IEEE Communications Magazine - PDF document

IEEE Communications Magazine ¥ October 2010 144 PERATIONOFAERFORMANCEOFAPACITYOF Data symbols carrythe joining CP-lineÕsreported error samples via the SOCbackchannel, extended to highercapacity. The DSLAMsets the error-sampleformat based on theactual SOC through-put and the requirederror precision. OKSMAN LAYOUT 9/21/10 11:45 AM Page 144 IEEE Communications Magazine ¥ October 2010 148 array of vendors, ensuring cost-effective availabil-ity of high-speed DSL access networks world-ACKNOWLEDGMENTSThe authors wish to thank Frank Van der PuttenREFERENCES[1] ITU-T Rec. G.993.5-2010, ÒSelf-FEXT Cancellation (Vec-toring) for Use with VDSL2 Transceivers.Ó[2] J. M. Cioffi and J. A. C. Bingham, ÒA Proposal for Con-sideration of a VADSL Standard Project,Ó ANSI contrib.T1E1.4/94-183, Dec. 1994[3] J. Cioffi et al., ÒVery High-Speed Digital SubscriberLines,Ó IEEE Commun. Mag., Aug. 2004.[4] A. Paulraj and T. Kailath, ÒIncreasing Capacity in Wire-less Broadcast Systems Using Distributed Transmis-ption (DTDR),Ó U.S. Patent5,345,599, Sept. 6, 1994.[5] J. M. Cioffi and G. D. Forney, Jr., ÒGeneralized Decision-Feedback Equalization for Packet Transmission with ISIper, Eds., Communication, Computa-tion, Control, and Signal Processing: A Tribute to[6] G. Ginis and J. M. Cioffi, ÒVectored Transmission forDigital Subscriber Line Systems,Ó IEEE JSAC, vol. 20, no.5, June 2002, pp. 1085Ð1104.[7] R. Cendrillon et al., ÒA Near-Optimal Linear CrosstalkPrecoder for VDSL,Ó ., May 2007.[8] J. Cioffi and K. Song, ÒLevel 3 DSM Results: Vectoringof Multiple DSLs,Ó ANSI T1E1.4 contrib. 2002-059, Van-couver, Canada, Feb. 18, 2002.[9] T. Starr, J. Cioffi, and P. Silverman, Understanding DSL,Prentice Hall, 1999.[10] ATIS pre-ublished tech rep. ATIS-PP-0600024, ÒMulti-ple-Inut Multiple-Outut Crosstalk Channel Model,Ó2009.[11] M. Mohseni, G. Ginis, and J. M. Cioffi, ÒDynamic Spec-trum Management for Mixtures of Vectored and Non-44th Annual Conf. Info.Sciences and SysBIOGRAPHIESVLADIMIROKSMAN(vladimir.oksman@lantiq.com) receivedhis M.S. and Ph.D. from the Leningrad Radio and Telecom-ping the firstlong-distance digital transmission systems in USSR, earlypartici-pates in multiple standard development organizations.HEINRICHSCHENK(heirich.schenk@lantiq.com) received Dipl.Ing. and Dr. Ing. degrees from Friedrich Alexander Universi-p-ment Division. Since 1997 he has worked in the field oft and algorithm development in the SemiconductorDeartment of Siemens and later at Infineon. Presently heis a senior pal at Lantiq in Munich.AXELCLAUSEN(axel.clausen@lantiq.com) received his M.S.E.and Ph.D. in electrical engineering from Arizona State Uni-versity in 1995 and 1999, respectively. He held differentpositions in system and concept engineering for broad-band access, developing algorithms and system architec-ture for ADSL, VDSL, and home networking products.Currently, he is a princial system engineer in Lantiq, lead-ing the development of system architecture and algorithmsfor vectored VDSL. He was actively involved in standardiza-JOHNM. CIOFFI(jcioffi@assia-inc.com) is chairman and CEOof ASSIA Inc. and Hitachi Professor Emeritus, Stanford Uni-versity. He received his B.S.E.E. in 1978 from the Universityp.,1991Ð1999, and the designer of the world's first ADSLpatents. Affiliations andawards include the Boards of Alto Beam and ClariPhy; IEEEMEHDIMOHSENI(mmohseni@assia-inc.com) joined ASSIA,Inc., in 2006 as a systems architect, focusing on DSL net-work optimization and improving management systems.His research interests incorporate multi-user informationtheory and optimization of wireless and broadband net-works using convex optimization techniques. He holds aBachelorÕs degree in electrical engineering from Sharif Uni-GEORGEGINIS(gginis@assia-inc.com) is currently vice presi-dent of Expresse Engineering at ASSIA, Inc., overseeingdeveloment and deployment of the DSL Expresse manage-ment product. Between 2002 and 2005 he was a DSL sys-tems engineer at Texas Instruments. He holds a Diploma inelectrical and computer engineering from the NationalTechnical University of Athens, and M.S. and Ph.D. degreesin electrical engineering from Stanford University.CARLNUZMAN(nuzman@research.bell-labs.com) is a Distin-guished Member of Technical Staff in the Enabling Com-puting Technologies Department, Bell Labs, Alcatel-Lucent.Recent research interests include vectored DSL systems,power control, and optical networking. He holds aPh.D. in electrical engineering from Princeton University,JOCHENMAES(jochen.maes@alcatel-lucent.com) received hisM.S. and Ph.D. in physics in 2000 and 2004 fromKatholieke Universiteit Leuven, Belgium, for his research oncharge coupling in nano-structured semiconductor lasers of1.3 m and 1.55 m wavelengths. Since 2006 he has beenresearching next-generation broadband access systemspper at Bell Labs Alcatel-Lucent. He acts as apub-lished around 35 papMIGUELPEETERS(miguel.eeters@broadcom.com) is a gradu-ate in engineering from the Universite Libre de Bruxellesporationin 2000. For more than 10 years his activities have beenKEVINFISHER(kfisher@ikanos.com) has more than 23 yearsof experience in the development of firmware, software,and complex silicon devices. His experience includes execu-tive and senior-level engineering positions including CEO ofVector Silicon, VP of IC and DSP engineering at 2Wire, anddeartment manager of advanced recording channels atQuantum. He earned his Ph.D. E.E. and M.S.E.E from Stan-puter engineeringat the University of Illinois. He holds 20 U.S. pPER-ERIKERIKSSON(per-erik.s.eriksson@ericsson.com) is cur-rently a senior research engineer at Ericsson Researchworking with next-generation DSL technologies involvingpper. He actively particiates in theDSL standardization in ITU-T SG15/4. He holds an M.S. in G.vectorÕs large DSLbenefit is expectedto accelerate video,voice, wireless(through backhaul ofincreasingly smallercells offering morebandwidth to mobileusers) and other highrevenue-generatingtelecommunicationsservices, at a timewhen such servicesare of particularinterest. OKSMAN LAYOUT 9/21/10 11:45 AM Page 148 TANDARDS Vladimir Oksman, Heinrich Schenk, and Axel Clausen, Lantiq Inc.John M. Cioffi, Mehdi Mohseni, and George Ginis, ASSIA Inc. The ITU-TÕs New G.vector Standard OKSMAN LAYOUT 9/21/10 11:45 AM Page 140 IEEE Communications Magazine ¥ October 2010 143 divides the downstream frequency spectrum intoup to eight non-overlapping bands for flexibleGrouping error samples into blocks savesbackchannel bandwidth, since a common 4-bitThe VCE configures the bands and format oferror sample reporting, and can adjust the errorBackchannelÑ The backchannel conveys errorsamples from the VTU-R to the DSLAM. Threetransport mechanisms are specified for flexibleand over the embedded operations channelLater we describe the O-P-VECTOR 2 stageinitializationÕs use of the SOC-based transportA DMT symbolÕs repetition of the same infor-mation on several subcarriers achieves SOC pro-VDSL2 uses EOC to convey OLR and opera-tions, administration, and maintenance (OAM)The Ethernet backchannel has a flexible datarate: the error reports are encapsulated in Eth-By specifying both Ethernet and EOC mecha-nisms, G.993.5 offers a choice of cancellationFigure 3.Symbol alignment, sync-symbol alignment, and pilot bits in a vectored group. (Same in upstream and downstream; line is atthe beginning of the joining procedure). Line 1Data Data Data SyncPilot signalbit m Data Sync Data DataPilot signalbit m+1 Data Data Sync Data DataPilot signalbit m Data Line NTimeData Data Data Sync DataVDSL2 superframe VDSL2 superframe VDSL2 superframe Sync Data Data Data Data Sync Data Data Data Line 2Data Data Data Sync Data Sync Data Data Data Data SyncLine k(joining) Sync Sync Sync Data Data Data OKSMAN LAYOUT 9/21/10 11:45 AM Page 143 IEEE Communications Magazine ¥ October 2010 140 0163-6804/10/$25.00 © 2010 IEEE ABSTRACTThis article explores the recently issued ITU-T G.vector (G.993.5) [1] that allows expanded IEEE Communications Magazine ¥ October 2010 147 ¥Enabling or disabling a particular lineÕs¥Selecting FEXT cancellation frequency¥Assigning each lineÕs FEXT cancellation¥Assigning each lineÕs target data rateONCLUSIONS 5000 50 60 70 80 90 100150 1000 1500 2000 2500 3000 3500 4000 45005000 Distance (ft)Downstream and upstream reach performance(average over 24 lines)200Bit rate (Mb/s) 40 60 80 100120 1200 1000 1400 1600 1800 2000 2200 2400 2600 2800 3000 Vector DSVector US Measured rate: VDSL2 w/o xtalkMeasured rate: VDSL2 with xtalkMeasured rate: vectoring Figure 7.Setting of line priorities and data rate configuration parameters. Max data rateNot allowed Not allowed Allocate vectoring resources according to line priority (high/low). Do not exceed max data rate. Allocate vectoring resources to achieve target data rate. If resources are insufficient, sacrifice data rate up to min data rate. Target data Min data rate OKSMAN LAYOUT 9/21/10 11:45 AM Page 147 IEEE Communications Magazine ¥ October 2010 142 REPORTINGOF DownstreamFEXT VCE- Overall control- Update channel TX line 1SymbolencodingDownstreamDSLAM(a)(b) TX line 2Symbolencoding TX line NSymbolencodingDownstreamFEXT pre-coder IDFT IDFT IDFT RX line 1Symboldecoding RX line 2Symboldecoding RX line NSymboldecodingUpstream FEXTcanceller FEXT pre-coderof line k DFT DFT DFT Downstream Upstream AFEUpstreamFEXTCPVTU-R CPVTU-R CPVTU-RCable AFE AFE TX line kSymbolencodingEncoded symbolsof all other N-1vectored linesexcept line k IDFT AFE OKSMAN LAYOUT 9/21/10 11:45 AM Page 142 IEEE Communications Magazine ¥ October 2010 146 1.Successive single-line activation to measure2.All lines activated with no vectoring3.All lines activated with vectoring to recordHALLENGESFORFORthe narrow bandwidth they use.MANAGEMENTINTERFACEOF Data rates 17a Annex A; Annex A down, 26awg1000 60 80 100 120 140 160180 200 300 400 500 600 700 800 9001000 FEXT-freeVectored OKSMAN LAYOUT 9/21/10 11:45 AM Page 146 INTRODUCTIONModern life increasingly depends on faster Inter-net access for applications such as email, voice,Figure 1 shows digital subscriber line (DSL)technology as the current undisputed leader in for DSL in 1999 and 2000 [6]. A reduced com-plexity near-optimal implementation of linearvectoring was proposed in [7]. Complexity ofvectoring is still challenging, especially for largenumbers of vectored subscribers.Vectored DSL proposals first appeared in theAmerican National Standard Institute (ANSI) inthe 2001 DSM project [8] as the highest of threelevels of crosstalk noise control methods for cop-per loop management. Vectored DSLs havephysically separated customer premises (CP)locations that cannot be coordinated directly bya common customer-side controller. Recommen-dation G.993.5 describes the necessary interop-erable line coordination functions at the DSLaccess multiplexer (DSLAM) and the individuallinesÕ training protocols within the coordinatedvector group.G.993.5 STANDARDECHNOLOGY IEEE Communications Magazine ¥ October 2010 141Figure 1.Broadband access-connection for 2006Ð2009 (source: Point Topic). Q306 50 Subscribers (in millions) 0 100 150 200 250 300 350 Q406 Q107 Q207 Q307 Q407 Q108 Q208 Q308 Q408 Q109 Q209 FTTx OKSMAN LAYOUT 9/21/10 11:45 AM Page 141 IEEE Communications Magazine ¥ October 2010 145 ¥The bit rate as a function of loop length,¥The number of a binderÕs crosstalking lines¥The impact of adjacent-binder FEXT G.994.1 Handshakephase G.993.2 Channeldiscovery phaseVECTOR-1Sec VECTOR-1: SNR on line 26, frequency 2.95 MHz0.2 0 40 db 38 42 44 46 48 50 52 54 56 58 0.4 0.6 0.8 1 1.2 Sec VECTOR-2: SNR on line 32, frequency 2.95 MHz5 0 40 db 38 42 44 46 48 50 52 54 56 58 10 15 20 25 30 G.993.2 Channelanalysis and exchangephase Showtime G.993.2 Training VECTOR-2 Ideal coefficientsNo cancellation Ideal coefficients Practical vectoredperformance gainsdepend heavily onimplementationspecifics and service-provider deploymentpractices. Preserva-tion of vectoring performance gainscritically involves allactive cable pairs atleast to ascertainwhich pairs activelycontribute to anyspecific linesÕcrosstalk. OKSMAN LAYOUT 9/21/10 11:45 AM Page 145