a later date when demand arie. For thi, MAC rotocol that hel to coordinate the haring of common reource hould be made indeendent of the number of uer. Toology Selection hould be aimed toward increaing the uer bae. Tranmitter and Receiver can be made variably tuned to uort WDM tranmiion, but attention mut be given to the fact that rotocol need to be modified deending uon the number of channel covered by the tunable range.

17 9 Receiver Senitivity, i.e, the average otical ower required to achieve a certain bit error rate at a articular bit rate ha to be taken ecial care of. Error Control Coding can be done to imrove coding gain reulting in an imrovement in receiver enitivity, thereby imroving the ower budget. Choice of receiver a -I-n FET or Avalanche Photo Diode (APD) govern thi iue. Similarly choice of tranmitter like ditributed feedback laer with externally modulated or directly modulated govern the ower budget. Modulation cheme have a ay in BER limitation. Other imairment like crotalk, nonlinearitie in fiber bound the ytem arameter. Tye of fiber like Single Mode Fiber, Multi Mode Fiber or Large Effective Area Fiber (LEAF) decide the dierion factor, choice of amlifier like EDFA or Semiconductor Otical Amlifier (SOA) and their lacement trategie lay a ignificant role in the deign of modern day network. Cacaded amlifier imairment like Amlified Sontaneou Emiion (ASE) noie and interchannel acing iue need to be conidered. Interchannel acing mut be ket otimum o a to maximize the number of channel tranmitted a well a to reduce the Inter Symbol Interference (ISI) roblem. Bandwidth narrowing of cacaded multilexer/demultilexer lace a tringent requirement on laer wavelength tability and accuracy. Equalization for the variation of ignal ower and Signal-to-Noie Ratio (SNR) mut be dealt with. Lat but not the leat; the roer wavelength lanning i required. 1.6 Thei Outline Chater Two deal with the need of Sub-Carrier Multilexing and the concet of All Fiber Video Ditribution (AFVD) and the brief decrition of Erbium Doed Fiber Amlifier along with the bird eye view of detection cheme, architectural conideration, the detail of which can be found in tandard textbook. Chater Three give the reult for all analog broadcat, all digital broadcat and hybrid ervice. It give the analyi of the number of ubcriber along with the number of EDFA and number of WDM channel that make toward convergence of ervice. Chater Four deal with the onet of new technologie in thi arena like the emloyment of Orthogonal Frequency Diviion Multilexing (OFDM). Thi chater

18 10 alo account for the ractical iue that cro u during the deign of a ractical ubcriber acce network like dierion, non-linearitie and other imairment like laer hae noie. Chater Five deal with concluion and exlore the future coe of extenion that can be carried out in thi work. They can be taken a imrovement/extenion of thi work and can be carried out a a equel to it. Aendice deal with the detail of the derivation and the ummary of relevant toic, to get oneelf familiarized with and to get tarted, the reference of which are aroriately tated a and when.

19 11 Chater Overview of the Sytem The alication of broadband SCM ytem to both aive and otically amlified ditribution network ha enviaged the All Fiber Video Ditribution concet. The advent of new technologie in the field of otical communication ha made the concet of broadband network good layer for ractical deloyment and not confining it to the four wall of reearch laboratorie..1 Subcarrier Multilexed Lightwave Sytem: Subcarrier Multilexing, when alied in conjunction with aive otical litting make u a a cot effective aroach for broadband network. Broadband Network are the one that uort the tranfer of high bandwidth data, being a good contender of ervice like Video on Demand. The concet behind the SCM technique i the tranmiion of microwave ignal over otical link. SCM ytem have the ability to accommodate both analog and digital modulation, to handle voice data, video, digital audio, high definition video and any future combination of ervice. Thi enormou flexibility make them attractive otion for broadband alication, for ervice originating from different ervice rovider, each uing different modulation format and requiring varying amount of bandwidth. The elementary concet behind thi SCM technique i hown in Fig..1. Digital or Analog Video Broadcat / Switched Service Combiner Laer Tranmitter Photodiode Fig..1 Baic Schematic of Subcarrier Multilexing

20 1 Otical amlification can be uccefully ued in SCM ditribution ytem to increae the number of ubcriber without otical regeneration. Good number of ubcriber can be accommodated by making reaonable aumtion about the amlifier gain characteritic and link budget. SCM network have the inherent caability to cater to the need of enhanced bandwidth, utilizing the enormou bandwidth in ingle mode fiber. SCM core high over conventional Time Diviion Multilexing a the referred aroach for broadband ubcriber ditribution. In a nuthell, SCM offer a technique for tranmitting data of many gigahertz bandwidth on each otical carrier, and allow for flexible allocation of bandwidth in reone to ubcriber demand. The key All Fiber Video Ditribution (AFVD) technologie are wideband otical tranmitter/receiver, multichannel video tuner and otical amlifier. Otical amlifier erformance i the key to advanced AFVD ytem baed on the tree branch toology becaue of the higher level of branching lo comenation that i required. The whole ytem can be exlained crudely a, the emitted light i amlified by an EDFA and ditributed by the otical litter. At the uer remie, the received light i converted back to electrical ignal. Tuner are needed to elect the deired channel, along with filter o a to earate out analog and digital ignal.. Erbium Doed Fiber Amlifier (EDFA): The advent of otical amlifier ha brought a henomenal revolution in high eed All Otical Network, thu removing the contraint on E/O and O/E converion, reulting in the removal of bottleneck at thee node. Semiconductor Otical Amlifier (SOA) and EDFA have been intrumental in making the cot effectivene of bandwidth today, with contribution to telecommunication and future data network a well a convergence of ervice. EDFA are uerior to SOA, although their weak oint lie in their large ize and difficulty in integration. The five uerioritie are: High Outut Light Power. Low Noie. Multichannel Amlification imultaneouly. Low Ditortion. Low Polarization deendency. The hyic behind the amlification i given in Aendix-B.

21 13 More detail about thi amlifier can be found in tandard reference [], [3], [4] and [7]. Erbium ion in the EDFA are excited by a uming light at a wavelength of 1480 nm or 980 nm. The weak inut light, whoe wavelength i 1550 nm, timulate emiion and amlified light i the outut. Reflection induced CNR degradation i taken care of by inut and outut iolator. The long wavelength a filter i neceary in forward uming cheme to iolate the uming light ource. Forward uming lead to a lower noie figure and a lower uming light to ignal light converion efficiency than backward uming. In forward uming cheme, both ignal and um light roagate in the ame direction through the fiber wherea in the backward uming, they roagate in the ooite direction. The forward uming direction rovide the lowet noie figure. In fact, the noie i enitive to the gain and the gain i the highet when the inut ower i the lowet. Backward uming rovide the highet aturated outut ower. Bidirectional uming cheme ha a higher erformance than the other two by combining the lowet noie figure and the highet outut ower advantageou although it require two um laer. The ignal gain i uniformly ditributed along the whole active fiber. The flatne of EDFA gain over wavelength i imortant for multiwavelength imultaneou amlification. The gain flatne of aturated EDFA, however degrade due to gain hole burning. Thi gain flatne degradation limit the multiwavelength tranmiion ytem erformance. The deloyment of EDFA in Subcriber Acce Network require the otimal deign of EDFA in term of maximization of ot amlification lo, that maximize litting ratio while maintaining a target CNR at the receiver. The length that maximize the gain i cloe to the length that maximize the ot amlification lo, o that they can be taken to be equal [10]. The comutationally imlified model i aumed o a to facilitate the deign roce, that make for the uer bound in the erformance. The um ower and the ignal ower are aumed a inut arameter for the deign of EDFA. The outut ignal ower can be comuted from the following et of equation out out [10]. i in hoton/econd, while P i meaured in watt ( P = hνq ) Q Q α ln( Q α α + ( IS IS Q Q out in Q ) α ln( Q )( Q in th in Q α ) + ( IS Q th ) = 0 α IS Q )( Q out Q in ) (.1) where the known wavelength deendent contant are α, α.

22 14 Q, Q, Q are abortion contant and intrinic aturation ower of ignal and IS IS th um, and threhold ower of um (ower required to invert an infiniteimally hort iece of fiber) reectively. inut ower reectively. Q, Q, Q in out in are ignal inut and outut ower and um Q th IS IS α Q Q = (.) IS IS ( α Q α Q ) The equation (.1) that i trancendental in nature can be imlified in linear form a: α ( Q th ln(0.7) 0.3 α Q ln(0.7) Q in in α ) Q Q in out α = Q th Q 0.7 ln(0.7) α 0.3 in α + Q th ( Q in Q th Q ) + α ln( Q th in ) (.3) The value of Q out can be ued to obtain the otimum length of EDFA a l G max 1 Q = [ln( α Q in th out ( Q Q ) Q IS in in ) ( Q Q + IS Q th ) ] (.4) Given the um and ignal ower, uing Equation (.1)-(.3) can give gain a Q G max. = (.5) Q out in Deigning of Network require the knowledge of amlifier ontaneou noie coefficient that can alo be aroximated a Equation (.6). The minimum value of the ontaneou noie coefficient, n can be obtained by letting the um ower tend to infinity, in P, thu achieving contant inverion, and yielding n α γ (min) = (.6) ( α γ α γ ) The arameter taken are P IS = mW, P α = 0.944m 1 = mW, α = 0.886m γ, γ are ignal and um gain coefficient reectively, can be evaluated from γ = ζ Q ζ = Q E IS A τ α QE i the denity of erbium atom, A IS 1 i the area of the active region, (.7) τ i the ontaneou lifetime of the inverted atom in the uer level. ζ i taken a.60e m.

23 15 The model of EDFA conidered i linear, wherea it reult in aturation. Thi aumtion along with the lower bound on amlifier ontaneou noie coefficient reult in the uer bound in the erformance of Subcriber Acce Network..3 Comariion between different Detection cheme: Detection at the receiver end i an imortant conideration, ince it lay a role in the receiver noie conideration. Detection cheme are comared on the bai of the receiver enitivity, that i the number of hoton required to detect a bit correctly, that ha a direct relation to bit error rate. The following cheme will be comared [8],[16]: 1. Otical Preamlification with Direct Detection (for both two level ignal and non two level ignal).. Otical Preamlification with Avalanche Photodetection. 3. Otical Preamlification with Coherent Detection. The comariion cheme involve the tudy of the effect of both zero and non-zero extinction ratio and ignal ower hall be lotted againt the EDFA gain for all cheme of detection..3.1 Otical Preamlification with Direct Detection (Not alicable for two level ignal): The electrical SNR can be exreed a: SNR e = hν ηg B ( P + P ) + 4P P e where; P = n o o hν B o o Be Bo P + P Be o Bo ( B o Be ) + 4kTBe R ( ) hν eηg (.8) The above equation i olved for 16 db electrical ignal-to-noie ratio, which i decent for many alication. P o i an equivalent noie ower, with arameter fixed a T = 300K, B = B = 40GHz, R = 00Ω, k = 1.38e 3J K. Aumed RZ coding with o e / data rate of 10 Gb. Otical Preamlification can imrove the detector electrical SNR by 3 db inite of the fact that otical SNR i necearily degraded by 3 db in the roce. P hvbe 4Po Be Po hν (Be ) Po (Be ) Be 4kTBe hν ( SNRe )[ η G + B ] P + ( SNRe )[ ηg + ( Bo ) + R ( ηeg) ] = 0 (.9) o Bo

24 16 The above quadratic equation can be lotted for ignal ower with the reamlifier EDFA gain in Figure (.). Fig.. Comariion between different Detection cheme with Otical Preamlification..3. Otical Preamlification with Avalanche Photodetection: A in the reviou cae, the electrical SNR w.r.t. EDFA gain can be exreed a: 0.01 ) )( (1 : ; ) ( ) ( 4 ) ( 1 4 = = + = = = = = e h i M i i M M ex o o G e h M R ktb B o B B o B B o ex o e G h e k k M k ExceNoieFactor F AvalancheGain M B h n P where B P P P F P P B P SNR e e o e o e α α ν η ν η ν (.10)

25 17 ki i the ionization ratio (auming electron injection). The above equation i lotted for inut ignal ower with reect to the EDFA gain, and for M = G, M = G and lotted in Figure (.). M i the average value of Avalanche Gain and G i the reamlifier gain. However, at high data rate, receiver with otical reamlifier are coniderably more enitive than coherent and APD receiver..3.3 Digital Coherent Detection: The IM-DD ytem i more ucetible to ontaneou emiion from the otical amlifier, o that the ue of otical filter, which limit the bandwidth of the ontaneou emiion, become neceary, wherea in a coherent ytem with otical amlifier, the beat noie between the local ocillator light and the ontaneou emiion i redominant over the beat noie between the ontaneou emiion. A a reult, beat noie limited tate between the local ocillator light and the ontaneou emiion can be achieved, which greatly reduce the requirement for the bandwidth of otical filter. Alo, the waveform ditortion due to fiber chromatic dierion, which limit the tranmiion ditance at high data rate ytem, can be comenated by uing otical equalizer in the IM-DD ytem. However, it can be electrically comenated more imly by uing delay equalizer in the intermediate frequency tage of a heterodyne receiver. Otical data i modulated in ASK, FSK or PSK format. The total noie variance can be exreed by et of equation (.11). σ = σ d LO ASE ASE hot where; σ where; I σ σ σ ASE ASE where; I hot N = 4η I + σ = en LO ASE = 4η GI = ηe.b.( I = η I Be LO N Bo Be N Bo GB I I o Be N Bo ep =, I hν LO + σ e ASE LO eplo = hν + GI + σ ASE ASE + I N ) + σ th (.11) If local ocillator ower i high enough, local ocillator hot noie overcome the detector thermal noie, then

26 18 SNR SNR o e e where; F P = n o = SNR = ηf o hν B hot GP 1 GPo hν σ th o( 1+ P ) + ( G + ηn LO ) P + LO ηe.begplo hot = n o ηp = hν B + 1 ηg e SNR (.1) P i the equivalent noie ower. BER =.5ex( αsnr ) (.13) 0 e where the different value of α a 0.5, 0.50 and 1 correond to ASK enveloe detection, FSK dual filter detection and CPFSK or DPSK delay demodulation, o that SNR can be obtained accordingly for BER of 1e-9. The lot of inut ignal ower with the reamlifier EDFA gain can be exreed a Figure (.3). Fig..3 Detection Performance in Coherent Detection Analyi. The ower enalty i defined a the receiver enitivity degradation in decibel at the fixed BER a comared to the enitivity without the amlifier. Higher outut ower in

27 19 IM-DD ytem will lead to undeired henomena by the nonlinear effect of an otical fiber. However, coherent ytem are ued in thi low amlifier outut ower mode to alleviate thi roblem of fiber nonlinearitie. However, the effect of fiber dierion i the leat detrimental oing roblem in IM-DD ytem that accrue for it wide uage, a comared to other rooition. The above reult are evaluated without the ue of the olarizer at the detector. The analyi doe not account for any ditortion of the received data attern caued by otical fiber. There i large frequency deviation and otical bandwidth greater than data rate cauing negligible ditortion..3.4 Digital Direction Detection (Two Level Signal are conidered with Finite and Infinite Extinction Ratio): Fig..4 Performance Analyi in Digital Direct detection when both finite and infinite Extinction ratio are conidered.

28 0 The analyi remain the ame a it wa for reviou cae, however, the extinction ratio are conidered to be finite and the analyi are valid for two level ignal. BER = Q : Peronick' factor Aume _ ASK _ cheme N 1 1 Q ex( π Q (1 + K1 0) = Q ( NF). kr. (1 K ) ) 1 0 Q.( NF) + [ * (1 K ) 1 0 4K 1 0 Q k r + (1 K 1 0 ) k. k. k r th (.14) N 1 i the hoton er bit for 1. e r r kr,min = 0.5, B = k. B where i the data rate and B ot = K. Br where K = min 1, 1 ith k ) th = ( +, NF = n that give i fundamental limit on the erformance. K1 0 i the extinction ratio and different value of extinction ratio are conidered in thi work. σ1 i0 + σ 0 i1 D( deciion _ threhold) = σ + σ Q = i i σ σ 1 σ 0 SNR(1) = ( + ) σ σ 0 0 σ 1 1. Q (1/ η) +. n Otical _ noie _ figure = G 0 1 eq. G ASE B r Since the electronic bandwidth B e of hoto detector, i.e., of the -I-n tye i fixed by carrier tranit time and intrinic caacitance, the reduction in the otical aband caue enhancement in receiver enitivity..4 Conideration of Architecture emloyed in Subcriber Acce Network: Architecture lay a key role in the deign of the Subcriber Acce Network. The choice of architecture lay a cardinal role in the better cot effective, rotection enabled and retoration caable deign of network that can hel u in a long way to attain the objective of enhanced ubcriber bae couled with better throughut. There

29 1 are many architecture that can be emloyed, each one leading to different erformance with different ubcriber bae. The architecture can be drawn a: Central Office Fig..5 Architecture emloyed in thi work for the deign of Subcriber Acce Network. The couler ued in the deign of the ubcriber acce network i wavelength flattened 3-dB couler that i caable of giving 50:50 diviion at all wavelength in the deired window of tranmiion that make it a uitable candidate for ue in the WDM environment. The dual bu toology i ued in which two earate fiber are ued for utream and downtream tranmiion; the other variant of thi architecture i the ue of the ingle fiber for both utream and downtream traffic. Ring architecture [] ha alo gained the ufficient oularity for the deign of the network, due to it elf-healing characteritic. Fiber Ditributed Data Interface (FDDI) emloy fiber a a tranmiion medium, in thee elf-healing ring; creating alternate ath and roviding iolation of faulty link increae reliability. FDDI ue two counter rotating ring. It emloy a loo back mechanim for high reliability in cae of node failure. Paive otical ring can rovide high reliability becaue failure of a node

30 doe not break the ring. But the ower budget oe a limitation on the number of uer that can be alleviated uing otical fiber. Star toology i an old layer in thi area that ditribute otical ower equally at outut ort; therefore it reult in the uort of maximum number of uer without amlifier a comared to other toologie. Star couler can be imlemented in different toologie in itelf like tranmiive and reflective tar couler. The exiting conventional toologie can be modified o a to uort more number of uer and to uort different ervice. The architecture that i emloyed in thi work uit better for Subcriber Acce Network for convergence of ervice in term of imlicity in deign and better erformance. Other variant of the ame architecture can be emloyed to obtain different erformance, a er requirement. The above architecture can be modified to multilevel accordingly, o that more number of uer can be accommodated. However, for analyi uroe, the above architecture i conidered and the reult can be extraolated, to give much higher uer bound.

31 3 Chater 3 Performance Analyi of Subcriber Acce Network Thi chater give an uer bound on the erformance analyi of the Subcriber Acce Network. The reaon it i called uer bound i that all imairment are not taken into account, thu ractically; it will yield leer erformance, when all of them are accounted for. 3.1 Salient Feature regarding the architecture: Thi i a WDM environment, i.e., many channel of different wavelength can be tranmitted imultaneouly and each wavelength can contain many channel mounted on it, like, one wavelength may be containing broadcat data and the ret may contain witched data and all thee channel are ubcarrier multilexed. Thi architecture form the bai for ubcriber acce network that can be ued in FFTC/FTTB/FTTH concet. On the Bu, we can emloy EDFA to maximize the number of branching out of it uing a 3-dB couler. We conidered the litting lo, exce lo, fiber lice lo in the network. However, we cannot emloy any number of EDFA in the network, a the uer limit i laced by the accumulation of the Amlified Sontaneou Emiion Noie that rogree by accumulation along the fiber. Inut to each EDFA on the bu that i emloyed a an inline amlifier i ket contant, o that gain of each EDFA i alo aumed contant. Thi i a very cot effective aroach, a more and more ubcriber can be adhered to without incurring too much of lo. Utream and Downtream traffic requirement are different, o, in utream traffic, there i no requirement of reamlifier a we did in downtream traffic, by uing EDFA a reamlifier in each branch.

32 4 For utream conideration, we can tranmit at 1310 nm window alo. Although, we are uing dual bu toology, the ame wavelength window can alo be ued in downtream tranmiion. Utream traffic will not be dene wavelength diviion multilexing cenario, o that we can conider even an emloyment of Semiconductor Otical Amlifier (SOA) in the utream bu art of the Subcriber Acce Network. Emloyment of SOA in the utream traffic will be more cot effective than emloying EDFA that are more uitable for dene WDM alication; a EDFA are more cotly. Remote Node contain tunable otical banda filter o a to earate out data of witched alication. At Otical Network Unit, we conider loe due to litting only, o that it ha a wavelength flattened 1 x N litter. Only downtream tranmiion will be conidered, although utream analyi will be imilar and lot more eay a comared to downtream tranmiion. Thi work aim toward convergence of ervice a in next generation broadband fiber acce network, like trile lay-voice, video and data. 3. Analyi of All Analog Broadcat along with Switched ervice: Analog Broadcat can be either AM-VSB or FM-SCM along with witched ervice. Firt branch ha to be deigned; for that we aim at deigning the EDFA a a reamlifier in order to maximize the ot amlification lo, while till achieving the target Carrierto-Noie Ratio (CNR) Deigning of a branch in Subcriber Acce Network The linearly aroximated model of EDFA, a given in Chater [10] i conidered. For AM-VSB ignal, the received ignal ower at the receiver (at the hoto detector) i: P ignal = in m R G ( P ) L. (3.1)

33 5 where m i the modulation deth er channel, G i the gain of EDFA a a reamlifier, R in i the reonitivity of the hoto detector, i the inut ignal to the reamlifier. Alo, R = ηe /( hν ), where η i the hotodetector quantum efficiency. The imairment we conider are σ = erglp B, where i the noie bandwidth (3.) hot in e B e σ = 4R ( hν n ( G 1) L) B B (3.3) ig in o e e σ = 4Rηen ( G 1) L GP B. (3.4) We aumed that there i no olarizer ued at the detector end; ele thi ontaneouontaneou noie term i reduced by half. σ = i B (3.5) thermal th e P Where i i the equivalent circuit noie current denity exreed in A/ Hz. th σ in RIN = ( RIN)( ) RGLP B e (3.6) Where RIN i the laer Relative Intenity Noie, given by RIN= < I > I exreed in / db/hz and B e i the receiver electronic bandwidth. We hall aume that nonlinear ditortion i due to the cliing alone. Nonlinear ditortion i caued by large excurion in eak modulation current. 1+ 6µ 1 π..ex( ) 3 µ CNLD = (3.7) µ Where CNLD i the carrier to nonlinear ditortion ratio and µ i the rm modulation index er channel related to number of AM-VSB channel and eak modulation index er channel, a µ = m N / (3.8) Nonlinear ditortion include intermodulation of all order. It i a tee function of rm modulation index, that i, ditortion rie raidly a the uable index i uraed. Thi accounting of nonlinear ditortion i better than emloying CSO (Comoite Second Order Ditortion) or CTB (Comoite Trile Beat Ditortion). σ + total = σ RIN + σ hot + σ ig + σ + σ thermal σ NLD (3.9)

34 6 Other ource of nonlinear ditortion are not conidered here, like Four Wave Mixing, Stimulated Brillouin Scattering and Self Phae Modulation. We hall find out the CNR by dividing the ignal ower received a in Eq (1) and the individual noie variance. The reaon we are auming thee noie variance a noie ower i that thee noie ource are conidered a zero mean random roce. We hall look at the invere of CNR, i.e., NCR noie to carrier ratio; we need to minimize the NCR, a that will reult in maximization of CNR. NCR NCR NCR NCR NCR CNLD hot ig RIN thermal 1 = hν B = = 8hν n = 4 B B B ( RIN = e B e i e c e o /( ηgl ( P ( hν n ) /( µ /( R ( G 1) B G in ( G 1)) / N ) L ( P / N ) µ ) /( Gµ ( P in ( µ / N )) 3 = µ.ex( 1 /( µ )) /( π.(1 + 6µ )) e ) /(( µ in / N )) / N ) G ( P in ) ) (3.10) We hall adot the grahical method to find out the minima oint of the total NCR. Firt the rm modulation index ha to be found out followed by the number of receiver and the number of channel. The inut ignal ower and the inut um ower are rior information required for the deign of the EDFA a a reamlifier in the branch. NCR 1 total = NCRhot + NCRig + NCR + NCRRIN + NCRthermal + CNLD (3.11) Signal Wavelength λ Pum Wavelength Pum Power Signal Power Target CNR CNR λ t arg et Receiver Bandwidth Be 1550 nm 1476 nm 50 mw 1, 3., 5, 10 mw 48 db 5 MHz Photodetector Quantum Efficiency η 0.8 Equivalent circuit noie current denity i c 10 A/ Hz Relative Intenity Noie RIN Otical Bandwidth Bo -15 db/hz 5 nm

35 7 The exce lo of the aive litter can be given aγ log N, o that even if there are ome large number of ubcriber, then ince the exce lo of ingle 3-dB couler i of the order of 0.1 db, the net exce lo can be neglected. The target CNR of 48 db i well uited for many CATV alication, although the notion of CNR i mainly ubjective, yet the value of CNR a 48 db correond to a icture ubjectively judged excellent. With the target CNR being fixed, the number of receiver and number of AM- VSB channel that can be tranmitted on one wavelength ha to be found out. Starting with the equation of NCR total, we need to find out the value of rm modulation index that reult in maximum CNR or minimum NCR. Differentiating that equation w.r.t. µ will be mathematically cumberome, o we hall reort to grahical aroach where we lot individual NCR with reect to the rm modulation index. The individual NCR are lotted for a inut ignal ower of 5 mw, that i hown in the following Figure 3.1, from which rm modulation index can be found out. Fig. 3.1 Individual Noie to Carrier Ratio with reect to rm modulation index for AM-VSB channel.

36 8 Similar lot were obtained for other value of inut ignal ower, a mentioned in Table 3.1; the analyi can be roceeded a: NCR total almot coincide with 1 CNLD for higher value of rm modulation index and NCR total coincide with ( NCR hot NCR NCR + NCR + NCR + ig + RIN thermal ) for lower value of rm modulation index. So, the otimum value for µ come out to be 1 when = CNLD (3.1) NCR total 1 that i when CNLD =( NCR NCR NCR NCR + NCR hot + ig + + RIN thermal ) (3.13) To ummarize, at the oerating oint, the total noie ower and the nonlinear ditortion ower are equal. Thu, given the target CNR, we can find the rm modulation index from equation (3.7) and (3.1) by making NCR =. For imulating the Figure 3.1,we aumed the value of ot amlification lo L a hot noie=ignal ontaneou beat noie + ontaneou-ontaneou beat noie. Thi choice of oint correond to increaing the ot amlification lo where hot noie tart to dominate uon which any further increae in the lo can rove detrimental to ytem erformance. In the cae conidered, ince the target CNR i 48 db, olution of the above trancendental equation (3.7) and (3.1) yield the value of µ = total NCRt arg et Now, from Equation (3.10) and (3.11), where L=1/M, the whole ot amlification lo i attributed to ignal diviion, emloying Equation (3.1) in it yield the relation between number of AM-VSB receiver (M) and the number of AM-VSB channel (N) a in M = G *( ) *(( P / N) 3.39e 5) (3.14) For different value of inut ignal ower a 1, 3., 5, 10 mw, different grah are obtained where N i lotted from 0 channel onward. Thi relationhi i lotted for different value of in P and hown in Figure 3.. A ignal ower increae, more and more number of receiver and number of broadcat channel can be accommodated, but thi reult in other comlication alo, a inut ower cannot be increaed freely. The um ower of EDFA can alo be increaed for more gain and better erformance. However, thi work i done by fixing the um ower at 50 mw, o a to avoid the change of all arameter imultaneouly.

37 9 Fig 3. Number of ubcriber in one branch w.r.t. Number of AM-VSB channel for All Analog Broadcat. The Subcriber Acce Network i for broadband ervice, o along with AM-VSB broadcat, there will be FM-Broadcat channel alo, at a different wavelength a a art of WDM cenario, although FM being economical in term of ower requirement, but core le in term of bandwidth efficiency. FM tranmiion ha leer CNR requirement a 17 db. Since, AM-VSB require more ower than FM channel, o their ower criterion will drive the choice of the number of receiver. For FM-SCM broadcat channel, the relationhi i given for the inut ignal ower and the gain of EDFA a a reamlifier, and the ame i lotted in Figure 3.3. The imairment conidered are hot noie, ignal ontaneou beat noie, ontaneou-ontaneou beat noie, thermal noie,

38 30 Relative Intenity Noie. Nonlinear ditortion can be neglected due to it diminutive effect. The exreion for in P FM can be exreed a P in FM κ 1+ ηn. LG = hν. Be.. 1+ η LG 1+ ( σ / e. B. B ) + (ηn th o e. κ (1 + ηn LG) LG) B. B o e (3.15) 1 where m κ = ( ). Be RIN (3.16) CNRd m i the eak modulation index er channel and taken a 0.1. Figure 3.3 how the variation of the inut ignal ower with the gain of an EDFA a a reamlifier for ome value of M (number of ubcriber in one branch taken a 750, 375, 300, and 50 from Figure 3. a the AM-VSB ower criterion drive the choice of the number of receiver and number of channel). From the lot, it can be inferred that for erving to a articular number of ubcriber in one branch, the ower required i very le in cae of FM broadcat, a i evident. Fig 3.3 Inut ignal ower with reect to gain of an EDFA for FM Broadcat video for different number of ubcriber er branch of Subcriber Acce Network.

39 Deigning of bu of a Subcriber Acce Network: 3 db couler Central Office N th EDFA Branch 1 Branch N Fig 3.4 Subcriber Acce Network conidering downtream traffic mode only. For N EDFA on the bu, it could uort (N+1)M number of uer in SAN where M i the number of ubcriber in one branch, thi many number of uer can be uorted without otical regeneration. The 3-dB couler i a wavelength flattened couler to oerate in the WDM cenario. The number of EDFA uorted on the bu deend on the ASE accumulation due to cacade of EDFA in the main bu. -db of lo i alo aumed to account for the attenuation in the otical fiber at the rate of 0. db/km and lice/connector lo and any other lo (if there are any). Let P be the ASE generated out of the firt EDFA, and the gain of each EDFA on o the bu i aumed contant, delivering contant ower. The 3-dB couler being non-ideal introduce ome lo into the litting, by not litting into equal halve. N R 1 P ASE P o R 1 N P = hν ( G 1) Boη o o uer limit ch where (3.17) R i the lo-gain roduct. N ch rereent the number of WDM channel a we average the ASE ower over one channel i averaged where G i the gain of an EDFA a a

40 3 reamlifier and G i the gain of an EDFA a an inline amlifier on the bu. calculated keeing in mind the CNR at the receiver end or it can alo be calculated a SNR for a articular BER by relation between SNR and BER. The latter method i emloyed for all digital broadcat to be analyzed in the next ection. The SAN i to be ued for convergence of ervice, o the total ower mut be incluive of broadcat a well a witched channel. Since, witched channel account for very le ower a comared to analog broadcat one, o, there will not be much change in the ower budget in the network, with the number of witched ervice channel. The limiting factor of the number of EDFA on the bu i the ASE accumulation to a oint that till hel in achieving the deired CNR of 48 db at the receiver end. However, 50 db CNR at the ONU level for calculation i emloyed. i that value of ASE noie ower, beyond which any further increae in ASE noie will make the CNR criterion wore than the one required for good tranmiion/ recetion of video at the remote node. Subtituting in the Equation (3.17), the following relation i obtained which give the relation between the number of EDFA uorted and the number of WDM channel tranmitted. o uer lim it P ASE uer lim it P ASE uer limit P.( R 1) 1 N log ASE N ch. P + (3.18) o log10 R Since, it i convergence of ervice concet, only broadcat ignal will be reent all the time and the ret of witched ignal may be there or not according to individual ubcriber requirement. But, ince the broadcat ower i more than witched ervice ower by a good margin, the gain of reamlifier doe not vary much and will be very much in the tolerable limit. The number of EDFA uorted on the bu can be enhanced by emloyment of equalization trategie like SNR equalization, ower equalization, and gain equalization [1], [13]. However, they take into account that gain varie with wavelength and neglected out ASE aturation. But thi work account for ASE accumulation, but aumed that variation of gain ( dg / dλ ) i very le over the articular range of interet, i.e., 1540 nm 1570 nm, that can be obtained by doing EDFA with fluoride gla. Equation (3.18) can be ued to lot the relation between the number of EDFA uorted and the number of WDM channel, a hown in Figure (3.5). The limitation on analog tranmiion i reflected in term of number of WDM i

41 33 channel that it can uort. It can be een that four WDM channel can be tranmitted at the mot and the number of EDFA uorted i omewhere around even and decreae a the number of WDM channel increae. Fig 3.5 Number of EDFA and number of ubcriber in SAN w.r.t. number of WDM channel for analog video broadcat along with witched ervice. From the above figure, it can be inferred that a the number of WDM channel increae, the number of ubcriber go down, becaue, increae in the number of WDM channel caue the reduction in EDFA uorted and conequently reduction in the number of ubcriber on a ubcriber acce network in downtream conideration. The above deigning of a SAN can be exreed in an algorithmic fahion that i hown in Figure (3.8) that will hel in an otimum deign with limited ower budget. Thu thi conclude the analyi of all analog broadcat along with witched ervice that ha got evere limitation in the number of WDM channel and number of ubcriber, thu aving the way for digital trend in communication ytem.

42 34 Fig 3.6 Number of uer uorted on the SAN with the number of EDFA uorted on bu for Analog Video Broadcat along with witched ervice. Fig 3.7 Number of EDFA uorted on the bu with reect to the number of WDM channel.

43 35 Inut Signal Power Inut Pum Power Evaluate Gain of an EDFA a a reamlifier 1 Evaluate Number of Subcriber in one branch and Number of channel in one wavelength. 1. Increae Pum Power (not done often).. Increae inut Signal Power. CNR requirement at the Receiver I the reult atifactory? No Reduce Target CNR Iue command to change arameter Ye Evaluate the number of EDFA uorted on bu and number of WDM channel that can be tranmitted. I the Reult otimized? No Ye, Deign Comleted Fig 3.8 Algorithm for the otimum deign of SAN for Analog Video Broadcat Service.

44 Analyi of All Digital Broadcat (M-QAM/QPSK) with witched ervice: For all digital broadcat, ince the ower involved i le, o the ytem will be le rone to nonlinear effect, thu jutifying the ue of a directly modulated laer diode for good linearity. Relative Intenity Noie become the main limitation to Subcarrier Multilexed ytem at high channel count. The nonlinear ditortion due to cliing that wa aumed earlier in cae of all analog broadcat till hold here becaue the exreion wa indeendent of modulation format [14]. It i aumed that all other noie ource are indeendent of each other. The ame equence of te ha to be followed for the deign of SAN, a wa done earlier. The reult can be dicued according to the figure. The target SNR wa calculated according to the achievement of BER of 1e-9 at the receiver end [15],[16]. BER = 0.5.ex( α. SNR) (3.19) where α =0.5, 0.5, 1 for ASK enveloe detection, FSK dual filter detection and CPFSK or DPSK delay demodulation. It can be given in tabular form a in Table 3.. Modulation Scheme SNR(dB) RMS modulation index QPSK QAM QAM QAM Following the ame et of mathematical exreion (3.1)-(3.18), we hall find the relation between the number of digital channel and the number of digital receiver. There i a henomenal increae in the number of channel on one wavelength in a ubcarrier multilexed fahion, thu making it technology for the future, where thouand of channel with an equal number of ubcriber in one branch may be required, a comared to the analog video broadcat cae. Figure (3.9) how the number of digital

45 37 receiver with the number of digital channel mounted on one wavelength for QPSK and M-QAM modulation cheme (Aendix C), where M take value a 16, 64 and 56. Fig. 3.9 Number of Digital Receiver with number of digitally modulated channel mounted on one wavelength for All Digital Broadcat with witched ervice. The whole ubcriber acce network ha been deigned for a tyical value of 3000 digital channel. From Figure (3.9), it can be inferred that QPSK uort enormou number of uer; the number of receiver that can be accommodated reduce a the contellation ize increae in M-QAM modulation cheme. It can alo be inferred that

46 38 the erformance of 56-QAM i not atifactory and to make it erform better, Forward Error Correction (FEC) coding cheme [17] have to be emloyed. Thu, digital broadcat of video i better than all analog broadcat in term of both number of ubcriber a well a the number of channel that can be tranmitted. The ame method ha to be alied for the bu deign alo. Figure (3.10) give the relation between the number of ubcriber on a SAN with the number of EDFA that can be uorted on a bu in different modulation cheme. Fig 3.10 Number of Uer in a SAN for all Digital Video Broadcat with Switched ervice. The number of EDFA that can be uorted in analog broadcat i more by a mall number. The reaon may be given a, ince ignal ower i reduced, o a ASE

47 39 ower, the limit might have reached earlier. However, there i a henomenal increae in the number of ubcriber with digital video broadcat catering to more than ten of thouand ubcriber in a general ene, while analog broadcat ervice catering to omething around five thouand ubcriber, inite of having more number of EDFA in the bu. However, the main concern i the number of ubcriber; in which indeed digital broadcat core high than the conventional analog video broadcat ervice. Figure (3.11) give the lot of the number of ubcriber with the number of WDM channel Fig Number of Subcriber in SAN with Number of WDM channel in all Digital video Broadcat along with witched ervice.

48 40 From Figure (3.11), it can be inferred that, uto 0 WDM channel can be tranmitted. Again, it can be noticed that QPSK i uorting more number of ubcriber and 56-QAM requiring the emloyment of FEC or recoding for better erformance. The key aumtion that wa carried out i the linearity of gain of an EDFA with the inut ignal ower that can be deicted in Figure (3.1). Fig 3.1 Deigning of a reamlifier. The linearity of the gain of EDFA with the inut ignal ower over the deired window of interet continue to be key aumtion in thi work. The algorithmic aroach remain the ame what it wa for Analog Video Broadcat, excet that CNR limitation ha to be relaced by the BER limitation at the receiver end. Thi digital video broadcat may be little bit exenive, in term of the infratructure, due to exenive et to boxe, but thi technology anwer to the ever increaing need for bandwidth, catering to more number of ubcriber, thu, it can be vouched a a technology for future. The reult were given for the cae when otical regeneration i not emloyed, the emloyment of which will give a more enhanced erformance. Thi will come into icture once graceful migration to all digital ervice i accomlihed.

49 Analyi of Hybrid Multichannel AM-VSB/M-QAM Video Lightwave Tranmiion ytem with witched ervice: The motivation behind thi art of the work i that multichannel AM-VSB/M- QAM SCM lightwave tranmiion ytem are being deloyed by telecom and cable televiion comanie for imultaneou delivery of both multichannel video (broadcat nature) and interactive digital video/data channel (witched nature). 64/56 QAM cheme offer a very high bandwidth efficiency in b//hz and are more robut than analog tranmiion w.r.t. random noie and nonlinear ditortion. The digital et to box, which oerate in either 64/56 QAM mode for downtream conideration enable a hot of emerging interactive ervice like Internet acce, Video-on-demand, Video treaming, IP telehony, HDTV. Now again, the uer bound on the erformance, in term of the number of uer and number of channel will given, along with an algorithmic aroach to the deign of SAN, i.e., to find the aroriate oerating oint. The iue in multichannel Hybrid AM-VSB/M-QAM are lot more different from the conventional deign iue. For hybrid analog/digital tranmiion over an otical fiber, the induced in-band cliing noie will ignificantly degrade the recetion erformance of M-QAM ignal. The BER of M-QAM i ignificantly degraded and even a BER floor occur. Thi BER degradation i mainly caued by cliing behaviour of laer diode. Thi cliing noie[18],[19] can be modeled a Gauian a well a non-gauian tatitic, however the imulive and non-gauian nature of cliing noie imoe more evere limitation for digital M-QAM ignal than the reviou Gauian aroximation. M- QAM channel ( in number) are located in higher frequency than AM-VSB channel ( in number) under the Gauian aroximation of the multichannel inut to the laer, N 1 N it i hown that the outut of the laer i the um of a ignal term related to laer inut and a noie term ariing from the cliing. Thi i the cliing noie and in the cae of a rare cliing, the cliing noie can be modeled a a Poion imule train with ule duration following Rayleigh ditribution and hae following a arabolic arc and, the mean ule duration i much maller than the ymbol duration of M-QAM. Utilizing the model, it can be hown that the aymtotic ditribution of cliing noie combined with the Gauian noie in the ytem aroache to the firt order aroximation, a

50 4 generalized Gauian ditributed model which i a um of Gauian ditributed art with the robability of no cliing occurrence and a non-gauian art with a robability of cliing occurrence. With a non-gauian tatitic for the cliing noie, the analytic exreion for the uncoded BER of M-QAM ignal in a hybrid multichannel AM- VSB/M-QAM i[0], [1]: 1 Pe = log where; i = 1,,3 Γ g = P φ ( z) = H k av 1 ( ) i 1. γ erfc M /( σ R ) k M = g 3Γ 1 1 ( M 1)(1 + iγ G ) ex( z ) ( z). π g + φ3( ).5γ. ( + γg) φ5( 3) + 5.5γ. (3 + γg) 3 + (1 γ ). erfc 1.5Γg ( M 1) (3.0) The lat term indicate the effect of Gauian noie alone while the ret of the term indicate the effect of both Gauian and cliing noie. Bandwidth of 64/56 QAM ignal i 6 MHz. (z) H k are Hermite olynomial (Aendix E). γ i the cliing index denoting the cliing robability er ymbol interval. σ g i the Gauian noie variance. in mq.( RGLP ) P av = (3.1) Gauian noie refer to the other kind of additive noie in the fiber due to Laer RIN, hot noie, thermal noie. in in σ = i + e( RGLP ) + ( RIN)( RGLP ) + 4Rηen g c where; L = 1/ M ( G 1) L GP in (3.) The firt term in Equation (3.) correond to thermal noie followed by Shot Noie, Relative Intenity Noie and Beat Noie reectively. Again, the litting at the ONU i aumed to be aive, the way it wa aumed in the earlier two cae. The cliing noie that wa modeled a Poion tatitic, the variance of which can be given a: σ = (4τ / 3). π I ( 3/ ) where; µ = N. m 1 a 5. µ.ex( 1/ µ ).( RGLP + N. m q in ) (3.3)

51 43 m a, m q are RMS modulation indice of AM and QAM channel., N N 1 are the number of AM and QAM channel reectively. µ i the RMS modulation index, τ i the mean duration of cliing imule [4], [5], that can be exreed mathematically a: τ = erfc 1/ µ ).(1/ γ. R ) (3.4) ( R i the ymbol rate (Baud Rate) of M-QAM. Equation (3.3) and (3.4) how that the cliing noie ha got contribution from both AM and QAM cliing ditortion. However, the dominant art will be from the cliing of AM ignal becaue the QAM ignal normally require much le ower than AM ignal. Both channel caacity and Outut Modulation Index (OMI) of M-QAM i limited by the reence of cliing noie which i non-gauian and imulive. At the receiver end, bet erformance for both analog and digital channel i deired, the analyi of which require ome arameter to be fixed a: The deired CNR at the receiver end i 48 db, o at the ONU, thi value i ket omewhere around 49.5 db, a the CNR of 48 db correond to icture judged ubjectively excellent a er CATV tandard. CSO and CTB lay a crucial role in the deign, they can be modeled collectively a CNLD. CNLD can be monitored individually alo, but a uch it i taken into effect along with CNR conideration, a the degradation in CNLD will ill into the CNR calculation, reulting in ubotimum erformance of SAN. So, thi erformance degrading imairment ha to be taken care of. BER ha to be better than 1e-9. Thi work carried out the deign of SAN with BER conideration of 1e-10. For lower BER, the non-gauian modeling of cliing noie require a much lower OMI for AM for the ame BER than doe the Gauian noie model. Becaue of the reence of non-gauian cliing noie, only a limited number of channel can be accommodated, if ignal of larger contellation ize like 64/56- QAM are ued. Large number of QAM channel can be accommodated if ower ratio of M-QAM channel to AM channel i mall enough.

52 44 Tranmitted ignal ower mut alo be high in the Gauian cliing noie modeling than non-gauian model for M-QAM ignal cae. RF ower ratio between AM and M-QAM ignal lay a vital role in ytem deign, a higher the ratio i, more the number of M-QAM channel can be accommodated without areciable change in the BER. CNRtotal = CNR + CNR + CNR + CNR + CNLD (3.5) hot thermal ig RIN Externally Modulated DFB Laer tranmitter i ued that ha a caability to rovide higher ower a broadband ubcriber acce network require higher ower budget. Beide thi, EM-DFB laer tranmitter have a built in mechanim for linearization and ureion of Stimulated Brillouin Scattering (SBS). Simulation have been carried out uing MATLAB to find out the otimum oerating oint for the deign of SAN, the reult of which are dicued below. Two cae are of interet in the deign of the framework, the ower from the tranmitter can be high or low, but there exit an otimum value below which CNR decreae with the increae in QAM modulation index. Although, at that oint, the required arameter are atified like BER=1e-10, but being le ower involved, le number of EDFA will be uorted on the bu and conequently, reduced number of ubcriber in SAN. Since QAM ignal ower i ket at lower level than AM ignal ower, o in a fully loaded hybrid AM- VSB/M-QAM, the AM CNR degradation i due to cliing ditortion from both AM and QAM channel uto the oint when the CNR tart to increae again, then ower of QAM channel become high enough (and QAM channel alo act a additional AM channel a at that oint and beyond that, QAM channel have got a retty high modulation index). Figure (3.13) deal with the digreion into thi argument, o a to make thing retty clear. The following concluion can be drawn from Figure (3.13): The key aumtion till hold here, i.e., Linearity in EDFA gain with the inut ignal ower. in out In the high ower cae; GP = P i the difference in both. Once the ower i low, thi out P decreae with increaing OMI, but in high ower cae, the increae reult in increaed CNR. Thi high ower regime will be the region of oeration, ince enhanced number of EDFA i uorted reulting in increaed number of

53 45 ubcriber in SAN. Thi ower that differentiate into two different regime of oeration ha to be found out through reeated imulation where the onet of CNR cree in, although thi decreae and increae in ower i very mall, a uch QAM channel being very le, it going to have very negligible effect. Fig Analyi of two regime of oeration in Hybrid AM-VSB/M-QAM tranmiion cae in SAN to find the otimum oerating oint. The deigning of a SAN can be carried out in imilarly in an algorithmic fahion. Since, the RF ower ratio between the QAM and AM ignal have a ay in the deign, thi contribute to the analyi of two cae, a when the QAM ignal ower i le comared to AM ignal ower and econdly, when they have a comarable ower. The analyi can be carried out on imilar line. Simulation Reult are hown for both the cae. The reaon thee two cae are conidered i that, the tranmiion of the QAM

54 46 channel in a hybrid multichannel AM-QAM ytem ha a trong deendence on the RF ower ratio between the AM and QAM channel. Starting with the former one: Ste 1: There are two aroache to the tart of the deign. Fix u the AM modulation index and find the number of AM channel accordingly. Thi maximum AM modulation deth i choen uch that CNLD criterion i atified, becaue a the RMS modulation index i increaed, CNLD ratio will degrade, o that the oint of AM modulation index correond to the maximum, where CNLD i jut atified, although anything le than that will have better CNLD. Once the AM modulation index i fixed u, let the number of AM channel be varied till the deired CNR of 49.5 db (a in thi cae) i obtained at the receiver end. If leer number of channel i incororated, then obviouly, there i a better erformance than the former cae. Fix the number of AM channel and find AM modulation index accordingly. Once the number i fixed, then otimum modulation index of AM ha to be found out, that give many number of AM channel with CNLD criterion fixed and CNR 49.5 db. That may come out to be the value of modulation index le than or equal to the maximum AM modulation deth, the way it wa found in other alternative. Thu, at the end of Ste 1, Number of AM channel, modulation index of AM for deired CNLD ha been determined when tarted by either of the two aroache. Ste : The number of ubcriber and the number of QAM channel have to be determined for BER< 1e-9 and SNR > 8 db. Both thee criterion determine the number of ubcriber that can be accommodated and the number of QAM channel that can be ent. Increaing the number of QAM channel imrove the SNR, but degrade the BER. However, increae in the number of ubcriber, reult in the degraded BER and SNR. So, the number of ubcriber and the number of QAM channel have to be choen in an otimum way, o a to atify the BER criterion. At the end of Ste, Number of QAM channel and number of ubcriber ha been fixed for a given Subcriber Acce Network. Ste 3: Figure (3.14) deict the reult obtained from reviou te, from

55 47 which, otimum modulation index of M-QAM ha to be found out at the minimum robability of error. BER i fixed at 1e-10 for the deign and the ret of the arameter have been found out for the reviouly mentioned criterion a CNR 49.5 db and in P 13 dbm (ower contraint being imoed arbitrarily, a ower conideration alo come into the icture in the deign of SAN. Ste 4: With modulation index of QAM being fixed a in te 3, the branch deign of the SAN i done with. The arameter already fixed will be ued for the bu deign of the SAN. P ASE uer limit ha to be found out in the imilar fahion a wa done in the reviou cae, with aive litting being aumed at the ONU tage. The method of analyi remain the ame, what wa carried out in the reviou analye, giving the number of EDFA uorted and conequently, number of ubcriber in the network. Fig Performance Analyi of AM-VSB/64-QAM in SAN when 64-QAM ignal i le a comared to AM-VSB ignal ower. From Figure (3.14), it can be inferred that the otimum QAM modulation index came out to be.5 for robability of error of 1e-10; thi arameter wa fixed, the lot

56 48 were taken o a to et the minimum to 1e-10 and ret of the arameter were found out from that oint. It can alo be inferred that 4 AM-VSB channel and 7 64-QAM channel are uorted. The AM modulation index wa fixed at 4.35% and the cliing index wa taken a 8e-3. The CNR correonding to that oint i db, which i a er the requirement of SAN, with ower alo 1.5 dbm and 000 ubcriber er branch. A the OMI of M-QAM ignal increae, the BER firt reduce, reaching a minimum oint, and then increae again, a the cliing of M-QAM ignal begin to take effect. The minimum BER increae when the channel number increae while the otimum OMI decreae. With uer er branch being 000, the imlementation of Ste 4 of the algorithm, give the maximum number of WDM channel uorted a 1 and number of EDFA uorted on the bu a 15, and conequently ome 30,000 number of ubcriber without otical regeneration in a SAN, a deicted in Figure (3.15). Fig Uer Bae Size and Channel caacity in AM-VSB/64-QAM in SAN when 64- QAM ignal i le a comared to AM-VSB ignal ower. When higher contellation ize, like 56-QAM i conidered, either the number of ubcriber be ket fixed and the number of QAM channel being found accordingly or

57 49 vice-vera, a number of QAM channel being fixed and the number of ubcriber, being reduced in number, found out. Starting with the former one, when the number of ubcriber i ket fixed at 000 and the te1-4 reeated, the number of 56-QAM channel being tranmitted turned out to be 18, which i far le a comared to 70 QAM channel a in the former cae, a hown in Figure (3.16). Fig 3.16 Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i le a comared to AM-VSB ignal ower when number of ubcriber i fixed. The otimum modulation index ha gone u to 5%, imlying more ower in the QAM channel, reulting in more cliing ditortion and hence increaed QAM SNR.

58 50 However, the number of channel and number of ubcriber remain the ame, becaue ower budget remain the ame with more uer to be accommodated being comenated by leer number of channel, with enhanced QAM OMI er channel. When the latter cae i conidered, i.e., when the number of 56-QAM channel i fixed at 70, and number of ubcriber earched for BER=1e-10, it reulted in 800 ubcriber at an otimum modulation index of. with CNR being db (a before) and SNR=34.79 db, with a minute difference in ower hown in Figure (3.17). Fig Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i le a comared to AM-VSB ignal ower when number of QAM channel i fixed. There i a dratic reduction in the number of ubcriber. So, either way, it can be adoted, a er requirement, going for more number of ubcriber or more number of QAM channel.

59 51 Fig Uer Bae Size and Channel caacity in AM-VSB/56-QAM in SAN when 56-QAM ignal i le a comared to AM-VSB ignal ower. The reaon for thi degraded erformance in QAM can be owed to reduction in immunity to ISI due to cliing induced imule train, reulting in QAM tate being cloe together. The modulation index of 56-QAM i alo much higher than that of 64- QAM for the ame number of ubcriber er branch to achieve the ame BER of 1e-10. Although, the BER erformance can be imroved by oerating the 56-QAM channel at higher QAM ignal level, the increaed QAM channel loading may caue the degradation of AM-channel quality. QAM ignal level for the tranmiion link wa found to be higher than that without EDFA. Thi i exected conidering the added ASE noie from EDFA [3]. The reaon, why only AM-to-QAM interference i conidered, i that dominant comonent of cliing noie come from the cliing of AM ignal, therefore QAM-to-QAM interference hall not be taken into account []. The model ha neglected the bandlimited effect of cliing noie by auming the noie to be white. However, the actual noie ectrum i bandlimited and not flat, that make the way for

60 5 aumtion a any finite bandwidth correction are mall. A the OMI of AM ignal increae, the robability of cliing increae and the ower ratio of Gauian noie comonent to cliing noie comonent reduce, thu the BER of M-QAM degrade ignificantly. For higher cliing robability, the emloyment of FEC become imerative, a alication of higher ower/ higher SNR of M-QAM doe not imrove BER very much. Thi work carried out the analyi by aumtion of low cliing robability denity. A the contellation ize increae, BER erformance degrade, ince higher the ize of M-QAM, the more tringent i the SNR required to obtain the ame Bit Error Rate. A iterated earlier, the tranmiion of QAM channel in a hybrid multichannel AM-VSB/M-QAM ytem ha a trong deendence on the RF ower ratio between AM and QAM channel, o, the other cae when, the QAM channel ower i comarable to AM channel ower will be conidered through imulation reult hown in Figure (3.19)-(3.4). Fig Performance Analyi of AM-VSB/64-QAM in SAN when 64-QAM ignal i comarable a comared to AM-VSB ignal ower.

61 53 For 64-QAM cae, emloying the ame et of te, otimum modulation index of.40 i obtained. The difference in the reult, from the earlier 64-QAM cae (in which the 64-QAM channel ower wa le a comared to AM channel ower) i that SNR ha gone u, becaue of the increae in ower) reulting in better erformance of SAN in term of the number of WDM channel that can be tranmitted and number of ubcriber that be accommodated. The otimum modulation index ha died a little becaue, a the number of QAM channel, that can be increaed by increaing the QAM ower, the otimum OMI of M-QAM decreae with the increae in BER. Ret of the analyi remain the ame, with the increae in QAM channel ower reulted in increment in the number of QAM channel uorted to76 in number. Figure (3.0) deict the uer bae ize and number of WDM channel obtained from the deign of the bu network. Fig.3.0 Uer Bae Size and Channel caacity in AM-VSB/64-QAM in SAN when 64- QAM ignal i comarable a comared to AM-VSB ignal ower.

62 54 When 56-QAM modulation cheme i analyzed, with the ubcriber ket fixed a before, and te followed in a imilar fahion a before, the number of QAM channel went down to 18 and otimum modulation index wa uto 5, o nothing much difference in the erformance excet that QAM SNR ha gone uto 35.4 db, with ubcriber bae ket at 000. Power ha gone above omewhere around 13 dbm owing to enhanced QAM channel ower. However, when the number of QAM channel i fixed at 76 with 4 AM-VSB channel, the number of ubcriber went down to 800 with otimum modulation index of.10 and ower being around 13 dbm a before. The above analyi i hown in Figure (3.1) and Figure (3.). Fig. 3.1 Performance Analyi of AM-VSB/56-QAM in SAN when 56-QAM ignal i comarable a comared to AM-VSB ignal ower when number of ubcriber i fixed.