Survey of Next-Generation Optical Access System Concepts
advertisement
Consortium groups the following Organizations: Partner Name Short name JCP-CONSULT SAS JCP DEUTSCHE TELEKOM AG DTAG INTERDISCIPLINARY INSTITUTE FOR BROADBAND IBBT TECHNOLOGY TECHNISCHE UNIVERSITAET MUENCHEN TUM KUNGLIGA TEKNISKA HOEGSKOLAN KTH ADVA AG OPTICAL NETWORKING ADVA ERICSSON AB EA ERICSSON TELECOMUNICAZIONI TEI ACREO AB. ACREO MAGYAR TELEKOM TAVKOZLESI NYILVANOSAN MAGYAR MUKODO RESZVENYTARSASAG TELEKOM RT SLOVAK TELEKOM AS SLOVAK TELEKOM UNIVERSITY OF ESSEX UESSEX Country FR D B D S D S I S H Survey of Next-Generation Optical Access System Concepts SK UK Deliverable 4.1 Abstract: Document to be considerate as Word Template for all deliverables OASE_D4.1_WP4_EAB_210612_v3.0.doc Version: 3.0 Last Update: 21/06/2012 “The research leading to these results hasDistribution received funding from the European Community's Seventh Level: Framework Programme (FP7/2007-2013) under agreement n° 249025” PU grant (Public) Distribution level PU = Public, RE = Restricted to a group of the specified Consortium, PP = Restricted to other program participants (including Commission Services), CO= Confidential, only for members of the OASE Consortium (including the Commission Services) Partner Name DEUTSCHE TELEKOM AG INTERDISCIPLINARY INSTITUTE FOR BROADBAND TECHNOLOGY KUNGLIGA TEKNISKA HOEGSKOLAN ADVA AG OPTICAL NETWORKING ERICSSON AB ERICSSON TELECOMUNICAZIONI S.P.A ACREO AB. MAGYAR TELEKOM TAVKOZLESI NYILVANOSAN MUKODO RESZVENYTARSASAG UNIVERSITY of ESSEX Short name Country DTAG D IBBT B KTH ADVA EA TEI ACREO MT S D S I S H UESSEX UK Abstract: This deliverable presents a survey of system concepts for next-generation optical access based on OASE requirements. For the considered systems, key evolving components/subsystems are identified and a survey for each component/subsystem presented. A broad range of system concepts and key technologies are covered, including key aspects such as maturity, cost, capacity and reach. “The research leading to these results has received funding from the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 249025” O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 2 of 168 Document Identity Title: Subject: Number: File name: Registration Date: Last Update: Survey of Next-Generation Optical Access System Concepts WP4 “System concepts for Next Generation optical access networks” D4.1 OASE_D4.1_WP4_EAB_210612_v3.0.doc th March, 12 2010 21/06/2012 Revision History No. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Version 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 0 Comments: 1 Comments: 2 Comments: 3 Comments: Edition 1 Author(s) A. Bianchi (TEI) Date 03/12/2010 2 B. Skubic (EAB) 04/10/2010 Sub-chapters brought out of document and edited individually 3 B.Skubic (EAB), A. Bianchi (TEI) 10/8/2010 Compilation of contributions from ADVA, ACREO, DTAG, EAB, IBBT, KTH, TEI, MT, UESSEX 4 B. Skubic (EAB), A. Bianchi (TEI) 10/12/2010 Version released for internal review 5 M. Parker (UESSEX) 10/21/2010 6 B. Skubic (EAB) 10/25/2010 7 M. Roppelt (ADVA) 10/25/2010 8 J. Chen (KTH) 10/25/2010 9 B. Lannoo (IBBT) 10/27/2010 10 B. Skubic (EAB) Included update from KTH and DTAG 11 A. Bianchi (TEI) 10/27/2010 12 J. Chen (KTH) 10/27/2010 13 M. Roppelt (ADVA) 10/27/2010 14 B. Lannoo (IBBT) 10/28/2010 15 B. Skubic (EAB) Final editing 0 E. Malésys (JCP) Formatting/ released version 0 B. Skubic (EAB) Incorporation of reviewers’ comments 0 B. Skubic (EAB) Incorporation of reviewers’ comments O O A S OA AS SEEE FP7 – ICT– GA 249025 10/27/2010 10/29/2010 10/29/2010 06/06/2011 21/06/2012 Page: 3 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table of Contents EXECUTIVE SUMMARY ................................................................................................................................... 6 REFERRED DOCUMENTS ................................................................................................................................ 8 FIGURES AND TABLES .................................................................................................................................. 16 ABBREVIATIONS ............................................................................................................................................. 21 1. INTRODUCTION ..................................................................................................................................... 29 2. TASK DESCRIPTION AND SCOPE ...................................................................................................... 29 2.1 3. ASSUMPTIONS AND METHODOLOGY ............................................................................................. 31 3.1 3.2 3.3 4. NGOA SYSTEM CONCEPTS.................................................................................................................. 31 KEY ASPECTS FOR THE ASSESSMENT ................................................................................................... 33 COST AND POWER ASSUMPTIONS ........................................................................................................ 35 SYSTEM CONCEPTS FOR NGOA ........................................................................................................ 36 4.1 4.1.1 4.1.2 4.1.3 4.1.4 4.2 4.2.1 4.2.2 4.2.3 4.3 4.3.1 4.3.2 4.4 4.5 4.5.1 4.5.2 4.5.3 4.5.4 4.5.5 4.6 4.6.1 4.6.2 4.6.3 4.7 4.7.1 4.7.2 4.8 4.8.1 4.8.2 4.8.3 4.8.4 5. RELATIONSHIP TO OTHER WORKPACKAGES ........................................................................................ 30 TDM-PON ......................................................................................................................................... 36 Serial 40G NRZ ............................................................................................................................. 40 Serial 40G QPSK .......................................................................................................................... 41 Serial 40G DQPSK ....................................................................................................................... 41 Serial 4x10G NRZ ......................................................................................................................... 42 WDM-PON ........................................................................................................................................ 43 WDM-PON with (Tunable) Lasers and Laser-Arrays ................................................................... 43 WDM-PON with seeded Reflective Transmitters .......................................................................... 50 WDM-PON with wavelength reuse ............................................................................................... 55 OFDM-PON ....................................................................................................................................... 57 Serial 40G CO-OFDM .................................................................................................................. 58 Serial 40G DDO-OFDM ............................................................................................................... 59 CDM .................................................................................................................................................. 61 WDM+XXM HYBRID ........................................................................................................................ 66 Hybrid WDM/TDM-PON .............................................................................................................. 67 Hybrid WDM/CDM-PON .............................................................................................................. 72 Hybrid WDM/OFDM-PON ........................................................................................................... 77 Hybrid WDM/SCM-PON............................................................................................................... 79 UDWDM ....................................................................................................................................... 80 HYBRID ACTIVE WDM ....................................................................................................................... 81 Hybrid active WDM with active P2P access ................................................................................. 81 Hybrid active WDM-PON ............................................................................................................. 82 Comparison of active/passive hybrids and WDM-PON ................................................................ 83 NEXT GENERATION AON: RPR, WDM, ETHERNET ........................................................................... 85 GbE access .................................................................................................................................... 85 10 Gb/s Ethernet access and distribution/aggregation ................................................................. 86 RADIO OVER FIBRE BACKHAULING ..................................................................................................... 87 Radio-over-Fibre (RoF) Technologies .......................................................................................... 89 CPRI .............................................................................................................................................. 90 OBSAI............................................................................................................................................ 90 Hybrid Optical-Wireless PON Architectures ................................................................................ 91 SURVEY OF EVOLVING COMPONENTS/SUBSYSTEMS ............................................................... 92 5.1 WDM COMPONENTS/SUBSYSTEMS .................................................................................................... 93 5.1.1 Wavelength selective components for WDM systems .................................................................... 93 5.1.2 Example of FTTH-PON-based on ASE-injected FP-LDs .............................................................. 94 5.2 ONT BASED ON RSOA, REAM, REAT .............................................................................................. 95 5.2.1 ONT based on RSOA ..................................................................................................................... 95 5.2.2 ONT based on REAM .................................................................................................................... 98 5.2.3 ONT based on REAT ................................................................................................................... 100 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 4 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 5.3 TUNABLE LASERS ............................................................................................................................. 101 5.4 WAVELENGTH SELECTIVE RECEIVERS............................................................................................... 115 5.4.1 Tunable Filters ............................................................................................................................ 115 5.4.2 Coherent Receivers ..................................................................................................................... 122 5.5 BURST MODE RECEIVERS .................................................................................................................. 133 5.6 ADC/DAC ....................................................................................................................................... 140 5.7 DISPERSION COMPENSATION ............................................................................................................. 143 5.7.1 Compensation at the transmitter ................................................................................................. 143 5.7.2 Compensation at the receiver ...................................................................................................... 143 5.8 PASSIVE WAVELENGTH SELECTIVE DEVICES ..................................................................................... 144 5.8.1 Thin film filter-based WDM components .................................................................................... 144 5.8.2 Fibre Bragg grating-based WDM components ........................................................................... 145 5.8.3 Etched Diffraction Grating-based WDM components ................................................................ 146 5.8.4 Arrayed Waveguide Grating-based WDM components .............................................................. 147 5.8.5 Technology constrains for EDG and AWG WDM components ................................................... 149 5.9 REACH EXTENDERS .......................................................................................................................... 149 5.10 SWITCHING COMPONENTS ................................................................................................................. 153 5.10.1 Power consumption ................................................................................................................ 153 5.10.2 Challenges on high speed switching process: ........................................................................ 154 5.10.3 Network stability ..................................................................................................................... 154 6. SUMMARY .............................................................................................................................................. 155 7. APPENDIX .............................................................................................................................................. 160 7.1 7.2 7.3 COST AND ENERGY CONSUMPTION DATA ........................................................................................ 160 EXAMPLES DEVICES ON THE MARKET BASED ON TFF ....................................................................... 161 EXAMPLES DEVICES ON THE MARKET BASED ON AWG..................................................................... 165 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 5 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Executive summary This document provides a survey of system concepts/technologies with potential of meeting the requirements of next-generation optical access (NGOA) as defined in OASE Deliverable 2.1. Key requirements include residential peak data rates of ≥1 Gb/s, support for 256 to 1024 customers per feeder fibre, support for 20 to 40 km passive reach, and support for 60 to 90 km extended reach for the protection path. For each next-generation system candidate, a detailed technical description is made as well as a coarse investigation with respect to key aspects such as bandwidth, number of customers per feeder fibre, reach, cost and power consumption. Furthermore, based on the presented survey of next-generation systems, key evolving components/subsystems are identified and investigated as a basis for an assessment for system maturity and time to market. A detailed technical description of each evolving component/subsystem is made as well as a coarse assessment based on key aspects. Key components include WDM components/subsystems, tunable lasers, reflective transmitters, wavelength selective receivers (such as tunable filters and coherent receivers), burst mode receivers, analogue-to-digital converters (ADC), digitalto-analogue converters (DAC), dispersion compensation and passive wavelength selective devices. Reach extension technology as well as switching technologies are also discussed. The main results for each considered system concept are summarized below: - Time division multiplexing (TDM) – passive optical networks (PON): Options based on higher rate on-off keying as well as advanced modulation and coherent detection are investigated. TDM-PON approaches present significant challenges with respect to simultaneously increasing bandwidth, reach and splitting ratio, whilst maintaining low cost and power consumption. - Active optical networks (AON): AON imply either large port count at the central office in a centralized architecture or low degree of node consolidation in a decentralized architecture. Both these factors have implications on cost and power consumption. - WDM-PON: The WDM-PON concepts are here categorized based on optical network unit (ONU) design, i.e. based on tunable lasers, reflective lasers or reflective lasers employing wavelength reuse. Ultra-dense WDM-PON based on coherent detection is also considered. Advantages of WDM-PON include long reach and large per customer sustainable bandwidth. A drawback is the limited fan-out of traditional WDM-PON approaches, limiting the number of customers per feeder fibre. - Orthogonal frequency division multiplexing (OFDM)-PON: OFDM is based on data transmission over several densily spaced subcarriers. The OFDM format, with its tolerance to chromatic dispersion, offers long reach and large resource flexibility. Several variants of OFDM-PON are described. OFDM-PON presents similar challenges as TDM-PON concerning limited total capacity and limited splitting ratio. System complexity and processing requirements present challenges with respect to power consumption and cost. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 6 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 - Code Division Multiplexing (CDM): Different CDM implementation options and fanout possibilities are discussed with focus on coherent optical CDM using onedimensional code words. CDM presents significant challenges with respect to fan-out requirements. - Radio-over-fibre technologies and digital interfaces for radio access network (RAN) transport are also discussed considering their potential role for future RAN architecture over the NGOA. - Hybrid concepts: Hybrid concepts are motivated by the fact that each of the aforementioned pure system concepts individually may have difficulties in fulfilling the complete set of NGOA requirements. Hence, hybrid concepts have been proposed that combine advantages of different concepts. Typically the advantages that are exploited are the increased overall capacity of WDM and the efficient resource sharing of TDM, OFDM or CDM. Concepts that involve different types of active remote nodes have also been considered. A preliminary cost and power consumption analysis of different variants shows that the most promising configurations are hybrid WDM/TDM-PON as well as various active hybrid variants (WDM-PON with AON access and two stage WDM-PON). As an outcome of the survey it is seen that the main candidates for further consideration within the OASE project based on the posed requirements are different variants of pure WDM-PON, hybrid WDM/TDM-PON, AON as well as various active hybrid variants based on WDM-PON and AON. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 7 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Referred documents Project Contract - Annex 1 “Description of the Work” (DoW) OASE deliverable D2.1 – “Requirements” Gigabit-capable passive optical networks (GPON), ITU-T G.984.x series of recommendations [Online]. Available: http://www.itu.int/rec/T-REC-G/e [4] IEEE Standard for Information technology-Specific requirements - Part 3 [Online], IEEE Standard 802.3, 2008. Available: http://standards.ieee.org/getieee802/802.3.html [5] 10-Gigabit-capable passive optical networks (XG-PON), ITU-T G.987.x series of recommendations [Online]. Available: http://www.itu.int/rec/T-REC-G/e [6] J. Kani and K. Suzuki, "Standardization Trends of Next-generation 10 Gigabit-class Passive Optical Network Systems," NTT Technical Review, Vol. 7, No. 11, November 2009. [7] Optical system design and engineering considerations, ITU-T Rec. G.Sup39 (12-2008). [8] J. Yu, et al., “Applications of 40-Gb/s chirp-managed laser in access and metro networks,” IEEE Journal of Lightwave. Technology, Vol. 27, No. 3, pp. 253–255, February 2009. [9] B. Mason, et al., “40Gb/s photonic integrated receiver with –. 17dBm sensitivity,” OFC2002, March 2002, FB10. [10] OFDM Introduction: Lecture Online, http://bwrc.eecs.berkeley.edu/classes/ee225c/ Lectures/Lec16_ofdm.pdf [11] Characteristics of a single-mode optical fibre cable, ITU-T G.652 series of recommendations (10-2000). [12] I. Van de Voorde, C. Martin, J. Vandewege, and X.Z. Qiu, “The SuperPON Demonstrator: an Exploration of Possible Evolution Paths for Optical Access Networks”, IEEE Communications magazine, pp. 74-82, February 2000. [1] [2] [3] [13] C. Antony, et al., “Demonstration of a Carrier Distributed, 8192‐Split Hybrid DWDM ‐TDMA Pon Over 124km Field‐Installed Fibers” OFC 2010, San Diego, March 2010, PDPD8. [14] F. Ponzini, et al., “Evolution scenario towards WDM-PON [Invited],” J. Opt. Commun. Network, Vol. 1, No. 4, September 2009. [15] K. Kanonakis, et al., “ACCORDANCE: A Novel OFDMA-PON Paradigm for UltraHigh Capacity Converged Wireline-Wireless Access Networks,” ICTON 2010, Munich 2010. [16] N. Cvijetic, D. Qian, J. Hu, and T. Wang, "44-Gb/s/λ Upstream OFDMA-PON Transmission with Polarization-Insensitive Source-Free ONUs," OFC2010, San Diego, March 2010, OTuO2. [17] W. Shieh, H. Bao, and Y. Tang, “Coherent optical OFDM: theory and design,” Optical Express, Vol. 16, No. 2, p. 841, January 2008. [18] B. J. C. Schmidt, A. J. Lowery, and J. Armstrong, “Experimental Demonstrations of Electronic Dispersion Compensation for Long-Haul Transmission Using DirectDetection Optical OFDM,” IEEE J. Lightwave Technology, Vol. 26, No. 1, January 2008. [19] D. Qian, N. Cvijetic, J. Hu, and T. Wang, “108 Gb/s OFDMA-PON with Polarization Multiplexing and Direct Detection,” OFC 2009, San Diego, March 2009, PDPD5. [20] T. N. Duong, et al., "Very High Bit Rate Transmission for NGPON using AMOOFDM Direct Modulation of Linear Laser," OFC 2010, San Diego, March 2010, OTuO3. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 8 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [21] X. Wang, et al., “Experimental demonstration of 511-chip 640 Gchip/s superstructured FBG for high performance optical code processing,” ECOC2004, Stockholm, September 2004, Tu1.3.7. [22] S. Etemad et al., “Spectrally efficient optical CDMA using coherent phase-frequency coding,” IEEE Photonics Technology Letters, Vol. 17, No. 4, pp. 929–931, April 2005. [23] Z. Jiang, et al., “Four-user 10-Gb/s spectrally phase-coded O-CDMA system operating at ~30 fJ/bit,” IEEE Photonics Technology Letters, Vol. 17, No. 3, pp. 705–707, March 2005. [24] B. Chen, et al., “Add/drop multiplexing and TDM signal transmission in an optical CDMA ring network,” OSA J. Optical Networking, Vol. 6, pp. 969-974, 2007. [25] S.P. Wan and Y. Hu, “Two-dimensional optical CDMA differential system with Prime/OOC codes,” IEEE Photonics Technology Letters, Vol. 13, pp. 1373–1375, 2001. [26] S.M.R.M. Nezam, et al., “Experimental Demonstration of OCDMA Transmission Using a Three-Dimensional (Time-Wavelength-Polarization) Codeset”, IEEE J. Lightwave Technology, Vol. 23, pp. 3282-3289, 2005. [27] K. I. Kitayama, X. Wang, and N. Wada, “OCDMA Over WDM PON – Solution Path to Gigabit-Symmetric FTTH”, IEEE/OSA J. Lightwave Technology, Vol. 24, pp. 16541662, 2006. [28] K. Fouli, and M. Maier, “OCDMA and optical coding: principles, applications, and challenges,” IEEE Commun. Mag., Vol. 45, pp. 27-34, 2007. [29] X. Wang, et al., “Demonstration of Over 128Gb/s Capacity (12-User x 10.71 Gb/s/User) Asynchronous OCDMA Using FEC and AWG-Based Multipoint Optical Ecoder/Decoder,” IEEE Photonics Technology Letters, Vol. 18, pp. 1603-1605, 2006. [30] N. Kataoka, N. Wada, G. Cincotti, K. Kitayama, ” Dispersion-Compensation-Free, Long-Reach OCDMA-PON System with Passive Remote Node at Arbitrary Position using Single Multi-port Encoder/Decoder”, CLEO, 2010. [31] C. S. Brès, I. Glesk, R. J. Runser, and P. R. Prucnal, “All-optical OCDMA code-drop unit for transparent ring networks,” IEEE Photonics Technology Letters Vol. 17, pp. 1088–1090, 2005. [32] D. Wake. Radio over Fiber Systems for Mobile Applications. In: Radio over Fiber Technologies for Mobile Communications Networks. Ed.: H. Al-Raweshidy, S. Komaki. Artech, Boston, pp. 217-241, 2002. [33] D. G. Moodie, et al., “Applications of electroabsorption modulators in high bit-rate extended reach transmission systems,” OFC 2003, Invited Paper TuP1, pp. 267-268, 2003. [34] H. Al-Raweshidy and S. Komaki (ed.), Radio over Fiber Technologies for Mobile Communications Networks, Artech House 2002. [35] Ken-Ichi Kitayama, et al., “An Approach to single component antenna base stations for broadband millimeter-wave fiber-radio access systems,” IEEE Transactions on Microwave Theory and Techniques, Vol. 48, No. 12, pp. 2588-2595, 2000. [36] CPRI Specification V4.1 (2009-02-18): “Common Public Radio Interface (CPRI); Interface Specification”, Online, http://www.cpri.info/ [37] IEEE Std 802.3-2005: "Part 3: Carrier sense multiple access with collision detection (CSMA/CD) access method and physical layer specifications", December 2005. [38] ISO/IEC 14165-115 – Information Technology – Fibre Channel Part 115 Physical Interface (FC-PI), February 2006. [39] INCITS (ANSI) Revision 8, T11/08-138v1– Fibre channel Physical Interface-4 (FC-PI4), May 2008. [40] Infiniband Trade Associationm, “Infiniband Architecture”, Rel. 1.1, Vol. 2, November 2002. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 9 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [41] Open Base Station Architecture Initiative BTS System Reference Document (Version 2.0). [42] L.G. Kazovsky, et al., “Next-Generation Optical Access Networks,” IEEE Journal of Lightwave Technology, Vol. 25, No. 11, November 2007. [43] C. Hoymann, “Analysis and performance evaluation of the OFDM-based metropolitan area network IEEE 802.16,” Computer Networks (Elsevier), Vol. 49, Issue 3, October 2005. [44] IEEE 802.11 WG, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification, 1999. [45] Y. Xiao, “IEEE 802.11n: Enhancements for Higher Throughput in Wireless LANs,” IEEE Wireless Communications, Vol. 12, Issue 6, pp. 82-91, December 2005. [46] W.-T. Shaw, et al., “Hybrid Architecture and Integrated Routing in a Scalable OpticalWireless Access Network,” IEEE Journal of Lightwave Technology, Vol. 25, No. 11, November 2007. [47] S. Sarkar, S. Dixit, and B. Mukherjee, “Hybrid Wireless-Optical Broadband Access Network (WOBAN): A Review of Relevant Challenges,” IEEE Journal of Lightwave. Technology, Vol. 25, No. 11, November 2007. [48] G. Shen, R.S. Tucker and C.-J. Chae, “Fixed Mobile Convergence Architectures for Broadband Access: Integration of EPON and WiMAX,” IEEE Commun. Mag., Vol. 45, Issue 8, pp. 44-55, August 2007. [49] S.-J. Park, et al., “Fiber-to-the-Home Services Based on Wavelength-DivisionMultiplexing Passive Optical Network,” IEEE Journal of Lightwave Technology, Vol. 22, Issue 11, pp. 2582-2591, November 2004. [50] J.-M. Kang, S.-K. Han, “A Novel Hybrid WDM/SCM-PON Sharing Wavelength for Up- and Down-Link Using Reflective Semiconductor Optical Amplifier,” IEEE Photonics Technology Letters, Vol. 18, No. 3, February 2006; E. Wong, K. L. Leem, T. B. Anderson, “Directly Modulated Self-Seeding Reflective Semiconductor Optical Amplifiers as Colorless Transmitters in Wavelength Division Multiplexed Passive Optical Networks,” IEEE Journal of Lightwave Technology, Vol. 25, Issue 1, pp. 67-74, 2007. [51] P. Healey, et al., “Spectral slicing WDM-PON using wavelength-seeded reflective SOAs,” Electronics Letters, Vol. 37, No. 19, pp. 1181-1182, September 2001. [52] Berrettini et al., “Testbed for experimental analysis on seamless evolution architectures from GPON to high capacity WDM-PON,” Int. J. Communication Networks and Distributed Systems, Vol. 5, No. 1-2, pp. 193-210, 2010. [53] M.S. Lee, B-T Lee, B.Y. Yoon and B.T. Kim, “Bidirectional amplification with linear optical amplifier in WDM-PON,” OFC2006, San Diego, March 2006. [54] Ultra-wideband Radio over Optical Fibre (UROOF). FP6 European integrated project. www.ist-uroof.org [55] Fully-converged quintuple-play integrated optical-wireless access architectures (FIVER). FP7 STREP project. www.ict-fiver.eu [56] J. Buus, and E.J. Murphy, “Tunable Lasers in Optical Networks,” IEEE Journal of Lightwave Technology, Vol. 24, No. 1, January 2006. [57] L.A. Coldren, “Tunable Semiconductor Lasers – a tutorial (Invited),” OFC2003, Atlanta, March 2003, ThF1, Online: http://www.ece.ucsb.edu/Faculty/Coldren/papers/TSL.pdf [58] L.A. Coldren, et al., “Tunable Semiconductor Lasers: A Tutorial,” IEEE Journal of Lightwave Technology, Vol. 22, No. 1, January 2004. [59] S. Dahlfort, et al., “Tunable Transmitters and receivers for WDM PON,” FSAN NGPON2 Workshop, San Francisco, February 2010. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 10 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [60] G. Busico, “Introduction to DS-DBR,” Oclaro technical slides, 2009, Contact: giacinto.busico@oclaro.com [61] A.J. Ward, et al., “Widely Tunable DS-DBR Laser With Monolithically Integrated SOA: Design and Performance,” IEEE J. Selected Topics In Quantum Electronics, Vol. 11, No. 1, January/February 2005. [62] Y. Liu, et al., “Directly-Modulated DS-DBR Tunable Laser for Uncooled C-band WDM System,” OFC2006, Anaheim, March 2006. [63] K. Grobe, and D. Nesset, “Low-cost Tunables for WDM-PON,” FSAN NG-PON2 Workshop, San Francisco, February 2010. [64] M.J. Wale, et al., “Technology Routes for Cost-Effective, High Performance, NextGeneration Access Components,” EPIC Workshop, FTTH Council Europe, Copenhagen, February 2009. [65] S. Zou, et al., “100mW phase-shifted 1550nm BH DFB arrays with a 10-micron pitch,” OFC2005, Anaheim, March 2005, JWA28. [66] G.W. Yoffe, et al., “Widely-tunable 30mW laser source with sub-500kHz linewidth using DFB array,” 21st LEOS Annual Meeting, Newport Beach, November 2008, ThBB2. [67] J.P. Wilde, G.W. Yoffe, and J.M. Kahn, “Frequency Noise Characterization of a Widely Tunable Narrow-Linewidth DFB Laser Array Source,” NFOEC2009, San Diego, March 2009, JWA33 [68] ChemOptics, “Preliminary Technical Specifications Tunable Laser Module”, January 2009, Contact: 104-11 Moonji-Dong, Yusung-Gu, Deajeon 305-380, South Korea, URL: www.chemoptics.co.kr, E-mail: sales@chemoptics.co.kr [69] H. Rohde, “Questions for Tunable/selectable TRx technology - Capability of Tunable/Selectable TxRx,” FSAN NG-PON2 Workshop, San Francisco, February 2010. [70] M. Jiang, et al., “Error Free 2.5 Gb/s transmission over 125 km conventional fiber of a directly modulated widely tunable vertical cavity surface emitting laser,” OFC2001, Anaheim, March 2001, TuJ3-1. [71] P. Debernardi et al. “Modal Properties of Long-Wavelength Tunable MEMS-VCSELs With Curved Mirrors: Comparison of Experiment and Modeling,” IEEE Journal of Quantum Electronnics, Vol. 44, No. 4, April 2008. [72] C.C. Renaud, et al. “100 GHz Spaced 10 Gbit/s WDM over 10 °C to 70 °C using an uncooled DBR laser,” OFC2006, Anaheim, March 2006, OWI76. [73] D. Sadot, and E. Boimovich, “Tunable Optical Filters for Dense WDM Networks,” IEEE Commun. Mag., Vol. 36, No. 12, December 1998. [74] Hyung-Kew Lee, et al., “A Wide-Range Linearly Tunable Optical Filter Using Lorentz Force,” IEEE Photonics Technology Letters, Vol. 16, No. 9, September 2004. [75] Hailiang Zhang, et al. “Novel tunable liquid crystal Fabry-Perot filters for Fiber-Optical system, in: Metro and Access Networks,” Wanyi Gu, Jianhui Zhou, Jin-Yi Pan, Editors, Proceedings of SPIE, 64 Vol. 4583, 2001. [76] D. Hohlfeld, et al. “Tunable Thermo-Optic Filter for WDM Applications,” Proc. IEEE, pp. 564-567, August 2002. [77] T. Nakazawa, et al. “Ti:LiNbO3 Acousto-Optic Tunable Filter (AOTF),” FUJITSU Sci. Tech. J., Vol. 35, No. 1, pp. 107-112, July 1999. [78] M. Ferchichi, et al. “Design of Temperature-Strain Tunable UDWDM, DWDM, WDM FBG filter for Passive Optical Network Access,” 2nd ICTON - Mediterranean Winter 2008, Marrakech, Morocco, December 2008, Fr1A.5 [79] M. Lequime, “Tunable thin-film filters: review and perspectives,” Proceedings SPIE Vol. 5250 (Advances in Optical Thin Films), February 2004. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 11 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [80] F. Havermeyer, et al. “Volume holographic grating-based continuously tunable optical filter,” SPIE Journal Opt. Eng., Vol. 43, No. 9, September 2004. [81] Fiber Fabry-Perot Tunable Filter | FFP-TF2, Data Sheet, Micron Optics, Inc. 2008, Online: http://www.micronoptics.com/uploads/library/documents/Datasheets/Micron%20Optics %20FILTERS%20-%20FFP-TF2%20%20-%200812.1.pdf [82] Thin-film Variable Filters, Data Sheet, Lightwaves2020, 2010, Online: http://www.lightwaves2020.com/publications/lightvision/newsletter-Feb-2010-4p.pdf [83] R. Murano, et al. “Tunable 2.5Gb/s Receiver for Wavelength-Agile DWDM-PON,” OFC2008, San Diego, March 2008. [84] R. Murano, et al. “Tunable GPON Receivers Enable Phased Migration to 1Gb/s per Subscriber,” OFC2009, San Diego, March 2009. [85] F. Riemenschneider, et al. “Low-Cost Electrothermally Tunable Optical Microcavities Based on GaAs,” IEEE Photonics Technology Letters, Vol. 14, No. 11, November 2002. [86] F. Derr, “Coherent Optical QPSK Intradyne System: Concept and Digital Receiver Realization,” IEEE Journal of Lightwave Technology, Vol. 10, No. 9, September 1992. [87] C.R.S. Fludger, et al. “10 x 111 Gbit/s, 50 GHz spaced, POLMUX-RZ-DQPSK transmission over 2375 km employing coherent equalisation,” OFC2007, Anaheim, March 2007, PDP22. [88] C.R.S. Fludger, et al. “Coherent Equalization and POLMUX-RZ-DQPSK for Robust 100-GE Transmission,” IEEE Journal of Lightwave Technology, Vol. 26, No. 1, January 2008. [89] F. Gini, G. B. Giannakis, “Frequency Offset and Symbol Timing Recovery in FlatFading Channels: A Cyclostationary Approach,” IEEE Trans. Communications, Vol. 46, No. 3, March 1998. [90] F. M. Gardner, “A BPSK/QPSK timing-error detector for sampled receivers, IEEE Trans. Communications,” Vol. COM-34, pp. 423-429, May 1986. [91] M. Oerder, and H. Meyr, “Digital Filter and Square Timing Recovery,” IEEE Trans. Commun., Vol. COM-36, pp. 605-612, May 1988. [92] S.J. Savory, et al. “Digital Coherent Receivers for Uncompensated 42.8Gbit/s Transmission over High PMD Fibre,” ECOC2007, Berlin, September 2007. [93] S.J. Savory, et al. “Electronic compensation of chromatic dispersion using a digital coherent receiver,” Opt. Express, Vol. 15, Issue 5, pp. 2120-2126, 2007. [94] C.R. Johnson, et al. “Blind Equalization Using the Constant Modulus Criterion: A Review,” Invited Paper, Proc. IEEE, Vol. 86, No. 10, October 1998. [95] A. Vgenis, et al. “Nonsingular Constant Modulus Equalizer for PDM-QPSK Coherent Optical Receivers,” IEEE Photonics Technology Letters, Vol. 22, No. 1, January 2010. [96] B. Widrow and M. E. Hoff, “Adaptive Switching Circuits,” IRE WESCON Conv. Rec., Part 4, pp. 96–104, 1960. [97] Guifang Li, “Recent advances in coherent optical communication,” Advances in Optics and Photonics, Vol. 1, Issue 2, pp. 279–307, February 2009. [98] Yan Han and Guifang Li, “Coherent optical communication using polarization multipleinput-multiple-output,” OSA Optics Express, Vol. 13, No. 19, pp. 7527-7534, September 2005. [99] A.J. Viterbi and A.N. Viterbi, “Nonlinear Estimation of PSK-Modulated Carrier Phase with Application to Burst Digital Transmission,” IEEE Trans. Information Theory, Vol. IT-29, No. 4, July 1983. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 12 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [100] M.G. Taylor, “Phase Estimation Methods for Optical Coherent Detection Using Digital Signal Processing,” IEEE Journal of Lightwave Technologies, Vol. 27, No. 7, April 2009. [101] M.G. Taylor, “Coherent Detection for Optical Communications using Digital Signal Processing,” OFC2007, Anaheim, March 2007. [102] N. Wiener, Extrapolation, Interpolation, and Smoothing of Stationary Time Series, Wiley, New York 1949. [103] Shaoliang Zhang, et al. “Novel Ultra Wide-Range Frequency Offset Estimation for Digital Coherent Optical Receiver,” OFC2010 (EDC Session), San Diego, March 2010. [104] T. Pfau, et al. “Coherent optical communication: Towards realtime systems at 40 Gbit/s and beyond,” OSA Optics Express, Vol. 16, No. 2, pp. 866, January 2008. [105] Ip, E., et al. “Coherent detection in optical fiber systems,” OSA Optics Express, Vol. 16, No. 2, pp. 753, January 2008. [106] D.S. Millar, et al. “Experimental Comparison of Nonlinear Compensation in Long-Haul PDM-QPSK Transmission at 42.7 and 85.4 Gb/s,” ECOC2009, Vienna, September, 2009, Paper 9.4.4. [107] P.J. Winzer, et al. “56-Gbaud PDM-QPSK: Coherent Detection and 2,500-km Transmission,” ECOC2009, Vienna, September 2009, PD2.7. [108] J.E. Simsarian, et al., “Fast-Tuning 224-Gb/s Intradyne Receiver for Optical Packet Networks,” OFC2010, San Diego, March 2010, PDPB5. [109] Akihide Sano, et al., “240-Gb/s Polarization-Multiplexed 64-QAM Modulation and Blind Detection Using PLC-LN Hybrid Integrated Modulator and Digital Coherent Receiver,” ECOC2009, Vienna, September 2009, PD2.2. [110] J.M. Fabrega, and J. Prat, “New intradyne receiver with electronic-driven phase and polarization diversity,” OFC2006, Anaheim, March 2006, JThB45. [111] M. Noda, S. Yoshima, K. Ishii, S. Shirai, M. Nogami, J. Nakagawa, “Dual-rate Optical Transceiver incorporating Fully Optimized Burst-mode AGC/ATC Functions for 10GEPON Systems” , ECOC 2010, Turin, September 2010, Mo.2.B.2 [112] Gigabit-Capable Passive Optical Networks (GPON): PMD Layer Specifications, ITU-T Recommendation G.984.2, 2003. [113] MAC Parameters, Physical Layers and Management Parameters for Subscriber Access Networks, IEEE Draft P802.3ah, 2003. [114] C. Su et al., “Theory of Burst-Mode Receivers and its Applications in Optical Multiaccess Networks,” IEEE Journal of Lightwave Technology, Vol. 15, No. 4, pp. 590-606, 1997. [115] N. Suzuki, et al.,“Single Platform 10G-EPON 10.3-Gbps/1.25-Gbps Dual-Rate CDR with Fast Burst-Mode Lock Time Employing 82.5 GS/s Sampling IC and Bit-Rate Adaptive Decision Logic Circuit,” ECOC 2010, Turin, September 2010, Mo.2.B.3 [116] J. Sugawa, H. Ikeda, D. Mashimo, “10.3Gbps burst-mode receiver capable of upstream transmission with short overhead for 10G-EPON,” ECOC 2010, Turin, September 2010, Mo.2.B.4 [117] M. Nakamura, et al., “A 10G burst-mode PIN-TIA module with 10 nsec response for PON systems,” LEOS, 2007. [118] A. Li, J. Faucher, and D. Plant, “Burst-Mode Clock and Data Recovery in Optical Multiaccess Networks using Broad-Band PLLs,” IEEE Photonics Technology Letters, Vol. 18, Issue 1, pp.73-75, 2006. [119] T. Ajmal, et al., “Design of a 10Gbps Optical Burst Mode DPSK Receiver for Data and Clock Recovery,” ECOC 2008, Brussels, September 2008, We.2.F.2. [120] J. Kwon, et al. “AC-Coupled Burst-Mode OLT SFP Transceiver for Gigabit Ethernet PON Systems,” IEEE Photonics Technology Letters, Vol. 17, No. 7, July 2005. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 13 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [121] Y. Ohtomo, et al., “A 10 Gb/s burst mode clock and data recovery IC with frequencyadjusting dual gated VCOs,” IEICE Trans. Electron., Vol. E91-C, No 6, 2008. [122] P. Ossieur, et al., “A 1.25-Gb/s Burst-Mode Receiver for GPON Applications,” IEEE Journal of Solid-State Circuits, Vol. 40, No. 5, May 2005. [123] P. Ossieur, et al., “DC-coupled 1.25 Gbit/s burst-mode receiver with automatic offset compensation,” Electronics Letters, Vol. 40, No. 7, April 2004. [124] P. Ossieur et al., “A 10 Gb/s Burst-Mode Receiver with Automatic Reset Generation and Burst Detection for Extended Reach PONs”, OFC2009, San Diego, March 2009, OWH3. [125] S Nishihara, et al., “10.3 Gbit/s burst-mode PIN-TIA module with high sensitivity, wide dynamic range and quick response,” Electronics Letters, Vol. 44, No. 3, 2008. [126] S. Nishihara, et al., “A burst-mode 3R receiver for 10-Gbit/s PON systems with high sensitivity, wide dynamic range, and fast response,” Journal of Lightwave Technology, Vol. 26, No. 1, 2008. [127] E. Hugues-Salas, et al., “Fast edge-detection burst-mode 2.5Gbit/s receiver for gigabit passive optical networks,” Journal of Optical Networking, Vol. 6, Issue 5, pp. 482-489, 2007. [128] Han Sun, et al., “Challenges for VLSI implementation of 100G digital coherent receivers,” OFC2009, March 2009, OTuN1. [129] P. Schvan, et al., “A 24GS/s 6b ADC in 90nm CMOS,” ISSCC 2008, February 2008. [130] J. Lee, et al., “High-Speed Analog-to-Digital Converters in SiGe Technologies”, CSIC 2007, October 2007. [131] Hiroshi Onaka, “Challenges for VLSI implementation of 100G digital coherent receivers,” ECOC 2009, Workshop on DSP & FEC, September 2009. [132] Fujitsu Microelectronics Europe, “Fujitsu Launches Second Generation Ultra-fast 65GSa/s 8-Bit ADC Technology for 100G Optical Transport,” Online : http://www.fujitsu.com/emea/news/pr/fseu-en_20100913-978.html [133] I. Dedic, “56Gs/s ADC Enabling 100GbE,” OFC2010, San Diego, March 2010, OThT6. [134] Maxtek, Online: http://www.maxtek.com/services6.html [135] Micram Microelectronic GmbH, Online: http://www.micram.de [136] SP Devices, Online : http://spdevices.com/index.php/products/high-speed-dataacquisition [137] Mobius Semiconductor, Online: http://www.mobius-semiconductor.com/ [138] IHP, “ADCs and DACs: High-speed ADC and DAC components with 1-50 GS/s in SiGe BiCMOS technology,” Online: http://www.ihp-microelectronics.com/174.0.html [139] M. Grözing, et al., “High-Speed ADC Building Blocks in 90 nm CMOS”, Fourth Joint Symposium on Opto- and Microelectronic Decices and Circuits (SODC 2006), Duisburg, Germany, September 2006. [140] National Semiconductor, ADC08B3000, Online: http://www.national.com/pf/DC/ADC08B3000.html [141] H. Bülow, F. Buchali and A. Klekamp, “Electronic Dispersion Compensation,” IEEE IEEE Journal of Lightwave Technology, Vol. 26, Issue 1, pp 158-167, January 2008. [142] A. Faerbert, et al., “Performance of a 10.7 Gb/s Receiver with Digital Equaliser using Maximum Likelihood Sequence Estimation,” ECOC2004, Stockholm, PD TH415. [143] T. Kupfer, “Measurement of the Performance of 16-states MLSE Digital Equalizer with Different Optical Modulation Formats,” OFC2008, San Diego, February 2008, PDP13. [144] S. J. Savory, et al., “Digital equalization of 40 Gbit/s perwavelength transmission over 2480 km of standard fibre without optical dispersion compensation,” ECOC2006, Cannes, September 2006, Th2.5.5. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 14 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [145] L. Wosinski, M. Dainese, H. Fernando and T. Augustsson, “Grating-assisted add-drop multiplexer realized in silica-on-silicon technology,” Proceedings of the Conference “Photonics Fabrication Europe”, Brugge, Belgium, Proc. SPIE 4941, pp. 43-50, October/November 2002. [146] L. Wosinski, M. Swillo and M. Dainese, "Imprinting of low dispersion Bragg gratings in planar devices for 40 Gbps DWDM systems," Proc. of International Congress on Optics and Optoelectronics, Warsaw, Poland, Proc. SPIE 5956, pp. OD1 - OD8, August/September 2005. [147] N. Zhu, et al., “Experimental demonstration of a cross-order echelle grating triplexer based on an amorphous silicon nanowire platform,” Optics Letters, Vol. 34, Issue 3, pp. 383-385, 2009. [148] L. Wosinski, “Silica-on-Silicon Technology for Photonic Integrated Devices,” ICTON 2004, Wroclaw, Poland, Vol. 2, pp. 274 – 279, July 2004. [149] Janz, S. et al., “Planar waveguide echelle gratings in silica-on-silicon,” IEEE Photonics Technology Letters, Vol. 16, Issue 2, pp. 503–505, February 2004. [150] D-link, Online: http://www.dlinkgreen.com [151] Zhang Yuanwang, “100G Ethernet Technology and Applications”, ZTE Communicatinos, Vol. 4, December 2009, Online: http://wwwen.zte.com.cn/endata/ magazine/ztecommunications/2009year/no4/articles/200912/t20091223_178948.html O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 15 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figures and tables Figure 1 : Figure 2 : Figure 3 : Figure 4 : Figure 5 : Figure 6 : Figure 7 : Figure 8 : Figure 9 : Figure 10 : Figure 11 : Figure 12 : Figure 13 : Figure 14 : Figure 15 : Figure 16 : Figure 17 : Figure 18 : Figure 19 : Figure 20 : Figure 21 : Figure 22 : Figure 23 : Figure 24 : Figure 25 : Figure 26 : Figure 27 : Figure 28 : Figure 29 : Figure 30 : Figure 31 : A generic and high level activity split between WP3 and WP4. The major areas of work are in “Network Functions and Scope” and software (SW) and hardware (HW) design implementation, simulation and emulation. ............... 30 NG-PON Roadmap .......................................................................................... 32 Systems concept tree ........................................................................................ 33 Wavelength plans [6] ....................................................................................... 37 Downstream serial 40G NRZ ........................................................................... 40 Downstream 40G QPSK .................................................................................. 41 Downstream 40G DQPSK ............................................................................... 42 Downstream stacked 4x10G NRZ.................................................................... 43 Generic WDM-PON with (fixed or tunable) laser array in OLT and tunable lasers in ONUs ................................................................................................. 44 WDM-PON with tunable lasers in ONUs supporting power-split ODN ......... 45 Tuning of tunable ONU via closed-loop control incorporating the OLT and an ECC .................................................................................................................. 46 Tuning of tunable ONU via closed-loop control incorporating a partial reflector in the RN ............................................................................................ 46 Cyclic AWG ..................................................................................................... 47 DS and US frequency division at the ONU...................................................... 48 Laser-based WDM-PON with 25 GHz spacing for enhanced channel count .. 48 Laser-based WDM-PON with 25 GHz grid and S-band extension for very high channel count.................................................................................................... 49 SSMF spectral attenuation. The dashed lines are the boundaries according to ITU-T G.652A [11]. ......................................................................................... 50 Generic WDM-PON with seeded reflective (REAM) OLT transceiver array and seeded reflective ONUs (based on RSOAs, REAMs, IL-FP lasers, or combinations thereof, e.g., REAM-SOA) ........................................................ 51 WDM-PON with seeded reflective (REAM) OLT transceiver array and tuneable ONUs. Redundant MFLs in the OLT are shared between several PONs. ............................................................................................................... 51 Increase of spectral efficiency in WDM-PON with seeded ONUs through use of constant-envelope FSK downstream modulation and intensity upstream modulation ........................................................................................................ 52 Decrease of seed/upstream Rayleigh crosstalk through end-to-end dual-fibre working............................................................................................................. 52 Decrease of Rayleigh crosstalk through use of dedicated upstream feeder fibre . .......................................................................................................................... 53 Decrease of Rayleigh crosstalk through RN-based MFL Seed accommodation .......................................................................................................................... 53 Super-PON [12] as an example for active-RN-based seed accommodation .... 53 DWDM-TDMA-PON [13] as an example for dedicated upstream feeder fibre .. .......................................................................................................................... 54 ONU transmitter variants ................................................................................. 54 Simple ONU transmitter................................................................................... 55 FSK generation by means of a directly modulated laser .................................. 56 Frequency to Amplitude Modulation by means of an optical filter ................. 56 Figure RZ/IRZ remodulation scheme .............................................................. 56 OFDM-Transmission on multiple orthogonal subcarriers ............................... 58 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 16 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 32 : Figure 33 : Figure 34 : Figure 35 : Figure 36 : Figure 37 : Figure 38 : Figure 39 : Figure 40 : Figure 41 : Figure 42 : Figure 43 : Figure 44 : Figure 45 : Figure 46 : Figure 47 : Figure 48 : Figure 49 : Figure 50 : Figure 51 : Figure 52 : Figure 53 : Figure 54 : Figure 55 : Figure 56 : Figure 57 : Figure 58 : Figure 59 : Figure 60 : Figure 61 : Figure 62 : Figure 63 : Figure 64 : Figure 65 : Downstream CO-OFDM [17]: a) direct down/up conversion, b) intermediate frequency .......................................................................................................... 59 Transmitter and receiver alternatives for DDO-OFDM [18]: a) b) and c) show different transmitter alternatives and d) a typical receiver. .............................. 60 DDO-OFDMA-PON scheme after [16]. .......................................................... 61 OCDM PON with a power splitter at the remote node .................................... 63 A device with capability of simultaneously encoding/decoding multiple timespread optical codes (OCs) [29]. ...................................................................... 63 (a) Configuration of OCDM-PON system with code splitter at RN and (b) experimental setups and results: (i) downlink and (ii) uplink [30]. ................. 64 Schematic of the all-optical 2-D OCDM code-drop unit [31] (D: Delay, HPF: High-pass filter, and TOAD: terahertz optical asymmetric demultiplexer). .... 65 Structure of OCDMA add-drop multiplexer for any coding technology [24] (TOAD: terahertz optical asymmetric demultiplexer). .................................... 66 Hybrid PON basic setup ................................................................................... 67 Hybrid WDM/TDM-PON ................................................................................ 67 WDM/TDM-PON based on an ONU with one tunable transmitter and one tunable optical filter, for a general remote node .............................................. 69 WDM/TDM-PON as in Figure 42 with cascaded power splitters for the remote node. ................................................................................................................. 70 WDM/TDM PON as in Figure 42 with a WDM splitter and power splitter for the remote node. ............................................................................................... 70 WDM/TDM-PON as in Figure 42 with a wavelength router and power splitter for the remote node. ......................................................................................... 71 WDM/TDM-PON as in Figure 42 with a wavelength selective switch and power splitter for the remote node. .................................................................. 72 Hybrid WDM/OCDM PON [27] (OC: optical code, TX: transmitter and RX: receiver). ........................................................................................................... 73 Hybrid WDM/E-CDM-PON ............................................................................ 74 Hybrid WDM/O-CDM-PON............................................................................ 75 An example of wavelength channel allocation of OCDMA over CWDM using 511-chip SSFBG en/decoder [27]. ................................................................... 76 OCDMA over WDM or hybrid WDM/TDM PONs ........................................ 77 Hybrid WDM/OCDM PON with code splitting at the remote node. ............... 77 Hybrid WDM/O-OFDM-PON ......................................................................... 78 Hybrid WDM/SCM-PON ................................................................................ 79 UDWDM-PON................................................................................................. 80 Hybrid active WDM-PON with active P2P access .......................................... 81 Hybrid active WDM-PON (“PON-in-PON”) .................................................. 82 Reference WDM-PON ..................................................................................... 83 2.5Gb/s WDM-PON employing RSOA-based ONUs and NRZ signaling ...... 96 BER versus received power with multiwavelength source placed at the OLT (a) and the RN (b)............................................................................................. 96 BER measurements versus received power with multi-channel near RN ........ 97 2.5Gb/s WDM-PON employing RSOA-based ONUs and RZ signaling ......... 97 BER performance for different DS extinction ratio values of 3dB (a) and 5dB (b) ..................................................................................................................... 98 10Gb/s WDM-PON employing REAM-based ONUs ..................................... 99 US and DS BER performance for REAM WDM-PON ................................... 99 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 17 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 66 : Figure 67 : Figure 68 : Figure 69 : Figure 70 : Figure 71 : Figure 72 : Figure 73 : Figure 74 : Figure 75 : Figure 76 : Figure 77 : Figure 78 : Figure 79 : Figure 80 : Figure 81 : Figure 82 : Figure 83 : Figure 84 : Figure 85 : Figure 86 : Figure 87 : Figure 88 : Figure 89 : Figure 90 : Figure 91 : Figure 92 : Figure 93 : Figure 94 : Figure 95 : Figure 96 : Figure 97 : Figure 98 : Figure 99 : Figure 100 : Figure 101 : Figure 102 : Figure 103 : Figure 104 : Figure 105 : Figure 106 : Cut-down (single-user) WDM-PON highlighting REAT-based ONU at CPE. 100 3-section DBR laser ....................................................................................... 103 5-section DS-DBR laser ................................................................................. 103 DS-DBR rear section wavelength comb (measured) ..................................... 104 DS-DBR laser pseudo-wavelength map ......................................................... 104 DS-DBR lasing spectrum – with side-super--modes ..................................... 105 Side mode suppression ratio ........................................................................... 105 DS-DBR reverse-biased SOA as a shutter ..................................................... 106 DS-DBR laser RIN (a), and line width (b) ..................................................... 107 DS-DBR laser temperature-depending output power (as of today) ............... 108 Cross-sectional diagram of a MEM-VCSEL ................................................. 109 VCSEL tuning. (B) shows tuning of an SCC type VCSEL, (A) and (C) for EC types. .............................................................................................................. 110 Schematic diagram of an ECL ....................................................................... 110 Principle of an FBG-ECL (left), and FBG-ECL from Xponent Photonics, Inc. (right) .............................................................................................................. 111 Packaging and pin configuration of commercial ChemOptics FBG-ECL ..... 111 Photo of DFB array chip ................................................................................ 112 Photo of DFB array chip [66] ......................................................................... 113 Schematic of tunable TFF with active substrate ............................................ 116 Schematic of tunable TFF for GPON WDM overlay [83] ............................. 116 Packaging of tunable TFF for GPON WDM overlay .................................... 117 Schematic of liquid-crystal filter .................................................................... 118 Schematic of (tunable) FBG ........................................................................... 118 Reflection peak of an FBG ............................................................................. 119 Schematic of tunable Volume Holographic Grating (VHG) .......................... 120 Schematic of MEM-tunable filter .................................................................. 120 Schematic of tunable microcavities................................................................ 121 Coherent detection (basic scheme) ................................................................. 123 Coherent intradyne DP-QPSK transmission .................................................. 124 Two realizations of 90° hybrids in polarization-diverse coherent receivers .. 124 Coherent intradyne receiver: digital realization ............................................. 125 Example of simple FIR filter response (blue: Re, red: Im part) ..................... 125 Phase estimation using Wiener filter .............................................................. 127 Phase estimation in presence of phase noise and combined phase plus amplitude noise .............................................................................................. 127 Cascaded FOE [103] ...................................................................................... 128 Linear-only or linear plus nonlinear compensation [106] .............................. 129 Homodyne detection without polarization diversity and 90° hybrids ............ 130 Reduced homodyne detector .......................................................................... 131 Digital implementation [110] ......................................................................... 132 Polarization-scrambled heterodyne detection ................................................ 132 Typical Passive Optical Network (PON) scheme with a BMR located in the OLT ................................................................................................................ 133 a) mass distribution function of differentiated signal level sampled immediately after the start of each bit b) mass distribution function of differentiated signal level sampled after passing MLEPW(set to 0.05 of bit length) after the start of each bit. SNR for both scenarios was set to 9 dB. ... 136 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 18 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 107 : Figure 108 : Figure 109 : Figure 110 : Figure 111 : Figure 112 : Figure 113 : Figure 114 : Figure 115 : Figure 116 : Figure 117 : Figure 118 : Figure 119 : Figure 120 : Figure 121 : Figure 122 : Figure 123 : Figure 124 : Figure 125 : Figure 126 : Figure 127 : Figure 128 : Figure 129 : Figure 130 : Figure 131 : Figure 132 : Mass distribution function of differentiated signal level sampled immediately after the start of each bit. SNR= 7 dB. ........................................................... 137 BER versus the SNR for different values of MLEPW. .................................. 138 a) the differentiated pulse with very large value of RC constant for differentiator b) the differentiated pulse with small value of RC constant for differentiator. .................................................................................................. 139 BER versus SNR for a pseudo-random sequence and a sequence of “0 1 0 1…”. ............................................................................................................... 140 The figure shows a set of possible modulation formats for 100 Gb/s systems and their requirements in terms of ADC performance (green circles) [128]. Also reported is the performance of bipolar and CMOS ADCs from publications [129], [130]. ............................................................................... 141 Multi-chip (upper figure) and monolithic (lower figure) configuration of ADC/DSP block in a 100 Gb/s receiver [131]. .............................................. 142 Overview of dispersion compensation methods............................................. 143 Bragg grating assisted optical Add-Drop Multiplexer based on 2x2 MMI coupler. 30dB extinction ratio between drop and transmitted channels and 3dB excess loss in the dropped channel have been obtained. ................................ 146 Add-Drop Multiplexer based on balanced Mach-Zehnder interferometer. Obtained crosstalk -25dB, insertion loss -3dB, switching time 2ms and power consumption 0.5W have been achieved. ........................................................ 146 Etched diffraction grating demultiplexer: signal from input waveguide with wavelengths λ1 ,λ2 ,λ3,… is diffracted by the planar concave grating, and refocused into different output waveguides. .................................................. 147 Schematic of an arrayed waveguide grating: (a) input-output waveguides are coupled through two slab (FPR) waveguides and an array of curved waveguides, (b) tilted phase front with an angle , focused at an output waveguide in the output FPR. ........................................................................ 148 Series of 32 x 32AWGs on a 4” Si wafer in SiO2/Si technology, 0.8 nm channel spacing (100 GHz), 25 nm band: a) Fabricated devices, b) AWG transmission spectrum. ................................................................................... 148 Example GPON OEO..................................................................................... 150 RPT approach ................................................................................................. 151 Comparison of different reach extender technologies for 10G TDM-PON (XGPON): OA based reach extender (L-band EDFA for the 1577 nm downstream + SOA for the 1270 nm upstream), O/E/O based reach extender and Remote Protocol Terminator (RPT)/ mini-OLT approach. ......................................... 152 RE variants ..................................................................................................... 153 Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels. ................. 161 Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels, low loss series. .............................................................................................................. 162 Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 100GHz ... 162 Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 200GHz ... 163 Oplink’s FTTX triplexer 1310/1490/1550 WDM (1x2) ................................ 163 JDSU Coarse Wavelength Division Multiplexer 4/8 channels ...................... 164 JDSU Dense Wavelength Division Multiplexer 4/8 channels ....................... 164 Oplink’s Dense Wavelength Division Multiplexer 100 GHz. ....................... 165 Oplink’s Dense Wavelength Division Multiplexer 50 GHz. ......................... 166 JDSU Dense Wavelength Division Multiplexer 100 GHz, Narrowband (Gaussian)....................................................................................................... 166 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 19 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 133 : Figure 134 : Figure 135 : Table 1 : Table 2 : Table 3 : Table 4 : Table 5 : Table 6 : Table 7 : Table 8. Table 9 : Table 10 : Table 11 : Table 12 : Table 13. Table 14 : Table 15 : Table 16 : Table 17 : Table 18. Table 19 : Table 20. Table 21 : Table 22 : Table 23. Table 24. Table 25 : JDSU Dense Wavelength Division Multiplexer 100 GHz, Wideband (Flat Top). ............................................................................................................... 167 JDSU Dense Wavelength Division Multiplexer 50 GHz, Wideband (Flat Top). ........................................................................................................................ 167 JDSU Dual Duplexer 1310/1550.................................................................... 168 Key system aspects............................................................................................... 34 Key component/subsystem aspects ...................................................................... 35 Main TDM-PON standards .................................................................................. 38 40G-PON variants for the DS .............................................................................. 39 40G-PON variants for the US .............................................................................. 39 Encoder/decoder technologies.............................................................................. 62 Energy consumption and cost figures for hybrid WDM/TDM-PON................... 68 Energy consumption and cost figures for hybrid WDM/E-CDM-PON............... 74 Energy consumption and cost figures for hybrid WDM/O-CDM-PON .............. 75 Energy consumption and cost figures for hybrid WDM/O-OFDM-PON ............ 78 Energy consumption and cost figures for hybrid WDM/SCMA-PON ................ 79 Energy consumption and cost figures for WDM/UDWDM-PON ....................... 81 Energy consumption and cost figures for hybrid active WDM-PON with active P2P access ............................................................................................................ 82 Energy consumption and cost figures for hybrid active WDM-PON .................. 82 Energy consumption and cost figures for reference WDM-PON ........................ 84 Selection of Ethernet PHY standards ................................................................... 85 Comparison of relevant parameters of tunable lasers ........................................ 114 Comparison of relevant parameters of tunable filters ........................................ 121 Linewidth requirements for single-polarization modulation at a target BER of 10−3 ..................................................................................................................... 129 Comparison of coherent WDM-PON schemes .................................................. 132 Table summarizing state-of-the-art DACs available. ......................................... 142 Example of supported GPON configurations ..................................................... 150 System concept comparison table ...................................................................... 157 Energy consumption and cost figures of relevant hybrid PON components ..... 160 Relevant optical parameters for hybrid PON performance analysis .................. 160 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 20 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Abbreviations 3R Reamplification, Reshaping, and Retiming ADC Analogue-to-Digital Converter AFC Automatic Frequency Control AM Amplitude Modulation AMO-OFDM Adaptively Modulated OFDM AON Active Optical Network AOTF Acousto-Optic Tunable Filter APD Avalanche Photodiode AR Anti-Reflection ARN Active RN ASE Amplified Spontanous Emission ATM Asynchronous Transfer Mode AWG Arrayed Waveguide Grating AWGN Additive White Gaussian Noise BER Bit Error Rate BMR Burst Mode Receiver BPF Band Pass Filter BPSK Binary PSK BS Base Station BtB Back to Back BTJ Buried Tunnel Junction BW Bandwidth CD Chromatic Dispersion CDM Code Division Multiplexing CDMA Code Division Mutiple Access CDR Clock Data Recovery CFO Carrier Frequency Offset CID Consecutive Identical Digits CMA Constant-Modulus Algorithm CML Chirp Managed Laser CO Central Office CO-OFDM Coherent Optical OFDM CP Cyclic Prefix O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 21 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 CPE Customer Premisses Equipment CPRI Common Public Radio Interface CPU Central Processing Unit CR Code Restorer CRAN Cloud RAN CW Continous Wave CWDM Coarse Wavelength Division Multiplexing DAC Digital to Analogue Converter DBA Dynamic Bandwidth Allocation DBR Distributed Bragg Reflector DCD Drop Code Decoder DCF Dispersion Compensation Fibre DCU Dispersion Compensation Unit DD Direct Detection DDO-OFDM Direct Detection Optical OFDM DFB Distributed Feedback DFE Decision Feedback Equalizer DIY Do It Yourself DL Downlink DML Directly Modulated Laser DP-QPSK Dual-Polarized QPSK DPSK Differential PSK DQPSK Differential Quaternary Phase-Shift Keying DS Downstream DSCA Dynamic Sub-Carrier Allocation DS-DBR Digital Supermode DBR DSP Digital Signal Processing DWDM Dense Wavelength Division Multiplexing EAM Electroabsorption Modulator EAT Electroabsorption Transceiver EC Extended Cavity ECC Embedded Communication Channel E-CDMA Electrical CDMA ECL External Cavity Laser E/D Encoder/Decoder O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 22 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 EDC Electronic Dispersion Control EDG Etched Diffraction Grating EDGE Enhanced Data rates for GSM Evolution EDFA Erbium-Doped Fibre Amplifier EG Echelle grating EML Electroabsorption Modulated Laser EONT Embedded ONT EOTF Electro-Optical Tunable Filter EPD Electronic Pre-Distortion EPON Ethernet Passive Optical Network EQ Equalizer E-UTRA Evolved Universal Terrestrial Radio Access FBG Fibre Bragg Grating FBG-ECL Fibre-Bragg-Grating External-Cavity Laser FDD Frequency Division Duplex FDL Fibre Delay Line FE Fast Ethernet FEC Forward Error Correction FF Feeder Fibre FFE Feed Forward Equalizer ffs For Further Study FFT Fast Fourier Transform FIR Finite-Impulse Response FM Frequency Modulation FOE Frequency Offset Estimator FP Fabry-Perot FP-LD Fabry-Perot LD FPR Free Propagation Regions FSAN Full Service Access Network FSK Frequency Shift Keying FTTB Fibre-to-the-Building FTTC Fibre-to-the-Curb FTTH Fibre-to-the-Home FTTx Fibre-to-the-x GbE Gigabit Ethernet O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 23 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 GCSR Grating-assisted Co-directional Coupler Laser with Sampled Grating Reflector GPON Gigabit-capable Passive Optical Network GPRS General Packet RAdio Service GSM Global System for Mobile Communications HFC Hybrid Fibre-Coaxial HPF High Pass Filter HSPA High Speed Packet Access HW Hardware IAD Integrated Access Device ICI Inter-Carrier Interference IEEE Institute of Electrical and Electronics Engineers IF Intermediate Frequency IFFT Inverse Fast Fourier Transform IL Interleaver IM Intensity Modulation IRZ Inverse Return to Zero ISI Inter-Symbol Interference LAN Local Area Network LCFP Liquid Crystals FP LD Laser Diode LED Light Emitting Diode LMS Least Mean-Square LO Local Oscillator LTE Long Term Evolution LX Local Exchange MAC Media Access Control MAP Maximum a Posteriori MBE Molecular Beam Epitaxy MDXM DWDM Multiplexer/Demultiplexer MEMS Micro-Electro-Mechanical Systems MFL Multi Frequency Laser MG-SGC Matrix-Grating Strongly Gain-Coupled MIMO Multiple Input Multiple Output MMF Multi Mode Fibre MMI Multimode Interference O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 24 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 MMLEPW Minimum Latch Enable Pulse Width MMSE Minimum Mean-Square Error MQW Multi-Quantum Well MSA Multi Source Agreement MZ Mach Zehnder MZI MZ Interferometer MZM MZ Modulator NGA Next Generation Access NGOA Next Generation Optical Access NLC Nonlinear Compensation NRZ Non-Return to Zero OA Optical Amplifier OADM Optical Add-Drop Multiplexer OBPF Optical Bandpass Filter OBSAI Open Base Station Architecture Initiative OC Optical Code O-CDMA Optical CDMA ODN Optical Distribution Network ODL Optical Delay Line OE Optical-Electrical OEO Optical-Electrical-Optical OFDM Orthogonal Frequency Division Multiplexing OFDMA Orthogonal Frequency Division Multiple Access OLT Optical Line Terminal ONT Optical Network Terminal ONU Optical Network Unit OOC Optical Orthogonal Codes OOK On Off Keying OPLL Optical Phase-Locked Loop OSNR Optical Signal-to-Noise Ratio OSSB Optical Single Sideband OTN Optical Transport Network OXC Optical Cross-Connect P2P Point-to-Point PAPR Peak-to-Average Power Ratio O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 25 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 PBC Polarization Beam Combiner PBS Polarization Beam Splitter PC Polarization Controller PCG Planar Concave Grating PD Photo Diode PDG Polarization Dependent Gain PE Phase Estimation PHY Physical Layer PIC Photonic Integrated Circuit PIN p-i-n receiver PLC Planar Lightwave Circuit PoP Point Of Presence PMD Physical Medium Dependent PMD Polarization Mode Dispersion PON Passive Optical Network PRBS Pseudo Random Bit Sequence PRN Passive RN P/S Parallel to Serial PSK Phase Shift Keying QAM Quadrature Amplitude Modulation QCSE Quantum Confined Stark Effect QPSK Quadrature Phase Shift Keying QoS Quality of Service RAN Radio Access Network RAU Radio Access Unit RBS Radio Base Station REC Radio Equipment Controllers RE Radio Equipment REAM Reflective EAM REAT Reflective EAT RF Radio Frequency RIN Relative Intensity Noise RN Remote Node ROADM Reconfigurable Optical Add-Drop Multiplexer RoF Radio over Fibre O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 26 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 RPT Remote Protocol Terminator RPR Resilient Packet Ring RSOA Reflective SOA Rx Receiver RXA Receiver Array RZ Return to Zero SCC Semiconductor-Coupled Cavity SCM Sub-Carrier Multiplexing SCMA Sub-Carrier Multiple Access SFI System Framer Interface SG-DBR Sampled Grating DBR SOA Semiconductor Optical Amplifier SFP Small Form-factor Pluggable SFF Small Form Factor SFW Single Fibre Working SLPM Spatial Light Phase Modulator SMF Single Mode Fibre SMSR Side-Mode Suppression Ratio SNR Signal to Noise Ratio S/P Serial to Parallel SSFBG Super Structured FBG SSG-DBR Super-Structure Grating DBR SSMF Standard Single Mode Fibre SW Software TDM Time Division Multiplexing TDMA Time Division Multiple Access TEC Thermo-Electrical Cooler TF Tunable Filter TFF Thin-Film Filter TOAD Terahertz Optical Asymmetric Demultiplexer TRx Transceiver TS-OCDMA Time Spreading OCDMA TW Travelling Wave Tx Transmitter TXA Transmitter Array O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 27 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 TXFP Tunable XFP UDWDM Ultra Dense WDM UL Uplink UMTS Universal Mobile Telecommunications System US Upstream UTRA Universal Terrestrial Radio Access VCSEL Vertical Cavity Surface Emitting Laser VHG Volume Holographic Grating VMZ Vertical Mach-Zehnder WAN Wide Area Network WDM Wavelength Division Multiplexing WDMA Wavelength Division Multiple Access WL Wavelength WSS Wavelength Selective Switch XFP 10 Gigabit Small Form Factor Pluggable XG-PON 10 Gigabit-capable PON O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 28 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 1. Introduction The aim of OASE is the development of Next Generation Optical Access (NGOA) architecture and system concepts for the “2020” time horizon, based on European requirements [1]. The focus is on fibre-to-the-home (FTTH) network scenarios. The purpose of this deliverable is to provide an overview of system concepts/technologies with potential of meeting the requirements of next-generation optical access (NGOA) as defined in Deliverable 2.1 [2]. Key requirements include residential peak data rates of ≥1 Gbit/s, support for 256 to 1024 customers per feeder fibre, support for 20 to 40 km passive reach, and support for 60 to 90 km extended reach for the protection path. This document presents a wide survey of proposed system concepts with potential of meeting these requirements. For each of these next-generation systems, a detailed technical description is made as well as a coarse investigation considering key aspects such as bandwidth, number of customers per feeder fibre and reach. In the continuation of the project a reduced number of concepts will be selected for more detailed techno-economic assessment. WP4 will take input from WP2 “General Requirements” and WP3 “Architectures” which provide boundary conditions for work in WP4 with respect to parameters such as: Power splitter, wavelength splitter configurations Upstream/downstream bandwidth per subscriber (peak/average), Number of users per CO (feeder) fibre, Total fibre distance from subscriber to CO, Number and possible locations of active equipment in the field Resilience requirements (availabilities, switch-over times) System design for minimized OpEx in telecom processes Other system parameters, such as complexity (which has a strong bearing on the cost analysis of WP5), and power consumption, will subsequently be analyzed in the assessment of different system concepts. 2. Task description and scope This document, D4.1, is the first deliverable of OASE WP4 “System concepts for nextgeneration optical access networks” and is the end-result of task T4.1 as defined in [1]. Task T4.1 is composed of the following activities: - A4.1.1 Survey of current concepts for next-generation optical access A4.1.2 Survey of evolving components/subsystems A4.1.1 includes a wide survey of possible next-generation technologies (from conferences, journal papers, vendors, operators, tradeshows, standardization bodies, etc). For each of the next-generation technologies, a detailed technical description is made. The technologies are coarsely investigated/ assessed by filling in a table with all the aforementioned aspects. Out of the survey of A4.1.1, key evolving components/subsystems are identified and a survey of possible next-generation technologies is made in A4.1.2 (starting from conferences, journal papers, vendors, operators, tradeshows, standardization bodies, etc). O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 29 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 2.1 RELATIONSHIP TO OTHER WORKPACKAGES This section aims to clarify the scope of WP4 “System concepts for next-generation optical access networks” and in particular to define the boundary to WP3 “Architectures”. Although there is a natural overlap between these two packages the following section aims to define areas of responsibility for the different work packages. WP3 focuses on network architectural work that meets the requirements of WP2, as well as protocol design and non-protocol based network element functionality activities. WP3 includes simulation based work that can be used as a hardware and software design input. WP4 has an implementation focus. Both hardware design and implementation work will be part of WP4, while only the implementation part of WP3 designed software will be done in WP4. This will be transferred into WP4 in task T4.4. WP4 will perform activities in data, control and management plane software implementation. This implementation can be used in order to control an emulated or physical data plane system solution. Figure 1 shows a high level view of the activity split between WP3 and WP4, which allows for additional input from WP2 on requirements. The figure also states that in both WPs there can exist activities that focus on a per network as well as a per network element level, i.e. both WPs can have an end-to-end network scope or “just” a node local scope. Figure 1 : A generic and high level activity split between WP3 and WP4. The major areas of work are in “Network Functions and Scope” and software (SW) and hardware (HW) design implementation, simulation and emulation. Figure 1 shows a number of areas in which work will be done in WP3. The “Network Functions and Scope” area is the main area which basically has the goal of creating the OASE functional network architecture. I.e. high level requirements supplied from WP2 should be transformed into corresponding functions needed in different geographical locations and functional planes of the network elements of the OASE “scoped” network. E.g. if the high level requirement is “network resilience” then there is need of control plane functions that disseminates information of resilience node capabilities which allows for path calculation that considers e.g. protective paths. The control plane should include the possibility to signal this type of information related to resource allocation. In the management plane the corresponding functions could be fault identification and localization, which can trigger protection or restoration actions. All of this depends, as well, on the type of data plane technology and the O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 30 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 network topology of this part of the network. The other two areas included in WP3 are SW design and simulation. The design part is mainly work focused on the design of protocols and other software functions that are needed in order to realize the functions defined in the network architecture, which is created in order to meet the requirements of WP2. In order to make efficient designs, simulations might be a useful tool but it can also be used for simulation activities on dynamic bandwidth allocation or how locality enabled peer-to-peer protocols will influence the access and distribution part of the network. Figure 1 shows a number of areas in which work will be done in WP4, as well. The output from WP2 and WP3 into WP4 is a set of functions that meets a number of requirements. The work of WP4 is to realize this input through the design and implementation of hardware and software. As a research and support tool for this realization of the OASE functional network architecture, hardware simulation or modelling and software emulation can be used. The output of WP4 is to be used in WP7. In all of the above interactions between the WPs there are of course feedback loops that allows for e.g. re-design, re-implementation, added network functions etc. 3. Assumptions and methodology This chapter outlines the assumptions and methodology for the survey of system concepts. Section 3.1 describes the rational behind the structure of the survey. Section 3.2 presents an overview of the relevant assessment parameters and section 3.3 provides a description of the assumptions for cost and power consumption used in the deliverable and in the continuation of the OASE project. 3.1 NGOA SYSTEM CONCEPTS The requirements for the OASE NGOA are defined in Deliverable 2.1 [2]. Two important trends are reflected in these requirements, i.e. the projected increase in residential and backhaul bandwidth requirements and the requirements resulting from an operator desire to reduce number of central office sites. Some of the main technical requirements from [2] are listed below: - FTTH residential peak data rates ≥1 Gb/s - Business, backhaul (fixed, mobile) peak date rate: ≥10 Gb/s - Average sustainable downstream based on peak-hour service usage of 500 Mb/s per Optical Network Unit (ONU)/customer - Support of more traffic symmetry, with ratio of at least 1:2 between upstream and downstream - Support from 256 to 1024 ONUs/customers per feeder fibre - Support of 128 Gbit/s to 500 Gb/s aggregate capacity per feeder fibre - Support of 20 to 40 km passive reach option for the working path, depending on the degree of node consolidation - Support of 60 to 90 km extended reach option for the protection path, depending on the degree of node consolidation - Legacy ODN compatibility desirable O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 31 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 NGOA system candidates must therefore support increased customer bandwidth, increased number of customers and increased reach. Candidates should additionally support smooth migration from today’s FTTH architectures to future architectures. In ITU-T and FSAN, work is centred along two main tracks for Next Generation PON (NGPON), one mid-term track (NG-PON1) and one long-term track (NG-PON2): NG-PON1: NG-PON coexisting on same Optical Distribution Network (ODN) as GPON based on G.984.5 approach NG-PON2: “disruptive” NG-PON with no requirement to coexist on same fibres as GPON Figure 2 shows the FSAN roadmap for NG–PON1 and NG–PON2, where NG–PON1 is viewed as a mid-term upgrade and NG-PON2 as a longer-term solution. The timeline is shown to reflect the nominal expected period for specification and publication of standards for NG-PON1 (2009-2012) and NG-PON2 (2013-2015). NG-PON2 Capacity Component R&D to enable NG-PON2 “Co-existence” enables gradual migration in the same ODN. G-PON E.g. Higher-rate TDM, DWDM, CDM, OFDM, etc. NG-PON1 incl. long-reach option WDM option to enable to overlay multiple XGPONs XG-PON (Up: 2.5G to 10G, Down: 10G) 1G-EPON Splitter for NG-PON2 (power splitter or something new) Power splitter deployed for Giga PON (no replacement / no addition) Now ~2010 Figure 2 : ~2015 NG-PON Roadmap For the OASE NGOA time horizon, emerging and disruptive technologies can be considered in order to meet the requirements of increased capacity, extended reach and increased number of subscribers per feeder fibre. Increased system capacity can be achieved for example by increased transceiver rates, advanced modulation, orthogonal frequency division multiplexing (OFDM), wave-length division multiplexing (WDM), etc. Multiplexing techniques in future systems could include TDM, WDM, CDM, SCM, etc. Figure 3 presents an attempt to systemize NGOA system candidates in a tree like structure. At the top-level systems are categorized with respect to topology (tree, ring, mesh). The different categories are further categorized based on fibre infrastructure and whether it is shared or dedicated and whether the remote node is active or passive. The concepts are further categorized based on remote node design. The passive tree based concepts can for example be based on power splitters and/or on wavelength filters. The categorization continues further with technology and realization. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 32 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 3 : Systems concept tree The analysis of system concepts for NGOA in this deliverable is based on systems identified in the system concept tree. These concepts have been grouped into a number of subchapters in Chapter 4. The TDM-PON section 4.1 covers pure TDM-PON variants as well as stacked TDM-PON solutions (few wavelengths). The WDM-PON section 4.2 covers different flavours of WDM PON (tunable ONUs, seeded reflective ONUs and systems based on wavelength reuse). The OFDM-PON section 4.3 covers different OFDM-PON variants. The CDM section 4.4 covers various CDM solutions with focus on optical CDM-PON. The WDM-XXM hybrid section 4.5 considers solutions based on a combination of WDM and TDM, CDM, OFDM and SCM. It also considers UDWDM which is based on the similar two stage ODN infrastructure as the hybrid concepts. Section 4.6, on active hybrids, covers variants with active remote nodes such as active P2P and active WDM-PON remote notes. The AON section 4.7 considers technologies used for tree, ring and mesh based active networks. Finally, section 4.8, Radio-over-fibre backhauling, covers both radio-over-fibre technologies and digital interfaces which are relevant mainly for RAN transport. 3.2 KEY ASPECTS FOR THE ASSESSMENT In the assessment of different system concepts performed in this document, a number of key aspects were selected that were considered more crucial for the selection of NGOA system. These key aspects are summarized in Table 1 and to the extent it is possible each of these key aspects is discussed for each of the system concepts covered in Chapter 4. The assessment of the different concepts performed in this document, is naturally complicated by the different levels of detailed information available for the different concepts. In addition to the list of key aspects, a list of optional aspects to consider was also made (Table 1). Based on the identified systems, key emerging components/subsystems are identified, where component/subsystem evolution is critical for the respective system. Some components are specific for certain systems, while others are key elements of several systems. The survey of these evolving components/subsystems is based on a similar list of key aspects (Table 2) as for the system concepts. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 33 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 1 : Key system aspects Key system aspects Supported topologies (tree, ring, etc.) - Architecture, RN design, Technology, Realization Reach (km) - Without reach extension - With reach extension (what technology) Power budget (dB) Optional system aspects Security (optical encryption…) Migration options - From splitter-based infrastructure - For higher bandwidth and/or reach Protocol complexity (MAC in time/wavelength domain, resource allocation, etc.) Bandwidth efficiency Bandwidth per sub (peak, sustain) - Data rate compared to line rate No of subscribers per FF - Spectral efficiency Cost [high, medium, low] - Dynamic allocation - Today - Projections to 2020 Power consumption - Today - Projections to 2020 Support for Re-configurable network - Wavelength switching Temperature ranges - Suitable for outdoor, or indoor use only Possibility for power saving (sleep mode, power shedding, etc.) Support for co-operation Maturity, time to market (technology roadmap) Quantitative cost indication (relative eg to GPON) Resilience Key hardware cost drivers (CapEx) Supported techniques (1+1, 1:1, shared...) Support/possibilities for unbundling Possibility for low OpEx Footprint, physical size, port density [high, Aggregation solution medium, low] Extension/boundary of access network CoS support QoS support Detailed Data plane performance/issues O O A S OA AS SEEE FP7 – ICT– GA 249025 - Delay - Jitter - BER - Packet loss Page: 34 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 2 : Key component/subsystem aspects Key component/subsystem aspects Cost [high, medium, low] - Today - Projections to 2020 - Optional: Absolute cost Power consumption - Today - Projections to 2020 - Possibility for power saving (sleep mode, power shedding, etc.) Maturity, time to market (technology roadmap) Physical size - Today - Projections to 2020 Temperature ranges Pluggability 3.3 COST AND POWER ASSUMPTIONS Cost and power consumption data to be used in the OASE assessment must be adjusted to the 2020 time horizon. Reasonable assumptions on evolution of technology as well as cost and power consumption must be made. This section presents the basic assumptions to be used within OASE for compilation of cost and power data. - Numbers are the best estimates that can be made at this point in time, based on the collected experience of the consortium. If new information is found, numbers are subject to change. - Numbers which are to be used in the assessment are aggressive assuming maturity in the 2020 time horizon and mass-production, ramped up approximately in its third year. The effect of learning curves is incorporated in the numbers. Part of the data will also be based on the most advanced components available today (e.g., low-power consuming SFPs, tunable XFPs, etc.). All data should be considered aggressive with respect to both cost and power consumption. To the best of our knowledge, different approaches will be treated equally aggressive. - Estimates for critical components are based on internal sources, discussions with component vendors and discussions within FSAN subject to NDA. - Relative differences between different components is based on a complexity analysis of the components and the manufacturing process. The complexity analysis should consider the functionality needed (e.g., number of sections in an integrated transmitter, O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 35 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 added XDMA, added DSP, etc.), the bandwidth, and the required power budget (e.g., TX-RX budget of 28dB vs. 36dB or even 45dB). 4. System concepts for NGOA This chapter presents a detailed description of the system concept variants for NGOA discussed in Chapter 3. The different variants are grouped under the subsections: TDM-PON, WDM-PON, OFDM-PON, CDM, hybrid WDM/XXM, active hybrids and RoF backhauling. Within each section the main alternatives are discussed including a short description of each concept as well as brief discussion on key aspects and optional aspects. 4.1 TDM-PON The state-of-the-art PON standards are based on TDM/TDMA transmission mechanisms. In the downstream direction the data allocation to the ONU is realised by TDM controlled by the OLT. TDMA is used for the upstream direction in order to prevent collisions on the PON. Gigabit–class PONs have been standardised in ITU-T/FSAN and IEEE. The G-PON (i.e., 2.488 Gb/s downstream and 1.2 Gb/s upstream) was specified in the ITU-T recommendation series G.984 [3] and the 1G-EPON (1.25 Gb/s downstream and 1.25 Gb/s upstream, with 8b10b coding giving 1 Gb/s data rate) has been standardised by the IEEE working group 802.3ah in 2004 [4]. Both PON technologies are now being deployed, the G-PON mainly in Europe and America whereas the 1G-EPON is widely used in Asia. PON technologies with a downstream data rate of 10 Gb/s are standardised in ITU-T/FSAN and IEEE. The IEEE working group 802.3av has approved the 10 Gb/s Ethernet Passive Optical Network (10G-EPON) standard in 09/2009. The downstream rate is 10 Gb/s whereas two upstream rates are available, 1 Gb/s and 10 Gb/s. In total, three types of EPON ONTs can be supported: 1G/1G (1G-EPON) 10G/1G 10G/10G 10G EPON and 1G-EPON can coexist on the same ODN. The downstream traffic in 10GEPON is based on broadcast Ethernet frames and the upstream is based on TDMA with each ONU transmitting in its own timeslot. XG–PON [5] in Figure 2 represents an NG-PON1 system with 10 Gb/s line rate in the downstream direction. Two types of XG-PON have been defined in FSAN addressing different upstream line rates. XG-PON1 has an upstream line rate of 2.5 Gb/s and XG-PON2 allows a symmetric line rate of 10 Gb/s. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 36 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 4 : Wavelength plans [6] Figure 4 shows the Wavelength plan of the Gigabit-class and 10 Gigabit-class PON specifications. Both 10 Gb/s PON specifications are using the same wavelength windows for the downstream and upstream. In contrast to G-PON and XG-PON the upstream wavelength windows of 1G-EPON and 10G-EPON are overlapping, i.e. WDM multiplexing in order to separate the two data rates at the OLT site is not possible. Instead TDMA sharing is performed. The wavelength window between 1550nm and 1560nm is available for downstream RF video. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 37 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 3 summarizes some key parameters of the main TDM-PON technologies currently standardized. Standardization body Standardization Group PON Family Name Standardavailability Data rate down [Gb/s] Table 3 : IEEE 802.3av G.984 G.987 Tbd 1G-EPON 10G-EPON G-PON NG-PON1 NGPON2 Tbd 1000Base -PX 2004 10GBasePRX 2009 10GBasePR 2009 1 (Line rate 1.25G) 1 (Line rate 1.25G) 1:16/32 10 (Line rate 10.3G) 1 (Line rate 1.25G) 1:16/32/64 10 (Line rate 10.3G) 10 (Line rate 10.3G) 1:16/32/64 10/20 10/20 10/20 15/20/29 15/20/29 15/20/29 Ethernet Circuit emulation Ethernet Circuit emulation Ethernet Circuit emulation Splitting rate Reach extension Power saving Supported topologies ITU-T (FSAN) 802.3ah Data rate up [Gb/s] Min. reach [km] Logical reach [km] Power budget [dB] Frame Encryption TDM support Main TDM-PON standards Star, tree, ring Star, tree, ring Star, tree, ring - XG-PON1 XG-PON2 2004 2009/2010 - 2.488 9.953 9.953 Ca. 2015 * 1.244 (2,5) 2.488 ~10 * 1:32/64/128 * * 20 60 1:32/64/12 8 20 60 * * * * 15/20/25/28 /30/32 GEM AES GEM encapsulati on G.984.6 29/31/33/3 5 GEM AES GEM encapsulati on ffs * * * * * * * * * * Star, tree, ring ffs Star, tree, ring * * * * (*) For further study With the OASE requirements for NGOA of 1 Gb/s per end-user, current TDM-PON standards are insufficient. Considering that serial NRZ schemes are used for current TDM-PON standards, the natural starting point for higher capacity (40G-PON) is to investigate higher rate serial NRZ. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 38 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 4 and Table 5 summarize the 40-G-PON variants covered in this section. In general, all combinations of mixing any 40G DS variant with 40G or 10G US variant are possible. Table 4 : 40G-PON variants for the DS Serial 40G Serial 40G QPSK DQPSK DS 40G Serial 40G NRZ Bit rate per WL 40 40 40 OLT Tx EML+EDFA DBF+MZ DBF+precoder + MZ ONT Rx Technology PIN-PD + OA Coherent receiver Coherent receiver Direct detection APD Disp. Comp. DCF at OLT side (all pon), EDC at ONU side (per drop) No No No WLs C/L-band C/L-band C/L-band Tunable Tx/Rx No No No Link budget (total reach+split) ~27 dB - - ~26 dB Long-reach Optical amp (EDFA), OEO Optical amp (EDFA), OEO Optical amp (EDFA), OEO Optical amp (EDFA), OEO OLT- RN resilience Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) Table 5 : NRZ 4 x 10G 10 EML (XG-PON1, G.987.2) C/L-band (XG-PON1 1575-80 nm) Tunable ONU Rx (fixed operational prohibited) 40G-PON variants for the US Upstream 40 or 10G Bit rate per WL Serial 40G NRZ Serial 40G QPSK Serial 40G DQPSK NRZ 4 x 10G NRZ 1 x 10G 40 40 40 10 10 ONT Tx DFB+MZ+OA DBF+MZ DBF+precoder + MZ DML (802.3av) DML (802.3av) Coherent receiver Coherent receiver TFF fixed filter + APD Direct detection APD No No No No O/C/L-band O/C/L-band 4 wl in O/C/L-band O-band No No No Tunable Tx No As XG-PON1 (2931 dB) - - 29-31 dB 29-31 dB Long-reach Optical amp (EDFA), OEO Optical amp (EDFA), OEO Optical amp (EDFA), OEO Optical amp (EDFA), OEO Optical amp (EDFA), OEO Resilience down to RN Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) Dual feeder fibre (2:N splitter) OLT Rx Disp. Comp. WLs Tunable Tx/Rx Link budget (total reach+split) PIN-PD + OA DCF at OLT side (all pon), EDC at ONU side (per drop) O/C/L-band O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 39 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 4.1.1 Serial 40G NRZ The main problem with 40G serial NRZ is the limited reach due to chromatic dispersion (CD). Using well-known formulas for the dispersion limit [7], the 40G limit is ~4 km in the C-band and ~30 km in the O-band. The conventional dispersion compensating method is by DCF, but different methods are indicated in Figure 5. Typical DCF has up to 1 dB/km loss and 1 km DCF compensates 5-10 km of standard fibre in the C-band. In a PON system, each drop fibre may be of different length and attaching DCF of different lengths to each drop fibre should be avoided. However, if the difference between the shortest and longest drop fibre (differential length) is within 4 km (the CD limit of 40G systems), a DCF of the average length of the PON could be placed at the OLT side to provide dispersion compensation for the whole PON. If the differential length is >4 km, adaptive dispersion compensation should be used and preferably low-cost since it has to be used at every ONU. One such solution is electronic dispersion compensation (EDC). Adaptive EDC can increase the differential length and/or relax the requirement on the OLT side with DCF being very precise in compensation. Another alternative for coping with dispersion which eliminates use of DCF is the use of a chirp-managed laser (CML) for phase correlation between adjacent bits [8]. The transmitter consists of a conventional 40G directly modulated DFB with a subsequent optical filter which performs the FM (frequency modulation) to AM (amplitude modulation) conversion. This transmitter allows for 20km reach without dispersion compensation over a SSMF. Another problem with 40G NRZ is the high optical power required at the receiver. Reported results for PIN-SOA receivers [9] states -17 dBm at BER=10-10 in the C-band. However, similar to XG-PON1, strong FEC can be used which reduces the received BER limit to ~10-3. In this case, the PIN-SOA 40G receiver sensitivity would be ~-20 dBm. Similar to the XGPON1 link budget class Extended 2, an optical amplifier can be used at the OLT Tx boosting the optical signal to ~+ 12 dBm. The resulting link budget would be ~27 dB (OLT Tx – DCF IL at 20 km compensation – CD penalty – ONU Rx sens: 12 dBm – 3 dB – 2 dB – (-20 dBm) = 27 dB. 27 dB in the C-band can be used for ~25 km ODN reach at 1:32 split. Due to limited power budget there are clear limitations to increasing reach and/or split ratio (beyond 1:32) without use of active equipment. Hence, reach as well as number of subscribers per feeder fibre are clearly restricted. Component availability and maturity is currently quite low due to the required high rate DFB laser. Cost is impacted mainly by the required high speed transceivers. Figure 5 : O O A S OA AS SEEE FP7 – ICT– GA 249025 Downstream serial 40G NRZ Page: 40 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 4.1.2 Serial 40G QPSK Another alternative for serial transmission is 40G QPSK using coherent detection. Compared to direct-detection schemes, coherent detection allows for increased power budget with increase reach and splitting ratio. The main drawback is the high cost associated with the coherent receiver. The basic setup for the downstream is shown in Figure 6 which includes a digital coherent receiver based on high speed ADC and DSP. At the transmitter side a DFB in conjunction with a MZ-modulator is used to generate the optical signal. At the receiver side the optical signal is detected by a coherent receiver consisting of a LO, an optical phase hybrid, a photoelectric conversion circuit and AD-converters. A digital signal processing unit (DSP) manages distortion equalization, carrier frequency offset (CFO) compensation, phase recovery and symbol detection. For QPSK small laser frequency offset introduces a static rotation of the constellation. Larger offsets require frequency offset estimation [10]. The CFO information can be used either as feedback through automatic frequency control (AFC) to tune the LO or to digitally reverse rotate the constellation in the DSP. Carrier phase estimation is used in a similar way to reverse rotate the constellation diagram digitally. Decision strategy (symbol detection) could involve a decision feedback with minor extra complexity. Practically, electronic dispersion compensation is performed at the receiver in the DSP, hence avoiding need for costly and bulky optical dispersion compensation. QPSK could also be combined with Polarization Multiplexing (PM) further reducing the symbol rate. These results in a 3dB penalty compared to single polarization at twice the rate. The reduced baud rate of QPSK results in increased CD tolerance but leads to decreased OSNR tolerance. By employing polarization multiplexing a rate of 10G would be sufficient to achieve 40G. Long reach, large fan-out and high bandwidth can be achieved with the scheme. The main drawback is the higher complexity related to coherent detection at the ONU side. Figure 6 : Downstream 40G QPSK 4.1.3 Serial 40G DQPSK An alternative to QPSK modulation is using the differential (D)QPSK format which relieves the receiver from detecting absolute phase as information is encoded in the changes of the phase. A schematic system diagram is shown in Figure 7. Synchronization with a local oscillator (LO) at the receiver is avoided. At the transceiver a precoder is introduced to provide the correct input to the MZ-modulator based on the current state. At the receiver side, instead of a LO the signal is multiplied with a version of itself delayed by one symbol period reducing complexity. Similar to QPSK the DQPSK format transmits 2 bits per symbol resulting in a baud rate which is half the bit-rate. DQPSK is tolerant to chromatic dispersion (CD), polarization-mode dispersion (PMD), and has high spectral efficiency. Although the configuration of a DQPSK system is less complex compared with a QPSK system, large size and high power consumption of the optical transceivers still pose challenges. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 41 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Long reach and high bandwidth can be achieved. Large fan out can be achieved. The main drawback is large complexity although lower than for the QPSK scheme. The main cost contributor for the system is the coherent detector at the ONU although the local oscillator is avoided compared to the QPSK scheme. Figure 7 : Downstream 40G DQPSK 4.1.4 Serial 4x10G NRZ Higher throughput can also be achieved by stacking multiple 10G TDM-PONs over different wavelengths. Here consider the downstream stacking of 4x XG-PON in order to realize a 40G-PON. Stacking requires support in the wavelength plan. For the considered system, one solution is to use the current extended wavelength plan in Figure 4 with 1575-1580 nm (190.3-187.9 THz) for the downstream using 200 GHz spacing for the four stacked PONs. The system scheme for the downstream of a stacked PON is illustrated in Figure 8. An EML (DFB integrated with EAM) per wavelength is used on the OLT side. A MUX/DEMUX is required on the OLT side for multiplexing the four downstream wavelengths and demultiplexing the upstream wavelengths if stacking is used also in the upstream. With the requirement of colourless ONUs, a tuneable filter is used to set the downstream wavelength of the ONU. Assuming parameters from XG-PON1 link budget class Nominal 2, with minimum mean launched transmitter power of +4 dBm and a minimum receiver sensitivity of -28 dBm at BER reference level ~10-3 (strong FEC), the resulting link budget would be ~26 dB (OLT Tx – MUX – 2 x diplexer - CD penalty – TF - ONU Rx sens: 4 dBm - 2dB -1 dB – 2 dB – 1dB (-28 dBm) = 26 dB. 26 dB in the C-band can be used for ~21 km ODN reach at 1:32 split (accounting for 3 splices and 4 connectors). At a 2:32 split the corresponding reach is 20 km. The additional insertion loss of the MUX/DEMUX and TF reduces reach slightly compared to 10G PON. The existence of a power splitter limits potential split and reach. Increased split or reach beyond 20 km requires use of active reach extenders. The dispersion limit is 61 km (1565 at 1dB optical path penalty) which is well beyond the attenuation limit. However for reach beyond 60 km some form of dispersion compensation may be required. A principle drawback with stacked PON compared to the other 40G PON alternatives discussed in this section is that the ONU peak rate is limited to 10G which in principle limits bandwidth flexibility. Statistical multiplexing across the stacked PONs can to some degree be exploited, although not to the same extent as in the corresponding 40G TDM-PON system. Another drawback compared to 40G PON is the need for four times more transceivers at the OLT. Component availability and maturity is high since most of the required components have already been used for 10G PON. The solution depends most critically on the availability of low-cost tuneable filters and lasers. Stacking could also be used in the upstream direction. For colourless ONU operation this requires a tuneable laser at the ONU side. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 42 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 8 : Downstream stacked 4x10G NRZ 4.2 WDM-PON The potential advantages of using WDM in fibre access networks are well known: Protocol and service agnostic end-to-end connectivity, on a per-wavelength basis; Bandwidth scalability, due to bit rate transparency and easy channel upgrade; Enhanced distance reach, due to the low optical path loss; Simplified network operation (one solution for business, residential and backhaul) and maintenance (splitting different services to different wavelengths); However, the high cost of some of the optical components today still is a major obstacle to the application of WDM techniques in the access network. This clearly calls for improvements of (components) technology, at least for some crucial components such as tunable lasers. In addition, there is a strong dependence between system concept – which in turn depends on the major components which become available – and the related network architectures. Part of the network architecture-related advantages listed above (e.g., simplified operations) heavily depend on the system concepts and the relevant components used, where in turn systemsrelated advantages like long reach depend on architecture prerequisites like the allowance of filters, rather than power splitters. As far as the system architecture is concerned, we can distinguish three main classes of WDM-PONs, all leading to one single, wavelength-agnostic variant of the ONU. This is a frequent requirement of network operators and service providers in order to simplify the network operation, reduce the inventory and leverage on high manufacturing volumes. In the first class, the upstream optical carrier is locally generated, e.g., using tunable lasers (WDM-PON with Tunable Lasers and Laser-Arrays). In the second class, the optical carriers are remotely generated and then distributed to the ONUs, where different types of devices (RSOAs, REAMs or injection-locked FP-LDs) can be used to modulate the upstream signal (WDM-PON with seeded Reflective Transmitters). For these two first classes, similar principles are applicable to both, the ONUs and also the OLT (where the OLT would have to make use of the respective array technology). In the third class, each downstream wavelength is amplified and re-used to also carry the upstream traffic: this is done on the same fibre using techniques able to avoid downstream/upstream interference (WDM-PON with Re-use). Basically, the third class is a derivative of the seeded reflective approach. 4.2.1 WDM-PON with (Tunable) Lasers and Laser-Arrays Coming from typical WDM transport systems as they are used in backhaul, metro core and backbone transport, lasers as transmitters in both, the OLT and the ONUs, are the first natural choice. However, the requirement for colourless or wavelength-agnostic ONUs leads to the O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 43 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 necessity to use tunable lasers in the ONUs. So far (2010), such lasers are prohibitively expensive in order to enable WDM-PON to become a widespread – residential – access solution. This leads to the seeded reflective WDM-PON approaches, and meanwhile there is work in its way which will also lead to ultra-low-cost tunable lasers which can be used in the PON access context. The most basic block diagram of a laser-based WDM-PON is shown in Figure 9. Three major elements can be identified which are necessary for any efficient WDM-PON solution: A cyclic AWG in the RN to allow single-fibre working (SFW) An array technology for the OLT transceivers An ONU transceiver based on a low-cost wavelength-agnostic transmitter, a diplexer, and the related receiver Cyclic AWGs are the most elegant way to provide SFW in WDM-PON. Other filter technologies like TFF can be used but they either lead to the requirement of additional diplexers for combining both directions of transmission (downstream, upstream), or they have higher insertion loss. Here, a C+L-band cyclic AWG is used. Other options exist (both example red and blue sub-band in the C-band). RN L-Band 100 GHz Tx Array 1 ... Mux PIC N Figure 9 : Cyclic AWG 1 ... C C-Band ~100GHz N Rx Array L Cyclic AWG OLT ONU T-LD SFF Rx Identical Cyclic AWGs Generic WDM-PON with (fixed or tunable) laser array in OLT and tunable lasers in ONUs Array technologies for the multiple OLT transceivers are a must. Such photonic-integrated technologies and their related components (PICs, Photonic Integrated Circuits) are required for reasons of cost (components’ CapEx), footprint, and energy consumption. An array of dedicated transceivers (e.g., SFPs, SFFs, XFPs) would not be acceptable as OLT solution specially with regard to accumulated form factor and energy consumption. This holds particularly when being benchmarked against EPON/GPON or XG-PON solutions. The OLT PIC can consist of two arrays, one for the receivers and one for laser transmitters. If the PIC contains all channels, the lasers need not be tunable, however, tuneability may increase the production yield of such devices and help decrease the cost of the PICs. The ONU transceiver needs to be low in cost, low in energy consumption, and wavelengthagnostic (“colourless”). Both, low cost and low energy consumption may lead to the requirement of a coolerless and wavelength locker-less component. This may lead to special tuning requirements in the WDM-PON context and is covered later. A variant of the basic, laser-based WDM-PON is shown in Figure 10. This variant supports ODNs with power splitters/combiners rather than WDM filters (AWGs). O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 44 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 OLT ONU RN Tx Array 1 ... Mux PIC T-LD L C SFF TF Rx Rx Array 1 ... N Cyclic AWG N Figure 10 : WDM-PON with tunable lasers in ONUs supporting power-split ODN Generally, this chapter focuses on filter-based WDM-PON. WDM support on splitter-based ODN may become a requirement for reasons of migration into existing (EPON/GPON) infrastructure. As compared to Figure 9, the basic WDM-PON now requires an additional wavelength-selective component for the receiver. This function is necessary because now all downstream channels are broadcasted via the power splitter. Generally, two options exist for a wavelength-selective receiver, tunable filters or coherent detection using a tunable local oscillator (i.e., a tunable local laser). As compared to the broadband direct-detection receiver used in Figure 9, both add cost to the WDM-PON ONU and must hence be seen a critical component. Considering system performance WDM-PONs should be operated via filters. In addition, one of the major advantages of WDM-PONs as listed at the beginning of this chapter also gets lost when using a splitter-based ODN: the long reach performance. Where AWGs typically have 4-5 dB of insertion loss, the loss of power splitters increases exponentially with fan-out port count. For example, 1:64 splitters have ~20 dB loss, and 1:1024 splitting would lead to almost 35 dB of insertion loss. The most important way to make tunable lasers low in cost is to reduce their complexity. Every sub-component which is not absolutely necessary must be omitted. In addition, most network operators in FSAN require coolerless transceivers due to both, reduced cost and reduced energy consumption. Hence, he respective tunables must be coolerless. Tuneability should not come at significant added cost (e.g., complicated external cavities etc.). Hence, lasers must be chosen which are inherently tuneable, for example 3-Section DBR lasers. Further, the onboard wavelength locker should be omitted because of added cost. The result would be a monolithically integrated (multi-section) laser without locker and cooler, a component which has true lowest-cost potential (because it is not more complex than grey, low-cost SFPs or EPON/GPON transceivers). However, since the resulting device will lack its own wavelength locker, system-wide means for tuning the ONU lasers must be established. Generally, two ways for ONU tuning in the PON system context are possible, i.e. tuning via the OLT and tuning via the RN. Both enable implementation of a distributed, shared wavelength locker, at lower cost than the respective number of dedicated lockers. The principle of tuning the ONU laser via the OLT is shown in Figure 11. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 45 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 OLT RN ... L C (C-Band) Y GHz Tune Embedded Communications Channel ECC µC2 Rx with ECC 1 Data AWG Tx Array ... µC1 N T-Tx AWG AWG 1 ONU (L-Band) X GHz Rx Array N Figure 11 : Tuning of tunable ONU via closed-loop control incorporating the OLT and an ECC Tuning is possible because two wavelength-selective elements are part of the optical path – the AWGs in the RN and in the OLT. Hence, tuning can be based on controlling the ONU laser in order to receive maximum power at the OLT. The respective control commands can be sent from the OLT to the ONU via an Embedded Communications Channel (ECC) which in turn can, e.g., be based on a pilot tone modulation. This ECC can always be established in the downstream because the ONU can have a broadband receiver. An alternative tuning scheme is based on autonomously tuning the ONUs with the help of a partial reflector which is placed in the upstream directly behind the AWG in the RN, refer to Figure 12. This partial reflector has a small influence on the power budget (<1 dB) Now, the AWG in the RN is traversed twice. In order to provide a unique tuning criterion, the ONUs must now add a specific pilot tone to their upstream signal. The respective portion of this pilot tone can be used in the ONU for exact tuning. Tx AWG AWG Rx (L-Band) X GHz A C L A (C-Band) Y GHz RN ONU Cyclic AWG OLT T-Tx Tune Tone µC Rx Partial Reflector (wavelength-selective) Data Figure 12 : Tuning of tunable ONU via closed-loop control incorporating a partial reflector in the RN Both tuning schemes described herein so far have in common that they do not need dedicated wavelength lockers per ONU laser. These lockers are replaced by a shared, common locker mechanism which can be implemented at lower cost. For example, the partial reflector shown in Figure 12 is part of this shared wavelength locker. This way, the ONU transmitters can be reduced in complexity and cost. It is worth noting that the same transmitters are useless in most other applications because of the missing tuning capability (and also because these transmitters are designed for maximum distances probably not exceeding 100 km). The DEMUX operation of a cyclic AWG is illustrated in Figure 13 (the MUX functionality is obtained by inverting the propagation direction): two combs of equally spaced frequencies enter into the common port of the device; a pair of corresponding wavelengths in the two combs, separated by a frequency gap , is present at each output port. Each pair of wavelengths is finally separated or coupled by wideband WDM splitters at the OLT and each ONT. Typical choices for the two upstream and downstream bands are C (1530 – 1565 nm) and L band (1565 – 1625 nm) or “blue” (1530 – 1547.5 nm) and “red” (1547.5 nm – 1565 nm) sub-bands in the C band. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 46 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 13 : Cyclic AWG With a simple, straightforward DWDM-PON, no more than 80…96 bi-directional channels are possible with SFW (single-fibre working). Since massively more channels may be required, ways of extending the channel count must be considered. With a straightforward DWDM-PON involving direct-detection receivers for lowest-possible cost, several such ways exist, extending the wavelength region, and decreasing the WDM grid to an extent where simple direct-detection receivers can still be used. Decreasing the WDM grid down to 25 GHz is considered feasible. This is due to the fact that only two (multi-stage) filters have to be traversed, one multiplexer and one demultiplexer. In a PON, there are no effects of multiple cascaded filters (which might have significantly narrower passbands). Also, tuning to within 25 GHz is still considered possible, as is proper passive demultiplexing. The opposite is already true for 12.5 GHz grid. Here, all tolerances decrease to a point where e.g., coherent detection might become necessary. If bit-rate and frequency stability are sufficiently lower than the AWG bandwidth, a straightforward technique is to use tunable lasers, to adjust the upstream optical carrier frequency so that it does not collide with the downstream one but remains close enough to pass through the same port of an optical AWG. A typical situation is illustrated in following Figure 14. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 47 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Downstream optical carrier Upstream optical carrier GbE BW GbE BW GbE BW GbE BW US Frequency stability DS Frequency stability 1.25 GHz 10 GHz 1.25 GHz 1.25 GHz 10 GHz 1.25 GHz 25 GHz AWG pass band Figure 14 : DS and US frequency division at the ONU. The only drawback of this solution is the cost of the tunable laser that currently limits its use just to FTTB applications. However, low cost tunable lasers are under study and could extend the applicability field of this simple and effective scheme. A simple extension of a WDM-PON to 25 GHz grid is shown in Figure 15. Like before, the PON is based on highly integrated PICs in the OLT, and low-cost tunables in the ONUs. In this example, two wavelength grids of 50 GHz each are interleaved to 25 GHz. OLT RN 2 C AWG L ONU RN 1 AWG L-Band 25 GHz 25 GHz Interleaver PIC 2 (50 GHz) 25 GHz Interleaver AWG AWG RXA TXA PIC 1 (50 GHz) T-LD Rx SFF C-Band ~25 GHz FSR = 100 GHz Cyclic 100 GHz 1:48 AWG Figure 15 : Laser-based WDM-PON with 25 GHz spacing for enhanced channel count Different architectures of cascaded filters are possible, which are for further study. Either interleavers (IL) plus AWGs or cascaded AWGs with suitable FSR can be used. Note that in the OLT shown in Figure 15, two ILs must be used due to the slightly deviating US/DS grid. It is also worth noting that the system concept according to Figure 15 is still based on the same PICs and low-cost tunables as described before. So far, no severe complexity has been added. Hence similar per-channel cost can be assumed. Further increase of the channel count is possible by extending the wavelength region. If the low-loss region is considered, only the S-band and part of the eXtended L-band (XL) are suitable candidates. The XL-band, however, should be avoided due to the dramatic increase of added fibre-bending loss. This leaves the Sband as the primary contender for wavelength extension. The block diagram of a DWDM-PON allocating the S-, C- and L-band and also using the 25GHz grid is shown in Figure 16. This WDM-PON is based on band splitters (C/L, S/C, and also red/blue within the S-band, R/B), interleavers, and also cascaded AWGs. This is one possible filter architecture, and subject to further investigation. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 48 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 OLT PIC 3 S, 50 GHz PIC 4 S, 50 GHz T-LD Rx C L/SR-Band 25 GHz C AWG AWG C/L 25 GHz Interleaver L SFF C S S C/SB-Band ~25 GHz AWG S 25 GHz IL PIC 2 C/L, 50 GHz ONU RN 2 RN 1 25 GHz IL 25 GHz Interleaver AWG AWG RXA TXA PIC 1 C/L, 50 GHz Cyclic 100 GHz 1:48 AWG R B FSR = 100 GHz Figure 16 : Laser-based WDM-PON with 25 GHz grid and S-band extension for very high channel count The PON shown in Figure 16 can still make use of the aforementioned components, and may thus maintain its inherent cost advantages. However, with such a massive DWDM-PON, first problems also arise. Due to the multiple splitter / interleaver / filter stages, severe insertion loss is accumulated. With lowest-cost (PIN) receivers, and without any added means for reach extension, maximum reach will be limited to something in the range of 5 dB fibre power budget. Hence, such a PON may requires high-power transmitters (which is generally possible at least with some of the tunables), pre-amplifiers in the OLT, and/or APD receivers. Nonetheless, up to a theoretical maximum of 384 channels (96 channels each in C-, L-, S/red and S/blue) are feasible. Further problems relate to laser safety, and Raman loss / crosstalk. Laser safety must be carefully considered due to the high channel count. Amongst others, OLT downstream Raman pumping for extended reach may become impossible due to strict Laser Safety Class 1M restrictions. Nonetheless, operation with >300 channels over >50 km is considered possible. Raman pumping must also be considered carefully due to the broad spectrum covered with channels. However, it can be shown that the S-band is well-suited to Raman-amplify the C/Lband. This can be used to have the downstream pumped by the upstream. Such a power transfer may become necessary because typically, the downstream in a PON is weaker than the upstream. This is due to the fact that the upstream can easily be amplified by means of a lumped amplifier (e.g., an EDFA) in the OLT. The opposite – a booster amplifier – is not true due to laser safety restriction. The spectral loss of a standard single-mode fibre is shown in Figure 17. Also shown are the lower and upper limits for the loss according to ITU-T G.652 [11]. It can be seen that even with use of the S-band, the water-peak region is still avoided and maximum fibre loss is limited to within 0.4 dB/km (assuming correct fibre deployment etc.). O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 49 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Attenuation [dB/km] 1.0 S-Band C 0.5 L ITU upper Limit 0 1400 1500 1600 Wavelength [nm] Figure 17 : SSMF spectral attenuation. The dashed lines are the boundaries according to ITU-T G.652A [11]. Many candidates for tunable lasers have been described in the literature so far. The two most important classes are DBR (Distributed Bragg Reflector) lasers and external-cavity lasers (ECL). However, few of these lasers have the potential for true lowest cost. For a complete overview of tunable laser candidates see chapter 5.3. In conclusion, the main advantages of a WDM-PON based on tunable laser are: High distance reach (50 km or more, depending on the cable attenuation), due to the low optical components loss (~5 dB per AWG, ~0.7 dB per WDM coupler). The performance is equal for downlink and uplink and can be further increased by using optical amplification at the OLT; No penalty from lumped or distributed reflections compared to reflected approaches (due to wavelength diversity between upstream and downstream) along the optical path, leading to predictable performance, easy network planning and minimization of troubleshooting in field, with consequent improvement of operational costs and customer satisfaction; 10Gb/s and beyond compatibility for future capacity upgrade, in order to allow a long term profitable investment to the operator, with no need of further infrastructure changes; Easy scalability and resilience to failures (channels can be added or removed with no impact on the other channels, PIC needs less to no cabling, for highest availability PICs can be duplicated). 4.2.2 WDM-PON with seeded Reflective Transmitters The basic alternative to using tunable transmitters in a WDM-PON consists of using seeded reflective transmitters. This approach holds for both, OLT PICs and ONUs, as can be seen from Figure 18. Unlike older approaches, the light sources used for seeding the reflective transmitters (transmitter arrays) are based on Multi-Frequency Lasers, MFL. Compared to spectrally sliced broadband sources like ASE sources or LEDs, MFLs have the advantage of higher power levels which translates to better reach performance. Regarding the OLT, compact state of the art components can be used with excellent performance but with little opportunities for cost reduction. Instead the single chip laser arrays is the most promising alternative. Arrays bring down costs, in particular those related to frequency stabilization, that can be shared for a whole WDM comb instead of being performed on a per channel basis. Tuning accuracy compatible with the 100 GHz ITU grid is one of major cost sources for O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 50 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 WDM transmitters. This is seen by comparing the price of a gray SFP (about 25 dollars) and a colored SFP (about 500 dollars). Some suppliers have a low cost solutions in the O band (10 USD for a 16 channels device), based on FPLD based on quantum dots in GaAs. Their technology, based on quantum dots, is extremely interesting, but difficult to extended to the C and L bands. The PON shown in Figure 18 is based on two MFLs, one for seeding a reflective (REAM, RSOA, or combination thereof) transmitter array in the OLT, and a second MFL for seeding the ONUs. In a first simple implementation these MFLs can work in two different wavelength bands, e.g., C-band and L-band. RN REAM Array L-Band 100 GHz MFL C-Band ~100GHz Cyclic AWG MFL Rx Array ONU Cyclic AWG Cyclic AWG OLT Rx RSOA/ IL-FP/ REAM Tx Data Figure 18 : Generic WDM-PON with seeded reflective (REAM) OLT transceiver array and seeded reflective ONUs (based on RSOAs, REAMs, IL-FP lasers, or combinations thereof, e.g., REAM-SOA) Due to MFL seeding, no tuning of transmitters is required. Seeds and reflected modulated signals must be separated by means of circulators. It must be noted that the seeded reflective approach can be implemented for the OLT or the ONUs independently. Some publications describe that better reach performance can be achieved with ONUs based on tunables, rather than seeded reflective ONUs. Independent from that, seeded reflective OLT arrays must be considered an attractive alternative to laser arrays. One reason is that in a PoP or local exchange which terminates more than one PON, the MFL for seeding the OLT PIC can be shared between the different PONs, thus decreasing cost. This is shown in Figure 19. Here, a multiple OLT is shown. Two MFLs are used for redundancy, and than shared (distributed) amongst the PONs. Since the MFL is a contributor to total cost, this sharing approach helps decreasing per-client cost. Note that the WDM-PON shown in Figure 19 is further based on (low-cost) tunable ONUs. Array L MFL 2xN … L MFL WDM-PON N C WDM-PON 2 Array REAM Array RN ONU C WDM-PON 1 L C Cyclic AWG 2 redundant MFLs Multi-OLT T-LD SFF Rx Rx Array Figure 19 : WDM-PON with seeded reflective (REAM) OLT transceiver array and tuneable ONUs. Redundant MFLs in the OLT are shared between several PONs. With regard to ONU seeding, various approaches have been described in the literature. This includes different reflective devices – RSOAs, REAMs, IL-FP lasers – as well as different O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 51 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 schemes for providing the seed. The latter applies to both, the fibres used for seeding as well as the seed wavelengths. In a WDM-PON using reflective ONUs, upstream and downstream can make use of the same wavelengths. This increases spectral efficiency, or the number of bi-directional channels. Two problems arise, reach limitation due to Rayleigh crosstalk (which is unavoidable), and the necessity to provide a modulated downstream seed which has constant envelope in order to allow proper re-modulation. The latter can be achieved, e.g., through using FSK modulation with constant envelope for the downstream, see Figure 20. This modulation is removed in the RSOA which then performs OOK modulation onto the same wavelength for upstream transmission. As already mentioned, this approach is severely limited by Rayleigh scattering, and also by crosstalk due to downstream/upstream modulation. 3dB Rx AWG OOK ONU RSOA Data A ... RX Array FM Disc. A AWG TX Array AWG FSK OLT Figure 20 : Increase of spectral efficiency in WDM-PON with seeded ONUs through use of constantenvelope FSK downstream modulation and intensity upstream modulation Rx AWG TRX Array AWG In order to significantly improve maximum reach of WDM-PONs with seeded ONUs, dedicated seed fibres have been proposed. These dedicated fibres can run all the way down to the ONU (end-to-end), see Figure 21, or they can be used between the OLT and the RN only, see Figure 22. Provisioning of end-to-end seed fibres (Figure 21) has the advantage of complete elimination of Rayleigh crosstalk. However, it contradicts the common PON requirement of support of SFW. SOA Data AWG ... ONU MFL Seed OLT Figure 21 : Decrease of seed/upstream Rayleigh crosstalk through end-to-end dual-fibre working Provisioning of dedicated seed fibres between OLT and RN assumes that the majority of total distance falls into the feeder fibres. In that case, most of the Rayleigh-related penalties are eliminated, and dual-fibre working on the distribution fibres is avoided. This is the common implementation used today for long-reach experiments with seeded reflective ONUs. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 52 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Rx AWG AWG Tx Array plus Seed ... Data AWG AWG Rx Array ONU RSOA REAM OLT Figure 22 : Decrease of Rayleigh crosstalk through use of dedicated upstream feeder fibre An alternative for reducing Rayleigh losses is to distribute the seed sources more geographically, refer to Figure 23. Here, an active RN is used to accommodate the seed source (e.g., an MFL). Such an active RN can still be much lower in complexity as compared to a fully-blown local exchange, however the obvious disadvantage is that the ODN is not fully passive anymore. In the case of the PON shown in Figure 23, dedicated seed distribution fibres are used. One alternative may be the use of band splitters in order to allow SFW. The variant shown in Figure 23 has long-reach capability. It should be noted that the multiwavelength source does not need to be managed and only needs electrical power supply. Rx AWG AWG TRX Array SOA ... AWG MFL Seed Data ONU OLT Figure 23 : Decrease of Rayleigh crosstalk through RN-based MFL Seed accommodation Different combinations of the generic variants for seeded WDM-PONs have been used for long-reach and high-capacity (user count) demonstrators in the recent past. One such example is the so-called Super-PON shown in Figure 24 [12]. Basically, this is an application of active RNs used for seed accommodation, together with a hybrid WDM-PON scheme used for wavelength sharing. In this early example, 89 km of feeder fibre could be spanned, which together with another 10 km of distribution fibres allowed for a total distance of 100 km. The Super-PON is an example for both, active RNs and a hybrid implementation. It is an early demonstration that the combination of very long reach (100 km) and very high user count may cost-effectively only be achieved with the help of active components in the ODN. 90 km Laser 5 km 5 km R/B OLT Active Node Rx SOA ONU ... CW Rx R/B PN R/B – Red/Blue-Band Splitter Figure 24 : Super-PON [12] as an example for active-RN-based seed accommodation O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 53 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Most recently, a highest-performing, seeded-ONU WDM-PON has been described, refer to Figure 25 [13]. This PON was based on 32 wavelength pairs, each running at symmetric 10 Gb/s. Seeded REAM-SOA combinations were used for the ONUs, and the ODN consisted of WDM filters and cascaded 2:64 and 1:4 power splitters. Total customer count was 8192, with each customer getting a symmetric bandwidth of approximately 37 Mb/s. Obviously, with lower power-split fan-out, guaranteed bandwidth could easily have reached several 100 Mb/s. PoP / Service Node Active RN / LX PRN PRN ONU BC Seed APD Rx BS REAM-SOA ... AWG DS Tx MDXM ... Seed ... DS Tx 2:64 1:4 PRN PRN ONU BC APD Rx BS REAM-SOA ... AWG MDXM BM Rx ... BM Rx 124 km 10 km 1 km 100 m Figure 25 : DWDM-TDMA-PON [13] as an example for dedicated upstream feeder fibre The hybrid WDM/TDMA-PON shown in Figure 25 is another example for a highperformance PON which involved an active RN. Here, this RN was used to accommodate amplifiers which were necessary in order to compensate the accumulated loss of the ODN. The implementation of a colourless transmitter at the ONT is a crucial point. RSOAs or FPLD are viable options for 1.25 Gb/s channels and some manufacturers are investigating if the RSOA bandwidth (currently about 1.5 GHz) can be used for 2.5 Gb/s. Future 10Gb/s solutions entirely based on out of the shelf optical components have been investigated (SOAs, circulators and EAM). Two variants of the ONU transmitter are shown in Figure 26. in out 2 3 1 ONU SOA EAM in out SOA SOA 2 3 EAM 1 10G ONU 10G Figure 26 : ONU transmitter variants For both variants, a CW signal enters into a circulator, is modulated by a conventional EAM and looped back into the same circulator. The signal can be amplified by an input bidirectional SOA, or by separated pre- and booster amplifiers within the ring containing the EAM. In a simpler scheme (Figure 27), a SOA precedes a REAM. An attractive feature is that the modulator can be properly designed to introduce a chirp which compensates for the fibre chromatic dispersion, without lossy and expensive external dispersion compensators. Using the REAM, 50 km of link distance without dispersion compensation at 10 Gb/s has been observed in the Ericsson Research Lab in Pisa. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 54 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 27 : Simple ONU transmitter In conclusion, the main advantage of seeded solutions lies in the low cost of implementing colorless ONU solutions while performance is less strong compared to solutions based on tunable laser. Performance penalties are mainly due to: Rayleigh backscattering and the remote distribution of high power non-modulated channels for the uplink traffic. The unavoidable trade-off between performance and distance of the remote channel distribution 4.2.3 WDM-PON with wavelength reuse WDM-PON systems where downstream wavelengths are reused in the upstream allow for wavelength bi-directional channels and as a result doubled system capacity as well as number of users. Such wavelength re-use requires downstream modulation to be cancelled before applying upstream data and leads to additional penalty to the upstream signal, caused by residual downstream modulation and optical reflections along the optical link. Different schemes for wavelength re-use are discussed in the following, and compared in terms of cost and performance. The most basic approach is based on RSOA saturation at the ONU. Part of the downstream signal is tapped and sent to an RSOA. If the power of the tapped signal is sufficiently high (>10 dBm for the best devices available on the market), the RSOA is saturated cleaning the downstream modulation so that the upstream data can be applied to the RSOA electrical input with only a small penalty due to the residual modulation. In practice, the input power value needed to saturate the RSOA limits achievable power budget in practical systems (typical reach is < 10 km). An alternative technique is based on FSK/ASK and DPSK/ASK coding. With this technique, the downstream traffic is FSK (Frequency Shift Keying) modulated by using a directly modulated laser diode. The DFB laser is biased at a high value and modulated with small modulation depth. This enables low extinction ratio (1.5 ÷ 2 dB) and high chirp (4.5 ÷ 5 GHz), in order to maximize the frequency shift caused by any variation in the driving current. At the ONT, the received wavelength is split into two parts: one part is sent to an optical filter before the downstream receiver while the other part is used as optical carrier by the upstream data, which is intensity modulated by a RSOA. The function of the filter is to convert the FSK modulation format into a regular ASK (Amplitude Shift Keying) format. A simple method is to position the optical carrier frequency on one of the slope edges of the optical filter. The working principle is illustrated in Figure 29. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 55 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Laser Spectrum P 1 1 0 0 t Chirp: 4.5 ÷ 5 GHz Figure 28 : FSK generation by means of a directly modulated laser P filter input 1 1 0 0 laser output λ t Figure 29 : Frequency to Amplitude Modulation by means of an optical filter Colorless ONU operation requires that polarization insensitive periodic filters are used with free spectral range equal to the ITU-T frequency grid spacing (100 GHz). The main drawback with this approach is the additional cost associated with the filter, which requires thermal control or dedicated circuitry to track any drift of the downstream optical carrier. A cost effective implementation is possible by integrating the filter, RSOA and photodiode in a single composite device. Another similar scheme is the RZ/IRZ coding technique which allows for remarkable performance improvement compared to the saturated RSOA approach without introducing any additional optical components (and related cost). The scheme is based on a combination of RZ (Return to Zero) and IRZ (Inverse Return to Zero) modulation formats for upstream and downstream transmission, respectively. The working principle [14] is illustrated in Figure 30, where the inset shows what happens with all the possible combinations of “1” and “0” bits in downstream and upstream. Figure 30 : Figure RZ/IRZ remodulation scheme The downstream signal is coded using Inverse Return to Zero (IRZ) with 50% duty cycle. The RZ upstream bit pattern is shifted by half a bit in time with respect to the downstream IRZ pattern, so that the RZ pulses in upstream always are transmitted when the IRZ signal is at its high level. When the ONU receives a logical ”1” (a dark pulse) it can either suppress the dark O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 56 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 pulse high rail (to re-transmit a ”0”) or amplify it (to re-transmit a ”1”). If a logical”0” (a constant power bit) is received, the RSOA will carve a pulse on it (to retransmit a ”1” ) or will suppress the whole bit (to retransmit a ”0”). This gives four possible 2-bit combinations. The performance is further improved if the RZ modulation format is applied to a RSOA properly designed to introduce a large chirp on the upstream signal. Due to the chirp, the frequency of the upstream signal is made to not coincide with the downstream one in order to mitigate effects of reflections in the optical link. Upstream sensitivity at the OLT depends on cross-talk induced by reflections. In absence of reflections, good sensitivity values can be achieved (about -32 dBm). About 1 dB power penalty is recorded for a cross talk value of 15 dB. These are exceptional results for this class of bi-directional systems, where significant outage is expected for a cross talk of about 25 dB. 4.3 OFDM-PON Orthogonal Frequency Division Multiplexing (OFDM) in optical access networks has been driven by the necessity to increase transmission rates without increasing bandwidth of optoelectronic transceivers. OFDM in the downstream could be combined with Orthogonal Frequency Division Multiple Access (OFDMA) in the upstream. OFDM is a multi-carrier modulation technique where the data of a single optical channel is transmitted on multiple narrow bandwidth orthogonal subcarriers each operating on a low data rate. In OFDM the frequency multiplexed sub-channels allow for transmission of a multi-carrier signal using a single optical modulator. The sub-channels are orthogonal and overlap in frequency in such a way that the peak of each sub-channel coincides with nulls of the other sub-channels. Figure 31 shows the principle setup of an OFDM channel. OFDM offers increased spectral efficiency and reduced baud rate compared to serial OOK. OFDM modulation is very adaptable (bit and power loading) and scalable to higher order modulation formats (8-QAM, 16-QAM, etc.) on the same optical hardware due to easy electronic signal processing. Several OFDM channels, each consisting of several sub-channels, could be carried over the system. An OFDM channel could be in the range of a few GHz. For downstream transmission a single laser module of very high bandwidth (10-25 GHz) is used [15]. Assuming 16-QAM modulation and multiple OFDM channels, aggregate rates of 40-100 Gb/s could be achieved. The bandwidth of the ONU receiver could typically have the full range of the OLT transmitter. The upstream is more challenging due to signal to signal optical beat interference (SS-OBI) which occurs if subcarriers from different ONU transmitters are overlapping in the frequency domain. The ONU laser bandwidth would most likely be in the range of a few GHz. A wavelength demultiplexer plus APD or PIN receivers operating at the same bandwidth could be used at the OLT side. One upstream solution is to operate the ONUs sufficiently far from each other in order to avoid overlap. This implies use of WDM technology with the associated challenges of designing a system for colorless operation. Another option is the use of hybrid ODFMA/TMDA which requires strict synchronization. Ref [16] demonstrates a solution based on ONU-side carrier suppression. Orthogonal subcarriers Guard-band Optical carrier 10G OFDM-channel O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 57 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 31 : OFDM-Transmission on multiple orthogonal subcarriers In general, OFDM data transmission is very tolerant with respect to signal distortions like chromatic dispersion (CD) or polarization mode dispersion (PMD) even at high single channel data rates. OFDM is also less susceptible to E/O band response limitation and distortion. This enables longer reach and increase split ratios without need for dispersion compensation compared to conventional TDM-PON solutions. OFDM is also more robust to Rayleigh backscattering compared to NRZ in typical PON architectures. OFDM allows for use of symbol-by-symbol decision for each sub-channel, rather than complex equalization algorithms to compensate for distortion (e.g chromatic and polarization mode dispersion). In the presence of selective fading, radio OFDM channels can achieve very good performance. However, this is not the case for a fibre optic channel, where the amplitude channel response is essentially flat. An advantage with OFDM is that access to subcarrier granularity enables flexible resource allocation. Resource allocation is for example implemented by dynamic subcarrier allocation (DSCA) or conventional TDMA approaches. Another potential degree of adaptability is the modulation format (QAM-levels) depending on channel. The main challenges of OFDM are the sensitivity to phase noise (PN) and uncompensated residual carrier frequency offset (CFO) resulting in inter-carrier interference (ICI). Also, the large peak-to-average power ratio (PAPR) in the OFDM signal implies large sensitivity to non-linearities in the optical path. OFDM also exhibits various technical drawbacks such as the requirement of high resolution analogue-to-digital converters (ADC) and digital-toanalogue converters (DAC). DACs with at least 7 - 8 bits are necessary at the transmitter, whereas ADCs with 8-bit resolution are needed at the receiver. ADCs are now becoming commercially available while DACs could be a bottleneck, depending on the channel rate. The number of sub-carriers per OFDM channel is limited by ADC resolution. Requirements of high resolution ADC may be migitated by increasing the number of QAM-levels. However this comes at the expense of increased SNR requirements. Other drawbacks are the energy consumption and spectral overheads needed to correctly recover information. In the frequency domain, information-less sub-carriers can be used and in the time domain, a cyclic prefix is used. For 20% of virtual sub-carriers, the speed of ADCs is equivalent to an over-sampling factor of 1.2 samples per symbol. In optical communications two main OFDM approaches exist, one being the coherent optical (CO)-OFDM which involves modulation of the optical field and the other being direct detection optical (DDO)-OFDM which involves modulation of the optical intensity. COOFDM has the best performance but also requires more components such as optical or electronic I/Q-Modulators, RF-mixers or coherent receivers. Each of the two approaches exists in several variants that will be discussed more in detail below. 4.3.1 Serial 40G CO-OFDM Figure 32 shows different variants of CO-OFDM with different transceiver and receiver structures [17]. Figure 32a shows a variant based on direct up/down conversion and Figure 32b shows a variant with up/down conversion via an intermediate frequency (IF). O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 58 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 32 : Downstream CO-OFDM [17]: a) direct down/up conversion, b) intermediate frequency In both cases at the transmit end, the original data signal passes through a parallel to serial converter which divides the channel to a number of parallel bit-streams. The digital time domain signal is obtained by IFFT and is subsequently inserted with guard interval and converted into real time waveform through DAC. The guard interval is inserted in order to prevent inter-symbol-interference (ISI) due to channel dispersion. The sub-carrier channels can be modulated through different modulation formats such as BPSK, QPSK, 16-QAM, 64QAM, etc. In the direct up-conversion architecture (Figure 32a), the optical transmitter uses an optical I/Q modulator which consists of two MZMs to up-convert the real/imaginary parts from the RF domain to the optical domain. The OFDM optical receiver uses two pairs of balanced receivers and an optical 90◦ hybrid to perform optical I/Q detection. The RF OFDM receiver performs OFDM base-band processing to recover the data. The main advantages of direct-conversion are (i) elimination of need for image rejection filter in both the transmitter and receiver, and (ii) reduction of required electrical bandwidth for both the transmitter and receiver. In the intermediate frequency (IF) architecture, the OFDM base-band signal is first up-converted to an intermediate frequency in the electrical domain, and the OFDM IF signal is further up-converted to the optical domain with one MZM. In the receiver, the optical OFDM signal is first down-converted to an intermediate frequency and electrical I/Q detection is performed. The signal is then sampled with an ADC, and demodulated by performing FFT and baseband signal processing to recover the data. The DSP unit manages distortion equalization, CFO compensation, phase recovery and symbol detection. With these two transmitter and two receiver alternatives there are four system design alternatives. 4.3.2 Serial 40G DDO-OFDM Perhaps the most attractive OFDM solution for PON access, considering cost, is DDOOFDM. DDO-OFDM allows for a simpler receiver structure than CO-OFDM at the cost of reduced spectral and energy efficiency. The main advantage of DDO-OFDM is the simpler direct detection receiver. Typically DDO-OFDM is realized by the transmission of an optical single sideband (OSSB) OFDM signal and a component at the optical carrier frequency. A frequency guard band separates the OFDM signal from the optical carrier. The signal is O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 59 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 received by detecting the carrier signal mixing products. Three variants of the DDO-OFDM transmitter (Figure 33) [18] are here discussed. These have different degrees of optical complexity, but all require a similar type of single photodiode direct-detection photo receiver (Figure 33). Figure 33 : Transmitter and receiver alternatives for DDO-OFDM [18]: a) b) and c) show different transmitter alternatives and d) a typical receiver. For the different transmitter alternatives the digital sections differ only in the way the input vector is mapped onto the IFFT input vector. For the first transmitter design, a single input optical modulator is used to generate a double sideband optical signal and then one sideband is suppressed using an optical filter. The electrical input to the optical modulator is a real, baseband signal and only one digital-to-analog converter (DAC) is required. For the second design, the single DAC of the first design is replaced by two DACs and an electrical RF upconversion stage. This allows the complex baseband OFDM signal to be mixed with an RF carrier before driving the single input optical modulator. For the third transmitter design, an optical single sideband signal is generated using a signal and its Hilbert transform to drive an optical I/Q modulator. Figure 33 shows a simple direct detection receiver for DDO-OFDM. As a result of the square law characteristic of the photodiode, the received signal consists of a number of mixing products. The mixing products are classified into the useful components from which the data is recovered, unwanted components which fall within band and limit the BER performance, and unwanted components which fall out of band. The useful components are the difference terms which result from the mixing of the OFDM sideband and the optical carrier. There is a frequency dependent phase shift of the optical signal due to chromatic dispersion. For single sideband optical OFDM systems, with linear field modulation, each OFDM subcarrier is represented by a single optical frequency, so phase shifts in the optical domain lead to phase shifts in the demodulated electrical OFDM subcarriers. These phase shifts can be corrected in the digital section of the receiver by applying a single complex multiplication for each subcarrier. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 60 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 OFDM-PON has been demonstrated by several research groups. In [19], 108 Gb/s downstream OFDM-PON transmission using polarization multiplexing and direct detection was demonstrated. In [20] a version of DDO-OFDM referred to as adaptively modulated optical (AMO)-OFDM was demonstrated at 32 Gb/s over 2 km and 25 Gb/s over 60 km. Regarding the upstream, the challenge is to avoid the broadband beating noise without assigning individual wavelengths to each ONU. In [16] a source-free OFDMA-PON architecture that solves this problem was proposed which uses ONU-side carrier suppression to achieve simultaneous upstream transmission from multiple ONUs over a single wavelength. 36 Gb/s upstream transmission over 100km combined with 1:32 optical split was demonstrated. Two main features of the described solution include the ONU-side carrier suppression that avoids OLT-side beating noise, and coherent detection at the OLT that enables full regeneration of the electrical OFDM signal. In Figure 34 the OLT uses two laser sources, one for downstream optical OFDM transmission, and one as a distributed carrier for upstream transmission. An electrical OFDM signal generated by the inverse Fast Fourier Transform (IFFT) is used to drive an intensity modulator (IM), while the upstream continuous wave (CW) laser is distributed downstream to all ONUs through the same fibre path. At the ONUs, the downstream OFDM signal is separated from the distributed CW laser source by an optical filter. The latter is re-used as optical carrier for the upstream of each ONU. For upstream transmission, the upstream signal is first generated in each ONU and then upconverted to an intermediate RF frequency. Each RF signal modulates an IM driven by the CW source. The upstream IM at each ONU uses carrier suppression to avoid the beating noise that would be generated at the OLT due to mixing of multiple optical carriers. As shown in Figure 34, coherent detection is used at the OLT receiver in order to fully regenerate the OFDMA signals. Since the upstream optical carrier was originally distributed from the OLT, it is re-used as the OLT local oscillator (LO). Figure 34 : DDO-OFDMA-PON scheme after [16]. 4.4 CDM Code division multiplexing (CDM) provides another way of multiplexing data. Code division multiple access (CDMA) is the use of CDM technology to arbitrate channel access among multiple network nodes in a distributed fashion. Optical code-division multiplexing (OCDM) is a promising technique, offering random access to the entire bandwidth along with advantages such as simplified network control, increased security, and increased flexibility. OCDM addresses capacity upgrade by adding a code-based dimension to the FTTH system. In this section, we will review the key technologies for optical systems based on OCDM/OCDMA. In an OCDM system, each channel is distinguished by a specific optical code. An encoding operation optically transforms each data bit before transmission while the reverse decoding operation is required to recover the original data at the receiver. The main encoder/decoder technologies are shown in Table 6. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 61 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 6 : Encoder/decoder technologies Coding domain Chip modulation 1D Time Power Phase-shift-keying (PSK) Wavelength Required Laser Devices source Incoherent FDL (Fibre delay line) Coherent PLC (planar lightwave circuit); SSFBG (Super structured Fibre Bragg grating) [21] Power Incoherent PSK Coherent Power Incoherent AWG; FBG (Fibre bragg grating) High resolution phase E/D [22]; SLPM (Spatial light phase modulator) [23]; AWG+FDL [24]; FBGs [25] 3D Time/Wavele Power ngth/Polarizat ion Incoherent AWG+FDL [26]; 2D Time/Wavele ngth Due to the limited coding space, incoherent 1D optical coding technology (either in time or wavelength domain) is not feasible for future access networks which are required to support large numbers of end users. Both 2D and 3D encoding technologies require multiple domains to realize optical codes. Therefore, it is difficult to smoothly upgrade capacity of an access network where 2D or 3D encoders/decoders are employed. 1D coherent coding seems to be the only viable technology. They can have a large coding space. For instance, 511-chip gold code could have in total up to 513 codes (corresponding to 513 ONUs). Paper [21] has already experimentally demonstrated 10-user, 511-chip, 640 Gchip/s (1.25Gb/s) OCDMA system based on 1D coherent coding in the time domain. Paper [27] proposed an architecture of OCDM over WDM, i.e. hybrid WDM/OCDM PON. In this way, the capacity could be easily upgraded by expanding the wavelength spectrum. The most expensive part of the coherent OCDM technology is the transmitter which must be able to generate ultra-short pulses. Usually, the price of the devices is market-driven. If a large amount is required by the market, the expected price of ultra-short pulse lasers could become comparable with tunable lasers which are required in WDM PONs. The OCDM solution could include a passive power splitter at the remote node. OCDM-PON using a tree topology with passive power splitters have been widely studied e.g. in [28] (Figure 35). The maximum number of users that can be supported in this configuration is limited due to the high insertion loss of the power splitter. The coding space cannot be fully utilized. In addition, similar as for TDM-PON, reach is influenced by introduction of power O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 62 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 splitters. Figure 35 : OCDM PON with a power splitter at the remote node An alternative is to introduce code splitting at the remote node. For coherent time-spreading (TS-) OCDM, the multi-port OCDM encoder/decoder (E/D) has the unique capability of simultaneously processing multiple time-spread optical codes (OCs) with a single device (Figure 36), which makes it a potential cost-effective device to be used in the optical line terminal (OLT) of an OCDM network in order to reduce the number of E/Ds [29]. Figure 36 : A device with capability of simultaneously encoding/decoding multiple time-spread optical codes (OCs) [29]. If such a device is employed at the remote node between the OLT and ONUs, coding splitting can be realized [30]. In this way, each ONU is codeless, since E/D is not required at the user side. Furthermore, the insertion loss of this code-splitting device is much lower than the power splitter and comparable with the AWG. Therefore, without reach extenders, OCDMPON with code splitting can support reach beyond what is achievable in systems based on power splitting. Paper [30] experimentally demonstrates a 4-user 10Gb/s system (Figure 37) where both up- and down-link are over 59 km of standard single mode fibre (feeder fibre: 36 km, distribution fibre: 23 km) without inline dispersion compensation. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 63 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 (a) (b) Figure 37 : (a) Configuration of OCDM-PON system with code splitter at RN and (b) experimental setups and results: (i) downlink and (ii) uplink [30]. The multi-port OCDM E/D device has not been commercialized so far. The price will rely on market demand. Since the fabrication of this device is exactly the same as the AWG, the expected price is also similar to the AWG if there is large demand. While an optical access network using OCDM technology usually assumes a tree topology, in particular for OCDM PON, much research work on ring topology has also been investigated in recent years. The OCDM add/drop multiplexer is a key element to support the OCDMA ring. Paper [31] introduced an all-optical 2D OCDM code-drop unit (Figure 38) and experimentally demonstrated its use in an OCDM ring. This code-drop unit has three elements: a drop code O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 64 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 decoder (DCD), a terahertz optical asymmetric demultiplexer (TOAD), and a code restorer (CR). The codes from all other transmitters, whom delays are not matched to the DCD’s delays, spread in time as a cross correlation of maximum height 1 according to the properties of the code. The auto- and cross-correlation, after passing the circulator, enters the TOAD by Port 1. The autocorrelation peak and the control pulse exit the TOAD at Port 2 while the rest of the signal, outside of the switching window, is reflected back and directed by the circulator to Port 3. At Port 2, an optical high pass filter separates the sampled autocorrelation peak from the control signal. The intensity of the dropped code is then detected by a photodiode and converted into the original 2.5-Gb/s data sequence. At Port 3, the remainder of the signal goes through a code recover (CR). The CR undoes the time shifts caused by the DCD to the codes passing through the code-drop unit in order to restore them to their original sequence. The complete architecture for an OCDM add–drop multiplexer based on this type of code-drop unit can be realized, by simply adding a 2x1 power combiner and tunable OCDM transmitter to the output of the node. Figure 38 : Schematic of the all-optical 2-D OCDM code-drop unit [31] (D: Delay, HPF: High-pass filter, and TOAD: terahertz optical asymmetric demultiplexer). However, it is difficult to find a smooth and cost-effective way to upgrade the capacity of a ring network with this OCDM add/drop multiplexer, which only works for incoherent 2D optical codes. For either bit-rate upgrading or wavelength spectrum extending, the existing add/drop multiplexers should be replaced by the new one. To address this problem, paper [24] introduced an OCDM add–drop multiplexer for any coding technology. The disadvantage is that the decoders and encoders for all the optical codes used in network are required at each node. This increases the cost and decreases security. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 65 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 39 : Structure of OCDMA add-drop multiplexer for any coding technology [24] (TOAD: terahertz optical asymmetric demultiplexer). Due to the complicated architectures for the OCDM add-drop multiplexers, OCDM ring might not be suitable for networks which are cost-sensitive. 4.5 WDM+XXM HYBRID The OASE NGOA system requirements of up to 100 km reach (large power budget) and up to 1000 subscribers per fibre feed (high splitting ratio) provide motivation for hybrid concepts. The TDM-PON concepts alone cannot fulfil these combined requirements. Pure WDM-PON is more attractive, although the relatively expensive WDM components such as stable laser sources prevent residential mass-market solutions. The idea of hybrid concepts is to combine the advantages of several different concepts. In a hybrid WDM/TDM-PON this would be the TDM-PON advantages (bandwidth sharing, colourless ONT) with the WDM-PON advantages (increased number of subscribers and total BW). Hybrid concepts are based on WDM where several WDM channels are transmitted in the main feeder fibre to a first remote node, which typically is λ-selective (e.g. AWG wavelength splitter). In the AWG the downstream data channels are de-multiplexed and each wavelength fed into a second feeder fibre to a second remote node which depends on the concept which WDM is combined with (Figure 40). Basically, underlying WDM can be combined with any of the multiplexing and multipleaccess technologies which were listed in Chapter 3. These schemes are TDMA, CDMA and sub-carrier multiple access (SCMA). In addition, dense or ultra-dense WDMA must be considered (because it most effectively runs over the same filter-plus-splitter ODN), and OFDMA may be considered a more recent version of dense SCMA. In addition, CDMA can be implemented in either the optical (O-CDMA) or electrical (E-CDMA) domain, with massive impact on cost, and functionality. It is worth noting that in particular OFDMA, SCMA, and UDWDMA with many densely spaced sub-carriers each do not differ significantly in their power spectra, nor in the resulting spectral efficiency. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 66 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 λ1 1 λ2 … ONT λDS-1 … ONT λ1, λ2, λ3 …λN λN λUS-1 λDS-1…λDS-N OLT … Passivesplitter λDS-N λUS-N … ONT … … λ3 N λ-selective RN (AWG) λUS-1…λUS-N ONT Figure 40 : Hybrid PON basic setup One important question is which scheme allows an optimization of either cost/performance or energy-consumption/performance metrics (as two very important evaluation parameters in the access context). Such an analysis must take the relevant components of each hybrid system (with regard to cost and/or energy consumption) into account. The related cost and energy consumption figures as well as ODN assumptions are summarized in Appendix 7.1. In the comparison, 50 km of fibre were considered. This relates to the fact that almost no PON system with very high fan-out can span 100 km (or similar distances) without active RNs (amplifiers, reach extenders). In order to enable passive ODN wherever possible, 50 km distance was chosen for comparison. Clearly, the active/passive hybrid variants always require active RNs, but at 100 km distance requirement, almost all other variants would have required active RNs as well. Such, we avoided considering over-engineered solutions. 4.5.1 Hybrid WDM/TDM-PON The first hybrid PON analyzed here is the hybrid WDM/TDM-PON. In order to be able to provide high bandwidth and high customer count, we considered DWDM with 40 bidirectional channels of symmetric 10 Gb/s bandwidth each (extendable to 80 channels). Due to the combined ODN insertion loss, this requires 10G burst-mode transceivers with 35 dB power budget. This can be identified as one of the main cost drivers according to Appendix 7.1. A schematic diagram of one WDM/TDM-PON implementation is shown in Figure 41. Note that the most common solution consists of several TDM PONs embedded in a WDM PON system. RN1 Tx/Rx Array AWG TDMA MDXM TDMA MDXM 1xN AWG OLT λD λU RN2 1:k 10G-TDMA ONUn APD FEC SOA Data CLK Rec. T-LD Burst M. FEC Data Figure 41 : Hybrid WDM/TDM-PON Due to the power-budget requirement (35 dB), transceivers need APDs and very likely also O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 67 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 booster amplifiers (SOAs) and added FEC. All these add to the system’s cost and also energy consumption. The resulting figures (excluding common baseline cost and energy consumption contribution) are listed in Table 7. Note that both, cost and energy consumption of the DWDM-tunable 10G 35-dB ONU burst-mode transceiver must be considered as very aggressive – optimistic! – assumptions. Nonetheless, they are the major contributor as could be expected from TDM PONs. Table 7 : Energy consumption and cost figures for the hybrid WDM/TDM-PON variant shown in Figure 41, excluding baseline (chapter 7.1) Cost 217$ total Energy 3.8 W AWG/splitter ports 12$ OLT port 0.25 W OLT port 22$ OLT switching 1.0 W OLT amplifiers 3$ OLT amplifiers 0.05 W ONU TRX 175$ ONU 2.5 W OLT switching 5$ We refer to the WDM-PON section (subsection 4.2.2) for the description of two WDM/TDMPON demonstrator systems: Super-PON (Figure 24, [12]) and DWDM-TDMA-PON (Figure 25, [13]). To increase the flexibility in terms of dynamic bandwidth allocation (DBA), four different hybrid WDM/TDM-PON variations are described below. The flexibility to offer any bandwidth (in the limit of the physical bit rate) to anybody could be a major advantage. A hybrid WDM/TDM-PON with dynamic wavelength routing can combine the virtues of both the TDM and WDM solutions. From the OLT multiple TDM-PONs can be set up, each at a specific wavelength. Each TDM PON serves a set of users, and within this set, capacity is shared. By means of wavelength selection or routing, the number of users within the set can be varied, and thus the capacity offered per user can be varied. Hence a flexible hybrid WDM/TDM-PON can offer capacity-on-demand, and by means of the wavelength selection or routing flexibility, the congestion probability can be significantly reduced with respect to the static WDM/TDM-PON configuration. For example, it could be convenient to offer two different needs: a number of 100 Mb/s connections for subscribers served by a particular set of wavelengths, another number of 1 Gbps connections for subscribers having more needs on another set, and why not a full 10 Gbps pipe if really required. So by combining WDM and TDM, a lot of demands can coexist on a same network infrastructure in a very efficient manner. This is a strong advantage when compared to other systems. The flexibility of the WDM/TDM-PON system is increased by changing the implementation of RN1. The conventional hybrid WDM/TDM-PON scheme presented in Figure 41, which offers low flexibility in terms of wavelength allocation, is used as reference system. Extra flexibility is then added, e.g. by replacing the AWG by a power splitter (to increase flexibility at the cost of power loss and security threats, see Figure 43), by adding extra AWGs used as wavelength combiners (to increase the number of wavelengths sent to each ONU, see Figure 44) or by adding a more complex switching architecture (to increase flexibility without loss of security and/or power loss, see Figure 45 and Figure 46). Note that in the different variations, RN2 consists of a passive power splitter. In the remainder of this section, we have classified the hybrid WDM/TDM PONs in four main categories. These four variations are compared in terms of simplicity, flexibility, power loss and security. We assume that there are Nu uplink and Nd downlink wavelengths while for some of the architectures Nu = Nd = N. The general OLT and ONU architectures are depicted in Figure 42 and consist of following components: O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 68 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 The OLT has Nu uplink line cards and Nd downlink line cards. Each uplink line card is associated with one photo detector (PD) and similarly each downlink line card is associated with one DFB LASER for data transmission. The ONU has a three port circulator for separating uplink and downlink wavelengths. A WDM splitter can also be used for this purpose. The uplink part has the uplink line card and the tuneable burst mode transmitter for tuning to any desired wavelength. The downlink part has one classical photo detector and a tuneable optical filter is used for selecting the desired wavelength. Note that when using only one tuneable Rx in the ONU, the OLT has to store the wavelength each ONU was tuned to for the last frame. If an ONU is tuned to a wrong wavelength or it goes to sleep mode, however, a solution is needed to indicate to the ONU to which wavelength it can listen to. This can be resolved by adding a fixed Rx (in addition to the tuneable Rx) to the ONU, or directly by the MAC protocol, e.g. by indicating one wavelength as fallback. Central office OLT RN1 Wavelength or power splitter RN2 Power splitter User ONU TDM-PON Uplink line card PD PD ONU control Uplink control Uplink line card Tunable Burst Mode Transmitter Nu PD Downlink control Remote Node Downlink line card DFB PD Downlink Tunable optical filter line card DFB Nd DFB Figure 42 : WDM/TDM-PON based on an ONU with one tunable transmitter and one tunable optical filter, for a general remote node In Figure 43 the remote node design consists of a two-stage passive power splitter (RN1 and RN2) and is a simple broadcast and select architecture, where all the wavelengths are broadcast to all ONUs, and the ONU then selects the desired wavelength. This solution has a serious security threat as the content of all the wavelengths is available to all ONUs. It also suffers from high power loss due to the high splitting ratio and therefore requires stringent power budget. This may further restrict the number of users. We assume here that there are Us TDM-PON infrastructures and each TDM-PON has s ONUs. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 69 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Central office OLT RN1 Power splitter RN2 Power splitter User ONU Uplink line card PD s PD ONU control Uplink control Uplink line card Tunable Burst Mode Transmitter Nu PD Downlink control Downlink line card PD Us Downlink Tunable optical filter line card DFB DFB Nd DFB Figure 43 : WDM/TDM-PON as in Figure 42 with cascaded power splitters for the remote node. In Figure 44 the remote node design consists of a WDM splitter (and some wavelength combiners) in RN1 followed by a passive splitter in RN2. This design is proposed in order to overcome the high power loss from the previous remote node design. Instead of transmitting all the wavelengths to all TDM-PONs, we provide Nd/Us number of downstream wavelengths and Nu/Us number of upstream wavelengths to each TDM-PON. Us wavelength combiners are added to offer multiple wavelengths per TDM-PON. Note that this implementation reduces to the reference system of Figure 41 if Us = Nd (Nu). With the reduced power loss of the WDM splitter compared to a passive power splitter, the power budget is increased and more users can be supported. However, the flexibility is somewhat restricted compared to the previous remote node design. Central office OLT RN1 Wavelength splitter PD Nu/Us PD Nd/Us Uplink line card s Nu PD User ONU ONU control Uplink control Uplink line card RN2 Power splitter Tunable Burst Mode Transmitter Us Downlink control Downlink line card Us DFB PD Downlink Tunable optical filter line card DFB Nd Us DFB Figure 44 : Nu/Us Nd/Us WDM/TDM PON as in Figure 42 with a WDM splitter and power splitter for the remote node. In Figure 45 the remote node design consists of a wavelength router in RN1 followed by a passive splitter in RN2. The wavelength router consists of a WDM splitter, a passive splitter stage, optical switches and a passive combiner stage. For downstream transmission, it first splits each wavelength channel into Us different paths for each TDM-PON. Each TDM-PON can get data from all Nd downstream wavelengths. There are Nd optical switches installed to control which wavelengths to be forwarded to which TDM-PON. Note that each wavelength O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 70 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 can be routed to one or more TDM-PONs, providing a selected and dynamic multicast environment. The upstream data is carried in a separate path. This architecture improves the broadcast nature and security concerns of the broadcast and selects WDM/TDM-PON. However, with slow switches based on micro-electro-mechanical systems (MEMS), the power loss will be even more for this broadcast and select flavour. On the other hand, fast-switching semiconductor optical amplifier (SOA) switches will compensate the high losses due to the couplers and splitters, but at the cost of an expensive solution. However, the amplified spontaneous emission (ASE) noise might still degrade the performance which can be addressed by optical suppression of noise. Central office OLT RN1 Wavelength router RN2 Power splitter User ONU Uplink line card PD Us PD s Nu ONU control Uplink control Uplink line card Tunable Burst Mode Transmitter PD Downlink control Downlink line card PD Nd Downlink Tunable optical filter line card DFB DFB Us Nd Nd Nd DFB Figure 45 : WDM/TDM-PON as in Figure 42 with a wavelength router and power splitter for the remote node. In Figure 46 the remote node design consists of a switching configuration with wavelength selective switches (WSS) in RN1 followed by a passive splitter in RN2. WSS are generally implemented in MEMS that provide low insertion loss wavelength switching capabilities. When a WSS is used as reconfigurable optical demultiplexer, a WSS can steer each optical channel present on its input common port towards one of its output ports, and multiple wavelengths can be switched to one output port of the WSS (which increases flexibility, but again requires a tunable Rx at the ONU). On the other hand, in the reverse direction it can be used as wavelength blocking device, where it can block some of the wavelengths from each of the ports to enter to the common port. However, it can be configured in such a manner that it will allow all the possible wavelengths from each of the output ports to enter into the common port. Due to the use of WSS the power budget in the downstream direction is much better than for a WDM/TDM-PON with a wavelength router using MEMS (Figure 45) or a broadcast and select WDM/TDM-PON (Figure 43). Compared to an AWG used for the WDM/TDM-PON with WDM splitter (Figure 44), a WSS enhances the optical reconfigurability of wavelength distribution among the customers according to their traffic demand, and in this way a partially flexible RN is built. The WDM/TDM-PON using WSS is a long-term solution that needs further research in a laboratory environment before it can be considered for commercial use. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 71 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Central office OLT RN1 Wavelength switch RN2 Power splitter User ONU Uplink line card PD WSS PD nWSS Nu s ONU control Uplink control Uplink line card Tunable Burst Mode Transmitter PD Downlink control Downlink line card NWSS PD Downlink Tunable optical filter line card DFB DFB Nd DFB Figure 46 : WDM/TDM-PON as in Figure 42 with a wavelength selective switch and power splitter for the remote node. 4.5.2 Hybrid WDM/CDM-PON Code division multiple access (CDMA) is one of the alternatives for per-wavelength fan-out in a hybrid PON. There are different splitting approaches at the remote node that can be combined with hybrid WDM/OCDM. The main option (Figure 47) is hybrid WDM/OCDM-PON with a WDM splitter (e.g. arrayed waveguide grating AWG) and several power splitters each of which connects to one output port of the WDM splitter [27]. In this system, the WDM splitter and power splitters can be located separately, e.g. the WDM splitter is at a street cabinet close to the central office while the power splitters are put at distribution (interconnection) points (see Figure 4.4.4) close to the users. OCDM channels can be overlaid on WDM wavelength channels. It means multiple users are individually assigned with different optical codes on each WDM wavelength channel. We assume in this system that there are in total N wavelength channels and each channel can accommodate M code channels. The total number of users that can be supported by this system is N × M. OCDMA over WDM might be viewed in a way that a WDM channel is shared with M users by equally dividing the bandwidth into different code channels. A clear differentiation from a conventional TDMA is that this channel division can be realized in an asynchronous manner without using time slots. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 72 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 47 : Hybrid WDM/OCDM PON [27] (OC: optical code, TX: transmitter and RX: receiver). Generally, two different versions of CDMA must be considered in fibre-optic transmission, Electrical CDMA (E-CDMA) and Optical CDMA (O-CDMA). E-CDMA refers to a system where the code multiplex and the related spectrum spreading (both performed through multiplication with a fast, client-specific chip or code sequence) is done in the electrical domain, prior to modulating the lasers, and after photo-diode detection, respectively. For reasonable code multiplexing, say in the range of 10 multiplexed customers, CDMA typically already requires, to first approximation, a 100-fold bandwidth. This leads to the fact that in ECDMA PON systems, very high analogue bandwidth is required for the transceiver optoelectronics. Here, we considered a bandwidth of 30 GHz which is sufficient to provide ~500 Mb/s per customer non-oversubscribed. The schematic diagram of an E-CDMA hybrid PON is shown in Figure 48. The ONU is based on a tunable (non-reflective) transceiver where the code is applied to the upstream signal, and the complex conjugate of the code sequence is applied to the downstream receive signal. Depending on bandwidth and per-wavelength multiple-access, multi-channel transceivers can be used in the OLT. The multi-channel signal is then constructed in the digital domain. Depending on the resulting bandwidth and the implementation effort required, such an approach can help saving cost (CapEx) over an implementation using discrete OLT transceivers. A multi-channel OLT transceiver design is indicated in Figure 48 on the lefthand side. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 73 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 OLT RN1 AWG Tx/Rx Array RN2 1xN AWG ECDMA MDXM ECDMA MDXM 10G-ECDMA ONUn λD λU 1:k ()* Code RF Modulation ADC DAC Digital SCMA with CDMA S+H Figure 48 : Hybrid WDM/E-CDM-PON The resulting cost and energy consumption figures (again without baseline numbers) are summarized in Table 8. Note that in particular the cost / energy consumption assumption for an OLT per-client port must be considered very optimistic. Table 8. Energy consumption and cost figures for hybrid WDM/E-CDM-PON excluding baseline (chapter 7.1) Cost 237$ total Energy 5.85 W AWG/splitter ports 14$ OLT port incl. digital Tx 0.75 W OLT port 22$ OLT switching 1.0 W OLT amplifiers 6$ OLT amplifiers 0.1 W ONU TRX 175$ ONU incl. CDMA (1.5 W) 4.0 W ASICs + switching 20$ According to the numbers in Table 8, a hybrid WDM/E-CDM-PON still has low-cost potential, although it is more costly than a hybrid WDM/TDM-PON. This must be related to the fact that the required signal processing (code multiplexing) is performed in the electrical domain, with respective cost assumptions for the key components. However, the energy consumption of the E-CDMA hybrid PON as compared to the TDMA hybrid PON is ~2 W higher, on a per-client basis. In a large-scale deployment scenario, this is an unacceptable increase of power consumption, irrespective of the question if the difference has to be paid by the network operator or the end customer. O-CDMA refers to a fibre-optic implementation of code-based multiple access. Application of the different Optical Orthogonal Codes (OOCs) and the related spectrum spreading is performed in the optical domain, using passive components like AWGs and FBGs (Fibre Bragg Gratings). One possible implementation according to recent literature is shown in Figure 49. Encoding and decoding (E/D) in the ONUs is performed by two identical chirped FBGs. Different chirp patterns provide different OOCs. In the OLT, simultaneous multi-port E/D is performed by an N x N AWG. Both components (FBG, AWG) are fully passive and no severe cost drivers. In particular, they enable low-energy design since the electronic domain can be restricted to the range of, say, 1 Gb/s. However, in the design shown in Figure 49, two circulators are required in the ONU. To date, these devices can not be integrated and are costly. In addition, the ONUs require a (balanced) delay demodulator. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 74 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 OLT RN1 AWG Tx/Rx Array RN2 1xN AWG OCDMA MDXM OCDMA MDXM λD λU 10G-OCDMA ONUn 1:k T Rx DPSK NxN AWG Multiport E/D DPSK Encoder / Decoder Figure 49 : Hybrid WDM/O-CDM-PON The resulting numbers for cost and energy consumption are listed in Table 9. Energy consumption is very low compared to the other hybrid PONs, making the WDM/O-CDMPON an attractive contender for NGA considerations. On the other hand, cost is very high (mainly due to the use of circulators). This leads to the requirement to find more cost-efficient O-CDMA implementations in order to allow wide-spread use. Table 9 : Energy consumption and cost figures for hybrid WDM/O-CDM-PON excluding baseline (chapter 7.1) Cost 394$ total Energy 2.75 W AWG/splitter ports 13$ OLT port 0.55 W OLT port incl. E/D (N x N AWG!) 60$ OLT switching 1.0 W OLT amplifiers 16$ OLT amplifiers 0.2 W ONU TRX incl. E/D 300$ (Circulators!) ONU (DPSK) 1.0 W OLT switching 5$ In addition, it must be noted that to date, combined WDM and O-CDMA fan-out has been very restricted, up into the range of a total of 128 customers. This is related to two reasons. First, all hybrid WDM/O-CDM-PON demonstrators so far were based on a small number of wavelengths (typically, less than 10). Then, (as with all CDMA) the number of useful OOCs is limited. Up to 16 per-wavelength customers have been demonstrated, each getting several 100 Mb/s dedicated bandwidth. This translated to optical chip rates in the range of up to 600 GChips/s. Hence, the main technical question is if more wavelengths can be used in an OCDMA scheme. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 75 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 50 : An example of wavelength channel allocation of OCDMA over CWDM using 511-chip SSFBG en/decoder [27]. Super-structured Fibre Brag Grating (SSFBG) is a promising device for a long optical code (OC) because of the relatively low insertion loss. A 511-chip SSFBG enconder/decoder has been demonstrated. The chip length of the grating and the total length are 0.156 and 80 mm, respectively, which corresponds to the chip rate of 640 Gchip/s with the duration of the generated optical code of about 800 ps. The measured peak reflectivity is up to about 92%. Note that this characteristic is insensitive to the polarization state of the input signal, if the fabrication progress has guaranteed good cylindrical symmetry of the SSFBG. The notches of the spectrum spread appear at integer steps of 640 GHz (= 5 nm) from the central peak. If the neighbouring wavelength channel is allocated at the spectrum notch, WDM interchannel crosstalk can be almost neglected. Figure 50 shows an example of wavelength channel allocation where the space between two adjacent wavelength channels is 10 nm (2 x 5 nm) in hybrid WDM/OCDM-PON using the 511-chip SSFBG encoder/decoder (with 640Gchip/s). It should be noted that the shorter the length of the code is, the lower chip rate and the smaller space of neighbouring wavelength channels is required. Figure 51 illustrates two other possible options of the remote node to support capacity upgrade by adding OCDMA channel [28]. The original splitter may be replaced by employing WDM multiplexers (e.g. AWG) and wavelength/waveband selectors (WSs). Here, WS devices mean passive WDM multiplexers with lower wavelength granularity and are typically used to separate data and signalling wavebands (e.g. red/blue filter). To maintain broadcast signals in the WDM multiplexer configurations, the broadcast wavelengths for TDM and/or CDM must be separated from the point-to-point wavelengths for WDM so as to split their power among all users. WS devices separate or merge the broadcast signals. Note that the PON configuration in Figure 51 allows both TDM and WDM ONU to remain untouched and co-exist with the OCDMA upgrade of the individual ONU. Furthermore, this approach suffers from high power loss due to the insertion loss of the power splitter and therefore requires large power budget. This may restrict the number of users and maximum reach between the OLT and the ONU. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 76 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 51 : OCDMA over WDM or hybrid WDM/TDM PONs Another option for the remote node design (Figure 52) is a WDM splitter (e.g. arrayed waveguide grating AWG) combined with several code splitters (e.g. AWG based multiport E/D [21]) each of which connects to one output port of the WDM splitter. The WDM splitter and the code splitters can be located in separated places. For instance, the WDM splitter is at street cabinet close to the central office while the power splitters are close to the users. The signals can be encoded/decoded at the code splitter so that the ONUs do not need any encoder/decoder. In addition, the insertion loss of the code splitter is similar as for the WDM DEMUX which is much smaller than the power splitter. Therefore, this approach can overcome the high power loss of the previous approach (i.e. WDM splitter and power splitter) and may have larger number of users and longer reach between the OLT and the ONU. Figure 52 : Hybrid WDM/OCDM PON with code splitting at the remote node. 4.5.3 Hybrid WDM/OFDM-PON OFDM has recently gained much attention for optical transmission because of its spectral efficiency and its flexibility and robustness against dispersion, especially if implemented in the digital domain. For these reasons, it is now also considered for PONs, both singlewavelength and also hybrid WDM-PONs. Generally, OFDM could be combined with other O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 77 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 multiple-access mechanisms (i.e., TDMA, WDMA, or CDMA). Because OFDM is already based on multiple sub-carriers, SCMA however seems to be the most natural fit. The respective SCMA variant is then called OFDMA, and since it is implemented in fibre-optic systems, it is also referred to as O-OFDMA. Different implementations of O-OFDM have been described in the literature, including coherent intradyne, coherent heterodyne, and incoherent direct-detection variants (the latter with Single-SideBand modulation, SSB). They all have in common that the OFDM signals are constructed in the digital domain, making use of Inverse Fast Fourier Transform (IFFT) in the transmitter and FFT in the receiver. Figure 53 shows an implementation of an incoherent hybrid WDM/O-OFDM-PON. Since OFDMA with potentially very tight sub-carrier spacing is used, the different ONU transmitters of one nominal upstream wavelength must be tightly locked to that wavelength in order to avoid interference between different sub-carriers. Either tunable lasers with very precise tuning / wavelength locking or seeded (reflective) transmitters must be used. Here, an implementation based on seeded SOAs is shown. In order to allow SSB modulation, a biased MZM is used. The seed wavelength is de-coupled in a band splitter and then fed into a non-reflective (standard) SOA which is followed by the MZM. The modulated upstream signal is then coupled into the distribution fibre by means of a circulator. This way, the SSB signal is built in the analogue electrical domain. In the OLT, optical SSB modulation is shown as an alternative. Whether this is more efficient in its implementation remains for further investigation. 1xN AWG AWG cos RF SOA -sin cos SSB Filter MFL 90° + RF Modulation ADC Digital OFDM DAC OFDM MDXM 1:k DC Bias Tx/Rx Array 10G-OFDM ONUn λD λU RN2 RF sin ADC Serial-Parallel Ext. Removal FFT M-ary QAM RN1 OFDM MDXM DAC Parallel-Serial Cyclic Extens. IFFT M-ary QAM OLT Figure 53 : Hybrid WDM/O-OFDM-PON The cost and energy consumption figures are listed in Table 10, according to the assumptions stated in 7.1. As compared to the other hybrid WDM-PONs, cost and energy consumption both increased. In the ONU, this again is related to the cost of the circulator. Hence, other implementation alternatives should be considered. For both ONUs and OLT, the digital signal processing adds to cost and energy consumption significantly. This is due to the fact that per client digital OFDM signal processing at the accumulated per-wavelength bandwidth of 10 GHz must be performed. From today’s perspective, this will be the dominant contributor in particular to energy consumption, together with the contribution coming from the necessary high-performance 10G transceivers. In this respect, the numbers stated in Table 10 can not be regarded as pessimistic. However, they lead to unacceptable energy consumption, and also comparatively high cost. It may however be useful in cases where variable bandwidth or bandwidth × reach requirements must be supported. The latter can easily be achieved in digital OFDMA by adaptation of either the per-client number of sub-carriers, or the per-subcarrier modulation depth (e.g., QPSK vs. 16QAM). Table 10 : O O A S OA AS SEEE FP7 – ICT– GA 249025 Energy consumption and cost figures for hybrid WDM/O-OFDM-PON excluding baseline (chapter 7.1) Page: 78 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Cost 362$ total Energy 7.8 W AWG/splitter ports 12$ OLT port incl. digital Tx 0.75 W OLT port 22$ OLT switching 1.0 W OLT amplifiers 3$ OLT amplifiers 0.05 W ONU TRX 275$ (Circulator!) ONU 6.0 W ASICs + switching 55$ 4.5.4 Hybrid WDM/SCM-PON SCMA is similar to OFDMA in that it assigns different sub-carriers to different clients. In the OLT, a similar – digital – implementation can be chosen. The main difference is that the ONUs can be based on narrow-band electrical (RF) transceivers. Only the optical transceivers (lasers, photo diodes) need to support the accumulated bandwidth (here again 10 GHz). This has significant impact on potential cost and energy consumption. On the other hand, flexibility with regard to per-client bandwidth is lower as compared to OFDM. Different SCMA variants have been proposed, again including coherent and incoherent versions. Figure 54 shows an incoherent hybrid WDM/SCM-PON where the sub-carrier modulation is done in the digital domain (similar to OFDM). In the OLT, all sub-carrier channels of one wavelength can be modulated and de-modulated within a single transceiver. In the ONUs, only narrow-band sub-carrier processing is required because each ONU-specific sub-carrier is multiplexed / de-multiplexed to / from the 10-GHz signal by means of an analogue RF section. Similar to the WDM/O-OFDM-PON shown in Figure 53, the hybrid WDM/SCM-PON shown in Figure 54 is based on seeded ONU transceivers (in this case REAM-SOAs). This automatically allows tight wavelength locking. As compared to OFDM, this leads to much simpler design. Like most other hybrid high-performance PONs, WDM/SCM also requires high-performance 10G transceivers. 1xN AWG AWG Tx/Rx Array SCMA MDXM RN2 λD λU 10G-SCMA ONUn 1:k RF RF Modulation ADC Digital SCMA DAC REAM SOA RF MFL ADC Modulation RN1 SCMA MDXM DAC Modulation OLT Figure 54 : Hybrid WDM/SCM-PON Cost and energy consumption numbers are stated in Table 11. It can be seen that cost and energy consumption are both significantly lower than the respective O-OFDM numbers. Cost so far is only undercut by WDM/TDM, but energy consumption is still comparatively high. Hence, means for reducing energy consumption on WDM/SCM-PONs should be further investigated. Table 11 : O O A S OA AS SEEE FP7 – ICT– GA 249025 Energy consumption and cost figures for hybrid WDM/SCMA-PON excluding baseline (chapter 7.1) Page: 79 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Cost 235$ total Energy 4.8 W AWG/splitter ports 12$ OLT port incl. Digital Tx 0.75 W OLT port 20$ OLT switching 1.0 W OLT amplifiers 3$ OLT amplifiers 0.05 W ONU TRX 175$ ONU 3.0 W ASICs + switching 25$ WDM/O-OFDM and WDM/SCM both provide dedicated per-client sub-carriers over an ODN which consists of filters (AWGs) and power splitters/combiners. Similar multiple access is provided by coherent ultra-dense WDM-PONs (UDWDM-PONs), with the exception that carriers are assigned to customers, not sub-carriers. From that, a comparison of hybrid PONs with UDWDM-PON makes sense although UDWDM-PON typically is not regarded as hybrid PON. 4.5.5 UDWDM UDWDM requires coherent receivers in order to allow wavelength spacing as close as 2 or 3 GHz. Ideally, the local oscillator in the ONU is also re-used for upstream. For SFW, this requires heterodyne detection, where the RF frequency is also the frequency shift between downstream and upstream. One possible implementation is shown in Figure 55. Coherent detection requires alignment of the polarization planes of receive signal and local oscillator. Typically, polarization diversity is implemented in coherent receivers, which leads to dual balanced receivers and the related cost. (In homodyne detection, it leads to 4 balanced receivers due to further split into in-phase and Quadrature components.) Balanced receivers can be reduced to single-photo-diode receivers at the cost of 3 dB penalty regarding sensitivity. This penalty may be acceptable for ODN with filters and splitters. It will not be acceptable for splitter-only ODN because of the very high accumulated insertion loss. RN1 AWG Tx/Rx Array 1xN AWG UDWDM MDXM UDWDM MDXM DS – + RF US Pol. Scr. CLK Rec. 3dB 3dB 3dB RF Modulation DAC ADC 1G-UDWDM ONUn 1:k 3dB Digital SCMA λD λU RN2 3dB OLT Pol. Scr. / Diversity T-LD Figure 55 : UDWDM-PON Figure 55 shows an alternative implementation with regard to polarization. It is based on a polarization scrambler which is clocked at twice the symbol rate. Such, both orthogonal polarizations of the local laser are superimposed with the receive signal within each symbol duration. This scheme only requires one balanced detector for heterodyne receivers, however it leads to 3 dB penalty (because only one half of the local laser signal is aligned with the receive signal in polarization). The OLT UDWDM transceiver is based on a broadband, multi-carrier implementation. Several tightly spaced carriers are multiplexed in the digital domain and then modulated onto broadband optics. This way, the number of transceivers can be reduced, at the cost of broader O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 80 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 bandwidth required. Note that this approach is similar to the one chosen for WDM/SCMA. Table 12 lists the cost and energy consumption numbers for UDWDM. Cost is relatively high which relates to the effort which is required for coherent detection and (OLT) wavelength stabilization. Energy consumption is comparable to WDM/SCM-PON. No broadband (10 GHz) components are required in the ONUs, but higher effort is required for balanced receivers, and OLT transmitter stabilization. Table 12 : Energy consumption and cost figures for WDM/UDWDM-PON excluding baseline (chapter 7.1) Cost 312$ total Energy 4.5 W AWG/splitter ports 12$ OLT port incl. digital Tx 1.5 W OLT port 100$ OLT switching 1.0 W ONU TRX 175$ ONU 2.0 W OLT switching + ASICs 25$ Comparing WDM/SCM-PON with UDWDM leads to interesting results. From the spectral viewpoint, both look very similar, especially if UDWDM wavelengths are grouped in order to be able to pass through WDM filters. Both are similar with regard to energy consumption. Both can make use of (partly-) integrated, broadband digital OLT transceivers. Both can support a very high customer number. A hybrid DWDM/SCM-PON with 96 wavelengths of 10 GHz each and 1:16 per-wavelength split could scale to 1536 customers, with per-customer bit rates of up to 1 Gb/s which then depends on the sub-carrier modulation depth. UDWDMPONs with 3-GHz spacing can scale into the same region (if no guard bands for WDM filters have to be used). The main differences relate to cost and reach performance. From today’s perspective, UDWDM requires higher implementation effort and is more costly. On the other hand, it offers better sensitivity and hence better reach performance without amplifiers. Over short distances, it can also support splitter-only ODNs. Since both approaches have similar reach/cost performance, the most important question relates to the reach requirement. If very high reach really is required, UDWDM may be the preferable choice. If reach in the 50…60 km range is regarded sufficient, WDM/SCM-PON is the more cost-efficient choice. 4.6 HYBRID ACTIVE WDM Since OASE aims at access solutions enabling high reach, high fan-out and high (total or perclient) bandwidth, we must also consider passive/active hybrid solutions. 4.6.1 Hybrid active WDM with active P2P access TXFP PoP PRN TXFP ... TXFP SFP Scalable Universal Switch AWG TXFP L2 AWG ... SFP Figure 56 shows a combined access/backhaul hybrid active/passive WDM-PON. It is based on 10G DWDM for backhaul and broadband users, and active P2P fan-out for residential access. WDM-PON interfaces are based on Tunable XFPs (TXFP). ONU SFP ARN / LX Figure 56 : Hybrid active WDM-PON with active P2P access O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 81 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 It is obvious that an approach according to Figure 56 can support high reach (easily up to 100 km) and very high fan-out, at the cost of active RNs. However, it can be shown that this approach has very low cost (CapEx) as well as low energy consumption. This can be seen from the respective numbers in Table 13. Table 13. Energy consumption and cost figures for hybrid active WDM-PON with active P2P access excluding baseline (chapter 7.1) Cost 147$ total Energy 3.3 W AWG ports 2$ PoP switching 1.0 W 10G TRX 100$ (!) LX switching 1.0 W Switch (PoP) 5$ LX TRX (2 x TXFP) 0.3 W Switch (LX) 10$ CPE TRX (2 x grey SFP) 1.0 W CPE TRX (2 x grey SFP) 30$ Both very low cost and low energy consumption can be related to the possible use of cheapest, simplest 1G transceivers for the active fan-out – grey 10-dB SFPs. These transceivers easily undercut every other transceiver type and lead to lowest cost, even though integrated backhaul with TXFPs has been considered. In addition, low-energy SFPs can go down to ~0.4 W which again makes them the most efficient solution for dedicated 1G services. Also note that energy consumption is comparatively low because for all other solutions, a first aggregation switch is also included (mostly, in the OLT). 4.6.2 Hybrid active WDM-PON Scalable Universal Switch AWG AWG ... TXFP TXFP TXFP ARN / LX PoP ONU TSFF AWG OLT TXFP WDM PIC A derivative of the active/passive hybrid access solution mentioned before is the PON-inPON concept, see Figure 57. It requires the OLT of an access WDM-PON to be accommodated in an active RN. These OLTs are then backhauled by a WDM-PON running 10G wavelengths. This solution obviously requires other distribution fibres in the ODN, as compared to the approach with active P2P access. It has very high reach and capacity performance as well. PRN2 PRN1 Figure 57 : Hybrid active WDM-PON (“PON-in-PON”) Cost and energy consumption numbers of the PON-in-PON approach are listed in Table 14. There is already a significant difference against the approach with active P2P access. Higher cost and also energy consumption of PON-in-PON can be attributed to the higher performance of the access WDM-PON, as compared to grey SFPs. The access WDM-PON already has 25-dB transceivers, which in addition are DWDM-tunable. Table 14 : O O A S OA AS SEEE FP7 – ICT– GA 249025 Energy consumption and cost figures for hybrid active WDM-PON excluding baseline (chapter 7.1) Page: 82 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Cost 262$ total Energy 3.6 W AWG ports 12$ PoP switching 1.0 W 10G TRX 120$ LX TRX (2 x TXFP) 0.35 W Switch (PoP) 5$ OLT Array port 0.5 W WDM-PON OLT port (LX) 50$ ONU TRX 1.0 W CPE TRX (2 x grey SFP) 75$ OLT Switching 0.75 W As compared to all other (hybrid, UDWDM) access solutions discussed in this chapter, the PON-in-PON approach performs very well. Only the hybrid WDM/O-CDM-PON has lower energy consumption. However, this approach so far only supports up to ~128 customers. Only WDM/TDM-PON and WDM/E-CDM have lower-cost potential. E-CDM leads to prohibitive energy consumption, and fully passive WDM/TDM has much lower reach performance. Hence, the hybrid active/passive PON-in-PON approach must be considered one of the more powerful alternatives for NGA with high reach / fan—out / bandwidth requirements. 4.6.3 Comparison of active/passive hybrids and WDM-PON From the discussion given before it becomes clear that active P2P access has advantages with respect to cost and energy consumption if it can be based on cheapest, lowest-power SFPs. However, it requires dedicated distribution fibres and has relatively poor foot print in the OLT. Both can translate to added cost, or may prohibit this approach. It also became clear that any added per-wavelength fan-out adds to cost and energy consumption, in particular if dedicated bandwidths in the range of 500 Mb/s to 1 Gb/s are required. It should also be noted that the cost and energy consumption numbers for the hybrid WDM/TDM-PON to first approximation also hold for XG-PON (in particular XG-PON2 which is the 10G symmetric variant). Hence, it is necessary to compare these numbers also to a simple, stringent DWDM-PON which may come out as the ultimate reference and benchmark with regard to systems CapEx and perclient energy efficiency. Figure 58 shows the schematic block diagram of a DWDM-PON which is based on OLT PICs and low-cost tunables. This approach has been discussed in Chapter 4.2. It must be noted that there are clear design goals for the tunables in the sub-100$ and sub-1-W region. Current status of the related components research indicates that these numbers are feasible (see the tunable lasers chapter). RN L-Band 100 GHz Tx Array 1 ... Mux PIC N Cyclic AWG 1 ... C C-Band ~100GHz N Rx Array L Cyclic AWG OLT ONU T-LD SFF Rx Identical Cyclic AWGs Figure 58 : Reference WDM-PON Table 15 states the relevant cost / energy consumption performance numbers of the WDMPON shown in Figure 58. From these performance figures, it can be seen that a simple WDMPON is superior over all other access concepts with regard to system CapEx and per-client energy consumption if dedicated broadband services (500 Mb/s and higher) are required. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 83 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 15 : Energy consumption and cost figures for reference WDM-PON excluding baseline (chapter 7.1) Cost (per client) 150$ total Energy (per client) 2.5 W AWG ports 20$ OLT array port 0.5 W OLT array port 50$ OLT switching 1.0 W ONU TRX 75$ ONU 1.0 W OLT switching 5$ From all other approaches analyzed in sections 4.5 and 4.6, only the hybrid active/passive WDM-PON with active P2P access has similar cost (it is actually marginally lower). However, no other solution has similarly low energy consumption, in particular for dedicated 1-Gb/s services. It must be noted that there is a single enabler for the cost / energy consumption performance of a stringent WDM-PON: the possibility to use lowest-cost, 1-Gb/s, small-power-budget transceivers. Here, small power budget (at 1 Gb/s!) refers to the 25-dB class. A simple WDMPON without any further per-wavelength fan-out is the only approach where such transceivers can be used. This does include the UDWDM-PON approach. It also runs at 1 Gb/s per wavelength, however it requires highly stabilized laser sources and coherent receivers. All other approaches require added effort – on the per-client level! – which also adds to complexity, cost, and energy consumption. In most cases, high-performance (32…35 dB power budget) 10G transceivers are required, with additional means for per-wavelength fanout. It is doubtful that these 10G transceivers will ever have the cost and energy consumption potential of simpler 1G transceivers. It should once again be noted that equally aggressive cost and energy consumption figures have been assumed for the different transceivers in 7.1. From the discussion in this chapter it becomes clear that a stringent WDM-PON approach has good potential to outperform any other access approach with regard to per-client CapEx and energy consumption. This holds in particular when being compared against single-wavelength (40G, 100G) next-generation PON approaches. It is extremely doubtful that such approaches can ever develop low-cost or low-energy potential. In addition to that, a WDM-PON has good reach performance. The lowest-cost version analyzed herein so far – 25-dB class – can easily span some 50 km in urban areas. On good (low insertion loss) fibres, it already supports some 60+ km. It is likely that this reach covers the majority of applications and hence avoids an over-engineered solution. Nonetheless, a WDM-PON can also easily be extended in its maximum reach be means of OLT-based amplifiers. It can then approach the 80…100 km reach domain at reasonable added cost and energy consumption (caused by the amplifiers). There is one obvious disadvantage of the WDM-PON: its maximum client count. A simple WDM-PON is likely to be restricted to 80, possibly 96, bi-directional channels, using 50-GHz grid and cyclic C/L-band AWGs. This is an order of magnitude away from OASE’s target fanout number. It may be possible to extend the WDM-PON – refer to Chapter 4.2 – to both, the 25-GHz grid and also into the S-band. This may enable a theoretical maximum of up to 384 channels. The respective system concept will still be relatively simple, low in cost, and low in energy consumption. However, it will still not have the potential for 1000 customers. Hence, in order to avoid over-engineered – too costly and energy-consuming – solutions it must be clarified very carefully where the optimum per-feeder-fibre client number is. This has to include the consideration that beyond a certain per-feeder-fibre client count, at least the feeder fibre itself (and possibly also part of the OLT equipment) will require resilience mechanisms in order to avoid excessively high failure penetration ranges in case of failures where key components (single points of failure) are affected. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 84 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 4.7 NEXT GENERATION AON: RPR, WDM, ETHERNET Point-to-point (PtP) optical links provide the basis for active optical networks (AON) that can support a variety of tree-, ring- or mesh-based FTTx topologies. This section discusses PtP technologies relevant for next-generation access. Any of the defined IEEE 802.3 physical layers are relevant for AONs. Table 16 shows a rough outline of some of the relevant PHY standards that are either finalized or in development. Table 16 : Selection of Ethernet PHY standards Standard Fibre Range Bitrate 100BASE-FX MMF/SMF 400m-10 km 100 Mb/s 100BASE-BX SMF 40km 100 Mb/s 1000BASE-SX MMF 550m 1 Gb/s 1000BASE-LX10 SMF 10 km 1 Gb/s 1000BASE-ZXX SMF 70 km 1 Gb/s 10GBASE-LR SMF 10 km 10 Gb/s 10GBASE-ER SMF 40 km 10 Gb/s 40GBASE-LR4Y SMF 10 km 40 Gb/s 100GBASE-LR4Y SMF 10 km 100 Gb/s 100GBASE-ER4Y SMF 40 km 100 Gb/s X : de facto standard, Y: standard not yet finalized 4.7.1 GbE access Gigabit Ethernet (GbE) is defined in the IEEE 802.3-2008 standard and refers to various technologies for transmitting Ethernet frames at a rate of gigabit per second. The specification allows for half-duplex gigabit links connected through hubs, but full-duplex with switches are more common on the market. For single-mode fibre, three main physical layer variants are defined. The first is 1000BASE-LX, with 1310 nm for both downstream and upstream transmission, specified to work over a distance of ≤5 km over 10 µm single-mode fibre. An extended version – 1000BASE-LX10, with a reach of 10 km – was standardized six years after the initial gigabit fibre versions as part of the Ethernet in the First Mile task group. The extended version is similar to 1000BASE-LX, but with increased reach due to higher quality optics. In practice 1000BASE-LX10 was already in widespread use before standardization as vendor proprietary extensions referred to as 1000BASE-LX/LH or 1000BASE-LH. The second variant is the duplex solution 1000BASE-BX10, with 1490 nm for downstream and 1550 nm for upstream transmission, specified to work over a distance of ≤10 km over a single strand of single-mode fibre. Because of the different wavelengths in the downstream/upstream there is an asymmetry between the terminals on each side of the fibre. A third variant is 1000BASE-ZX, with 1550 nm for both downstream and upstream transmission, specified to work at a maximum distance of 70 km to 120 km. 1000BASE-ZX is not an IEEE standard, but rather an industry-accepted term to refer to gigabit Ethernet transmission using the 1550 nm wavelength to achieve distances of at least 70 km over single-mode fibre. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 85 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Transceivers for GbE are mature and available on the market (e.g. Oesolution, Delta, Lasermate, etc.). Transceivers with 1310 nm for both transmission and reception are based on FPs or DFBs with a range of up to 40 km. Transceivers with 1550 nm for both transmission and reception can have a reach of up to 120 km. Transceivers using different wavelengths for transmission and reception, e.g. transmission at 1310nm and reception at 1550nm or vice versa, have a range up to 60 km. All these transceivers are typically packaged in SFP packages. The power consumption of a GbE transceiver for a 10 or 20 km link is within 1W. Cost varies widely, indicatively an SFP will cost from 10 USD to 100 USD (for 100 units purchases) depending on distance and quality. 4.7.2 10 Gb/s Ethernet access and distribution/aggregation The 10 Gigabit Ethernet (10GbE) standard defines Ethernet with a nominal data rate of 10 Gb/s. The standard was first published in 2002 as IEEE Std 802.3ae-2002 and over the years two variants for 10GbE over fibre have been defined: 802.3ae-2002 (fibre -SR, -LR, -ER and -LX4 PMDs) and 802.3aq-2006 (fibre -LRM PMD with enhanced equalization). The 802.3ae2002 amendment was consolidated into the IEEE 802.3-2005 standard. IEEE 802.3-2005 and the other amendments have been consolidated into IEEE 802.3-2008. 10 Gigabit Ethernet is based on full duplex links which can be connected by switches. Half duplex operation and CSMA/CD (carrier sense multiple access with collision detection) are not supported in 10GbE. The 10 Gigabit Ethernet standard encompasses a number of different physical layer (PHY) standards. At the time the 10 Gigabit Ethernet standard was developed there was much interest in 10GbE as a WAN transport which led to the introduction of the concept of the WAN PHY for 10GbE. This operates at a slightly slower data-rate than the LAN PHY and adds some extra encapsulation. The WAN PHY and LAN PHY are specified to share the same PMDs (physical medium dependent) which means that 10GBASE-LR and 10GBASE-LW can use the same optics. In terms of number of ports shipped, the LAN PHY greatly outsells the WAN PHY. The different 10GbE PHY standards are summarized in the following. 10GBASE-LR ("long range") uses the IEEE 802.3 Clause 49 64B/66B Physical Coding Sublayer (PCS) and 1310 nm lasers. It delivers serialized data over single-mode fibre at a line rate of 10.3125 Gb/s. 10GBASE-LR has a specified reach of 10 kilometres (6.2 mi), but 10GBASE-LR optical modules can often manage distances of up to 25 kilometres (16 mi) with no data loss. Fabry–Pérot (FP) lasers are commonly used in 10GBASE-LR optical modules. Fabry–Pérot lasers are more expensive than VCSELs but their high power and focused beam allow for efficient coupling into the small core of the single mode fibre. 10GBASE-ER ("extended range") uses the IEEE 802.3 Clause 49 64B/66B Physical Coding Sublayer (PCS) and 1550 nm lasers. It delivers serialized data over single-mode fibre at a line rate of 10.3125 Gb/s. 10GBASE-ER has a reach of 40 kilometres. Several manufacturers have introduced 80 km range ER pluggable interfaces under the name 10GBASE-ZR. This 80 km PHY is not specified within the IEEE 802.3ae standard and manufacturers have created their own specifications based upon the 80 km PHY described in the OC-192/STM-64 SDH/SONET specifications. The 802.3 standard will not be amended to cover the ZR PHY. Optical modules are not specified in 802.3 but by multi-source agreements (MSAs). The relevant MSAs for 10GbE are XENPAK, X2, XPAK, XFP and SFP+. XENPAK was the first MSA for 10GE and has the largest form factor. X2 and XPAK which were introduced later have smaller form factors, but have not been as successful in the market as XENPAK. XFP which was introduced after X2 and XPAK provides further reduction of the form factor. The newest module standard, SFP+, developed by the ANSI T11 fibre channel group is O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 86 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 smaller than XFP and allows for reduced power dissipation. SFP+ is now the most popular socket for 10GbE systems. SFP+ modules only do optical to electrical conversion, without clock and data recovery, putting higher burden on the host channel equalization. SFP+ modules share a common physical form factor with legacy SFP modules, allowing higher port density than XFP and re-use of existing designs for 24 or 48 ports in a 19" rack width blade. Optical modules are connected to a host by either a XAUI, XFI or SFI interface. XFP modules use a XFI interface and SFP+ modules use a SFI interface. XFI and SFI use a single lane data channel with the encoding specified in IEEE 802.3 Clause 49. SFP+ is an upgrade of SFP for 10 Gb/s transmission. SFP transceivers are available with a variety of different transmitter and receiver types, allowing users to select the appropriate transceiver for each link to provide the required optical reach over the available optical fibre type. Optical SFP modules are commonly available in several different categories: 850 nm 550m MMF (SX), 1310 nm 10 km SMF (LX), 1550 nm [40 km (XD), 80 km (ZX), 120 km (EX or EZX)], and DWDM. SFP dimensions are: 8.5 x 13.4 x 56.5 mm. The XFP (10 Gigabit Small Form Factor Pluggable) is a hot-swappable, protocol-independent optical transceiver, typically operating at 850nm, 1310nm or 1550nm, for 10 Gigabit per second SONET/SDH, Fibre Channel, gigabit Ethernet, 10 gigabit Ethernet and other applications, including DWDM links. It includes digital diagnostics similar to SFF-8472 but more extensive, and provides a robust management tool. The XFI electrical interface specification is a portion of the XFP Multi Source Agreement specification. XFP was developed by the XFP Multi Source Agreement Group. The physical dimensions of the XFP are slightly larger than the standard SFP. One of the reasons for the increase in size is to allow for on board heat sinks for greater cooling. XFP dimensions are: 8.5 x 18.3 x 78 mm. Transceivers for 10 GbE are mature and available on the market (Brocade, Oesolution, JDSU, Delta, Lasermate, etc.). They use the same wavelength for transmission and reception, and as for GbE transceivers, the wavelength used depends upon the target reach: for up to 10km (standard 10GBASE-LR/LW), 1310nm is used, while transceivers for longer haul (standard 10GBASE-EW/ER for 40km and 10GBASE-ZW/ZR for 80km) use 1550nm. Common form factors are SFP+ and XFP. The power consumption lies around 2.5W and 3.5W for link lengths of 10km and 40km, respectively. Cost varies widely, indicatively an SFP will cost from 100 USD to 1000 USD (for 100 units purchases) on distance and quality. 4.8 RADIO OVER FIBRE BACKHAULING Radio over Fibre (RoF) has up to now been conventionally considered as an analogue technology, where an optical wavelength which has been modulated by an RF signal is transmitted along a fibre. Considering the scope of OASE we do not consider Fibre-to-theAntenna technologies, e.g. spanning the last few metres between a Mobile Access Node (NodeB, Base Station, etc.) and an antenna by fibre instead of coaxial cable. Instead, we consider RoF technologies, where analogue RF signals are transmitted from an optical source located within a CO or other appropriate active site via SMF towards a suitable environmentally resilient optical-electrical (OE) converter, filter or amplifier unit in the field within a typical radius of 15 km. In addition to RoF technologies, we also consider digital interfaces (CPRI, OBSAI) with similar advantages as the RoF technologies but where IQsamples are carried digitally over the system. We also discuss the more general topic of integration of wireless technologies in next-generation optical access. The development of RoF systems has generally been motivated by the desire to replace a central high power antenna with a low power distributed antennas system (DAS). RoF O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 87 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 systems therefore frequently consist of multiple base stations (BS’s) which are connected to a single central office (CO). A key aspect to RoF development has been the effort to reduce BS cost and migrate complexity up the network towards the CO [32]. More centralized radio access architectures are being considered as possible candidates for future radio access networks (RAN) with concepts such as main-remote and cloud-RAN (CRAN). Typically this involves partitioning the traditional monolithic base station to a central base band unit serving several remote radio units over for example RoF or CPRI. For CPRI a maximum reach of at least 10 km is required. However, maximum reach up to 100 km have been discussed which would enable even larger degrees of node consolidation in the RAN and potential cost and energy savings. In the access context, there are various useful inherent advantages to RoF analogue transmission and the mentioned digital interfaces. Using analogue transmission avoids the requirement for complex hardware (e.g. CPU, FPGA, DACs etc.) in the field, and therefore offers simplified and cheaper operations and maintenance costs. It also offers a simplified modular concept for hardware in the field which reduces repair times. Fewer spare parts are required, of which many are universal for a lot of applications especially when it comes to convergent access networks, so reducing inventory and storage sizes, simplifying logistics, whilst also optimizing the recycling management of used parts. Analogue transmission also offers low group-delay, in contrast to the processing delay associated with the CPU’s of distributed Mobile Access Nodes, allows effective MIMO implementation i.e. with up to 50 antennas in an array on a single mobile access node located, for example in the CO. A centralized Mobile Access Node allows the computation of ideal values for all the connected antennas situated in the field, so as to achieve maximum space diversity and receiving power aggregation. In this way an optimum carrier-to-interference ratio can be reached on the fly without the need for comprehensive signalling messages between several Mobile Access Nodes placed close to each antenna site. In contrast to femtocells inside the home, a Microcellular Antenna system close to customers can aggregate many more subscribers (due to the higher coverage). This means that fewer Network Elements are required and/or need to be managed. Costs for many small antenna sites (i.e. telephone poles) could be cheaper than one big antenna site with many antennas on a roof-top (e.g. due to rental fees). Also there are potential aesthetic advantages, since such a small antenna can possibly be better integrated within a facade. Remote fixing of complex problems or upgrading the software on-the-fly via a second Ethernet port is possible from “on-site” at the CO. This means that O&M manpower doesn’t need to go into the field anymore. The outdoor unit can be mounted on many small places as it can be small in size and cheaper (fewer parts inside) than classical Mobile Access Nodes with their comprehensive features and functionalities. As the number of outdoor units builds up in many places, the microcells coverage area can be made small. With the coverage area being small, the distance to end-user equipment (i.e. mobile phone or device) is smaller so that less power needs to be transmitted (e.g. due to field-path attenuation affects). Another advantage is the flexibly installation to almost any place, so as to avoid natural obstacles such as trees, buildings etc., which cause high attenuations. In that way there is less interference in the field, so allowing the possibility of a higher effective throughput. An Optical Line Termination, Mobile Access Node and Cable Mode Termination System virtualisation on a distributed computing cloud (Blade Servers in the CO) allows the shutting down of parts of the computing cloud (CPU sleepmode) during times of low traffic intensity and therefore contributes to saving electrical energy. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 88 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 4.8.1 Radio-over-Fibre (RoF) Technologies RoF is characterized by up-conversion from the RF to the optical domains, and can be realized either by direct laser modulation or external modulation methods [33]. Direct methods have the advantages of simplicity and low cost. However, limitations to this approach include the relatively limited bandwidth (10 GHz) of low-cost lasers, e.g. VCSELs, as well their high chirp, non-linear and inter-modal distortion, and an SNR limited by relative intensity noise (RIN). In contrast, common external modulation methods include: - Mach-Zehnder (MZ) interferometer based approaches, but which are still characterized by limited bandwidths (2-3 GHz), high linearity, low chirp, and high bias voltages. However, travelling wave configurations of the MZI-based modulator allow the bandwidth limitations to be overcome. - Electroabsorption modulator (EAM) approaches [33], which are characterized by higher operating bandwidths. EAMs based on the quantum confined Stark effect (QCSE) in quantum wells can also exhibit advanced performance. Down-conversion from the optical down to the RF domains can be implemented by PIN and avalanche photodiode (APD) photodetectors. These are characterised by their relative simplicity, and typically offer bit-rates of few 10's of Gb/s (e.g. using a surface illuminated PIN) [34]. Travelling wave (TW) PIN based receivers employing QW technology can also be used to generate microwave power, with the 3dB frequency response of such a TW photodetector being approximately 100 GHz. An electroabsorption transceiver (EAT) can act as a receiver for the downlink direction, as well as a modulator for the uplink. This approach tends to be most appropriate for mobile and metro systems, where the cost of the BS needs to be significantly reduced with the complexity migrated up towards the CO. In this case, an optical amplifier such as an EDFA is often required so as to compensate the link losses. EATs operate mainly in the 1500 nm wavelength window, making them compatible only with SMF, Fabry-Perot and/or DFB lasers. However, EAT technology with its multi-quantum well (MQW) III–V semiconductor active waveguide exhibits bandwidths of up to 60 GHz, and thus is a highly promising candidate for BS’s in future broadband wireless access systems [34],[35]. In general, analogue transmission is applicable for the following signal types, including amongst others: - DVB-C (Digital Video Broadcast for Cable TV, also DVB-C2) - EuroDOCSIS3.0 (incl. older and future variants and other coax cable RF modulated signals) - GSM, GPRS, EDGE, UMTS, HSPA, LTE and LTE enhanced signals - Satellite TV signals (analogue and QAM modulated – Local Satellite Distribution Service) - Analogue TV signals (i.e. PAL or SECAM QAM modulation) Using a dedicated spectrum (Lambda) for RoF analogue transmissions provides more capacity for IP-based (e.g. GPON, ATM, Gigabit Ethernet, etc…) access technologies (e.g. out-of-band transmission, optionally combined on a single fibre) in a highly scalable way. For example, whereas many subscribers may be satisfied with a low bandwidth capacity (e.g. in the Mb/s region as provided by EuroDOCSIS) other customers in the same area may prefer high bandwidth services (e.g. 3D TV) and can get their service via FTTH. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 89 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 4.8.2 CPRI The Common Public Radio Interface (CPRI) standard [36] defines the interface between Radio Equipment Controllers (REC) and Radio Equipment (RE) and covers both the physical and datalink layers. It specifies the necessary items for transport, connectivity and control, including user plane data, control plane transport mechanism and means for synchronization. The specification complies with 3GPP UTRA FDD (release 8), WiMax (release 1) and 3GPP E-UTRA (release 8). The physical layer must fulfill the requirements relating to clock stability and noise, as well as the requirements to satisfy a BER < 10-12. Several physical layer line rate options are defined. These are defined as multiples of 614.4 Mb/s (1x, 2x, 4x, 5x, 8x, 10x). The line rates are chosen such that the UMTS chip rate (3.84 Mb/s) can be easily recovered. The physical layer supports both electrical and optical interfaces. Regarding the optical interfaces, for each line rate a short reach and a long reach option has been defined. The CPRI standard recommends reuse of optical transceivers from the following standards: - Gbit Ethernet: Standard IEEE 802.3-2005 [37] clause 38 (1000BASE-SX/LX) 10 Gbit Ethernet: Standard IEEE 802.3-2005 [37] clause 53 (10GBASE-LX4) Fibre channel (FC-PI) – Standard ISO/IEC 14165-115 [38] Fibre channel (FC-PI-4) – INCITS (ANSI) Revision 8, T11/08-138v1 [39] Infiniband Volume 2 Rel 1.1 (November 2002) [40] Serial transmission is based on 8B/10B line coding according to IEEE 802.3-2005 [36], clause 36. The length of a basic frame is 1 TC = 1/fc = 1/3.84 MHz = 260.416667 ns. Each frame consists of 16 words and the length of each word (8 bits-80 bits) depends on the line rate. 4.8.3 OBSAI The Open Base Station Architecture Initiative (OBSAI) defines an open standardized internal modular structure for the Base Transceiver Station (BTS). Four main blocks are defined: radio frequency block, baseband, control and clocking block, and transport block. Relevant for OASE is the definition of the external network interface of the transport block of the BTS. At this interface according to OBSAI [41] user data and control data should be transported over a suitable communications path such as T1, E1, DS3, OC1, OC3, Ethernet, or DSL. For 3GPP systems, the logical interface to the Radio Network Controller (RNC) is designated by the Iub, described in the 3GPP Series 25.4xx specifications. For GSM/EDGE specification of the transport interface is given in the Series 3GPP/08.xxx. For 3GPP2 systems, the logical interface to the Base Station Controller (BSC) is specified by the Abis interface in the TIA/EIA IS-828-A specifications. For 802.16/WiMax systems, the logical interface is specified as R6 to the ASN GW (centralized GW) or R3 to the CSN (distributed GW). The transport block performs interworking functions between the external network interface and the interfaces as defined by the OBSAI Reference Point 1 Specification and the OBSAI RP2 Specification. Regarding upper protocol layers (Radio Network Layer, RNL), these may be RAN/BSS vendor proprietary or defined by standardization bodies dedicated to the specification of RAN/BSS network interface functions. OBSAI states that the transport block shall only support Transport Network Layer (TNL) functions and not any RNL related tasks; it does not terminate any RNL related protocols. Hence, all RNL traffic (User Plane, Control Plane and Management Plane) is passed transparently through the transport block, whilst still handled O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 90 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 according to associated QoS attributes. Differing RAN/BSS topologies (e.g. tree, star, and ring) are also supported. 4.8.4 Hybrid Optical-Wireless PON Architectures In general, discussion of integration of wireless technologies in next-generation optical access networking has been a topic of particular research interest, e.g. [42]. In particular, a key aspect of research is the exploitation of WDM-PON topologies to act as backhaul infrastructures, e.g. for IEEE 802.16 [43] and IEEE 802.11 [44] networks, where wireless links can also be considered to form the basis for a robust and resilient multihop wireless networks (mesh networks). Considering wireless broadband access in particular, favoured technology options up to now have included the UWB, IEEE 802.16 and IEEE 802.11 technologies, whilst with regard to the two IEEE standards, the IEEE 802.16m and IEEE 802.11n [45] amendments have been created to help maximise available end-user data rates to above 100 Mb/s and beyond. Fixed and/or mobile wireless access is also a key research theme, with significant recent effort invested to create solution platforms, e.g. WiFi, (802.11x), and WiMax for both fixed access (802.16-2004 or 802.16d) and mobile provision (802.16e-2005.) Ultra wide band (UWB) radio with the ECMA-368, ECMA-369, as well as ISO/IEC 26907 standards has also opened up important new opportunities, with a particular impact on the overall wireless perspective of future 4G systems also expected. Associated with these standards is the frequency allocation 3.1 GHz to 10.6 GHz, whilst there is also the generic “60 GHz” option. For this latter frequency band, the recent release of the 57–64 GHz band in IEEE802.11ad for commercial applications such as wireless HDTV connectivity is an important development. IEEE802.11n has been created to offer up to 300Mb/s at 40 MHz (with short guard interval), with WiMax offering a somewhat lower bandwidth of up to 72 Mb/s but potentially over distances of up to 10 km, in comparison to the 100m distances of WiFi. We note that WiFi 802.11a can also offer a longer range lower bit-rate option too (802.11-2007 standard), indicating the overall flexibility of the wireless approach and the possibility of various service platforms. Integrating a wireless mesh network with an optical backhaul topology has been especially studied in [46] describing a hybrid architecture concept. Particular aspects considered include: a solution for a reconfigurable optical backhaul topology so as to load balance capacity amongst the various optical links in the backhaul side, whilst simultaneously providing an optimized routing approach from end-users to the RAU across the nodes of the wireless mesh network. As noted in the study, such a hybrid approach also assists in the deployment of optical fibre, since cost-sharing among operators of fixed and wireless networks is intrinsically enabled. In addition, open access topologies are also supported; in contrast to the separate infrastructures in operation today. Other recent studies have also analysed hybrid access network architectures consisting of optical back-end and gateway-routers, which connect end-users using either devices based on the IEEE 802.11 or IEEE 802.16 standards [47]. In particular, the challenges arising from deploying these hybrid PON-wireless access networks topologies with appropriate basestation placements and the associated routing issues are also considered. Improved versions of routing algorithms designed to improve network efficacy are presented; although Quality of Service (QoS) issues have been neglected in these studies, as has been appropriate Medium Access Control (MAC) algorithms to optimise the performance of the overall access network. An important aspect to consider is the avoidance of outage of a large number of end-users in the event of individual node failure – this can be achieved by appropriate load-balancing techniques in the WDM and/or TDM domains. For example, traffic prioritization capability is already inherent to the IEEE 802.16 and IEEE 802.11e/n standards, via a dynamic bandwidth O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 91 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 allocation (DBA) algorithm for application in WDM-PONs. In the context of both connection-orientated and connectionless data traffic current research issues also include appropriate prioritisation algorithms and load-balancing within a converged optical-wireless access network. Other approaches that have been described which consider the integration of optical networking with wireless technologies, both within the metro as well as the access contexts, are considered in [48]. There are several advantages with optical-wireless optical architectures that should be considered. In areas where it is difficult to bring fibre into a living space, or the acceptance of a new installation is difficult at least from the customer point of view (e.g. when the final few metres would otherwise be a coax access solution), an LTE-enhanced or similarly based access technology provides an important broadband access solution (e.g. Gigabit Level) as well as significant flexibility in the early years when a FTTH take-up rate may not be particularly high (e.g. only a 5-20% take-up rate in the first few years.). DIY kits allow subscribers to plug into existing access (LTE enhanced, coax wall-outlet) provision to allow them to initially sample a limited broadband service until they decide to go for the full installation of fibre into their living space. Where a geographical area has completely migrated from enhanced LTE and HFC (EuroDOCSIS) the active part in the external hardened box can be shut down (the redundant parts removed) and only the optical power or colour splitters retained within the box. A single Operations and Maintenance system for all access technologies (EuroDOCSIS, LTE, GSM, FTTH) provides better failure traceability and reduces costs. (Naturally, this is addition to a further transport network O&M system.). 5. Survey of evolving components/subsystems For each of the covered NGOA system concepts in Chapter 4 the system description contains a presentation of the main system specific components/subsystems. Some of these can be further identified as key evolving components/subsystems for the NGOA. In this chapter key components/subsystems are identified. These are then described with respect to a number of key aspects defined in Chapter 3. Starting with the WDM based systems including pure WDM-PON, as well as the passive and active hybrids; these rely on the evolution of a large number of different components/subsystems depending on system variant. In section 5.1, general WDM components/subsystems such as wavelength selective components are discussed. Section 5.2 discusses different reflective transmitters used at the ONU in several WDM based PON concepts. Sections 5.3 and 5.4 cover tunable lasers and wavelength selective receivers which are basic building blocks for a large number of WDM system concepts. Section 5.6 discusses analogue-to-digital converters (ADC) and digital-to-analogue converters (DAC) used in system based on advanced modulation. Some system concepts, and especially 40G TDMPON, require dispersion compensation to meet reach requirement. Different methods for dispersion compensation are discussed in 5.7. Section 5.8 contains a survey of various passive wavelength selective devices such as thin film filter-based WDM components, fibre Bragg grating-based WDM components etched diffraction grating-based WDM components and arrayed waveguide grating-based WDM components. To meet the largest reach requirements of ~100 km most systems require some form of reach extension discussed in 5.9. The chapter is concluded with a discussion on switching components in section 5.10, where evolution in terms of cost and power consumption will have large impact on AON solutions. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 92 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 5.1 WDM COMPONENTS/SUBSYSTEMS 5.1.1 Wavelength selective components for WDM systems Wavelength Division Multiplexing (WDM) has been introduced to increase transmission capacity without having to install new fibres. WDM is a technique that allows signals from different wavelength channels to be multiplexed into one optical fibre, transmitted together and demultiplexed. Today Time Division Multiplexing (TDM) systems are widely used in optical communication networks. In TDM technology multiplexing and demultiplexing (MUX/DEMUX) are realized with help of electronics. In each node optical-to-electronic conversion occurs to process the signal in electronic way and then signals are converted back to optical domain. In these systems speed is limited by electronics. On the other hand WDM systems are based on multiplexing and demultiplexing in the ptical domain. Moreover, this treatment is based on passive wavelength selective devices, that if needed can be used in a massively parallel way providing high speed and high throughput. These wavelength selective devices can be designed to have different channel count and channel spacing depending on their application in the WDM system, such as: - WWDM: Wide Wavelength Division Multiplexing uses normally only few channels with channel spacing ≥ 50 nm, - CWDM: Coarse Wavelength Division Multiplexing uses a relatively small channel count with channel spacing < 50 nm, - DWDM: Dense Wavelength Division Multiplexing uses a large number of channels with channel spacing ≤ 8 nm. Wide WDMs can be used as triplexers, for example in PONs for FTTx applications. In this case they are designed to combine or separate three channels at wavelengths of 1310 nm, 1490 nm, and 1550 nm carrying data, video and voice signals onto one single optical fibre. Coarse WDMs combine usually up to 16 wavelengths onto a single fibre. It is a cheap solution due to the fact that the large channel separation eliminates expensive stabilized narrow-band lasers. CWDM technology uses an International Communication Union (ITU-T) standard 20-nm spacing between the wavelengths, from 1310 nm to 1610 nm. Dense WDMs combine 40, 80 or 160 wavelengths onto a single fibre. DWDM technology uses an ITU-T standard channel spacing of 12.5, 25, 50 or 100GHz (0.1, 0.2, 0.4 or 0.8 nm respectively) arranged in several bands in optical communication window 1500-1600 nm. This is the most expensive solution, but gives the highest network capacity, suitable for applications, where there are many users sharing costs. DWDMs are today mainly used for long haul and metro networks. There are several kinds of MUX/DEMUX components available. Those based on thin film filters and fibre bragg gratings consist usually of discrete components for each wavelength. The second class is Planar Lightwave Circuits (PLCs) also called Photonic Integrated Circuits (PICs) represented mainly by arrayed waveguide gratings (AWGs), but including also more seldomly used etched diffraction grating-based multiplexers (EDGs). This is the class that can manage large numbers of channels and treat them in a parallel way. This class can also be integrated with other passive and active components on a common platform. During the last decade, PLC-based components have greatly improved core network solutions and capacity and are expected to have similar influence on next generation access networks. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 93 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Photonic integration based on silicon platform is now under a very intensive development and due to its compatibility to standard CMOS electronics promises low cost mass production allowing for deployment of FTTx infrastructures more cost effectively than ever before. Possible components to be characterized: - thin film filter-based ADD/DROPS fibre bragg grating-based ADD/DROPS arrayed waveguide grating-based MUX/DEMUX and more advanced architectures etched diffraction grating-based MUX/DEMUX, triplexers and other devices transmitters: FP lasers, DFB/DBR lasers, VCSELs receivers: p-i-n, avalanche photodiodes 5.1.2 Example of FTTH-PON-based on ASE-injected FP-LDs WDM technology has been successfully used in backbone network systems for quite a long time and presently there is a trend to move this technology for application in access networks. There have been several solutions described in literature. In [49] the authors proposed a WDM-PON system that employs the ASE-injected Fabry-Perot Laser Diodes as light transmitters, both in CO (OLT) and ONTs. The system has 32 wavelength channels with channel spacing 0.8nm (100 GHz) and 2 athermal AWGs, one in a remote node (RN) to distribute and collect signals to and from ONTs (L band was used for downstream signals and C band for upstream) and one in the OLT for distribution of upstream signals and multiplex downstream signals. Downstream and upstream signals are separated by thin-film filters and AWGs, due to their cyclic ability can treat L band and C band simultaneously. As light sources, wavelength-locked FP-LD lasers were used, as they are expected to be economical and practical. In this scheme the broadband light source based on erbium-doped fibre (ASE) was spectrum sliced by an AWG and injected into the FP-LD. The light output from the laser was spectral filtered by the same AWG. The upstream (downstream) transmitted wavelength of (to) each subscriber is determined by the wavelength of the injected light, sliced by the AWG. In this way a common type of FP-LD for upstream (downstream) data transmission of (to) all subscribers can be used, and the ONT is “colorless”. Components used in this architecture: - AWG: cyclic, 32 channels, channel spacing 0.8nm (100 GHz), for L and C bands - Thin film filter separating L band and C band - ASE: amplified spontaneous emission broadband light source based on erbium fibre amplifiers for C band and L band - Multimode Fabry-Perot laser diode with possibility for wavelength locking - PD: p-i-n photodiode - Optical circulator to distinguish between directions of injected light from ASE sources: C band for upstream and L band for downstream. In this or similar architectures, instead of using wavelength locked FP lasers one can apply: - tunable DFB laser - tunable DBR laser O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 94 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 - Tunable VCSEL - Spectrally sliced superluminescent light-emiting diode - Spectrally sliced erbium-doped fibre amplifier - External modulator - Reflective SOA The last two devices are used with centralized light source at the OLT and external modulator or reflective SOA at ONU for upstream modulation. In this case to transmit downstream and upstream in a full duplex manner in the same wavelength, various data modulation schemes have been proposed: ASK-ASK using different power offset values, frequency- shift keyingASK, and SCM-SCM. 5.2 ONT BASED ON RSOA, REAM, REAT 5.2.1 ONT based on RSOA In absence of low cost tunable lasers, other kinds of optical devices must be used to implement colorless ONTs. The RSOA (Reflective Semiconductor Optical Amplifiers) appears as one of the most promising solutions, at least for bit-rates equal to or lower than 1.25 Gb/s, due to the capability to amplify and modulate the input signal in the same device. One of the major limits of the RSOA is the modulation bandwidth. 10Gb/s is not feasible at the moment. For that reason WDM-PON architectures based on RSOA aim at increasing the number of subscribers with respect to GPON solutions, rather than offering higher bandwidths. The main features of RSOA based WDM-PONs are the possibility to modulate and amplify the signal in one step with a switching time in the ns range, wide optical bandwidth (more than 60 nm) which can be centred in a wide range (1200-1550nm) and low noise (down to 6dB). Reflective SOA with NRZ Signaling Various system architectures based on RSOAs have been proposed in literature [50]. Compared to a broadband light source [51], the use of a comb of coherent wavelengths to generate CW carriers allows for longer distances, due to the lower noise induced on the US signal. Moreover it permits 2.5Gb/s symmetric transmission, which is well beyond the nominal modulation bandwidth of the device. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 95 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 GPON OLT Optical Network Unit … Feeder Fiber L band expansion AWG … AWG Data DL 2.5G Multi λ 32Ch ONT WDM C/L L band expansion 1:8 WDM Data UL 2.5G Multi λ CW Carrier 32Ch Service ONU AWG . . . AWG . . . Optical Line Terminal Multi λ 32Ch IL . . . IL BERT RSOA C/L SMF CW Carrier BERT IL ONU Remote Node Figure 59 : 2.5Gb/s WDM-PON employing RSOA-based ONUs and NRZ signaling The system architecture based on RSOA [52] is shown in Figure 59. At the OLT, DS modulated channels (λ1D,···, λ32D) and CW carriers utilized for US re-modulation are generated by two different multiwavelength sources each producing 32 100GHz-spaced channels. The two combs are offset by 50GHz in order to obtain a set of 64 wavelengths aligned to a 50GHz spaced ITU-T grid. In particular the channel plans are 1534.250nm-to1558.983nm and 1534.643nm-to-1559.389nm for the two multiwavelength sources respectively. The use of different DS and US carrier wavelengths allow for better tolerance to back-reflections (caused by fibre connector, splices, etc…), which is one of the most prominent degradation effects in single feeder fibre schemes. One of the two generated combs is demultiplexed by a conventional 100GHz AWG and each wavelength is individually modulated with a 231-1-long PRBS at 2.5Gb/s. The modulation format is the conventional NRZ-OOK. The 32 modulated channels carrying the DS information are then recombined in a similar AWG, coupled with the CW comb and sent downlink through the SMF. A booster amplifier allows fixing the channel power to 5dBm. At the RN, DS channels and adjacent CW optical carriers are routed to each end users by means of a 50GHz spaced AWG. Low-cost and low-insertion loss (about 1dB) interleavers (ILs) are used in order to couple the correct DS signal and CW carrier to the user, keeping a single fibre connection between the RN and the ONU. The DS signal is detected using an APD receiver, while the CW carrier is sent to a RSOA for US modulation. The used RSOA is a commercial device optimized for 1.25Gb/s operation with 20dB small signal gain, 3dBm saturated output power, low polarization-dependent gain (PDG) (1.5dB), ultra-low front facet reflectivity (<10-5) and 1.5GHz electrical bandwidth. The modulated US optical signal (λ1U,···, λ32U) is then transmitted by the ONU to the OLT across the same feeder fibre as used for the downlink. At the OLT an optical circulator followed by a 100GHz spaced AWG is used to route the US signal toward the receiver. To improve the US system performance the CW multiwavelength source can be moved to a service-ONU (S-ONU) closer to the user terminations and connected to the RN common port as shown in Figure 59 (dashed line). Figure 60 : BER versus received power with multiwavelength source placed at the OLT (a) and the RN (b) O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 96 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Results show that the architecture with the multi-channel source at the RN is less affected by power budget restrictions. The transmission channel is extended to 35 km with a maximum uplink penalty of 3 dB. An extension of the transmission channel up to 45.2km reach can be achieved by using the multi-channel at the RN and an optimum RSOA input power with maximum uplink penalty of 2.5 dB. BER measurements are shown in Figure 61. Figure 61 : BER measurements versus received power with multi-channel near RN Reflective SOA with RZ Signaling The setup used for experimental tests on a WDM-PON system based on RSOA with RZ modulation is shown in Figure 62. GPON OLT Optical Network Unit … Feeder Fiber L band expansion AWG … ODL AWG Data DL 2.5G Multi λ 32Ch ONT WDM C/L L band expansion 1:8 WDM C/L BERT ODL Data UL 2.5G . . . BERT . . . AWG . . . AWG Pulse Generator Optical Line Terminal RSOA SMF ONU Remote Node Figure 62 : 2.5Gb/s WDM-PON employing RSOA-based ONUs and RZ signaling At the OLT a 32-channels multiwavelength source is used to generate the DS carriers (λ1D,···, λ32D). A MZ-modulator driven by an electrical pulse generator acts as pulse carver in order to obtain a train of 70ps-width pulses. The WDM comb is then demultiplexed by a 100GHz AWG and each wavelength individually OOK modulated by an electrical 2.5Gb/s 231-1 PRBS signal using a MZ-modulator. An optical delay line (ODL) before the AWG is used to synchronize the RZ pulses with the electrical data pattern. The wavelengths are then recombined in another AWG and sent into the feeder fibre with a power of 5dBm per channel. The DS signal extinction ratio is controlled by varying the pulse carver modulator bias voltage and performance was investigated at 5dB and 3dB of its value. At the receiver of each O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 97 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 ONU the DS channel is divided by a 90%-10% splitter. The smallest part of the signal is photodetected by a PIN receiver with an optimized threshold. The largest part of the signal is sent to the RSOA, which erases the DS data, amplifies the recovered pulsed carrier and generates the US signal. In order to optimally cancel the DS modulation, the RSOA should operate near to the gain saturation regime. Best performance is obtained by applying a DC bias current of 80mA, which permits maximizing the RSOA gain, and a RF voltage of 3.3V. The US data is 2.5Gb/s 231-1-long PRBS, applied to the RSOA RF electrical input by means of a tee-bias coupling. At the RN all the US channels are multiplexed using a 100GHz AWG and sent towards the OLT, where they are demultiplexed and detected. The RSOA input power is - 4.5dBm in the BtB case and decreases when the optical fibre is introduced. 3 BER Evaluation Downstream Extintion Ratio 3dB 3 DL BtB UL BtB DL 20Km UL 20Km DL 35.2Km UL 35.2Km 4 6 7 8 9 10 11 12 DL BtB UL BtB DL 20Km UL 20Km DL 35.2Km UL 35.2Km 4 5 -LOG(BER) -LOG(BER) 5 BER Evaluation Downstream Extintion Ratio 5dB 6 7 8 9 10 11 12 -26 -25 -24 -23 -22 -21 -20 -19 -18 -17 -16 -15 -26 -25 -24 -23 -22 -21 -20 -19 -18 -17 -16 -15 Received Power [dBm] Received Power [dBm] Figure 63 : BER performance for different DS extinction ratio values of 3dB (a) and 5dB (b) The US and DS BERs curves are reported in Figure 63(a) and (b) for the 1550.1nm channel in the case of 3dB and 5dB DS extinction ratio respectively. Performance obtained exploiting 20km and 35.2km of feeder fibre lengths are compared with the BtB case. In particular, for an extinction ratio of 3dB, the power penalty is less than 1dB up to 20km. The testing of longer lengths is not possible due to the sensitivity limitations of the PIN photodiode. A proper comparison of these results with those obtained for the NRZ format, would require use of an APD detector in the latter case. Looking at the BtB curves in Figure 60 and Figure 62 one may note that the sensitivity difference between the NRZ and RZ formats is given by the horizontal distance of the corresponding curves. Therefore the expected improvement when using an APD is 9.7dB and 8.1dB for an RZ format with 5dB and 3dB of extinction ratio respectively. It is important to point out that such performance is tolerable if applying conventional Forward Error Correction techniques to the US signal. It can be also noticed that increasing the DS modulation depth improves the DS performance but impairs the US. 5.2.2 ONT based on REAM An example of REAM at the ONT operating at 10Gb/s per wavelength have been reported in [52]. These devices are less bandwidth-limited as compared to RSOAs and are good candidates for low cost deployment. Figure 64 shows a WDM-PON system based on wavelength remote distribution. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 98 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 GPON OLT … Feeder Fiber L band expansion Optical Network Unit APD + TIA L band expansion 1:8 High Sensitivity CDR BERT IL AWG … Multi λ 32Ch AWG Data DL 10G ONT WDM C/L WDM C/L SOA SMF Multi λ CW Carrier 32Ch REAM Data UL 10G Multi λ 32Ch Service ONU . . . CW Carrier . . . Optical Line Terminal AWG . . . APD + TIA High Sensitivity CDR BERT AWG IL IL ONU Remote Node Figure 64 : 10Gb/s WDM-PON employing REAM-based ONUs In order to reduce the impairments due to Raleigh back-scattering (RBS), the multiwavelength source for CW distribution is located at the RN as in the 2.5Gb/s case. DS signals within the 100GHz ITU-T grid, are generated and modulated at the OLT. Conventional NRZ-OOK modulation is used. The data stream is a 231-1-long PRBS at 10Gb/s. US CW channels have the same wavelength spacing but they are 50GHz shifted to avoid superposition. A couple of ILs, at the RN and at the ONU, separate the DS and US channels as in the RSOA-based scheme. The US data stream is obtained through 10Gb/s NRZ-OOK 231-1-long PRBS used to directly modulate the REAM. A cascaded SOA is used bidirectionally [53] to compensate for the REAM losses. 3 BER Evaluation DL BtB UL BtB DL 25km UL 25km DL 50km UL 50km 4 -LOG(BER) 5 6 7 8 9 10 11 12 -32 -31 -30 -29 -28 -27 -26 -25 -24 -23 -22 Received Power [dBm] Figure 65 : US and DS BER performance for REAM WDM-PON US and DS BER performance are shown in Figure 65 for above mentioned SOA at optimum bias current values and for a REAM at a bias voltage of -1.2V and a VRF=4V peak-to-peak. Results referring to 25km and 50km standard SMF are compared with the BtB case. A power penalty of 2dB at BER=10-9 has been measured for both US and DS using a 25km-long feeder fibre link without any dispersion compensation. The same US performance has been obtained for an up to 50km-long fibre link without any dispersion compensation, due to the chirp induced by the REAM, whilst a tolerable degradation of 4dB is measured for the DS in the last case. In fact, whilst the DS degradation can be imputed to the higher bit rate impact on the fibre transmission, the US performances are strongly related to the ONU configuration. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 99 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 However, this demonstrates the possibility of 10Gb/s upgrade over 2.5Gb/s installed WDMPON. 5.2.3 ONT based on REAT Another emerging technology offering colourless ONU functionality is that based on the reflective electro-absorptive transceiver (REAT). This offers similar possibilities as the RSOA and REAM, but with the additional benefit of requiring a significantly lower component count. In particular, the REAT located within the ONU acts as an integrated photo-receiver for the downstream data (at a wavelength in the 1300 nm window), and also as a reflective modulator for upstream data transmission (with the upstream reflected wavelength within the 1550 nm window.) This has the advantages of not requiring additional splitters, AWGs, optical circulators or ILs as are required for the RSOA and REAM architectures. However, while RSOA and REAM devices tend to be used for digital transmission systems, the high linearity of the REAT allows analogue operation, so that it has currently been most demonstrated in radio-over-fibre (RoF) systems, such as that developed within the UROOF project [54], and is currently being studied within the EU FP7 FIVER project [55]. In addition, the REAT has been optimised for use with DS wavelengths in the 1300 nm region, while US data is modulated onto wavelengths within the 1550 nm window. This is a consequence of the REAT acting as a PD at 1300 nm, while acting as a reflector at the longer wavelengths of 1550 nm. When used as an analogue device, the REAT is employed to transmit UWB (3.1 – 10.6 GHz RF), LTE (2.4 GHz), and WiMax (3.5 GHz) signals in both US and DS directions, while the intrinsic bandwidth of the REAT operating in receive/photodiode (DS) and modulation (US) modes is at least 10 GHz. While the UROOF project was only able to demonstrate distances between CO and ONU of up to about 1 km, the FIVER project is aiming at distances of up to 100 km between CO and customer premises. However, due to the intrinsic losses associated with locating a CW 1550 nm signal source within the CO, so that light has to travel effectively twice the overall length; as well as the fact that the downstream light at 1300 nm experiences higher attenuation compared to the 1550 nm window, REAT devices can only realistically allow a maximum distance of up to 20 km between CO and ONU. Figure 66 : Cut-down (single-user) WDM-PON highlighting REAT-based ONU at CPE. Figure 66 depicts the generic architecture for a REAT-based RoF access architecture, showing a 1300 nm source at the central office, with DS data modulation, with a 1550 nm CW source being multiplexed onto the same fibre via a broadband 1300/1550 nm WDM multiplexer onto O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 100 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 the same fibre. For upstream data transmission, within the CO light at 1500 nm from a CW laser diode passes through an optical circulator, is subsequently multiplexed with the downstream 1300 nm data signal, and is passed to the 20 km optical fibre link towards the client end. Within the CPE, the REAT directly receives the DS data on the 1300 nm wavelength, but in full duplex mode modulates (with upstream data) the CW 1500 nm light in reflective mode. The reflected upstream 1550 nm wavelength passes back towards the CO where the optical circulator routes the upstream 1550 nm signal away from the 1550 nm CW laser source, and into the photoreceiver where the upstream data is received. 5.3 TUNABLE LASERS Tunable lasers are one of only two ways to achieve “colourless” ONUs. Here, “colourless” refers to the requirement that in a mass roll-out scenario, no wavelength-specific ONUs will be allowed. (Hence, the fixed-wavelength WDM-PON ONU is not seen an alternative.) The other alternative to tunable laser sources for ONUs are seeded reflective ONUs, based on reflective components like RSOAs, REAMs, or IL-FP lasers. These are described in the previous chapter. There is a number of reasons to not allow wavelength-specific ONUs. These can be split inventory reduction (i.e., cost reduction), and network operations enhancements (at least partly again cost reductions). Inventory reduction refers to two aspects. The first one is supplier- (components vendor-) driven: tuneability obviously reduces product codes (or variants) and hence ultimately helps decreasing cost, rather than increasing it. For example, the C-band has some 96 channels spaced 50 GHz which would translate to 96 different products, with all associated logistic impact. Tuneability reduces this to one product; given full-band tuneability is achieved (which is hence the design goal). The second aspect is customer- (service provider-) driven. It relates to much simpler inventory and spare parts handling. Here, tuneability reduces OpEx. This is one of the main reasons for service providers to demand tuneability. Network operations enhancements are obviously driven by the respective service provider. Here, tuneability drastically eases ONU installation, in particular remote service provisioning and configuration. For example, the field staff only needs to consider a single type of ONU. In addition, protection mechanisms based on wavelength protection become simpler, or more powerful (or even enabled at all). This is complemented by potential advantages when operation over a flexible (access ROADM) infrastructure or ODN (should this be a serious alternative to static passive ODNs). These aspects all refer to operations and hence OpEx savings or optimization. Obviously, these target advantages of tunables – which hereinafter are referred to as low-cost tunables – must not come at an increase of cost / price. This would directly contradict the first inventory aspect, and also the cost / price goal for residential access where cheap EPON transceivers are the benchmark. Further, tuneability must not compromise on system performance or reliability. Otherwise, tuneability will be, and stay, a minority interest. This, amongst other aspects, would translate to a no-profit situation for the components vendors and consequently to no such developments. In principle, tuneability of a laser can be achieved by [56]-[60]: • • Electronic Tuning Thermal tuning O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 101 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 • Mechanical Tuning In addition, arrays of DFB (Distributed Feedback) lasers must be considered as a quasitunable transmitter, especially since first mid-price products exist [66]. As a further potential laser source, Quantum-Dot lasers may be considered. However, at the time being, these lasers only support the wavelength range of 1050…1320 nm and hence are not considered hereinafter as a contender for low-cost tunables for WDM-PON. Electronic tuning refers to the principle of applying an electric field which directly tunes the laser frequency. Here, several different possibilities exist, with derivatives of the well-known DBR (Distributed Bragg Reflector) laser being the most important ones: • • • • • • • Three-section DBR Sampled Grating DBR Laser (SG-DBR) Super-Structure Grating DBR Laser (SSG-DBR) Digital Supermode DBR Laser (DS-DBR) Grating-assisted Co-directional Coupler Laser with Sampled Grating Reflector (GCSR) Y-Junction laser (Y3 laser, transverse interferometric type) Vertical Mach-Zehnder Laser (VMZ, vertical interferometric type) According to [60], only one relevant type of thermally tuned lasers exists, namely the MatrixGrating Strongly Gain-Coupled (MG-SGC) DFB Laser. For mechanical tuning, two relevant classes can be identified: • • Micro-Electromechanical Tunable Vertical Cavity Surface Emitting Laser (MEMVCSEL) External Cavity Laser (ECL) From this list, only very few have strong potential for lowest cost. Here, it must be remembered that the long-term cost goal for the tunable laser is in the 50$ range. The respective lasers are a multi-section DBR (namely the DS-DBR and SG-DBR lasers), ECL, and MEM-VCSEL. In addition, DFB arrays must be considered due to their early availability. DS-DBR Laser Multi-section, tunable DBR lasers have been described in various papers, e.g., [56]-[58], [61], [62]. They are in mass production today. In the form of the DS-DBR laser they are followed in the low-cost WDM-PON context by Oclaro, Inc. (formerly Bookham, Inc.). The clear advantage of these lasers is the fact that they are monolithically integrated which directly translates to low-cost potential. Here, we focus on the concept of the DS-DBR laser because of its high potential as a lowest-cost source. However, similar considerations hold for other multi-section DBR lasers as well. The basis of any tunable, monolithically integrated DBR laser is the so-called 3-section DBR laser [60]. Its schematic cross-sectional diagram is shown in Figure 67. Not shown here are two sub-components which are typically necessary for a laser: a Thermo-Electrical Cooler (TEC, also referred to as Peltier cooler), and a wavelength locker for wavelength tuning or stabilization. The important three sections are the Gain, Phase, and Rear sections. The Gain section, as with O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 102 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 every other laser diode, provides the lasing gain. In itself, it is not wavelength-selective and provides gain in a relatively broad spectrum. The Rear section is able to select a single laser line by means of the Bragg grating which is etched into the substrate. The Phase section can provide some degree of fine tuning. Therefore, the 3-section DBR laser can not provide fullband tuning capability. It is, however, relevant as the basis for improved tunable DBR lasers. p contacts Igain Iphase Irear p InP light output QW gain region Tuning Regions Grating n InP substrate Reflection coating n contact AR coating Figure 67 : 3-section DBR laser One relevant derivative of the 3-section DBR laser is the so-called Digital-Supermode (DS-) DBR laser. Its cross-sectional schematic diagram is shown in Figure 67. The DS-DBR laser is a more recent derivative within the family of monolithically integrated, electronically tuned lasers. It was developed by Oclaro, Inc., and has been described in various papers, e.g., [61]. Unlike most other tunable lasers, the DS-DBR laser has the potential for uncooled operation, and such uncooled operation has already been demonstrated, e.g., [62], [63]. Since in a WDM-PON context the wavelength locker can also be omitted, the DS-DBR laser has unique potential as a true low-cost tunable source. On the other hand, the lack of TEC and wavelength locker makes it useless in most other applications so that it will not cannibalize other laser markets, despite its low cost. Today, it must be seen one of the most promising lasers for WDM-PON. Ifront ISOA p contacts Igain Iphase Irear p InP light output QW gain regions Tuning regions Grating n InPsubstrate AR coating n contact AR coating Figure 68 : 5-section DS-DBR laser The DS-DBR laser shown in Figure 68 is a 5-section DBR laser. As compared to the 3-section DBR laser, on section has been added for improved tuning capability. The fifths section is a monolithically integrated SOA. This section provides both, broadband modulation capability (up to 10 Gb/s), and high per-channel output power (up to 13 dBm). Due to monolithic integration, this does not lead to severe cost increase. Wider tuning range comes from a grating – in the rear section – with many reflection peaks. This is combined with another grating in the front section for selecting between these peaks. Partial cancellation between different parts of the rear grating leads to a top-hat like envelope on the grating comb which is shown in Figure 69. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 103 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 1.0 Reflectivity 0.8 0.6 0.4 0.2 0.0 1500 1520 1540 1560 1580 1600 Wavelength (nm) Figure 69 : DS-DBR rear section wavelength comb (measured) Wavelength tuning caused by changes of the front and rear grating currents against their normalised settings can be depicted in the so-called wavelength map. For the DS-DBR laser, this wavelength (or tuning) map has wide supermode bands, corresponding to the front grating settings which are approximately 8nm (~ 1 THz) apart. Each such supermode in a DSDBR laser behaves just like a standard 3-section DBR laser. Since the tuning map contains supermode jumps, it is also referred to as a pseudo-wavelength map. Additionally the DSDBR has longitude modes which are approx. 37 GHz apart. Within one longitude mode the frequency of the DS-DBR laser can be fine tuned via the phase section. By setting the operation point to the middle of a longitude mode a tuning range of approximately +/- 20 GHz can be achieved. An example for a C-band DS-DBR laser is shown in Figure 70. Figure 70 : DS-DBR laser pseudo-wavelength map Shown here are the normalized front and rear grating current settings and, in colour code, the resulting wavelength. The tuning behaviour is relatively insensitive to actual front grating currents, which allows a broad tuning range per supermode. If this range is to be exceeded, a supermode jump becomes necessary. These supermode jumps are indicated in Figure 70 where the thick black lines cross one of the thin, vertical, black lines. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 104 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 The resulting single-mode lasing spectrum of a DS-DBR laser is shown in Figure 71. It can be seen that output power (per WDM channel, coupled into the fibre) is as high as 15 dBm. Super-Side-Mode Suppression Ration (SMSR) is in the range of 45 dB (super-side modes can still be seen in Figure 71). Since a comb with 7 reflection peaks is generated in the rear section in order to cover a full wavelength band (here the C-band). Hence, they are additionally suppressed in the AWGs used for a WDM-PON. The corresponding detrimental effect is negligible. More problematic are although longitude side-modes as they are closer to the desired frequency. As seen in Figure 72 the SMSR is normally ~ 50 dB. Figure 71 : DS-DBR lasing spectrum – with side-super--modes Figure 72 : Side mode suppression ratio The DS-DBR laser as used today for tunable laser sources and as intended as low-cost source for WDM-PON is a 5-section DBR laser including a SOA. The SOA provides modulation bandwidth and high output power. Here, it can also be used as a fast shutter in cases of switching events or even in cases of the – predictable! – (super-) mode jumps. Then, the SOA is reverse-biased which is shown in Figure 73. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 105 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 20 Laser Power (dBm) 10 0 -5 -4 -3 -2 -1 1563.33nm 1556.15nm 1549nm 1528.3nm 0 1 -10 -20 -30 -40 -50 SOA Voltage (V) Figure 73 : DS-DBR reverse-biased SOA as a shutter Relevant optical characteristics over the entire tuning range (>100 channels in the 50-GHz grid) are shown in Figure 74. Shown here are the Relative Intensity Noise (RIN, in dB/Hz) and the laser line width (in MHz). It can be seen that RIN is <-150 dB/Hz for all channels and that line width is in the range of 500 kHz for all channels. Not displayed here are output power and SMSR. They are in the range of 15+/-1 dBm (power) and 42…56 dB (SMSR), respectively. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 106 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 74 : DS-DBR laser RIN (a), and line width (b) For WDM-PON, the respective transmitter must have lowest-possible power consumption. For this reason – and in order to reduce CapEx – most service providers require TEC-less transmitters. Generally, the DS-DBR laser has the potential to work without TEC. However, temperature drift will not only cause wavelength drift (which can be counter-acted through grating current tuning) but also variations of the maximum output power. This is demonstrated in Figure 75. At 25°C, the output power stays above 14 dBm for all channels. However, at 45°C, output power of some channels drops to slightly below 12 dBm. Hence, avoiding the TEC may lead to somewhat reduced output power (where the 12-dBm range is still very good for WDM-PON applications). Other parameters like SMSR are less affected by temperature drift. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 107 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 75 : DS-DBR laser temperature-depending output power (as of today) Application of an uncooled DS-DBR laser for WDM-PON has already been demonstrated [62]. In order to tackle the temperature drift problem, the most recent approach for the DSDBR laser is to be based on new high-temperature materials, e.g., AlGaInAs/InP [64]. This approach could support uncooled operation well into the 70°C region, possibly supported by a small low-power heater (rather than a TEC). Target power consumption is in the 1-W range. The DS-DBR laser has good potential as a low-power-consuming, low-cost transmitter for WDM-PON. It must be noted that similar (multi-section) DBR laser derivatives with similar potential exist, for example the Sampled-Grating DBR (SG-DBR) laser or the Super-Structure Grating DBR (SSG-DBR) laser. More detailed comparisons between these DBR types can be found in [56], [57]. MEM-VCSEL Micro-Electro-Mechanically (MEM) tuned VCSELs are a promising alternative the monolithic (multi-section, DBR) lasers. Generally, VCSELs have low-cost potential, and by adding MEM means, they can be tuned. Tunable MEM-VCSELs have already been described for metro-area applications in 2001 [70]. In the last years, they have gained attraction with regard to WDM-PON due to their cost characteristics. The basic concept of the MEM-VCSEL has not changed significantly over time. It is based DBR-VCSEL with an attached micro-mirror which is thermally tunable. (In [70], application of an electrostatic force to the mirror led to tuning.) A schematic diagram of a more recent MEM-VCSEL is shown in Figure 76. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 108 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 76 : Cross-sectional diagram of a MEM-VCSEL MEM-tunable VCSELs are a hybrid two-chip assembly of a MEM system (MEMS) with concave AlGaAs-GaAs mirror membrane and an InP-based semiconductor cavity with tunnel junction aperture. Using electro-thermal MEMS actuation, the included air-gap can be expanded and the cavity resonance be tuned to longer wavelengths. The gold p-contact also serves as an efficient heat-sink. The active part (according to Figure 76) is formed by 7 compressively strained InGaAlAs quantum wells embedded between an n-type and a thin ptype InGaAlAs confinement layer grown by molecular beam epitaxy. A centrally limited buried tunnel junction (BTJ) defines the active size (diameter) and reduces the inefficiently conducting p-type region to a minimum extend forming the junction. The BTJ is overgrown with a binary n-InP current- and heat-spreading layer in a subsequent MBE process, followed by an evaporated dielectric backside distributed Bragg reflector (DBR) composed of 3.5 pairs of calcium fluoride and zinc sulfide with gold coating and an electroplated substrate. Such back mirror has a limited size, slightly larger that the BTJ. The semiconductor cavity has an optical length of according to coupled cavity considerations and the corresponding limitation on the tuning range. The top DBR consists of MBE-grown GaAs-AlGaAs quarter wavelength layers with refractive index contrast. The inclusion of a graded indium content induces compressive strain, which is released after membrane processing and shaping, resulting in a curved mirror membrane with radius of curvature of 2…3 mm, suspended on four arms. The upward movement is achieved by Joule heating of the upper membrane. Depending on the cavity configuration, tuning ranges up to 60 nm and single-mode output power up to 2 mW with high sidemode suppression and stable polarization have been demonstrated. Slightly different design schemes for MEM-VCSELs exist, namely the semiconductorcoupled cavity (SCC, type B) and the extended cavity (EC, types A, C) designs. From the basic SCC design (bottom Bragg mirror cavity air–gap and curved mirror), the EC scheme can be obtained quite easily by placing an antireflection coating layer (AR-c in Figure 76) at the semiconductor-air interface of the lower half VCSEL. By comparing the overall performances of the two schemes, it turns out that an extended tuning range of EC devices has to be paid at the expense of lower delivered powers and higher thresholds. The different tuning performances can be compared in Figure 77 by the spectra for different tuning currents. One can clearly see that the laser can be continuously tuned over almost the whole free spectral range (FSR) and it stays always single mode with high SMSR. From such measurements one can also extract the FSR, i.e., the distance between two adjacent longitudinal modes, which is related to the length of the air-gap. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 109 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 77 : VCSEL tuning. (B) shows tuning of an SCC type VCSEL, (A) and (C) for EC types. The FSR sets the upper limit for the tuning range. The latter can be limited by the material gain spectrum, which has to compensate for the wavelength-dependent threshold conditions. The limitations due to the DBR bandwidth are negligible in both designs. Tunable ECL External-Cavity Lasers are an obvious choice for tunable transmitters because of their inherent capability to be tuned via the external cavity. They are described in [56]-[58]. The basic ECL consists of a broadband gain section to which an external cavity is coupled via an Anti-Reflection (AR) coating and a collimating lens. This setup is shown in Figure 78. The gain section can be a simple Fabry-Perot laser diode. Wavelength tuning is possible by tuning the external grating. This can be done mechanically or thermally. One contributor to cost is the optical coupling between the components which typically has to be done via a lens. This is a disadvantage over monolithically integrated solutions. It also leads to lower output power due to the loss of the free-space region. Figure 78 : Schematic diagram of an ECL There are two basic tuning mechanisms, thermal and mechanical tuning. Mechanical tuning often is implemented using MEM components. Thermal tuning is often applied to etalons. It can also be applied to FBGs which in addition offers the chance for better and more costO O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 110 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 effective integration. The corresponding devices are also referred to as FBG-ECLs, a principle diagram and an early implementation are shown in Figure 79. + Lens FBG Gain HR – AR Figure 79 : Principle of an FBG-ECL (left), and FBG-ECL from Xponent Photonics, Inc. (right) FBGs are also the tuning means for a first mid-price commercial component which is available [68]. Though not yet being in the price / cost region which is considered to be necessary for WDM-PON mass roll-out, this product is already cheaper than any other tunable laser available on the market today. As an early implementation of a tunable laser, it still has somewhat limited performance specifications. At the time being, maximum tuning range is ~26 nm, and the device has ~6 W power consumption which is mainly driven by two TECs, one for the laser (gain) chip, and one for the grating. The device is suitable for up to 2.5 Gb/s bit rate, and it comes in a butterfly packaging, see Figure 80. Figure 80 : Packaging and pin configuration of commercial ChemOptics FBG-ECL The per-channel output power is in the range of 6.2+/-0.4 W, and temperature drift over the range of -10…+70°C is ~0.14 nm. This makes it suitable for the 100-GHz grid, for 50 GHz some improvements would have to be made. Alternatives to thermally tuned (FBG-) ECLs are mechanically tuned ECLs. A proposal for a MEM-tunable ECL which is suitable for UDWDM-PON application has recently been made O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 111 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 [69]. This device is based on a gain chip and an external cavity which consists of a dielectric filter which is controlled (tuned, turned) by a micro-motor. According to [69], this transmitter has sufficient availability (failure resistance, having in mind that a motor with moving parts is used), full-band tuneability with a precision of a few 10 MHz, and a line width of <200 kHz (the latter being necessary for use in a coherent UDWDM-PON). It is, however, doubtful that a laser with tuning precision requirements of some 10 MHz – which are necessary in a coherent receiver – can be implemented without a TEC. Tunable DFB-Array DBR lasers, ECL, and (MEM-) VCSELs are the only lasers followed today for WDM tuning. This also holds for the latest tunable XFP generation which is based on DS-DBR and SGDBR lasers, respectively. The well-known DFB laser in itself is not tunable because gain and grating sections by definition are not decoupled (unlike, in particular, multi-section DBR lasers). However, integrated arrays of DFB lasers can be made quasi-tunable and hence are an alternative to DBR, ECL, and VCSEL. It is also worth noting that first commercial products already exist [66]-[67]. A photo of the respective device is shown in Figure 81. Figure 81 : Photo of DFB array chip The DFB array shown in Figure 81 consists of 12 DFB lasers, spaced 10 µm apart in position and 450 GHz apart in frequency. The entire chip size is 500 µm by 800 µm, which is comparable in size to high performance single element devices. It requires four separate epitaxial growths and two layers of metallization. First, a quaternary InGaAsP grating layer is grown on an n-type InP substrate, and multiple phase shifted gratings are fabricated by direct write electron beam lithography and wet etching. The grating is then overgrown with the strained InGaAsP multi-quantum well active region, and 2 µm wide mesas are formed by wet etching. A third growth adds the p-n InP blocking regions with selective area growth, and the fourth growth the top p-cladding and the heavily p-doped InGaAs contact layer. Isolation regions that are 3 µm wide are then etched between the elements of the array, the sidewalls are coated with silicon nitride, and a first gold metal line is then deposited and plated on top of each stripe. An intermediate silicon nitride layer is deposited on the chip. Wafers are then thinned and polished, n-contact metallization is applied to the back, and the wafers are then cleaved, coated AR/AR and bonded onto submounts. A great advantage of this geometry is that each stripe is contacted in the middle, rather than the end, so that uniform current injection along the stripe can be applied. Furthermore, there is no real-estate wasted in routing electrical contacts around the array. This results in a compact O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 112 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 geometry. To maintain high yield, the lasers are phase-shifted with AR coating on both sides, which necessarily implies lower efficiency, since some power is wasted from the backside. The grating and phase shift were designed such that twice as much power exits the front facet than the rear. A chip where the laser elements were varied in frequency by about 450 GHz was soldered in a butterfly package with a MEMS mirror for selection. One advantage of a widely tunable laser based on an array is that redundancy can be built in at the edges of the array, so that any error of a specific laser means that a subset of the entire tuning range can still be used. This way, the array yield can also be improved. The optical performance of the DFB array is within the range which can be expected from a DFB laser. Linewidth is below 500 kHz for all channels, and RIN is better than -153 dB/Hz. Figure 82 : Photo of DFB array chip [66] In addition, fibre-coupled output power in the range of +15 dBm and SMSR of >55 dB has been demonstrated. These optical parameters can be considered sufficient for highperformance WDM-PON applications. A disadvantage is the relatively high power consumption which is in the range of 5 W for a 12-channel device. Since future improvements regarding power consumption are expected to be relatively low [56], this parameter is considered too high for wide-spread deployments. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 113 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Comparison of relevant tunable lasers From the analysis provided herein so far, it becomes clear that only few promising contenders for low-cost and low-power-consuming tunable lasers suitable for WDM-PON exist. Here, it should be noted that focus is on lowest possible cost and energy consumption, rather than best optical specifications like line width, RIN, or modulation bandwidth. For many WDM-PON applications, laser output power however will be a relevant parameter. The most relevant parameters of the lasers discussed so far are listed in Table 17. Table 17 : Output Power [dBm] Power Dissipation [W] Components in Package Line Width RIN SMSR Tuning Range Bandwidth Future Cost Potential Comparison of relevant parameters of tunable lasers DS-, SG-DBR >13 CW >4 10G Tx <2.5 w/ TEC <1 w/o TEC Monolithic Laser with integrated SOA, no Locker, no TEC possible, plus Receiver ~500 kHz <-150 dB/Hz FBG-ECL ~15 CW <3.5 Laser, Lens, FBG, TEC? Receiver MEM-ECL ~15 CW <3.5 Laser, Lens, MEMS, TEC? Receiver >45 dB C-/L-band <100 MHz <-140 dB/Hz >30 C-/L-band >100 kHz <-150 dB/Hz >50 dB C-/L-band 10 Gb/s Very low 2.5 Gb/s Low 10 Gb/s Low MEM-VCSEL <3 DFB-Array >13 CW <1 W (?) <4.5 VCSEL, MEMS, TEC? Receiver 8-12 Lasers, Combiner plus SOA or MEMS, TEC? Receiver <100 MHz <-120 dB/Hz >30 dB C-Band (>40 nm) 2.5 Gb/s Very low <500 kHz <-150 dB/Hz >50 dB S-/C-/LBand 10 Gb/s Low From Table 17 the most promising candidates for future tunable WDM-PON lasers can be identified. From viewpoint of lowest power consumption, multi-section DBR lasers without TEC (uncooled operation of both, the DS-DBR and the SG-DBR laser has been demonstrated [62], [72]) and MEM-VCSELs are the most promising devices. From viewpoint of RIN, SMSR, line width and also tuning range, all lasers can be considered as applicable, with only slight disadvantages for FBG-ECLs. Regarding laser output power, XX-DBR (XX = DS, SG), ECL and DFB arrays have advantages over VCSELs. This is narrowed down to XX-DBR, MEM-ECL and DFB arrays in view of future high-speed (10 Gb/s) applications. The last relevant parameter is the future cost estimation, or the potential for lowest-cost future mass production. This parameter, to a good degree, correlates with the components which are necessary within the transceiver package. Here, advantages can be identified for the monolithically integrated XX-DBR lasers which have the potential for TEC-less design, and again for VCSELs. From this, we can conclude that XX-DBR lasers are the most promising transmitter variant for future WDM-PON. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 114 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 5.4 WAVELENGTH SELECTIVE RECEIVERS Wavelength-selective receivers are necessary whenever specific wavelengths have to be assigned to specific ONUs in a WDM-PON running via a splitter-based – i.e., broadcast – ODN. In general, they are not necessary in a WDM-PON with filters. This holds for most of the hybrid PONs where the splitters are used in order to allow wavelength re-use (or broadcast). An exception is the UDWDM-PON where in a filter-plus-splitter ODN, wavelength-selective receivers still are required. Since splitter-based ODN may become an important migration path for WDM-PON, wavelength-selective receivers must be considered a potentially important component. An example of early WDM capabilities in a splitter-based ODN is the stacked-XG-PON approach. However, at the time being, importance of similar approaches is not fully clear since, for example, splitters in a WDM-PON eliminate many of the advantages (for example: security, low insertion loss, and hence part of the potential cost advantages). In general, two approaches exist for wavelength-specific receivers, receivers with tunable filters and coherent (heterodyne, intradyne) receivers which make use of a tunable local oscillator. It must be kept in mind that both approaches are not in massive use in the PON context today. This has an impact on cost of the respective components which of course will add to total systems cost. 5.4.1 Tunable Filters Tunable filters for optical transmission have been investigated since the early WDM days. An early overview is given by [73]. In general, tuneability of a WDM filter can be based on one out of the following mechanisms (where WDM in most cases refers to DWDM): Fabry-Perot (FP) Interferometer Liquid-Crystal (FP) Filter MEM-tunable Devices FBG, temperature- or strain-tuned AOTF, Acousto-Optic Tunable Filter EOTF, Electro-Optical Tunable Filter AWG (thermally tuned) Mach-Zehnder Interferometer (cascaded, with heaters) Active Filters incorporating laser diodes Tunable Ring Resonator filters From this list, only few technologies have the respective cost, complexity, and also formfactor potential for future applications in (next-generation) PONs. The majority of the literature [74]-[85] indicates that relevant contenders for low-cost WDM-PON tunable filters come out of the first four technologies listed above. Tunable FP-TFF Most FP interferometers are based on Thin-Film Filter (TFF) technology. TFF is also used for many fixed DWDM filters (where the alternatives are etched diffraction gratings, AWGs and FBGs, see § 5.8). A good overview on tunable TFF is given in [79]. Fabry-Perot etalons are comprised of optical glasses with high-reflection coatings, the two high-reflection surfaces are separated by a certain distance. When light passes through the O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 115 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 etalon, the interference arising from the multi-reflection will let certain wavelengths pass through and reflect the other wavelengths. FP interferometers are the basis of most tunable filters. They consist of (TFF, liquid-crystal) etalons which are either angle-tuned, linearly tuned (i.e., a variable TFF addressed by translation), or which are tuned with active substrates or active layers. A simplified schematic diagram of an active-substrate TFF-FP interferometer is shown in Figure 83. Figure 83 : Schematic of tunable TFF with active substrate The basic idea of the tuning scheme shown in Figure 83 is to transfer a mechanical strain of the substrate to the spacer thickness of a DWDM bandpass filter deposited at its surface, and to achieve a change in the resonant wavelength of this all-dielectric FP structure. The mechanical strain can be induced either thermally, or via the piezo-electric effect. For thermal tuning, two effects exist, first the increase of the refractive index of the layers created by a rising temperature, and secondly the change of their thickness induced by a differential thermal expansion between the layers and the substrate. These effects can be used to partly compensate temperature sensitivity or, at the opposite, to maximize thermal sensitivity for thermal tuning. If temperature is kept constant, a mechanical deformation of the substrate, i.e. a strain, can also be used to induce a shift in the wavelength: this strain can be produced for instance by a compressive force applied on a standard substrate, or by an electric field applied on a piezo-electric substrate. A first commercial thermally tuned TFF for WDM overlay applications in GPON ODN is available on the market today [83],[84]. The device is specifically intended for use in stacked XG-PONs and similar environments, where several (few) wavelengths are added to a standard GPON. Hence, the device has limited tuning range (~5 nm, or 600 GHz). This corresponds with the stacked XG-PON approach, where only four bi-directional PONs are intended to be stacked. An interesting parameter is the maximum tuning power which is <200 mW. A schematic diagram of this filter is shown in Figure 84. Figure 84 : Schematic of tunable TFF for GPON WDM overlay [83] A major advantage, next to low power consumption, of the thermally tunable TFF according to Figure 84 is its compactness. This is shown in Figure 85. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 116 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 85 : Packaging of tunable TFF for GPON WDM overlay At the time being, the commercial tunable TFF [83], [84] has limited tuning range, sufficient for GPON overlay with few wavelengths. If future variants of this product allow broader tuning ranges is unclear. According to the manufacturer, broader tuning ranges contradict part of the technology advantages (size, power consumption). Tunable Liquid-Crystal FP filters FP etalons based on Liquid Crystals (LCFP) are a promising alternative to TFF [75]. As compared to other FP filters, main advantages of LCFP filters include low driving voltage (i.e., low power consumption), broad tuning range, high resolution, and also low insertion loss. LCFP filters can be used in spectroscopy, laser radar, and in optical networking. Nematic liquid crystals have optical anisotropy, or birefringence. They exhibit double refraction, light polarized parallel to the director has a different index of refraction than light polarized perpendicular to the director. The director is along the same direction as the surface rubbing directions when both the alignment surfaces are rubbed in the same directions. Nematic liquid crystals also show dielectric anisotropy, the dielectric constants parallel and perpendicular to the nematic director are not the same. The dielectric anisotropy introduces body torque on the molecules in the presence of external field, which in turn gives rise to the director re-orientation. Under the external field, the director of the liquid crystal with a positive dielectric anisotropy tends to be aligned parallel to the external field, while the director of the liquid crystal with a negative dielectric anisotropy tends to be aligned perpendicular to the external field. Thus by filling the etalon cavity with the nematic liquid crystal, the refraction index of the extraordinary component of light will change with the applied voltage. The wavelength of transmission peak is tunable with different voltages applied. The structure of an LCFP etalon is shown in Figure 86. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 117 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 86 : Schematic of liquid-crystal filter Normally, a polarizer needs to be placed parallel to the alignment direction of the liquid crystal to pass through the extraordinary mode of light and block the ordinary mode that is mot tunable. For tunable laser and system monitoring, the polarization dependence is not a key problem. For WDM system, polarization independence is challenging because the state of polarization of the channel signal may be unknown in the system. Several methods have been discussed to make polarization-insensitive FP devices. One of them is to use a calcite crystal to split the light source (polarization direction unknown) into two components with orthogonal polarization directions, then each component is passing through a tunable LCFP filter with an alignment direction of the liquid crystal the same as the polarization direction. After passing through the LCFP filters, the two components are recombined through another calcite crystal. An alternative is to use two neighbouring LC pixels to match the two polarized components of one channel. The two pixels need to have alignment layers perpendicular to each other, and LCD production technology is a cost-effective solution to achieve this. Tunable FBG Fibre-Bragg Gratings are an alternative to FP etalons when it comes to tunable filters. According to [78], they have excellent central-peak transmission and sidelobe suppression, and they can be designed to cover WDM grids from 200 GHz down to less than 25 GHz. In addition, they have a certain low-cost potential. This is, however, contradicted by the lack of integration potential. FBGs consist of periodic refractive-index variations along a piece of SMF which can be imprinted into the fibre e.g., by means of transversal UV laser radiation. The periodic grating reflects specific wavelengths, and it can be thermally tuned. The schematic diagram of an FBG is shown in Figure 87. Figure 87 : Schematic of (tunable) FBG In order to obtain a good tuned response, an FBG must present a spectral answer characterized by narrow bandwidth (BW, to support any DWDM grid), weak side lobe peaks to avoid BW overlapping, and suitable temperature or strain tuned response, which should not affect BW or passband amplitude. In case a silica fibre (e.g., a standard SMF) is used, several parameters must be taken into account: refraction index (core, cladding), index modulation, O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 118 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 and the FBG period Λ as shows in Figure 87. The FBG period is related to the Bragg wavelength by: λB = 2neff Λ For an SMF28, the effective index is neff = 1.447559. The resulting FBG can be tuned thermally, or via mechanical strain. Heating can be achieved through a metal layer which covers the fibre and which is traversed by an electric current. The reflected central wavelength then varies linearly with temperature, and BW and central peak transmission do not change significantly. For DWDM with grids of 1.6…0.2 nm, the temperatures ranges to be covered can easily be achieved. Assuming the assumption that strain is uniform along the FBG, it can be used for tuning as well. Then, strain does not affect BW and peak transmission of reflected wavelengths. Figure 88 shows the spectral characteristics of an FBG transmission peak. Figure 88 : Reflection peak of an FBG Volume Holographic Gratings (VHG) is in many ways similar to FBGs except that the recording medium for the grating is not a (single-mode) fibre but a volume medium. Consequently, the incident and diffracted light are not confined to the modes of the fibre, but can be assigned to any mode that can propagate in the bulk material. This opens new possibilities for device design and for example allows removal of the circulator, which is essential for FBGs. VHGs as means for tunable filters have been described in [80]. Wide tuning range of 1510…1590 nm and low insertion loss of ~1 dB can be achieved. Figure 89shows a schematic representation of the angle-tunable volume holographic filter. The Bragg wavelength of a holographic grating is determined by λB = 2neff Λ cos θ, where θ is the angle of incidence inside the holographic material. By changing θ, the Bragg wavelength can be tuned continuously. For practical application, the tuning range can easily cover the entire C band, limited mainly by the higher insertion loss at larger angle θ. The critical requirement to make this a practical tunable filter is to collect the drop signal into a fibre during angle tuning without an expensive tracking mechanism or feedback control system. This is achieved with a self-reflector architecture that recombines the reflective holographic grating with a wideband IR mirror, as shown in Figure 89. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 119 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 89 : Schematic of tunable Volume Holographic Grating (VHG) A retroreflector consists of two mirrors with a fixed angle between the mirror surfaces. This also applies to the structure shown in Figure 89, where one mirror is replaced by the VHG. At the Bragg wavelength, this VHG works as a reflective mirror with an effective depth into the material. When rotating the mirror/grating structure around the crossing point of the mirror and grating surfaces at the effective reflection depth, the drop signal beam is fixed spatially while the wavelength is tuned by the angle θ. With the addition of a temperature sensor and suitable control system, the device can be made athermal by compensating θ to offset temperature-induced variations of the grating period. MEM-tunable Filters MEM-tunable devices are the next promising technology for low-cost tunable filters. Basically, similar principles are applied which are used for the tunable MEM-VCSEL (which also has low-cost potential). MEM-tunable filters are described e.g., in [74],[85]. Several MEM-tunable optical filters based on electrostatic actuation or thermal tuning have been reported. Electrostatic actuation is adequate for low-power operation, but its tuning range is limited due to the so-called pull-in phenomenon, and its actuation voltage is relatively high. Also, electrostatic actuation intrinsically has a nonlinear behaviour. On the other hand, thermal actuation can achieve a wide tuning range but it consumes large power and its response time is slow. A more recent alternative is magnetic actuation [74]. A schematic diagram of one such filter is shown in Figure 90. Figure 90 : Schematic of MEM-tunable filter The filter consists of a FP cavity with two dielectric Distributed Bragg Reflectors (DBR) similar to other cavity-type MEM-tunable filters. One DBR is formed on the fixed part of the filter, and the other is located on the actuator part. The two DBRs are separated by a spacer. Since the actuator part consists of suspended conductor bridges, it receives Lorentz force by electrical current flowing through it in a magnetic field. The magnetic field is applied in plane with the actuator. Displacement of the actuated DBR changes a cavity length and hence allows tuning. This design allows very wide tuning range in excess of 200 nm and also very low power consumption in the range of 25 µW. In general it is therefore suitable for future WDM-PON applications. Various derivatives of MEM-tunable filters have been described. One of these approaches is based on tunable microcavities [85]. It is applicable to both, VCSELs and tunable filters and has low-cost potential. The basic idea is shown in Figure 91. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 120 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 91 : Schematic of tunable microcavities The device shown Figure 91 is based on a micromachined two-chip concept with a movable GaAs DBR mirror membrane on one chip and a fixed DBR mirror on a second chip which together form the microcavity. The mirror membrane is movable due to four flexible suspension beams. Deflection of the membrane as a consequence of electro-thermal heating is achieved by injecting a small electrical current (<4 mA, equalling 10 mW of electrical power) through the suspension beams. The second chip can be a simple dielectric DBR on a plane glass substrate for filter applications (in case of a tunable VCSEL, it is one half of a VCSEL consisting of DBR plus active region). It is possible to fabricate a membrane with rotation-symmetric concave curvature in order to obtain a stable low-loss cavity. The doped mirror material itself conducts the current for electro-thermal actuation. There is no need of additional metallization. Monolithic approaches are preferable in terms of fabrication costs. However, the two-chip concept has two main advantages. The two chips can be optimized independently which is relevant for complex devices like VCSELs. It is also possible to implement a quite long cavity. For filter applications, a longer cavity automatically reduces the bandwidth, which is important for DWDM with small channel spacing. Usually, a two-chip concept has the disadvantage that assembly costs are high. Here, assembly is done by placing the membrane chip directly on top of the second chip. A permanent connection between both chips can then be obtained by bonding, gluing, or using flip-chip technique. Comparison of tunable-filter technologies An overview on the most relevant characteristics of the filter technologies described so far is provided in Table 18. Table 18. Comparison of relevant parameters of tunable filters Insertion Loss TFF-FP 3 dB LCFP 3 dB MEM-FP 1 dB FBG VHG 0.1 dB (w/o circulator) 3 dB AOTF 4 dB EOTF 4 dB O O A S OA AS SEEE FP7 – ICT– GA 249025 Ch. Spacing / Isolation 50 GHz / 30 dB 50 GHz / 30 dB 50 GHz / 30 dB 25 GHz / 22 dB Bandwidth (3 dB) Tuning Range Tuning Speed Power Consumption <0.5 nm ~40 nm ~50 nm ~60 nm <10 nm ms Medium µs Low 100 µs Low ms Medium 100 GHz / 25 dB 200 GHz / 30 dB 200 GHz / 25 dB <0.7 nm ms? Low µs Medium ms Medium <0.5 nm <0.5 nm <0.2 nm <1.5 nm <1.5 nm ~40 nm ~60 nm ~50 nm Page: 121 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Insertion Loss Cascaded MZI AWG (tunable) Ring Resonator 19 dB (LiNbO3) 5 dB 3 dB Ch. Spacing / Isolation 25 GHz / 22 dB 50 GHz / 22 dB 50 GHz / 30 dB Bandwidth (3 dB) Tuning Range Tuning Speed Power Consumption <0.2 nm <10 nm ~40 nm ~25 nm 50 ns Medium 10 ms Medium ms Medium <0.2 nm ~0.2 nm From Table 18, suitable technologies for future WDM-PON applications (WDM running via splitter-based ODN) can be derived. In this context, low cost, low energy consumption, and channel spacing and bandwidth suitable for full-band DWDM tuning (C- or L-band) should be achieved. Compactness is also required, especially since it translates, to a certain degree, to cost. From this, TFF-FP, LCFP, MEM-FP, and VHG filters can be identified as promising technologies. Note that a first FBG is commercially available, however with the stated disadvantage of limited tuning range. 5.4.2 Coherent Receivers Coherent receivers are the only alternative to (tunable) WDM filters when it comes to wavelength selectivity. In addition, they provide superior sensitivity (or OSNR performance) through the use of a local oscillator (laser). This is one of the main reasons why they are now commonly considered as means for all ultra-high-speed transmission. In the (U) DWDMPON context, coherent receivers may provide the power budgets which are required for splitter-only ODN with very high splitting ratios. Direct detection as performed by a slow (as compared to the optical carrier) photo diode leads to envelope detection where all phase information is lost. In addition, it can only make use of the responsivity of the (PIN) photo diode, or the added gain and sensitivity of an avalanche mechanism (in cased of APDs) or an optical pre-amplifier. Coherent detection, on the other hand, leads to a beat term between received input (in) and Local Oscillator (LO) signals which can be detected by the photo diode and which also preserves the phase information. The beat term is centred on the so-called Intermediate Frequency ωIF (sometimes also referred to as RF which adopts the traditional Radio Frequency view). Regarding ωIF, we must divide three cases: Homodyne, ωIF = 0 Intradyne, ωIF ≈ 0 (typically <0.5 GHz) Heterodyne, ωIF > receive signal bandwidth Homodyne fibre-optic receivers were heavily investigated in the late 1980s in order to increase transmission span distances. They require an Optical Phase-Locked Loop (OPLL) which can track the receive signal carrier’s frequency and phase. An OPLL requires substantial effort, and the respective receivers have been made redundant with the invention of the Erbium-Doped Fibre Amplifier (EDFA, also invented in the late 1980s). Today, no major work is done with regard to the traditional homodyne detection scheme. Heterodyne detection has been considered in fibre-optic transmission with regard to several applications. Since it only requires Automatic Frequency Control (AFC) of the local laser – phase tracking is performed in the electrical IF domain – it is less complex. This advantage comes at the cost of a 3-dB penalty over the respective homodyne (de-) modulation scheme. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 122 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Today, heterodyne detection is considered a possible solution for coherent UDWDM-PON. Here, it can provide channel selectivity with regard to ultra-densely (~3 GHz) spaced channels, together with very good sensitivity which becomes necessary in certain splitter-only ODN. Intradyne detection is the most recent implementation of coherent receivers. The principle was first described in [86]. It is based on tuning the LO such that the intermediate frequency ωIF becomes close to zero. Then, analogue-to-digital conversion is performed, and exact phase tracking is done in the digital domain. Advantages include the avoidance of analogue OPLL and also the possibility to implement polarization-diverse receivers and (linear) dispersion compensators (equalizers, digital filters) which lead to almost zero penalty with regard to polarization and (chromatic) dispersion effects. Digital intradyne detection is now the intended quasi-standard solution for all future ultra-high-speed transmission. In coherent detection, the detected power consists of the contributions from the input and LO signals, plus the beat term between the two which is caused by the square-law detection: Pt Pin PLO 2 Pin PLO cosRFt in LO Only the beat term is used for demodulation. It allows coherent phase demodulation (because the cosine argument contains the input signal’s phase), and high sensitivity (because the term is weighted with the square root of the LO power). After down-conversion, the (intradyne, IF) signal is low-pass filtered (LPF) and demodulated according to the modulation scheme used. A schematic diagram (simplified, since it does not yet consider polarization diversity) is shown in Figure 92. I t R Pin PLO 2R Pin PLO cosRFt in LO E in (t ) + Photo Diode Detection + Shot Noise E LO (t ) LPF Envelope Detector Homodyne Out Heterodyne Out Electrical Amplification + thermal Noise Figure 92 : Coherent detection (basic scheme) A coherent (digital) intradyne system usually is more complex than indicated in Figure 92. This is due to the necessity of a polarization-diverse receiver which takes into account that the polarization planes of the input (receive) signal usually are not known. Hence, input and LO signal must be split into orthogonal polarizations which are then mixed and detected in independent receivers. The resulting system configuration can be found in the literature and is shown in Figure 93for coherent QPSK transmission [87]-[93]. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 123 of 168 90° QPSK Coder Driver Filter LO PC 90° 90° Hybr. 0° Client I/F (CFP) PBS 90° FEC, Framing, Monitoring PBC PC CDR, EQ., Carrier Rec., Decoder PC PBS CW LD 0° ADC 90° 90° Hybr. ADC Driver Filter ADC FEC, Framing, Monitoring Client I/F (CFP) QPSK Coder ADC Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 93 : Coherent intradyne DP-QPSK transmission The system shown in Figure 93 makes use of polarization multiplexing. The transmission format is hence called Dual-Polarized QPSK, or DP-QPSK. Similar schematic system diagrams result for other modulation schemes, but today much focus is being put on QPSK due to its superior OSNR characteristics. For polarization multiplexing, the transmit laser (CW LD) is split into two orthogonal polarization planes by means of a combined polarization controller / beam splitter (PC PBS). Both polarization signals are independently modulated and re-combined by a polarization beam combiner (PBC). At the receiver, the input signal is first split into orthogonally polarized signals by another PBS. The LO signal is also split and then combined with the input signal by means of two 90° hybrids (these are basically passive combinations of 3-dB couplers and 90° phase shifters). Each 90° hybrid has dual output ports for the respective inphase and quadrature components. In order to avoid OSNR penalties, the output signals are detected by four balanced receivers, i.e., a total of 8 photo diodes. The photo diodes are followed by fast Analogue-to-Digital Converters (ADC) which are then followed by the digital receiver. The digital receiver consists of clock data recovery (CDR), equalizer (EQ) for dispersion compensation (and possibly nonlinear impairment mitigation), carrier recovery, and finally decoding (here, QPSK decoding). This is followed by means for framing, monitoring, and error correction (FEC). The interface towards the application is provided through the client interface (I/F) which, in the case of high-speed transport, is implemented as a CFP pluggable. In case of low-speed (residential access) services, the client interface could consist of an optical or electrical SFP instead. Input PBS 90° LO PBS 90° I+(TE) I-(TE) Q+(TE) Q-(TE) I+(TM) I-(TM) Q+(TM) Q-(TM) Signal Ex i E y j PIN PIN PIN Elo PIN ImE E ReE E ImE E Re Ex Elo* x * lo y * lo y * lo Figure 94 : Two realizations of 90° hybrids in polarization-diverse coherent receivers For ultra-high-speed transmission, the intradyne system requires ultra-fast ADC, which is a significant contributor to cost. In order to decrease the symbol rate and consequently the sampling rate of the ADC, polarization multiplexing is used which increases the transmitter complexity and cost. This can be reduced in (coherent UDWDM-PON) access systems. Therefore, digital intradyne must be considered an attractive way of implementing a costeffective coherent receiver for UDWDM-PON. The digital receiver is described hereinafter in more detail. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 124 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 95 shows the digital receiver of the transmission system of Figure 93 in detail. It consists of several stages where different functionalities are implemented [87], [88]. Cubic Interpolator M Y90° Y0° M |x|2 56GS/s A/D X90° X0° X 1 arg 2 Y-X exp(-j2fst) X-Y Clock Recovery Y Equalization Decider, Decoder YI XQ 1 4 X4 Carrier Recovery YQ Adaptation Algorithm X-X arg . c0 c1 c2 c3 c4 Y-Y Decider e-j(k) Delay XI Figure 95 : Coherent intradyne receiver: digital realization After detection and sampling at approximate Nyquist rate, the inphase and quadrature components have orthogonal but arbitrary planes of polarization each. The four components are now digitally processed in several succeeding filter stages. In a first stage, clock recovery is performed. Next, equalization of chromatic and polarization-mode dispersion is provided. This filter stage also allows adjustment of the arbitrary receive-end polarization planes to those of the transmit signal, i.e., proper polarization demultiplexing is enabled. The last filter stage performs carrier recovery, i.e., the digital phase tracking. The clock recovery is based on a digital filter-and-square timing recovery. It also performs resampling to 2 samples/ symbol. Ideally, the ADC would already sample at an exact multiple of the Baud rate. More details on digital timing recovery can be found in [89]. It is worth noting that some references state a somewhat different structure of the digital filter stages, for examples refer to [106], [107], [108]. Equalization is often split into two stages. In the first stage, Chromatic Dispersion (CD) compensation is done. This stage can trial any other filter stage, including clock recovery. It is based on a simple n-tap Finite-Impulse Response (FIR) filter. Here, the sampled input data is convoluted with a vector which represents the inverse of the CD transfer function. An example of a simple 7-tap FIR filter is shown in Figure 96. 0.6 0.4 0.2 0 -0.2 -0.4 -0.6 -0.3 -0.2 -0.1 0 time (ns) 0.1 0.2 0.3 Figure 96 : Example of simple FIR filter response (blue: Re, red: Im part) According to [92], the upper bound for the tap number at a Baud rate of B GBd is given by 0.032∙B2 per 1000 ps/nm of chromatic dispersion. At 10.7 GBd (43G DP-QPSK), this O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 125 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 translates to 3.7 taps per 1000 ps/nm which can reduced to 2.2 taps per 1000 ps/nm with minimal penalty. For 28GBd (112G DP-QPSK), the number of taps increases by a factor of almost 7 as compared to 42.8 GBd (CD increases with the square of the Baud rate). PMD equalization and polarization recovery (demultiplexing) is performed by four complexvalued FIR filters. (These are also referred to as Feed-Forward Equalizers, FFE, transversal filters, or tapped-delay-line filters.) These filters are implemented in a butterfly structure, as multiple-input/multiple-output adaptive equalizer between the polarization planes, see Figure 95. The filter taps are spaced T/2 in time, where T is the symbol duration. Their tap number correlates with the target dispersion-compensation capability. The tap weights have to be optimized, usually using blind equalization techniques in order to avoid the necessity of training sequences. Blind adaptation can be done, for example, with a Least Mean-Square (LMS) algorithm. The earliest work on such adaptive filters can be traced back to the late 1950s. From this early work, the LMS algorithm emerged as a simple, yet effective, algorithm for the design of adaptive transversal filters. The LMS algorithm was devised by Widrow and Hoff in 1959 [96]. It is a stochastic gradient algorithm in that it iterates each tap weight of the transversal filter in the direction of the instantaneous gradient of the squared error signal with respect to the tap weight in question. An alternative to LMS is the so-called Constant-Modulus Algorithm (CMA) [93],[94]. One of the most important features of CMA is that it can equalise constant-modulus signals (such as QPSK, 8PSK) as well as non-constant-modulus signals (like 16QAM). CMA seeks to minimize a cost defined by the CM criterion. The CM criterion penalizes deviations in the modulus (i.e., magnitude) of the equalized signal away from a fixed value. In certain ideal conditions, minimizing the CM cost can be shown to result in perfect (zero-forcing) equalization of the received signal. CMA employs a cost function that does not discriminate between the two equalized signals. Hence, it is common that this algorithm converges to a tap-weight setup that produces the same transmitted signal at both equalizer outputs, usually the one that arrived with higher power at the receiver. The equalizer matrix in this case becomes singular. This problem can be circumvented by frequently monitoring the equalizer’s matrix determinant and reinitializing the tap-weights when singularity is approached, a solution that is practical, but could cause discontinuity issues at the equalizer output. A computationally demanding equalization algorithm based on the independent component analysis method has also been proposed. A more recent improvement is multiuser CMA, which is an extension of the conventional CMA to the multiple input case, such that singularity is avoided. Multiuser CMA employs an enhanced cost function, penalizing the correlation between the two signals at the equalizer’s outputs [95]. Frequency and phase offset between local laser and receive signal is corrected by a 4 th-power, Viterbi-and-Viterbi carrier recovery stage [99]. The approach is to calculate an optimal phase estimate, and to implement this on a parallel DSP. The best-possible phase estimate is derived through the Maximum a Posteriori (MAP) estimate where phase (n) and data d(n) are jointly estimated [101]. The MAP estimate can be calculated by applying a per-survivor method to a group of symbols, and calculating the phase by successive approximation for each symbol group instance. Neglecting higher-order noise terms, small-angle approximation is applied which yields a phase angle θ = 2 + (additive noise component). Estimation theory says that the best linear estimate of is derived if a Wiener filter applied to θ. The Wiener filter solves the signal estimation problem for stationary (or cyclo-stationary) signals. It was introduced by Norbert O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 126 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Wiener in the 1940s [102]. The filter is optimal in the sense of the Minimum Mean-Square Error (MMSE). (The Kalman filter solves the corresponding problem in greater generality for non-stationary signals.) A Wiener phase estimator is shown in Figure 97. Complex Signal d ei+p ( )2 ei2+p’ arg( ) Wiener Filter ÷2 exp( ) Phase Estimate Figure 97 : Phase estimation using Wiener filter The phase is estimated and applied to the signal before making the 1/0 or multi-level decision. A smoothing (low-pass) function is needed to reduce effects of additive noise and pass the actual phase change. Errors in the phase estimate lead to an increase in number of bit errors. They can lead to so-called cycle-slip errors, i.e., data inversion in case of BPSK or QPSK. This is the reason why even in coherent intradyne systems differential pre-coding (i.e., DQPSK) is applied although it is not necessary from the viewpoint of differential delay demodulation using self- or incoherent MZM. Figure 98shows the effects of phase noise and phase plus added amplitude noise on BPSK, respectively. No Noise Phase Noise only Phase + Amplitude Noise Figure 98 : Phase estimation in presence of phase noise and combined phase plus amplitude noise One key function is to recover the carrier phase using DSP-based phase estimation (PE) rather than optical phase-locked loops, thus allowing for a free-running LO laser. Some popular phase estimations, such as Mth-power [99], require that the frequency offset between transmitter and local oscillator (LO) laser should be quite small compared to symbol rate. The frequency offset between transmitter and LO lasers, however, can be as large as ±5 GHz. As a result, an additional DSP-based Frequency Offset Estimator (FOE) is required to ensure that subsequent PE algorithms can accurately recover the phase of received signals. A feed-forward FOE is preferred to avoid performance degradation when being implemented in parallelism. Mth-power is generally performed to remove data modulation in feed-forward FOEs. The maximal estimation range is limited to [-Rs/2M, Rs/2M], where Rs refers to the system symbol rate and M is the number of constellation states. This leads to an estimation range is ±0.125 Rs for QPSK. In a standard intradyne receiver, the frequency offset between the receive signal and the local oscillator lasers should be no more than 10 percent of the symbol rate [98]. For 43G DPQPSK transmission, this leads to ~1 GHz allowable frequency offset. In order to allow larger offsets, more tolerant Frequency Offset Estimators (FOE) are required. In [103], an ultrawide-range FOE has been described. It is based on a double-stage FOE architecture as shown in Figure 99. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 127 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Re-sampling to 2 sample/symbol I (k ) & Q(k ) Resampling Down-sampling to 1 sample/symbol jk fˆcTs e e CMA ↓ Delay NCO NCO MPE Sweeper Coarse FOE Look-Up Table Sign Identifier jk fˆ f Ts fˆc FFT-based FOE fˆ f 1/-1 Fine FOE Figure 99 : Cascaded FOE [103] The dual-stage, cascaded FOE consists of a coarse FOE and a fine FOE. In 43G DP-QPSK systems, the estimation range of the dual-stage FOE has been shown to be 4 times what can be achieved with the Mth-power algorithm. This is the largest range of FOEs reported so far in the literatures. Timing recovery, such as Gardner algorithm, is required to correct for the timing phase error between the transmitter and receiver clocks in coherent receivers. The simple Gardner algorithm can be used to generate a phase error output when only two samples per symbol are available. However, the performance of the Gardner algorithm suffers from performance degradation in the presence of frequency offset. Such offset makes the Gardner algorithm less insensitive to the sampling offset in coherent receivers. The relationship between frequency offset and MPE of the Gardner algorithm can be used as a measure to estimate the frequency offset. The absolute MPE varies with OSNR and is normalized to the one at zero frequency offset (FO). Then, the normalized MPE show almost the same trend within ±9 GHz for different OSNR. A polynomial fit is applied to the MPE/FO relationship. This polynomial fit can offer a coarse FOE (ΔfC) which is limited to ±1 GHz. This is within the estimation range of the Mth-power algorithm which is cascaded as fine FOE in a second step (with ΔfF). The polynomial fit of the MPE/FO relationship can be realized using a look-up table. Cascaded FOE is capable of estimating a frequency offset up to the range [-0.5Rs, 0.6Rs], which is approximately 4 times the theoretical limit of single FOE using Mth-power. Ultra-narrow linewidth lasers are required to implement a phase-locked loop for carrier recovery [86], and these are widely believed to be too expensive in today’s cost-sensitive economic environment. In contrast, a feed-forward carrier recovery scheme relaxes the sum linewidth requirement to about 0.001…0.0001 times the symbol rate, which is in the reach of normal, low-cost DFB or DBR lasers. Phase noise is an important impairment in coherent systems as it impacts carrier synchronization. In non-coherent detection, the carrier phase is unimportant because the receiver only measures energy. In DPSK, information is encoded by phase changes, and Δν only needs to be small enough such that the phase fluctuation over a symbol period is small. The receive signal is modulated by ejφ(t). In the absence of other impairments, this leads to a rotation of the received constellation. Carrier synchronization is required to ensure φ(t) is small so the transmitted symbols can be detected with low power penalty. (φ(t)= φS(t)- φLO(t)) With the phase error variances for FF carrier synchronizer, the power penalty can be determined. In Table 8, we compare the linewidth requirements for receivers that use FF carrier synchronizer, assuming a 1-dB power penalty at a target BER of 10−3 [105]. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 128 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Table 19 : Linewidth requirements for single-polarization modulation at a target BER of 10 −3 QPSK 16-QAM OSNR per bit [dB] 7.79 11.52 Max. σφ Max. [ΔνTb] Linewidth 4.91° 1.3×10−4 2.70° 1.5×10−5 In OFDM, laser phase noise destroys the orthogonality of the sub-carriers and causes InterChannel Interference (ICI), which has noise-like characteristics. ICI grows linearly with NC, with NC the number of sub-carriers. Thus, an important parameter is the linewidth-to-subcarrier-spacing ratio ΔνTC/NC. Phase noise considerations favor a smaller number of subcarriers. At higher phase noise, the carrier phase can no longer be assumed as constant over an OFDM symbol. The use of an MMSE equalizer which considers ICI and AWGN can improve system performance [Ip-111]. However, this technique does not estimate or compensate for phase noise, unlike feed-forward carrier synchronization for single-carrier modulation. Carrier synchronization in OFDM may require an iterative algorithm. Suppose at the beginning of an OFDM symbol, an initial estimate of phase kφ ~ is known from the previous symbol. We can first de-rotate the entire OFDM symbol (NC chips), and then perform the FFT, equalization and symbol detection. We can multiply the symbol decisions by the channel’s frequency response and then take the IFFT to compute what the time samples should have been without phase noise. This allows the receiver to compute estimates of the carrier phase for each chip period. MMSE filtering can then be employed to find more reliable phase estimates. We can de-rotate the OFDM symbol again, and a second iteration follows. This process can be repeated until convergence is achieved. The performance of such an algorithm has not yet been characterized. In DSP-based coherent detection, Nonlinear Compensation (NLC) allows the mitigation of nonlinear intra-channel effects. This is possible since these effects – self-phase modulation – are deterministic and can hence be compensated. Compensation of inter-channel effects (XPM, FWM) would require a much more complex MIMO equalizer. NLC can be implemented, together with chromatic-dispersion compensation, as a fixed equalizer which trials the adaptive (CMA) equalizer [106]. In order to reduce computational effort, it is desirable to compensate CD with a frequency-domain filter, followed by timedomain instantaneous NLC which compensates the nonlinear phase shift. In multi-span longhaul transmission, linear and nonlinear intra-channel compensation can be done iteratively, switching between time and frequency domain by means of FFT and IFFT, respectively, see Figure 100. ADC Resampling Linear only Dispersion Compensation FIR Filter Fixed Compensation Disp. Compensation FOE IFFT N Spans FFT CMA Equalizer Carrier Phase Est. NL Phase Shift Complex Decisions Figure 100 : O O A S OA AS SEEE FP7 – ICT– GA 249025 Linear and Nonlinear Linear-only or linear plus nonlinear compensation [106] Page: 129 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Using intra-channel NLC, the optimum long-haul launch power can be increased by ~1 dB, and maximum reach for 42.7 Gb/s and 85.4 Gb/s can be increased to more than 11,000 km and 8000 km, respectively [106]. An alternative homodyne / intradyne detection scheme which avoids both, polarization diversity and the necessity for 90° hybrids, has been proposed in [110]. The concept was designed for low-cost applications, for example in coherent ultra-dense WDM-PON systems but is not restricted to such applications. Polarization diversity and 90° hybrids are used in coherent detection in order to avoid polarization control (of both, the receive signal and the local laser) and to provide the Inphase and Quadrature components of the down-converted signal, respectively. A fully-blown polarization-diverse I/Q demodulator requires 4 branches with 8 photo diodes if balanced receivers are used, refer to Figure 93. In the PON context, this leads to prohibitively high cost. Avoidance of polarization diversity / control and 90°-hybrid-based I/Q demodulation according to is based on polarization and phase scrambling. The idea is to have the orthogonal states for phase and polarization in the same bit sequentially. Thus, for the first half of the bit the signal relative to one orthogonal component (I or H) will appear, whereas for the second half of the bit the signal relative to the other orthogonal component (Q or V) will be seen. The principle diagram is shown in Figure 101. Downstream + 3dB Tb Pol. Scr. PM Tb/2 Tb/2 Freq. Doubler RZ50 CLK Rec. Upstream 3dB Q H Data Figure 101 : LD t0 Q V I H I V Q H t0+Tb Q V I H I V t0+2Tb t Data Homodyne detection without polarization diversity and 90° hybrids Within each symbol duration Tb, the phase of the local laser is scrambled by means of an RZ50 (Return-to-Zero with 50% duty cycle) signal which is synchronized with the receiver clock recovery (CLK Rec.). The RZ50 signal generates, within each symbol interval, two time slots of duration Tb/2 with fixed 0° to 90° phase modulation. One slot represents the I component, the second slot represents the phase-shifted Q component. This is demonstrated in the insert in Figure 101 (together with the sub-slots resulting from polarization scrambling). By gating the photo receiver output with the data clock and its inverse, the I and the Q components are obtained separately in the two branches, now with RZ shape. The signal power fluctuates between the I and the Q branches randomly, due to phase noise, at a rate of the order of the laser linewidth, and the combination of both outputs can assure its recovery. This operates like a phase-diversity system. The local laser does not need to be phase coherent with the incoming optical carrier, although an automatic wavelength controller is convenient to maintain the two wavelengths close each other. This can be regarded as an intradyne receiver with near-zero intermediate frequency. As such, dual-fibre working is required in order to avoid downstream/upstream cross talk. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 130 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Using delay demodulation in the electrical domain, this scheme is suitable for detecting PSK. (Specially in the PON context, this is no restriction. PSK at 1 Gb/s per channel gives sufficient bandwidth efficiency to allow for more than 1000 bi-directional channels.) Electronic signal processing performs differential demodulation of both I and Q components, with a delay time equal to Tb, and the synchronous combination of the I and Q components. In Figure 101, polarization is scrambled as well. The polarization scrambler is run at duplicated symbol rate in order to produce the orthogonal polarization states per I and Q time slot. This now leads to 4 time slots of duration Tb/4 which contain all combinations of the orthogonal states of both, phase and polarization. This has been shown in the insert in Figure 101 already. This also means that part of the receive electronics must operate at an increased bandwidth. Several derivatives of the sequential phase / polarization diversity demodulator have been proposed in [110] and further work by the same authors. These mainly aim at providing different detection schemes with different performance (or penalty against a fully-blown digital polarization- and phase-diverse intradyne receiver), and at different hardware effort. One such configuration is shown in Figure 102. Downstream + 3dB Tb/2 Tb PM VQ VI RZ50 Tb Upstream 3dB CLK Rec. LD I Data t0 Figure 102 : Q t0+Tb I Q Data t0+2Tb t Reduced homodyne detector The receiver shown here uses phase diversity by means of scrambling only. This configuration assumes that polarization switching is performed in the OLT. Given that synchronization between the polarization scrambler and all downstream channels can be achieved in the OLT, the polarization scrambling can be done on the optical multiplex section in a single centralized component, thus reducing cost. Note that the same will not work for the upstream direction since synchronicity between the upstream channels can not easily be achieved due to uplink length differences. For cost efficiency and in order to achieve optimum performance, the configuration shown in Figure 102will be implemented as a digital intradyne receiver. The resulting configuration is shown in Figure 103. This is again the reduced variant without dedicated polarization scrambling. Like the other variants discussed in here, it requires two fibres for downstream and upstream. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 131 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Downstream + 3dB Upstream 3dB Data ADC RZ50 PM I/Q Proc. Data CLK Rec. LD Figure 103 : Digital implementation [110] The disadvantage of dual-fibre working can be avoided by using a heterodyne detection scheme instead of (digital) intradyning. Then, the local laser is tuned to an offset against the receive wavelength which is larger than the payload bandwidth. This offset (the intermediate frequency in heterodyning) allows to directly re-use the laser wavelength for upstream on he same fibre. This scheme also does not require 90° hybrids since I/Q processing is done in the electronic (RF) domain. In addition, it can be combined with polarization scrambling, which is shown in Figure 104. Downstream 3dB Data + 3dB RF Detection Upstream 3dB Pol. Scr. CLK Rec. LD Data Figure 104 : Polarization-scrambled heterodyne detection The coherent receivers intended for UDWDM-PON discussed so far can reduce cost, as compared to a fully-blown intradyne receiver. This cost reduction, however, comes at a certain signal penalty. Neglecting further implementation penalties (which is valid for digital realizations), phase scrambling leads to a penalty of 3 dB, as compared to homodyning or intradyning using 90° hybrids and twice the number of photo diodes. Polarization scrambling leads to another 3 dB penalty, compared to polarization diversity. Heterodyning also leads to 3 dB penalty as compared to homodyning / intradyning, given the modulation scheme (here: PSK) stays the same. Altogether, up to 6 dB penalty (for digital implementation) can result. This is summarized in Table 20. Table 20. Comparison of coherent WDM-PON schemes Phase Handling 90° Hybrid 90° Hybrid Scrambling Heterodyning Polarization Handling Diversity Scrambling Scrambling Scrambling Sensitivity Penalty 0 dB 3 dB 6 dB 6 dB Linewidth Tolerance 5 MHz 5 MHz 5.4 MHz 5 MHz Cost Estimation High Medium-High Medium Medium With 6 dB added penalty, a sensitivity of better than -45 dB can easily be achieved, assuming that PSK is used at a bit rate of 1 G/s. This can be considered enough for PON applications, even at long access distances and high customer numbers (splitting ratios). The question however remains if the various coherent receivers reduce complexity and cost sufficiently in order to allow large-scale access application. Even the polarization and phase scrambling requires substantial effort. For polarization scrambling, an additional component (the scrambler) is required. For phase scrambling, an additional phase modulator, together with the O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 132 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 synchronized I and Q delay demodulators running at duplicated bit rate are required. 5.5 BURST MODE RECEIVERS Burst Mode Receivers (BMRs) are a key component in next-generation optical access packetswitched PON networks. Located within the central office (CO) BMR designs have to overcome two major challenges: maintaining data recovery with large power fluctuations between adjacent packets and rapid (ideally instant) clock recovery on packet arrival. These variations in incoming data packets are due to the different paths that the bursts follow through a PON network. Moreover, additional issues such as poor extinction ratio, noise accumulation and random burst arrival times also have to be considered when designing an appropriate BMR. In addressing all these challenges, future BMR design also needs to deliver fast-response peak-detection to deal with threshold variations and nanosecondswitching. The system additionally needs to cope with variable gain transimpedance amplifiers; preset minimum guard time between consecutive packets and a timing–alignment preamble in front of each packet, so as to adapt the power intensity and bit phase of each incoming burst to a reasonable level able to be detected. The variation in power between adjacent packets is characterised by the loud-soft ratio (see Figure 105) with the associated extinction ratio parameter also indicated in the figure. 1 ONU 1 2 ONU 2 OLT BMR n Loud/soft ratio ONU n 1 n Extinction ratio 2 Figure 105 : Typical Passive Optical Network (PON) scheme with a BMR located in the OLT The maximum dynamic range of the loud/soft ratio is typically up to the order of 20 dB, with a sensitivity of -28 dBm. Recent BMR designs, e.g. [111] have achieved a receiver sensitivity of -30.8 dBm at 10G, and -35.5 dBm at a bit rate of 1.25 Gb/s. Conventional ON–OFF keying (OOK) receivers also experience a sensitivity penalty associated with the adaptive decision threshold used to handle large inter-packet amplitude variations. Fortunately, differential coding schemes employing balanced detection, such as differential phase-shift keying (DPSK) demonstrate a large tolerance to signal power fluctuations, fibre nonlinearities and polarization mode dispersion [112]- [113] making DPSK coding very attractive for optical packet/burst switching. Two major issues related to clock recovery in NRZ-DPSK BMR design are firstly: acquisition of a clock from the NRZ signal which as a rule doesn’t contain a clock frequency component [114] and secondly, to maintain this clock signal during guard times between the bursts/packets. For a BMR with adaptive decision thresholds, determination of the required decision threshold needs to be rapid, i.e. measured in the order of a maximum of tens of nanoseconds. Recent examples of 10G BMR designs for EPON applications achieve lock in times of 37 ns at 10.3 Gb/s, and 64 ns at 1.25 Gb/s [115]; an overhead of 280 ns at a bit-rate of 10.3 Gb/s has been demonstrated in [116], and 10 ns O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 133 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 response time for a 10G BMR has also been reported by Nakamura et al. [117]. Clock Recovery & Timing Extraction As already discussed above, rapid (as near instant as possible) clock recovery on packet arrival is essential for successful BMR receiver operation. A phase lock loop (PLL) finds widespread application in many BMR designs to recover synchronisation from a randomly arriving packet, e.g. [118] In order to achieve rapid acquisition of timing information and quick lock-in the PLL must feature a wide bandwidth of operation. However, a wide bandwidth associated with the various PLL multiple feedback paths (both linear and nonlinear in operation) means that a lot of noise is also admitted into the PLL, so leading to high jitter levels. This means that PLL-based clock recovery has difficulty in simultaneously achieving both swift and accurate lock-in times. An alternative method for rapid synchronisation with a low timing latency is based on the use of passive high-Q resonance filtering, e.g. [119]. Such a resonant circuit based approach is highly scaleable, being equally appropriate for clock frequencies from as low as the sub-GHz regime up to the multi-100’s of GHz range. A high-Q resonant circuit, where the Q value is greater than 1000, means that 1000 clock cycles have to elapse before the amplitude of the resonant response exponentially falls below the 1/e level, and de-synchronisation occurs. This indicates the stability of the high-Q resonator approach - in particular, from classical filter and signal theory, the phase response is closely related to the amplitude response of the resonator (i.e. they form a Hilbert transform pair) such that phase and amplitude stability are mutually ensured. For a Q>1000 resonator, provided that the intervals between incoming packets don’t exceed, for example, 1000 bits, then once the circuit is already resonating, any phase misalignment of the arriving packet is passively corrected for within as few as 3 bits, e.g. [119]. The advantages of this passive resonant high-Q scheme compared to PLL-based schemes are the simplicity and cost efficiency and the phase acquisition time (at the centre frequency) which are both critical for operation in a BMR. Current techniques to BMR design can be divided into two main categories: DC and AC coupling. Both techniques require a minimum guard time between two consecutive packets and a preamble for adjustment in front of each packet, which results in a reduction in the actual efficiency of channel capacity exploitation. It’s also worth noting that the length of the guard band is ambiguous due to the lack or no knowledge of the arrival of the subsequent packet. Within the overall concept of AC coupling, is another important emerging BMR technique based on edge detection. This is emerging as an important technology to achieve the ever-higher burst mode bandwidths and speeds that are a feature of next-generation optical access networking. In the following, we consider these three main approaches, discussing their strengths and weaknesses, as well as their appropriateness to meet the needs of future NGA optical networking. 5.5.1 AC Coupling AC coupling technique removes the input DC bias and biases the signal at the average signal power level, with no feedback and decision level setting requirements required, which makes the AC coupling technique relatively easy to implement. As such it is a well-established technology in the field, being cheap to implement, and requiring no feedback or decision level setting. Reference [120] describes an AC-coupled BMR proposed for Gigabit Ethernet PON systems. Conventional AC coupled receivers have long time constants (typically of the order of ms), which, however, is rather too long as bursts become shorter; i.e. a long time constant exceeds the burst duration. Smaller capacitors at the front end can reduce the time constant, but tend to filter out low frequency components of the signal, resulting in ISI. As such, line coding (e.g. 8B/10B) is required to reduce long strings of consecutive identical digits (CIDs). Scrambling is an alternative to line coding with its associated redundancy, but requires a O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 134 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 longer time constant (and preamble) since the protection against long CIDs is still only on a statistical basis. In AC coupled BMRs, long CIDs cause level drifting and clock and data recovery (CDR) malfunction, in which case the time constant needs to be longer than the longest CID acceptable for a given CDR and BER. Additional requirements for AC-coupled systems are generally determined by the acceptable BER for the system. For example, the BMR needs to settle down in less than the preamble in order to receive the data within the required BER, whilst the signal also needs to be held for longer than the maximum CID associated with the acceptable BER. By way of example, at 2.5 Gb/s the Guard Time (64 bits) is 25.6 ns, whilst the Preamble (108 bits) lasts 43.2 ns, and the conventional maximum CID is 72 bits corresponding to 28.8 ns. Assuming an even mark-space ratio for the data, the AC threshold is conventionally set to the midpoint. A large change in burst amplitude requires a finite settling time during which data will not be received. For example, the maximum time constant is determined by the settling time between loudest and softest bursts, in particular settling to within the upper threshold level. The minimum time constant is determined by the maximum CID period whilst remaining within the upper threshold level. Unfortunately reducing the maximum CID period is difficult, e.g. reducing from a maximum of 72 bits to a maximum of 12 bits is desirable. That said, maximum CID length is often built into most physical layer formats, e.g. GigE employs 8B/10B encoding which ensures no than a maximum of 6 CIDs. However, even a well chosen time constant will still result in some ISI due to the relative filtering effects at lower frequencies, with the situation worst for the soft to loud case. Achieving such performance is becoming increasingly difficult in AC-coupled BMRs, particularly with GPON and LRPON specifications, although a 10G BMR AC-coupled design has recently been demonstrated by NTT [121] featuring a 10.3 Gb/s data rate, with 1001 bits of CID tolerance. 5.5.2 DC Coupling In a DC coupled BMR, the decision threshold changes according to each burst amplitude, so that the drifting affects associated with AC-coupled schemes due to varying average power levels are avoided. In this case, the decision threshold level is determined using only the amplitude of the current bursty traffic, without any memory effects. Unfortunately, although it avoids the time constant issues of the AC-coupled approach described above, implementation of the DC-coupled technique within practical systems is more complex proposition compared to AC coupling. Reference [122] describes a DC-coupled BMR receiver which employs peak detectors to extract a decision threshold from a sequence of 12 successive non return-to-zero (NRZ) 1’s and 12 successive NRZ 0’s received at the beginning of each packet, whilst a DCcoupled 1.25 Gb/s burst-mode receiver with automatic offset compensation has been demonstrated in [123]. A DC-coupled BMR allows a feedforward architecture to be employed in the basic front end design, with amplitude recovery performed in a differential pre-amplifier setup. Such a feedforward design allows a faster settling time between bursts, and is important for reset functionality; however, feedforward is inherently less stable than a feedback-based design and more prone to oscillation. It also makes the design more complicated and expensive as compared with the AC-coupled version. A specific design consideration for a DC-coupled BMR is that the peak detector may need to detect both high and low levels to prevent markspace distortion, especially when the extinction ratio is poor. A fast reset is also required in order to recover from bursts arriving within the guard period – in this case, a feedforward design will provide the required speed. For systems employing fixed packet length formats (e.g. ATM) a fast reset is less of a problem, but variable burst length standards such as O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 135 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Ethernet (which are generally now the preferred option) present more of a problem and require fast reset functionality. Particular issues with the DC-coupled BMR approach include the need to reset the threshold level with each packet, and the need to reset in less than the guard time period. Decision settling is not assisted by the guard period duration, and must be completed within the preamble time (e.g. 43.2 ns for a GPON system at 2.5 Gb/s). Feedback control is also still necessary to reduce pulse width distortion (PWD) which increases the BER due to the reduced sampling duration (time setting) and may also lead to CDR malfunction since the CDR assumes equal ones/zeroes pulse widths. A PON compatible DC-coupled BMR developed in the PIEMAN project has been reported in [124] and is capable of operating at 10 Gb/s, featuring a guard time as short as 25.6 ns, a preamble of only 23.8 ns, and able to withstand up to 72 CIDs. NTT have recently demonstrated a 10G DC-coupled BMR featuring a 74 ns lock-in time at a sensitivity of -18 dBm [125][126]. 5.5.3 Edge Detection In an attempt to address the inherent problems associated with DC and conventional ACcoupled schemes, edge detection is emerging as an attractive solution, overcoming many of the key technical issues in BMR design. Here, advantage is taken of the short time-constant differentiation process to produce delta functions of alternating polarity at the start and finish of a bit. Once the edges have been detected, a high-speed comparator discriminates the received mark/spaces. However, this come at the cost of a still more complex design compared to the DC and AC coupling approaches, requiring the optimisation of a range of different sub-system parameters: differentiator constant, comparator holding time, RC constant and the impact of consecutive bits. Clock acquisition is achieved by introducing a low-level temporal nonlinearity in the DPSK demodulator, thus generating a spectral component at the clock frequency. This allows data recovery and clock acquisition to be achieved from the same data stream. The clock is then extracted in the second stage by using an elegant high-Q cavity resonator filter technique. Clock recovery using the phase-locked loop technique has also been demonstrated at a bit rate of 2.5 Gb/s [127]. Experimental results supported by simulations demonstrate the validity of the edge-detection technique for high speed (>10 Gb/s) [119] optical PON networks. These results show the advantages of the edge-detection techniques over the other techniques and its main limitations to be used in future PON networks. Figure 106 : a) mass distribution function of differentiated signal level sampled immediately after the start of each bit b) mass distribution function of differentiated signal level sampled after passing MLEPW(set to 0.05 of bit length) after the start of each bit. SNR for both scenarios was set to 9 dB. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 136 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Specific issues for edge detector BMRs related to the optimisation of the above-mentioned system parameters include: • The pulse must exceed the comparator threshold voltage prior to the decision time • The noise present on the signal must not trigger a false reading following the first trigger and prior to the decision time In addition, multiple bits provide a special problem for this receiver design because if bit m from a sequence of n bits triggers a false level on the comparator then the subsequent n-m bits will be in error as well (unless another positive error is triggered). However, as analysis (described below) indicates, the overall BER penalty is actually quite minimal. The basic principle of hysteretic edge detection is shown in Figure 106 and Figure 107. We assume that the hysteresis thresholds are set optimally so that as soon as the differentiated signal exceeds the hysteresis thresholds the BMR receiver responds to it. In practice, there is a minimum holding time for the signal to be stable so that the receiver doesn’t miss it. Figure 106(a) highlights the well-known base-line variation inherent in conventional AC-coupled BMRs. It is clear that long sequences of 1’s and/or 0’s will remove the decision reference threshold. If the AC-coupling time constant is reduced to below 20% of the bit duration, then inter-pulse influences are reduced asymptotically to zero as shown in Figure 106(b). The information contained in the original data sequence may be recovered completely by noting that the short time-constant differentiation process eventually produces delta functions of alternating polarity at the start and finish of a bit. Standard signal theory shows that an integrator can then be employed to remove the differentiation, with such functionality contained in a latest-generation SiGe comparator (e.g. Analog Devices ADCMP580 series). As shown in Figure 107, a positive impulse sets the comparator output to high with this condition persisting indefinitely until a negative impulse resets the output to low. By such a means an exact regenerated copy of the input pulse stream is produced. Figure 107 : Mass distribution function of differentiated signal level sampled immediately after the start of each bit. SNR= 7 dB. Ideally a small value of RC constant for the differentiator should result in improved performance by avoiding the effects of baseline drift. In other words, a smaller differentiator RC constant allows the receiver to respond to the changes more quickly, while at the same time the capacitor can be discharged faster to avoid the possibility of base line drifting. However the maximum speed of the comparator puts an upper limit to the differentiator RC constant, i.e. a small RC constant results in short pulses that may not be captured by the comparator. In this case, the Minimum Latch Enable Pulse Width (MLEPW) is the minimum time that the latch enabling signal must be high in order to acquire an input signal change. If O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 137 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 the signal level decays below the threshold level before this minimum holding time has passed, then the comparator may miss the pulse. Such an event has a dramatic effect on BER performance since if a pulse is missed, provided that the subsequent bit has the same level (i.e. no value change) the subsequent bit will be also received in error until an edge is detected. The value of MLEPW depends on the capabilities of the comparator, but a higher value of MLEPW results in inferior BER performance. Figure 108 shows the graph of BER versus the signal SNR for different values of MLEPW. The figure clearly shows the effect of increasing the value of MLEPW. Figure 106 gives a better understanding of the role of MLEPW. Figure 106(a) shows a distribution of the differentiated signal level immediately after the beginning of each bit. To show the mass distribution function more clearly, the input signal level is assumed to be 5 volts for all bursts. The mass distribution function is calculated over 20,000 samples and the RC constant of the differentiator was set to 10% of the bit length. Figure 106(b) shows the result of another simulation run with the same setting but the distribution function is plotted for the differentiated signal level when a time slot equal to the MLEPW is passed after the start of each bit. In this case, the value of MLEPW was exaggeratedly set to 5% of the bit length. 0 10 -2 10 -4 BER 10 MLEPW MLEPW MLEPW MLEPW -6 10 -8 = = = = 0 0.01 * Bit Length 0.02 * Bit Length 0.03 * Bit Length 10 -10 10 2 4 6 8 10 SNR (dB) 12 14 16 18 Figure 108 : BER versus the SNR for different values of MLEPW. The first graph of Figure 106, can be interpreted as the ideal case where MLEPW=0, this evidently shows the impact of MLEPW as the distribution in Figure 106(b) is more concentrated towards the centre as a result of the capacitor discharge during the MLEPW interval. Clearly this results in a poorer BER performance. The effect of AWGN noise on the distribution of the differentiated signal levels after start of each bit is illustrated in figure 3. This shows that a decreasing channel SNR makes the peaks more difficult to be distinguished. Optimum RC constant of differentiator The value of the RC constant of the differentiator plays an important role in the BER performance of the receiver. If the RC constant is too high, this results in very narrow differentiated pulses which the comparator may not be able to capture. On the other hand, a high RC constant will result in baseline drift effect. As such, the optimum value of the RC constant needs to be chosen with regard to the value of the MLEPW. By way of example, Figure 109 shows a simulation of the differentiated pulses with high and low RC constants. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 138 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 1.5 1.5 1 1 0.5 0.5 0 0 -0.5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1.5 -0.5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1.5 1 1 0.5 0.5 0 0 -0.5 -0.5 -1 -1 -1.5 -1.5 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 (a) (b) Figure 109 : a) the differentiated pulse with very large value of RC constant for differentiator b) the differentiated pulse with small value of RC constant for differentiator. Impact of consecutive bits (CID) Since the receiver is responding to the bit changes rather than the individual bits, in case of a wrong decision, if there is no bit change immediately after the erroneous bit, subsequent bits will therefore tend to be erroneously interpreted. Fortunately, analysis and simulation results show that this does not severely degrade BMR receiver performance. Figure 110 shows the graph of BER versus SNR for a pseudo-random sequence and a succession where no consecutive bits can be found. The successive bits degrade the BER performance of the receiver in a way that if an erroneous bit is received, then all other subsequent bits are received incorrectly until a bit-change has occurred. As an example, in the case of two successive bits, the BER will be doubled. However, the probability of two consecutive bits in a bit stream is 0.5 (‘00’ or ‘11’). In order to calculate the average BER penalty due to such an event, the penalty introduced by having specific number of repeated bits needs to be multiplied by the probability of occurrence of that specific number of bits in the bit-stream sequence. Using the Gabriel's Staircase series we have: (1) In this case r=1/2 meaning that the total BER will change by factor (0.5/0.25)=2 due to the consecutive bits. Figure 110 shows the BER performance of the PRBS sequence and compares it with the case where no consecutive bits exist in the bit-stream. The result satisfactorily verifies the above analytical discussions. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 139 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 0 10 -2 BER 10 -4 10 PRBS No Consequtive Bits (Analytical) No Consequtive Bits (Simulation) -6 10 -8 10 0 5 10 15 SNR (dB) Figure 110 : BER versus SNR for a pseudo-random sequence and a sequence of “0 1 0 1…”. Note that the sequence of one and zero is an extreme case where there are no successive 0 or 1 bits and still the performance is not very far from the PRBS sequence. However this can cause major problem both for AC or DC [114] coupled receivers. 5.6 ADC/DAC The most advanced uses of Analogue-to-Digital Converters (ADC) today may be found in a typical 100 Gb/s DP-QPSK coherent receiver implementations. In the receiver, the two polarizations are split and fed into two balanced 90 degrees hybrid mixers, from which the output is fed into four parallel 56 GS/s ADCs for decoding. Figure 111 shows the ADC requirements for different 100G modulation formats and ADC performance that has been demonstrated and/or published in scientific literature. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 140 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 111 : The figure shows a set of possible modulation formats for 100 Gb/s systems and their requirements in terms of ADC performance (green circles) [128]. Also reported is the performance of bipolar and CMOS ADCs from publications [129], [130]. SiGe has been the technology of choice for high speed circuit design. Regarding DAC, there is however a trend in implementing them in CMOS in order to enable integration with digital logic, such as DSP functionality, while keeping power consumption low. As an example Fujitsu and Micram use different technology approaches. While Micram’s analogue front-end is implemented in SiGe technology which is integrated together with a CMOS interface to create a multi-chip System-in-Package (SiP), Fujitsu uses CMOS technology which allows for a monolithic configuration (Figure 112). Table 21 summarizes state-of-the-art DAC available. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 141 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 112 : Multi-chip (upper figure) and monolithic (lower figure) configuration of ADC/DSP block in a 100 Gb/s receiver [131]. Country Commercial Research Sampling rate Resolution Bandwidth Interleaving Technology Power consumption (typ) Table 21 : Table summarizing state-of-the-art DACs available. Fujitsu Maxtek [134] Micram [135] [132][133] UK USA Germany / Commercial Commercial Commercial Country Commercial / Research Sampling rate Resolution Bandwidth Interleaving Technology Power consumption (typical) O O A S OA AS SEEE FP7 – ICT– GA 249025 55-65 GS/s 8-bit 15 GHz CMOS 40 nm 1.5 W Mobious Semicond. [137] USA Commercial 63 GS/s 12.5 GS/s SP Devices [136] Sweden Commercial 30 GS/s 6-bit 20 GHz Yes, 60 GS/s SiGe 11.5 W 6.4 GS/s 8-bit 3 GHz Yes, 12.8 GS/s Ihp Gmbh [138] U Stuttgart [139] Germany Mostly Research 1-50 GS/s 3 - 12-bit Germany Research National Semicond. [140] USA Commercial 50 GS/s 3 GS/s 8-bit 3 GHz SiGe CMOS 90nm CMOS 1.6 W 6.25 GHz Yes, 25-50 GS/s SiGe Page: 142 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 5.7 DISPERSION COMPENSATION Dispersion in optical communications systems leads to inter-symbol interference (ISI) and limits the transmission distance and/or data rate. In optical single-mode fibres different dispersion effects occur: group velocity dispersion, also known as chromatic dispersion and polarization mode dispersion (PMD), a highly dynamic effect. Compensation of chromatic dispersion is usually required for optical transmission systems operating on bitrates of 10 Gb/s and above, while PMD-compensation is required for 40 Gb/s and beyond. Several compensation methods exist, Figure 113 shows an overview. The inline dispersion compensation by dispersion compensating fibres (DCF) along the fibre link was a very popular approach during the last decade but because it is a static approach, it is now about to be replaced by more dynamic solutions based on electronic signal processing [141] at the transmitter or at the receiver. EDFA Tx Rx SSMF •Electronic pre-distortion •Pre-chirping •PMD-scrambling •Inline dispersion compensation (DCF) •Dispersion shifted fibers •Optical dispersion compensation •Optical PMD compensation •Electronic equalization Figure 113 : Overview of dispersion compensation methods 5.7.1 Compensation at the transmitter At the transmitter, the electronic pre-distortion (EPD) approach is an effective method to avoid inline-DCF along the link. The real (I) and imaginary (Q) parts of the data signal are generated using digital signal processing (DSP) using two FIR-filters (finite impulse response). The I and the Q part of the optical modulator (e.g. a double Mach-ZehnderModulator) is then driven by the electronic pre-distorted I and Q signals, where the amount of pre-distortion depends on the dispersion of the link and therefore the fibre length. The goal is to pre-distort the signal so that the pre-distortion and the fibre dispersion compensate each other at the target fibre length. Therefore, the signal can only be received at the receiver placed at the target fibre length and in a small window around, depending on the modulation format. There are some other methods for transmitter-sided dispersion compensation. One is prechirping, where the chirp of the MZM is adjusted to increase the transmission length. Another other is polarization-scrambling, where the input polarization into the fibre is permanently scrambled which should increase the PMD tolerance of the system. 5.7.2 Compensation at the receiver At the receiver many approaches for dispersion compensation exist which were intensively studied during the last years. The simplest method may be to use a DCF at the receiver to compensate for all link dispersion, but this approach is limited by nonlinear fibre effects and is also only a static solution. Nowadays the system designers focus on more adaptive solutions like electronic equalizers. Many types of equalizers were proposed during the last years. The first analog equalizers realized for dispersion compensation were FFE and DFE structures (feed forward and decision feedback equalizer). The FFE typically consists of a tap delay-line filter with 5 or more forward taps, while the DFE has one tap in the feedback path. Usually O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 143 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 these two equalizers are combined to increase the compensation performance. The FFE/DFE uses a feedback criterion, e.g. the minimum square error (MSE) to minimize the ISI of the direct detected received signal. Therefore it operates independently of the signal distortion and can also be used for joint CD and PMD compensation. For NRZ-modulation the FFE/DFE can increase the tolerance to chromatic dispersion by about 60%. Practical implementations of FFE/DFE operating on 40 Gb/s were shown. The MLSE equalizer (maximum likelihood sequence estimation) is a digital equalizer which uses the Viterbi-algorithm to calculate the most likely transmitted bit sequence. Since the processing is done on a bit sequence, the performance of the ISI equalization is getting better with an increased bit sequence. For optical communication systems MLSE equalizer with 4states and 16-states were experimentally reported for 10 Gb/s [142], [143]. It also operates independently of the signal distortions and the joint compensation of CD and PMD was proven. The 4-states MLSE doubles the CD-tolerance and shows a PMD-tolerance of about 1Bit duration. However, the compensation performance is limited by the complexity of the MLSE at higher Viterbi-states and the clock-recovery at large signal distortions. The most sophisticated electronic equalization uses coherent reception and digital postprocessing the complex e-field of the signal. In the coherent receiver the optical field is converted into the electrical domain without loss of information. The e-field of the signal can then be equalized much better then after direct detection where the phase information is lost. The equalization of CD is then again done by using FIR-filters. With 128 filter taps 42000ps/nm CD can be compensated [144]. If a polarization diversity receiver is used the compensation of PMD is also possible. Other approaches to compensate for dispersion are using multi-carrier modulation formats such like OFDM (orthogonal frequency domain modulation). In these schemes the signal is modulated onto multiple low bandwidth and low data rate sub-carriers which are transmitting only a fraction of the total data. Because of the low data rate of each sub-carrier the transmission is only slightly influenced by dispersion effects and can also be equalized much easier. 5.8 PASSIVE WAVELENGTH SELECTIVE DEVICES 5.8.1 Thin film filter-based WDM components As it was described in § 5.4.1, thin film filters (TFF) or interference filters are made by depositing of thin alternating layers of two materials with different refractive index on a transparent substrate and their functioning is based on the interference of multiple reflections. The transmitted wavelength and filter shape depends on layers thicknesses, refractive indexes, angle of incidence on the filter (normal in the fibre-based filter) and the numbers of layers in the stack. Usually the more layers, the finer the resolution, and the narrower the range of wavelengths selected. There are three types of filters used in WDM devices: line filter, band filter, and cutoff filter, each with its own transmission characteristic. Line and band filters either reflect or transmit light in a selected range of wavelengths. If the range of wavelengths is narrow, we have to do with line filters, for example a filter to select one 100-GHz optical channel. Filters that select a broader range of wavelengths are called band filters, for example a filter that selects a 10-nm CWDM channel. Cutoff filters are designed to make a sharp transition between transmitting band and reflecting band at a certain wavelength. For example, a filter for separation of optical channels belonging to C-band and L-band of erbium-fibre amplifier has a cutoff wavelength at 1567 nm. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 144 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Thin film filters can be used to combine and separate wavelengths in WDM systems by taking light out of the fibre and passing it through a set of filters that sorts the light out by wavelength. Typically a lens system collimates or focuses the light emerging from the input fibre, which then passes through one or more filters and focus them into individual output fibres. Several interference filters can be cascaded to pick off a series of four or eight wavelengths in such a way that the first filter transmits channel 1 and reflects all other channels. The remaining channels hit the second filter, which transmits channel 2 and reflects the remaining channels and so on. In this way n – 1 filters are needed to separate n optical channels. Such a demultiplexer works fine for 4 – 8 channels, but the losses increase considerably for 16 channels or more. The solution in this case is to divide optical signals into groups of channels and treat them with band pass filters before they are then split up individually. This method does not reduce the total number of filters needed, but the number of filters for each channel is reduced. Interference filters are widely used for WDM due to the fact that the technology is well developed. Interference filters have been known for many years, although the extremely narrow-line filters used in DWDM systems were developed only recently. Filters can be made very small, either deposited on the fibre facet or separately with a cross section of a few millimeters. They have good performance, have modular construction and can be upgradeable, but it is always need to have as many filters as the number of channels to be separated. Although for add-drop multiplexers a single optical filter is needed to drop one channel with the remaining channels reflected and collected for transmission through the rest of the system. Examples of the devices on the market based on TFF are reported in Appendix 7.2. 5.8.2 Fibre Bragg grating-based WDM components Fibre (or planar waveguide) Bragg gratings work similarly, but they selectively reflect a narrow range of wavelengths, while interference filters selectively transmit a narrow range of wavelengths. In the multi/demultiplexer each reflected channel of light must go through an optical circulator to be separated from the input light. Using Bragg gratings as optical filters and optical circulators one can build different architectures with spectral characteristics depending on the used gratings spectral profiles with channel separations of 100GHz and 50GHz easy obtained and large number of channels, but using the same number of circulators and at least the same number of gratings. A single Bragg grating with circulators on both sides can act as an add-drop multiplexer. Another possibility for this kind of device can be a Bragg grating assisted optical Add-Drop Multiplexer based on 2x2 MMI coupler shown in Figure 114 [145] or Add-Drop Multiplexer based on balanced Mach-Zehnder interferometer (MZI) with two identical Bragg gratings imprinted in both arms of the interferometer shown in Figure 115 [146]. In both cases the devices were fabricated in planar technology. In the second configuration two 3dB couplers on both sides of the MZI redirect the reflected wavelength to the correct output, which makes that circulators are not necessary in this architecture. Moreover, an additional MZI equipped with a thermal switch and connected to the first one with help of two Bragg grating reflectors (shown in Figure 115 as Reflector 1 and Reflector 2) changes it to a Switchable Add-Drop Multiplexer, which is an important component for ring architectures. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 145 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 114 : Bragg grating assisted optical Add-Drop Multiplexer based on 2x2 MMI coupler. 30dB extinction ratio between drop and transmitted channels and 3dB excess loss in the dropped channel have been obtained. Figure 115 : Add-Drop Multiplexer based on balanced Mach-Zehnder interferometer. Obtained crosstalk -25dB, insertion loss -3dB, switching time 2ms and power consumption 0.5W have been achieved. In principle, fibre Bragg gratings (and circulators) can be arranged to obtain different functionalities with good performance although the cost and complexity of the components make that they are not commonly used. The described until now WDM components based on thin film filters and Bragg gratings split or combine optical signals one at a time. In the following we will describe more advanced integrated structures that split or combine a large number of wavelength channels in parallel due to a diffraction grating that spreads out a spectrum of incoming light in such a way that different wavelengths propagate in different angles. Both, etched diffraction gratings (EDGs) and arrayed waveguide gratings (AWGs) are usually designed as planar integrated components and despite advanced architecture can be fabricated in wafer scale allowing for mass production. Moreover, integration offers the advantages of compactness, reliability and potential possibility to have light sources, detectors and control electronics on the same chip. 5.8.3 Etched Diffraction Grating-based WDM components As it was explained above, the two main candidates for integrated multi/demultiplexer in WDM applications are etched diffraction gratings (called also Echelle gratings) and arrayed waveguide gratings. Both of them have been fabricated using different material platforms, including silica-on-silicon, III-V semiconductors-, as well as silicon large core- and nanowire -based technology. The principle of operation of an EDG multi/demultiplexer called also echelle grating (EG) or O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 146 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 planar concave grating (PCG) is based on a spectrometer with the Rowland Circle geometry (Rowland grating). The access waveguides are situated on a circle with a radius R (Rowland circle). The reflective concave grating with radius 2R, situated such that it is touching the circle, diffracts light coming from the input waveguide and focuses the reflected light at different wavelengths back to the points on the Rowland circle, where the output waveguides are situated (see Figure 116). Figure 116 : Etched diffraction grating demultiplexer: signal from input waveguide with wavelengths λ1 ,λ2 ,λ3,… is diffracted by the planar concave grating, and refocused into different output waveguides. The close to Littrow configuration, where the grating facets are almost perpendicular to the incoming light and reflect light back, almost in this same direction guarantees high efficiency for a chosen diffraction order. The first EDG-type demultiplexer was proposed 1979 by G.L. Tangonan and co-workers from Hughes Res. Labs, where the bulk grating was attached to an edge of circular profiled slab situated between two glass slides. The first monolithic device in application to DWDM systems with etched grating in SiO2 glass waveguide on Si was demonstrated in 1992 by P.C. Clemens et al. from Siemens AG. The main drawback of this construction at that time were relatively high insertion losses caused mainly by non-perfect verticality of deeply etched grating facets in a thick slab. Today, using Silicon-on-Insulator material platform the grating facets verticality is no longer the main constrain for EDGs, where shallow etching depth of 220-250nm is sufficient [147]. There are few companies like Enablance or Kotura making EDG-based PLC’s for commercial applications. Very recently (July 2010), Intel also announced 4-channel CWDM optical link using silicon Echelle gratings as MUX/DEMUX. 5.8.4 Arrayed Waveguide Grating-based WDM components An alternative solution was proposed by M. Smit from Delft Univ. in 1988, where the grating was replaced with a focusing and dispersive planar component based on phased array (called PHASAR) known today as arrayed waveguide grating. An arrayed waveguide grating is composed of input and output waveguides coupled to an array of waveguides (regularly arranged with increasing lengths) through two focusing slab waveguides, free propagation regions (FPRs) – where light is not confined laterally and therefore diverges or converges in the lateral direction, as shown in Figure 117a. An optical beam entering the 1st free propagation region (input FPR), through the input waveguide, spreads by diffraction and forms a circular wave front (equal phase) at the end of FPR. Due to the constant path length difference between adjacent waveguides, after passing through the array the signals that reach the output slab waveguide have different phase delay, and due to the mutual interference between the signals from each waveguide, all the wave O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 147 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 fronts are diffracted and for each specific wavelength are focused towards a uniform direction (Figure 117b). The correct positioning of the output waveguides allows the spatial separation of the different wavelengths. In this way it demultiplex the input signals from one of the input channels into different output channels according to the signal wavelength. Figure 117 : Schematic of an arrayed waveguide grating: (a) input-output waveguides are coupled through two slab (FPR) waveguides and an array of curved waveguides, (b) tilted phase front with an angle , focused at an output waveguide in the output FPR. a) b) Figure 118 : Series of 32 x 32AWGs on a 4” Si wafer in SiO2/Si technology, 0.8 nm channel spacing (100 GHz), 25 nm band: a) Fabricated devices, b) AWG transmission spectrum. Although it is not straightforward understood, the principle of operation of these two devices (EDG and AWG) is very similar. In the AWG at the end of the array of waveguides we see the same diffractive and focusing effect as in the case of concave diffraction grating. Additionally the front and back profiles of the free propagation regions in the AWG follow Rowland circle geometry as in the case of EDG. Silica-based AWG soon became a main multichannel multiplexing device on the market due to more simple and tolerant technology in comparison to EDG. Fig. 5 shows a series of of 32 x 32AWGs on a 4” Si wafer in SiO2/Si technology, 0.8 nm channel spacing and their transmission spectrum [148]. Even devices with high level of integration with cascaded AWGs and hundreds of channels are possible to fabricate with low loss and crosstalk. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 148 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 5.8.5 Technology constrains for EDG and AWG WDM components As it was mentioned earlier, when using Silicon-on-Insulator material platform the grating facets verticality is no longer the main constrain for EDGs, where shallow etching depth of 220-250nm is sufficient and so AWG superiority is no longer valid. In general EDGs offer instead smaller device size, especially for a large number of channels. When choosing the type of multiplexer one needs also to consider the parameters that limit the device performances that are specific to SOI technology. The devices are usually patterned directly by e-beam lithography or deep UV lithography using photomasks prepared with e-beam litho. In both cases we have pixelation errors due to finite step of the e-beam machine (10-50 nm) as well as stitching errors due to limited writing field (80-120 µm) and usually necessity to put together many writing fields to pattern a whole component. For EDG case the positions of grating facets are discretized by pixilation and distorted by stitching. The same errors in AWGs limit the accuracy of positioning of the beginning and the end of the waveguides in the array causing unexpected phase changes. Both two errors decrease the efficiency and increase crosstalk. In AWGs the phase errors can be additionally increased by changes in the effective path length of the waveguides in the array caused by non-uniform waveguide cross section, local strain, material composition, temperature gradients and other non-uniformities. In EDGs on the other hand quality of etching not only decreases the device efficiency, but also affects scattering from grating facets, increasing crosstalk and increasing overall insertion loss in comparison to AWGs. Both, arrayed waveguide grating- and etched diffractive grating-based wavelength selective devices are of interest for future highly integrated optical communication systems and computer interconnects. For applications, where a large number of channels should be treated and compactness is important, EDG-based components should be used, whereas for more relaxed applications the mature, high quality and low loss AWG devices are preferable. For example a 256-channel, silica-on silicon-based EDG with 25GHz channel spacing has an overall size of 20x40 mm [149]. The device has adjacent channel crosstalk of 30 dB, insertion loss of 10 dB, and a polarization-dependent wavelength shift less than 10 pm. The parameters are similar to those of AWG devices, but an AWG fabricated on the same material platform with 256 channels takes up about five times the area of the EDG device. Examples of the main devices on the market based on AWG produced by Oplink and JDSU are reported in Appendix. There are several other actors on the market offering WDM component although in some cases it is not clear whether they are offering real products or only design of future products. The companies include GEMFIRE, FITEL, NEL, Neo Photonics, NEC FiberOptech, Hitachi Cable, SANTEC, IGNIS, Enablance, Kotura, ANDevices, Fiberdyne Labs, Lightwaves2020, Go!Foton, Optelian, LightGAIN, GIGALIGHT and probably a few others. The number of companies and their names are changing constantly. 5.9 REACH EXTENDERS Reach extenders are an option to increase the optical link budget by using mid-span signal amplification. By using reach extenders, the network planning for an operator is made more flexible. Basic reach extension can be achieved by either Optical-Electrical-Optical (OEO) conversion or optical amplification (OA). The latter can be achieved by using either SOA or fibre amplifiers in the optical distribution network. Basic reach extension technologies are O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 149 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 summarized below. Optical-Electrical-Optical (OEO) RE OEO offers the most straight-forward RE solution. It builds on re-use of existing transceiver technology and offers flexibility and different implementation options. Figure 119 shows an example implementation using a reset-less OLT transceiver. Additional required building blocks are the embedded ONT (EONT), optical splitter for EONT, burst-mode clock/data recovery for upstream data, logic for reset generation, electrical switch for EONT and protection path and optical switch for protection or 1:2 splitter. Figure 119 : Example GPON OEO Measurements on an OEO for GPON show that it is possible to cover the complete logical reach, with a differential drop reach of 0-24 km and a trunk reach of 0-60 km. As a consequence the OEO offers simple network planning/deployment. OEO can be used for most system concepts. Power consumption, cost and reach are related to TRx blocks of each system concept. SOA The use of semiconductor optical amplifier (SOA) for reach extension for up to 50 km reach with a 1:32 split has been demonstrated for GPON. The configuration allows for a maximum reach of 37 km on the trunk side and 22 km on the drop side. SOA can be used for multichannel amplification offering co-existence with XGPON1 US (1270 + 1310 nm). However, analogue amplification makes planning/deployment more difficult than with OEO. As SOA is currently not a volume product, cost for the technology is high. Table 22 : Example of supported GPON configurations Drop reach (km) Trunk reach (km) Total reach (km) 1:32 split 15 37 52 1:64 split 10 34 44 1:128 split 5 28 33 Fibre amplifiers: Raman / EDFA Most long-reach options require active reach extenders. Raman could potentially offer reach extension without active equipment in the optical distribution network (ODN). Raman amplification in PON is however not straightforward. Pump lasers for the upstream would be in the cut-off region (i.e. < 1200 nm). GPON with XG-PON overlay would result in a complex wavelength plan with many pumps and signal wavelengths. In addition, Raman pumps require quite high power (~300 mW range), which stands in contrast to one of the motivations for PON, i.e. reduced power consumption. EDFA and similar doped fibre amplifiers could potentially be used for GPON and the XGPON downstream. Remote pump architectures have been shown. However, the upstream O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 150 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 requires complex control to handle burst transmission. Remote protocol terminator In addition to basic reach extension one should also consider the possibility of increasing the complexity of the active RE for performing basic tasks such as protocol termination. The remote protocol termination (RPT) approach for PON consists of extending the backplane of a PON OLT (Figure 120). The actual PON port can be separated from the OLT element. As a consequence, the system is no longer limited by the PON protocol. RPT enables traffic aggregation unlike the alternative L1 REs (OEO, OA, OTN). As a result conventional D/CWDM can be used for the uplink which saves fibres if needed. The difference between the RPT and a “mini-OLT” lies in the reduced management complexity since the RPT is managed as an OLT line card. There is no difference in the data plane. The network between the RPT and the OLT can be any packet network and the OLT can be an aggregation switch (AGS). Figure 120 : RPT approach Comparison between different reach extender technologies Figure 121 provides a comparison of different reach extender technologies when used in a 10G TDM-PON system. The total reach is divided into trunk reach (distance between OLT and reach extender) and access reach (distance between reach extender and ONTs). The orange shaded areas show achievable distances with 1:32, 1:64 and 1:128 split for a system without reach extender (Nominal 1 class, 29 dB link budget). Light green areas show achievable distances with an OA based reach extender (L-band EDFA for the 1577 nm downstream + SOA for the 1270 nm upstream). Performance is limited by the upstream SOA. Light blue areas show achievable distances with an O/E/O based reach extender. In dark green it is shown that the RPT / mini-OLT approach allows for possible extension of the trunk fibre up to 80 km. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 151 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 121 : Comparison of different reach extender technologies for 10G TDM-PON (XG-PON): OA based reach extender (L-band EDFA for the 1577 nm downstream + SOA for the 1270 nm upstream), O/E/O based reach extender and Remote Protocol Terminator (RPT)/ mini-OLT approach. With the O/E/O approach is possible to cover longer distances than for the OA approach. For a 1:64 split the difference is of the order 20 km. When several PONs are routed through the same remote node, it could make sense to use RPT instead of O/E/O. The benefit would be traffic aggregation and the possibility use WDM (i.e. CWDM) on the trunk lines. RE configuration With potential requirements on resilience there are different RE configurations the offer different trade-off between cost and resilience. Several potential RE options are displayed in Figure 122. Regarding the dual RE solution, development of low-power modes for the standby RE would be of interest. A promising solution is the one with a single RE with dual trunk. With this architecture it is possible to power off the protection path. For both the dualRE and the single-RE dual-trunk solutions a switch-over procedure needs to be defined. The dual trunk in the single-RE solution could be implemented either by a 1:2 splitter or dual uplink ports as illustrated in Figure 3. The exact configuration depends on the relative cost of the ONT TRx and the OLT TRx for the specific system concept. The implication of using a 1:2 splitter on the optical power budget should also be considered. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 152 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 122 : RE variants 5.10 SWITCHING COMPONENTS Ethernet is riding its success from LAN to access and metro markets with service providers migrating from circuit switched networks to Carrier Ethernet. A number of important reasons made Ethernet a winner, including low cost, simplicity, flexibility, and scalability. The most important factor was technological unification, because this guaranteed smooth interworking without the need for specialized gateways. Gigabit Ethernet (GbE) is commonly supported in most computer network cards, while in the access bandwidth is typically not larger than 100Mb/s. However, the evolution of applications and services such as triple play, HDTV, 3DTV requiring higher bandwidth to the subscriber drive GbE to be more and more used in the access as well as the home. GbE has been used in aggregation and backbone networks for many years and its use over either AON or PON has become a leading architecture for NGOA systems. The Ethernet switch unit is compulsory for both AON and PON architectures. The foreseeable major evolution of Ethernet may include: from Fast Ethernet (FE) to GbE by gradual conversion of low-cost CPEs, IADs, SOHO (small office and home office) switches, and access switches; deployment of 10G Ethernet (10GbE) in aggregation / distribution network; and 40GbE or 100GbE in the core network. The transitions are creating new requirement and research questions related to switching components. 5.10.1 Power consumption Recently market pressure and legislative action worldwide is demanding improvements in energy efficiency of networked systems. Reducing power consumption not only provides good ecological credentials for society, but also improves financial figures for operators since energy costs are a major component of operating cost. Unfortunately there is a clear relationship between faster switching speeds and the amount of power consumed. Hence switches are committed to new designs with lower power requirements. There are known avenues for lowering power, such as: - system level energy management techniques O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 153 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 - switch fabric design(hardware and software) to improving switch - thermal control and fanless cooling systems - more efficient power supplies - sleeping mode on optical transceivers - network level energy management techniques - network wide coordinated link and node sleep - bandwidth renegotiation High integration and low power dissipation IAD (Integrated access device) switches or CPEs are important as well. The largest portion of the per user power consumption is at the CPE. Several vendors already have “green Ethernet” products on the market [150] where low energy consumption mode has automatic entry/exit based on actual network traffic and Ethernet ports that can detect network cable length and unused ports to manage power consumption. 5.10.2 Challenges on high speed switching process: 10 Gb/s and ultra-high-speed transport Ethernet in aggregation/distribution network requires high performance switching hardware and software support. The increased interface data rate leads to a higher threshold for packet processing, storage, system switching, and backplane technology. Several examples of challenges are listed below [151]: - Content-Addressable Memory (CAM) of the network processor. In order to reduce switching latency, the bandwidth of search interface should be increased. - Data bus: There is a bottleneck for the data bus width and rate. - Multiple packet processing chipsets solution caused by bus interface conversion, the board area and power consumption are unacceptable. Single chip processing capability is limited; the solution based on FPGA-customization still need comprehensive technologies. - Switch fabric chipset: besides the non-blocking full duplex bandwidth capacity, switch fabric must actually provide additional bandwidth to accommodate cell overhead, buffering, and congestion-avoidance mechanisms. For supporting 100 Gb/s interface, the bandwidth of each line card should be upgraded to 200-500 Gb/s bandwidth. The requirements for backplane design, technique, material, and meeting the bus length are more critical than before. For the system meeting the carrier-class requirements, the Virtual Output Queue (VOQ) and Hierarchy Quality of Service (HQoS) should be met. It requires greater processing bandwidth, queue processing capability, and buffer, which greatly increases the difficulty of system design. 5.10.3 Network stability In current network systems all devices are managed by the control and management (C&M) system for automated monitoring, troubleshooting of network faults, QoS management etc. Normally the C&M function is implemented in individual devices (e.g. line card) apart from the switching unit. However it is necessary for switches to have hardware which commits the connectivity fault monitoring and the assignments from the C&M system. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 154 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 6. Summary Based on the survey presented in this deliverable a comparison table of potential NGOA system variants is presented in Table 23. The table covers a coarse comparison of the system aspects with respect to the main key system aspects defined in Chapter 3 and can be seen as a summary of the findings in the report. The focus of Table 23 is to present systems that are more or less realizable today. Cost and power consumption is given to indicate the order of magnitude today, without consideration of future potential for the different variants. A more detail assessment is to be provided in D4.2 and D4.3 [1] which will provide a consolidated view on reasonable technology evolution as well as cost and power consumption potential for the main candidates in the 2020 time horizon. The requirements in D2.1 provide the basis for the coarse assessment in Table 23. At current state of technology, all contenders have problems of fulfilling all requirements simultaneously, which makes it difficult to compare the systems on equal terms. The system configuration may be tailored depending on which requirements are prioritized (e.g. reach or splitting ratio). For this reason, for some systems, ranges have been specified and hence some care must to be taken when drawing conclusions based on this assessment. Several entries of Table 23 are listed only as rough estimates. TDM-PON approaches present significant challenges with respect to simultaneously increasing bandwidth, reach and splitting ratio, whilst maintaining low cost and power consumption. The configurations presented in Table 23 comply with reach requirements but do not meet the requirements on splitting ratio and in some cases bandwidth. Advantages of WDM-PON include long reach and large per customer sustainable bandwidth. A drawback is the limited fan-out of traditional WDM-PON approaches, limiting the number of customers per feeder fibre. OFDM-PON presents similar challenges as TDM-PON concerning limited total capacity and limited splitting ratio. The particular configuration presented in Table 23 allows for 64 users per feeder fibre but does not fulfill the requirements on sustainable bandwidth. The high cost and power consumption associated with OFDM and UDWDM in Table 23 is associated with the immature state of technology. The preliminary assessment of future cost potential of UDWDM (chapter 4.5.5) shows that it still is a potential candidate for NGOA despite large cost today. For OFDM the most interesting variant, in terms of meeting OASE requirements, is the hybrid WDM/OFDM-PON solution which showed slightly worse cost potential than the other hybrid alternatives (chapter 4.5.3). For OCDM-PON the scaling of the number of users per feeder fibre presents a serious problem which eliminates it as a serious contender for NGOA. Hybrid concepts are motivated by the fact that each of the aforementioned pure system concepts individually may have difficulties in fulfilling the complete set of NGOA requirements. Hence, hybrid concepts have been proposed that combine advantages of different concepts. Typically the advantages that are exploited are the increased overall capacity of WDM and the efficient resource sharing of TDM, OFDM or CDM. Concepts that involve different types of active remote nodes have also been considered. A preliminary cost and power consumption analysis of different variants (chapter 4.5 and 4.6) shows that the most promising configurations are hybrid WDM/TDM-PON as well as various active hybrid variants (WDM-PON with AON access and two stage WDMPON). These concepts have been included in Table 23. The considered WDM/TDM PON configurations prioritize high splitting ratio at the cost of not fulfilling the sustainable bandwidth requirement. Concerning AON the main drawback with respect to the OASE requirements concerns either large port count at the central office or low degree of node consolidation, depending on AON architecture. Both these factors have implications on total O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 155 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 cost and power consumption. As an outcome of the survey it is seen that the main candidates for further consideration within the OASE project based on the posed requirements are different variants of pure WDM-PON, hybrid WDM/TDM-PON, AON as well as various active hybrid variants based on WDM-PON and AON. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 156 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 0 1 10 4 PON ports per HU in rack “Quarter Pizza Box” Start of deployment <1 2.5 15 2 PON ports per HU in rack “Quarter Pizza Box” First demonstrator 2 8 4-16 slots of 5HU 25% 100% of 9HU Shelf “Quarter Pizza Box” First demonstrator 3 8 4-16 slots of 5HU 25% 100% of 9HU Shelf “Quarter Pizza Box” First demonstrator 2 0.04 1.5 32 20 28/30/32 3R 60 50+ 20 120 0.5 XG-PON 0.31 10 0.31 10 32 20 29/31/33 /35 3R 60 50+ 500 1000 40G serial TDM-PON 1.25 40 1.25 40 32 20 27 3R 60 50+ 5k 2k WDM-PON with lambda reuse 1-10 1 1-10 1 O O A S OA AS SEEE FP7 – ICT– GA 249025 1-10 1 1-10 1 40-400 96 15-50 40 5-17 for Fibre 12 - 60-100+ - 22-34 for Fibre - ~700 700 Page: 157 of 168 ~700 350 Time to market [years] Commercially available 2.5 4 4 Maturity Footprint ONU “Quarter Pizza Box” 0.78 EDFA, Raman SOA, TDF Power consumption today ONU [W] Power consumption today OLT per sub [W] Cost Today ONU [$] Cost Today OLT per sub [$] Power budget with reach extension [dB] Reach with reach extension [km] Reach extension – technology Power budget without reach extension [dB] Reach without reach extension [km] Typical No of subscribers per Feeder fibre Upstream bandwidth per sub - peak [Gb/s] Upstream bandwidth per sub - sustainable [Gb/s] Downstream bandwidth per sub - peak [Gb/s] 8 8 PON ports per HU in rack GPON State of the Art WDM-PON with tunable laser WDM-PON System concept comparison table Footprint OLT TDM-PON System concept Downstream bandwidth per sub - sustainable [Gb/s] Table 23. Hybrid WDM/TDM CDM-PON OFDM-PON Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Serial 40G OFDM-PON 0.63 40 0.63 20 - - - - 10k 50k 5 20 4-16 slots of 5HU 25% 100% of 9HU Shelf “Quarter Pizza Box” First demonstrator >5 “Quarter Pizza Box” Experiment in lab >5 4+ 64 59 - - - - 700? 700? - - 4-16 slots of 5HU 25% 100% of 9HU Shelf OCDM-PON 10 10 10 10 513 (theory) 12 (demons trated) Hybrid WDM/TDM broadcast & select 0.31 10 0.31 10 1024 NA/NA/ NA 31/34/37 EDFA, Raman, SOA 46/36/26 11/14/17 290 2500 1 10 Half Rack “Quarter Pizza Box” First demonstrator 2 Hybrid WDM/TDM with lambda split 0.31 10 0.31 10 1024 20/10/N A 22/25/28 EDFA, Raman, SOA 80/70/60 2/5/8 300 2500 1 10 Half Rack “Quarter Pizza Box” First demonstrator 3 Hybrid WDM/TDM with lambda switch 0.31 10 0.31 10 1024 NA/NA/ NA 32/35/38 EDFA, Raman, SOA 40/30/20 12/15/18 325 2500 1 10 Half Rack “Quarter Pizza Box” First demonstrator >3 Hybrid WDM/TDM with WSS 0.31 10 0.31 10 1024 20/10/N A 22/25/28 EDFA, Raman, SOA 80/70/60 2/5/8 350 2500 1 10 Half Rack “Quarter Pizza Box” Paper study + simulation >5 O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 158 of 168 UDWDM 1 1 1 1 5121024 30-100 - OA 100+ - 10k 10k 10 12 Half Rack “Quarter Pizza Box” First demonstrator >4 AON GbE access 1 1 1 1 - 70 - Higher quality optics 120 - 15 scalable 1 5 32 ports per HU “Quarter Pizza Box” Deployed 0 15+15 EDFA active sites required 8 Main OLT: Half Rack In-Field OLT: 1 Shelf “Quarter Pizza Box” Can be demonstrated today 3 15+10 EDFA active sites required 8 Main OLT: Half Rack In-Field AON: 2HU each “Quarter Pizza Box” Can be demonstrated today 1 WDMPON-AON PON-in-PON AON DWDM Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 PON-in-PON WDM-PON AON 0.1-1 0.1-1 1 1 O O A S OA AS SEEE FP7 – ICT– GA 249025 0.1-1 0.1-1 1 1 1k-10k 23049216 100 60 150+ 85+ 25+25 25+10 750 200 Page: 159 of 168 800 200 5 5 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 7. Appendix 7.1 COST AND ENERGY CONSUMPTION DATA This section contains data and assumptions used for the comparison of cost and energy consumption of different system concepts. System specific components are listed in Table 24. Table 24. Energy consumption and cost figures of relevant hybrid PON components Component 10G base TRX (SFF, coolerless, w/o Locker, 22 dB, not used hereinafter) 10G TXFP (TEC, Locker, 25 dB) 10G Burst-mode TRX, 35 dB (SFF, APD, SOA, FEC, coolerless, w/o Locker) 30GHz TRX (coherent, TEC, Locker, 16 Channels à 1G / 3 GHz) 30GHz TRX (32 dB, coolerless, w/o Locker, single channel) 10G REAM-SOA, incl. Fraction of MFL, 1G 26 dB Rx (!) 10G REAM-SOA, incl. Fraction of MFL, 10G 35 dB Rx (!) Energy Con. 1.25 W 3.5 W 2.5 W 8W 2.5 W 1W 2W Cost 100$ 1200$ 175$ 1600$ 175$ 85$ 175$ 1G coherent ONU TRX, pol.-diverse or w/ Pol. Scrambler 2W 175$ 1G tunable ONU TRX 1W 75$ 40x1G Laser/Rx Array 20 W 2000$ 40x1G REAM/Rx Array plus MFL and Circulators 20 W 2400$ 1G grey SFP, 10 dB 0.5 W 15$ ASIC 1G SCMA ONU 1W 10$ ASIC 10G OFDM ONU 4W 40$ ASIC 10G OFDM / SCMA OLT 16Ch 8W 160$ ASIC 50G UDWDMA OLT 16Ch 16 W 320$ ASIC CDMA OLT 8Ch 4W 120$ OLT EDFA Booster/Preamp Combo 25 W 2000$ Circulator -- 100$ AWG Port / Power Splitter/Combiner Port -- 20$ / 10$ OLT / PoP Switch per 1G 1W 5$ Baseline cost per client (CPE, OLT shelf, motherboards) 5W 100$ For the cost and energy-consumption calculations, common (constant) contributions are not listed in the following tables. These contributions come from ONU housing (PSU), ONU higher-layer equipment (i.e., it is assumed that all ONUs can perform the same higher-layer and subscriber management functionality), and OLT shelf / management (i.e., it is assumed that to first approximation, all variants require similar OLT cost and energy overhead on a per-client basis). Hence, the baseline per-client cost and energy consumption as per Table 24 (last row) must always be added to get the respective total cost and energy consumption. The individually differing numbers are also stated on a per-client, end-to-end basis. This does include the per-client portions of the RN equipment (filter, splitter ports), which often are not considered in similar calculations for EPON/GPON. Further, assumptions regarding the ODN must be made. This applies to the different filters and splitters which are required, as well as to the fibre per-kilometre insertion loss. Relevant numbers are summarized in Table 25. Table 25 : Relevant optical parameters for hybrid PON performance analysis 1:8 TFF / 1:40 AWG / 1:80 AWG 1:8 / 1:16 / 1:1024 Power Splitter/Combiner 50 km Fiber EOL incl. Patches etc. 3 / 4 / 5 dB 12 / 14 / 35 dB 16 dB Also note that we have chosen 16 dB per 50 km fibre as insertion loss. This is a trade-off O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 160 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 between minimum insertion loss (which could fall into the 10…12 dB range) and loss of very poor urban fibres (which would go into the 25 dB range). 7.2 EXAMPLES DEVICES ON THE MARKET BASED ON TFF Examples of devices on the market based on thin film filters (TFF) are shown below: Figure 123 : Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 161 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 124 : Oplink’s Coarse Wavelength Division Multiplexer 4/8 channels, low loss series. Figure 125 : Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 100GHz O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 162 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 126 : Oplink’s Dense Wavelength Division Multiplexer 4/8 channels, 200GHz Figure 127 : Oplink’s FTTX triplexer 1310/1490/1550 WDM (1x2) O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 163 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 128 : JDSU Coarse Wavelength Division Multiplexer 4/8 channels Figure 129 : JDSU Dense Wavelength Division Multiplexer 4/8 channels O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 164 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 7.3 EXAMPLES DEVICES ON THE MARKET BASED ON AWG Examples of devices based on AWG are shown below: Figure 130 : Oplink’s Dense Wavelength Division Multiplexer 100 GHz. O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 165 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 131 : Oplink’s Dense Wavelength Division Multiplexer 50 GHz. Figure 132 : JDSU Dense Wavelength Division Multiplexer 100 GHz, Narrowband (Gaussian). O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 166 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 133 : JDSU Dense Wavelength Division Multiplexer 100 GHz, Wideband (Flat Top). Figure 134 : JDSU Dense Wavelength Division Multiplexer 50 GHz, Wideband (Flat Top). O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 167 of 168 Survey of Next-Generation Optical Access System Concepts OASE_D4.1_WP4_EAB_210612_v3.0 Figure 135 : JDSU Dual Duplexer 1310/1550. ---End of Document --- O O A S OA AS SEEE FP7 – ICT– GA 249025 Page: 168 of 168
