TR (draft) V0.0.0 (2004-xx)

Transcription

1 Technical Report Transmission and Multiplexing (TM); Spectral management on metallic access networks; Part 2: Technical methods or perormance evaluations Work Item Reerence Permanent Document DTS/TM TM6(01)20 Filename m01p20a7.pd (rev 7) Date may 26 th, 2004 Rapporteur/Editor Rob F.M. van den Brink tel: (on behal o KPN) TNO Telecom ax: PO Box R.F.M.vandenBrink@telecom.tno.nl 2260 AK Leidschendam PO-Box 5050 Mark the new changes, valid since nov 24, GB Delt The Netherlands

2 2 Reerence DTS/TM Keywords spectral management, unbundling, access network, local loop, transmission, modem, POTS, IDSN, ADSL, HDSL, SDSL, VDSL, xdsl 650 Route des Lucioles F Sophia Antipolis Cedex - FRANCE Tel.: Fax: Siret N NAF 742 C Association à but non lucrati enregistrée à la Sous-Préecture de Grasse (06) N 7803/88 Important notice Individual copies o the present document can be downloaded rom: The present document may be made available in more than one electronic version or in print. In any case o existing or perceived dierence in contents between such versions, the reerence version is the Portable Document Format (PDF). In case o dispute, the reerence shall be the printing on printers o the PDF version kept on a speciic network drive within Secretariat. Users o the present document should be aware that the document may be subject to revision or change o status. Inormation on the current status o this and other documents is available at I you ind errors in the present document, send your comment to: editor@etsi.r Copyright Notiication No part may be reproduced except as authorized by written permission. The copyright and the oregoing restriction extend to reproduction in all media. European Telecommunications Standards Institute All rights reserved.

3 3 Contents Intellectual Property Rights...4 Foreword Scope Reerences Deinitions and abbreviations Deinitions Abbreviations Transmitter signal models or xdsl Generic transmitter signal model Cluster 2 transmitter signal models Transmitter signal model or "ISDN.2B1Q" Transmitter signal model or "ISDN.MMS.43" Cluster 3 transmitter signal models Transmitter signal models or "HDSL.2B1Q" Transmitter signal models or "HDSL.CAP" Transmitter signal model or "SDSL" Cluster 4 transmitter signal models Transmitter signal model or "ADSL over POTS" Transmitter signal model or "ADSL.FDD over POTS" Transmitter signal model or "ADSL over ISDN" Transmitter signal model or "ADSL.FDD over ISDN" Cluster 5 transmitter signal models Transmitter signal model or "VDSL" Generic receiver perormance models or xdsl Generic input models or eective SNR First order input model Second order input model (with residual distortion) Second order input model (with residual echo) Third order input model Generic detection models Generic Shited Shannon detection model Generic PAM detection model Generic CAP/QAM detection model Generic DMT detection model Generic models or echo coupling Linear echo coupling model Speciic receiver perormance models or xdsl Receiver perormance model or "HDSL.2B1Q" Receiver perormance model or "HDSL.CAP" Receiver perormance model or "SDSL" Receiver perormance model or "ADSL over POTS" (EC) Building blocks o the receiver perormance model Parameters o the receiver perormance model Receiver perormance model or "ADSL.FDD over POTS" Receiver perormance model or "ADSL over ISDN" (EC) Building blocks o the receiver perormance model Parameters o the receiver perormance model Receiver perormance model or "ADSL.FDD over ISDN Receiver perormance model or "VDSL" Transmission and relection models Summary o test loop models...32

4 4 8 Cross talk models Overview o dierent network topologies Validity limitations o cross talk modeling Generic cross talk models or two-node co-location Basic diagram or two-node topologies Models or cross talk cumulation FSAN sum or cross talk cumulation Models or cross talk coupling Basic models or equivalent NEXT and FEXT Models or cross talk injection Forced noise injection Current noise injection Generic cross talk models or multi-node co-location Examples o evaluating various scenarios Example scenario A Assumed coniguration Assumed conditions Evaluated perormance or scenario A Example scenario B Example scenario C Example scenario D...39 Annex A: Bibliography...39 History...39 Intellectual Property Rights IPRs essential or potentially essential to the present document may have been declared to. The inormation pertaining to these essential IPRs, i any, is publicly available or members and non-members, and can be ound in SR : "Intellectual Property Rights (IPRs); Essential, or potentially Essential, IPRs notiied to in respect o standards", which is available rom the Secretariat. Latest updates are available on the Web server ( Pursuant to the IPR Policy, no investigation, including IPR searches, has been carried out by. No guarantee can be given as to the existence o other IPRs not reerenced in SR (or the updates on the Web server) which are, or may be, or may become, essential to the present document. Foreword This Technical Report (TR) has been produced by Technical Committee Transmission and Multiplexing (TM). The present document is part 2 o a multi-part deliverable covering Transmission and Multiplexing (TM); Acces networks; Spectral management on metallic access networks, as identiied below: Part 1: Part 2: Part 3: "Deinitions and signal library". "Technical methods or perormance evaluations. "Construction methods or spectral management rules. NOTE: Part 3 is under preparation.

5 5 1 Scope The present document gives guidance on a common methodology or studying the impact on xdsl perormance (maximum reach, noise margin, maximum bitrate) in noisy cables when changing parameters within various Spectral Management scenarios. These methods enable reproducible results and a consistent presentation o the assumed conditions (characteristics o cables and xdsl equipment) and coniguration (choosen technology mixture and cable ill) o each scenario. The technical methods include computer models or calculating: xdsl receiver capability o detecting signals under noisy conditions; xdsl transmitter characteristics; cable characteristics cross talk cumulation in cables, originating rom a mix o xdsl disturbers; The objective is to provide the technical means or evaluating the perormance o xdsl equipment within a chosen scenario, such as calculations and measurements. This includes the description o perormance properties o equipment. Another objective is to assist the reader with applying this methodology by providing examples on how to speciy the coniguration and the conditions o a scenario in an unambiguous way. The distinction is that a coniguration o a scenario can be controlled by access rules while the conditions o a scenario cannot. Possible applications o this document include: Studying access rules, or the purpose o bounding the cross talk in unbundled networks. Studying deployment rules, or the various systems present in the access network. Studying the impact o cross talk on various technologies within dierent scenarios The scope o this Spectral Management document is explicitly restricted to the methodology or deining scenarios and quantiying the perormance o equipment within such a scenario. All judgement on what access rules are required, what perormance is acceptable, or what combinations are spectral compatible, is explicitly beyond the scope o this document. The same applies or how realistic the example scenarios are. The models in this document are not intended to set requirements or DSL equipment. These requirements are contained in the relevant transceiver speciications. The models in this document are intended to provide a reasonable estimate o real-world perormance but may not include every aspect o modem behaviour in real networks. Thereore real-world perormance may not accurately match perormance numbers calculated with these models. 2 Reerences For the purposes o this Technical Report (TR) the ollowing reerences apply: SpM [1] TR " Transmission and Multiplexing (TM); Spectral Management on metallic access networks; Part 1: Deinitions and signal library V1.3.1 ( ), dec [2] ANSI T1E1.4/ R6 "Spectrum Management or loop transmission systems" drat; revision 6, November 2000 (or a more recent version)

6 6 ISDN HDSL SDSL ADSL [3] TS (V1.3.2): "Transmission and Multiplexing (TM); Integrated Services Digital Network (ISDN) basic rate access; Digital transmission system on metallic local lines". [4] TS (V1.5.3): "Transmission and Multiplexing (TM); High bit-rate Digital Subscriber Line (HDSL) transmission systems on metallic local lines; HDSL core speciication and applications or combined ISDN-BA and kbit/s transmission". [5] TS , v1.1.3: "Transmission and Multiplexing (TM); Access transmission system on metallic access cables; Symmetrical single pair high bitrate Digital Subscriber Line (SDSL)". Nov [6] ITU-T Recommendation G (2001): "Single-Pair High-Speed Digital Subscriber Line (SHDSL) transceivers". [7] TS , v1.3.1, ( ): "Transmission and Multiplexing (TM); Access transmission systems on metallic access cables; Asymmetric Digital Subscriber Line (ADSL) - European speciic requirements", may [8] ITU-T Recommendation G (1999): "Asymmetric digital subscriber line (ADSL) transceivers". 3 Deinitions and abbreviations 3.1 Deinitions For the purposes o the present documents on spectral management, the ollowing terms and deinitions apply: Local Loop Wiring: Part o an access network, terminated by well-deined ports, or transporting signals over a distance o interest. This part includes mainly cables, but may also include a main distribution rame (MDF), street cabinets, and other distribution elements. The local loop wiring is usually passive only, but may include active splitterilters as well. Loop provider: Organization acilitating access to the local loop wiring. (NOTE: In several cases the loop provider is historically connected to the incumbent network operator, but other companies may serve as loop provider as well.) Network operator: Organization that makes use o a local loop wiring or transporting telecommunication services. (NOTE: This deinition covers incumbent as well as competitive network operators.) Access Port: An Access Port is the physical location, appointed by the loop provider, where to inject signals (or transmission purposes) into the local loop wiring. NT-access port (or NT-port or short): is an access port or injecting signals, labeled by the loop provider as "NT-port". (NOTE: Such a port is commonly located at the customer premises, and intended or injecting "upstream" signals.) LT-access port (or LT-port or short): is an access port or injecting signals, labeled as labeled by the loop provider as "LT-port". (NOTE: Such a port is commonly located near the telecommunication exchange, and intended or injecting "downstream" signals.)

7 7 Transmission technique: electrical technique used or the transportation o inormation over electrical wiring. Transmission equipment: equipment connected to the local loop wiring that uses a transmission technique to transport inormation. Transmission system: A set o transmission equipment that enables inormation to be transmitted over some distance between two or more points. Upstream transmission: transmission direction rom a port, labelled as NT-port, to a port, labelled as LT-port. This direction is usually rom the customer premises, via the local loop wiring, to the telecommunication exchange. Downstream transmission: transmission direction rom port, labelled as LT-port, to a port, labelled as NT-port. This direction is usually rom the telecommunication exchange via the local loop wiring, to the customer premises. Noise margin: the ratio by which the received noise may increase until the recovered signal does not meet the predeined quality criteria. This ratio is commonly expressed in db. Signal margin: the ratio by which the received signal may decrease until the recovered signal does not meet the predeined quality criteria. This ratio is commonly expressed in db. Max data rate: the maximum data rate that can be recovered according to predeined quality criteria, when the received noise is increased with a chosen noise margin (or the received signal is decreased with a chosen signal margin). Perormance: is a measure o how well a transmission system ulills deined criteria under speciied conditions. Such criteria include reach, bit rate and noise margin. Access Rule: Mandatory rule or achieving access to the local loop wiring, equal or all network operators who are making use o the same network cable that bounds the cross talk in that network cable. Deployment Rule: Voluntary rule, irrelevant or achieving access to the local loop wiring and proprietary or each individual network operator. (NOTE: A deployment rule relects a network operator's own view about what the maximum length or maximum bit rate may be or oering a speciic transmission service to ensure a chosen minimum quality o service.) Spectral management rule: A generic term, incorporating (voluntary) deployment rules, (mandatory) access rules and all other (voluntary) measures to maximize the use o local loop wiring or transmission purposes. Spectral management: The art o making optimal use o limited capacity in (metallic) access networks. This is or the purpose o achieving the highest reliable transmission perormance and includes: Designing o deployment rules and their application. Designing o eective access rules. Optimized allocation o resources in the access network, e.g. access ports, diversity o systems between cable bundles, etc. Forecasting o noise levels or ine-tuning the deployment. Spectral policing to ensure network integrity. Making a balance between conservative and aggressive deployment (low or high ailure risk). Cable management plan (CMP): A list o selected access rules dedicated to a speciic network. This list may include associated descriptions and explanations. Spectral compatibility: A generic term or the capability o transmission systems to operate in the same cable. The precise deinition is application dependent and has to be deined or each group o applications. Cable ill: (or degree o penetration): number and mixture o connected transmission techniques to the ports o a binder or cable bundle that are injecting signals into the access ports.

8 8 Signal Category: is a class o signals meeting the minimum set o speciications identiied in -TR (NOTE: Some signal categories may distinct between dierent sub-classes, and may label them or instance as signals or "downstream" or or "upstream" purposes.) 3.2 Abbreviations For the purposes o the present document, the ollowing abbreviations apply: ADSL BER CAP DMT FDD HDSL ISDN LT-port LTU NT-port NTU PAM PSD QAM REC SDSL SNR TRA VDSL xdsl 2B1Q Asymmetric Digital Subscriber Line Bit Error Ratio Carrierless Amplitude/Phase modulation Discrete Multitone modulation Frequency Division Duplexing/Duplexed High bit rate Digital Subscriber Line Integrated Services Digital Network Line Termination port (commonly at central oice side) Line Termination Unit Network Termination port (commonly at customer side) Network Termination Unit Pulse Amplitude modulation Power Spectral Density (single sided) Quadrature Amplitude modulation Receiver Symmetrical (single pair high bitrate) Digital Subscriber Line Signal to Noise Ration (ratio o powers) Transmitter Very-high-speed Digital Subscriber Line (all systems) Digital Subscriber Line 2-Binary, 1-Quarternairy (Special variant o a 4-level PAM linecode) 4 Transmitter signal models or xdsl A transmitter model in this clause is mainly a PSD description o the transmitted signal under matched conditions, plus an output impedance description to cover mis-matched conditions as well. PSD masks o transmitted xdsl signals are speciied in several documents or various purposes, or instance in Part 1 o Spectral Management [1]. These PSD masks, however, cannot be applied directly to the description o a transmitter model. One reason is that masks are speciicing an upper limit, and not the expected (averaged) values. Another reason is that the deinition o the true PSD o a time limited signal requires no resolution bandwidth at all (it is deined by means o an autocorrelation, ollowed by a Fourier transorm) while PSD masks do rely on some resolution bandwidth. They describe values that are (a slightly) dierent rom the true PSD, especially at steep edges (e.g. guard bands), and or modeling purposes this dierence is sometimes very relevant. To dierentiate between several PSD descriptions, masks and templates o a PSD are given a dierent meaning. Masks are intended or proving compliance to standard requirements, while templates are intended or modeling purposes. This clause summarizes various xdsl transmitter models, by deining template spectra o output signals. 4.1 Generic transmitter signal model A generic model o an xdsl transmitter is essentially a linear signal source. The Thevenin equivalent o such a source equals an ideal voltage source U s having a real resistor R s in series. The output voltage o this source is random in nature (as a unction o the time), is uncorrelated with any other transmitter signal, and occupies a relatively broad spectrum.

9 9 This generic model can be made speciic by deining: The output impedance R s o the transmitter. The template o the PSD, measured at the output port, when terminated with an external impedance equal to R s. This is identiied as the matched condition, and under these conditions the output power equals the maximum power that is available rom this source. Under all other (mis-matched) termination conditions the output power will be lower. 4.2 Cluster 2 transmitter signal models Transmitter signal model or "ISDN.2B1Q" The PSD template or modeling the "ISDN.2B1Q" transmit spectrum is deined by the theoretical sinc-shape o PAM encoded signals, with additional iltering and with a noise loor. The PSD is the maximum o both power density curves, as summarized in expression 1 and the associated table 1. The coeicient q N scales the total signal power o P 1 () to a value that equals P ISDN. This value is dedicated to the used ilter characteristics, but q N =1 when no iltering is applied ( L 0, H ). The source impedance equals 135Ω. P ( ) = P 1 ISDN 2 q X N 2 sinc X 1+ 1 H 2 N H 1 1+ L 2 [ W / Hz] 10 P ( ) = 2 ( Ploor _ dbm /10) 1000 [ W / Hz] P( ) = max Where: ISDN ( P ( ), P ( )) [ W / Hz] 1 ( 10 ) P /10 ISDN _ dbm 1000 P = [W] 2 R S = 135 [Ω ] sinc(x) = sin(π x) / (π x) Deault values or remaining parameters are summarized in table 1. Expression 1: PSD template or modeling "ISDN.2B1Q" signals. Dierent ISDN implementations, may use dierent ilter characteristics, and noise loor values. Table 1 speciies deault values or ISDN implementations, in case 2 nd order Butterworth iltering has been applied. The deault noise loor equals the maximum PSD level that meets the out-o-band speciication o the ISDN standard [3]. In case these deault values are not appropriated or speciic perormance studies, other values may apply as well (provided that they are speciied or these studies). Type X H L N H q N P ISDN_dBm P loor_dbm [khz] [khz] [khz] [dbm] [dbm/hz] ISDN.2B1Q 80 1 x Table 1: Deault parameter values or the ISDN.2B1Q templates, as deined in expression 1. These deault values are based on 2 nd order Butterworth iltering.

10 Transmitter signal model or "ISDN.MMS.43" <or urther study> 4.3 Cluster 3 transmitter signal models Transmitter signal models or "HDSL.2B1Q" The PSD templates or modeling the spectra o various "HDSL.2B1Q" transmitters is deined by the theoretical sincshape o PAM encoded signals, with additional iltering and a noise loor. The PSD template is the maximum o both power density curves, as summarized in table 2. The coeicient q N scales the total signal power o P 1 () to a value that equals P 0. This value is dedicated to the used ilter characteristics, but equals q N =1 when no iltering is applied ( L 0, H ), The source impedance equals 135Ω. P ( ) = P 1 HDSL 2 q X N 2 sinc X 1 1+ L H1 2 N H H 2 2 NH 2 [ W / Hz] 10 P ( ) = 2 ( Ploor _ dbm /10) 1000 [ W / Hz] P( ) = max Where: ( P ( ), P ( )) 1 ( 10 ) P /10 HDSL _ dbm 1000 P = [W] HDSL 2 R S = 135 [Ω ] sinc(x) = sin(π x) / (π x) Deault values or remaining parameters are summarized in table 2. Expression 2: PSD template or modeling "HDSL.2B1Q" signals. [ W / Hz] Dierent HDSL implementations, may use dierent ilter characteristics, and noise loor values. Table 2 summarizes deault values or modeling HDSL transmitters, and alternative values in case higher order Butterworth iltering has been applied to dedicated implementations. It is recommended to use the deault values or spectral management studies, unless motivated why alternative values are more appropriated. The deault power level P HDSL equals the maximum power allowed by the HDSL standard [4], since a nominal speciication does not exist. The deault noise loor P loor equals a value observed or various implementations. When these measurements were not available, the maximum PSD level was chosen here that meets the out-o-band speciication o the HDSL standard [4].

11 11 Deault Type X L P loor_dbm dbm/hz H1 N H1 H2 N H2 q N P HDSL_dBm khz khz dbm HDSL.2B1Q/ x 3 N/A N/A HDSL.2B1Q/ x 3 N/A N/A HDSL.2B1Q/ x 3 N/A N/A Alternatives Type X khz L khz H1 N H1 H2 N H2 q N P HDSL_dBm dbm H2.1 HDSL.2B1Q/ x 4 N/A N/A H2.2 HDSL.2B1Q/ x x P loor_dbm dbm/hz Table 2: Deault parameter values or the HDSL.2B1Q templates, as deined in expression 2. The alternative values are based on higher order Butterworth iltering. Choose H2 = and N H2 =1 when not applicable (N/A) Transmitter signal models or "HDSL.CAP" The PSD templates or modeling signals generated by HDSL.CAP transmitters are dierent or single-pair and two-pair HDSL systems. The PSD templates or modeling the "HDSL.CAP/2" and "HDSL.CAP/1" transmit spectra or two-pair and single-pair systems are deined in terms o break requencies, as summarized in table 3. These template are taken rom the nominal shape o the transmit signal spectra, as speciied in the HDSL standard [4] The associated values are constructed with straight lines between these break requencies, when plotted against a logarithmic requency scale and a linear dbm scale. The source impedance equals R s =135Ω. HDSL.CAP/2 2 pair HDSL.CAP/1 1- pair 135 Ω 135 Ω [Hz] [dbm/hz] [Hz] [dbm/hz] ,98 k -57 <TBD> <TBD> 21,5 k ,02 k ,58 k ,10 k ,62 k ,00 k -70 1,188 M M -120 Table 3. PSD template values at break requencies or modeling "HDSL.CAP/2" and "HDSL.CAP/1" NOTE: A PSD template or HDSL.CAP/1 is currently or urther study Transmitter signal model or "SDSL" The PSD templates or modeling the spectra o "SDSL" transmitters is deined by the theoretical sinc-shape o PAM encoded signals, plus additional iltering and a noise loor. The transmit spectrum is deined in three distinct requency bands, as summarized in expression 3 and the associated table 4. The break requency int is the requency where the curves or P 1 () and P 2 () intersect. This PSD template is taken rom the nominal shape o the transmit signal spectrum, as speciied in the SDSL standard [5]. The source impedance equals R s =135Ω.

12 12 < int : K sdsl P1 ( ) = R s X 2 sinc X NH ( ) 1+ ( L ) H 2 [ W / Hz] int 1,5MHz : P ( ) = K 2 x 0 1,5 [ W / Hz] > 1,5 MHz : P ( ) = [ dbm / Hz] R s = 135 Ω sinc(x) = sin(π x) / (π x) int = is the lowest requency above H where the expressions or P 1() and P 2() intersect Parameter values are deined in table 4 Expression 3. PSD Tempate values or modeling both the symmetric and asymmetric modes o SDSL Mode Data Rate R TRA Symbol Rate sym X H L 0 N H K SDSL K X [kb/s] [kbaud] [khz] [Hz] [V 2 ] [W/Hz] Sym < 2048 both (R+ 8 kbit/s)/3 sym X/ Sym 2048 both (R+ 8 kbit/s)/3 sym X/ Asym 2048 LTU (R+ 8 kbit/s)/3 2 sym x 2/ Asym 2048 NTU (R+ 8 kbit/s)/3 sym x 1/ Asym 2304 LTU (R+ 8 kbit/s)/3 2 sym x 3/ Asym 2304 NTU (R+ 8 kbit/s)/3 sym x 1/ Table 4. Parameter values or the SDSL templates, as deined in expression 3. <POWER BACK-OFF HAS BEEN LEFT FOR FURTHER STUDY> 4.4 Cluster 4 transmitter signal models Transmitter signal model or "ADSL over POTS" The PSD template or modeling the "ADSL over POTS" transmit spectrum (EC variant) is deined in terms o break requencies, as summarized in table 5. The associated values are constructed with straight lines between these break requencies, when plotted against a logarithmic requency scale and a linear dbm scale. The requency in this table reers to the sub-carrier spacing o the DMT tones o ADSL. The source impedance equals 100Ω. ADSL over POTS (EC) Up ADSL over POTS (EC) Down DMT carriers [k 1:k 2] [7:31] DMT carriers [k 3:k 4] [7:255] [Hz] P [dbm/hz] [Hz] P [dbm/hz] k k k k ( 28.03) ( 28.03) ( ) ( ) ( ) -90 x = <TBD> k M M M M M M M -112 = khz = khz Table 5. PSD template values at break requencies or modeling "ADSL over POTS" <POWER BACK-OFF HAS BEEN LEFT FOR FURTHER STUDY>

13 Transmitter signal model or "ADSL.FDD over POTS" The PSD template or modeling "ADSL.FDD over POTS" transmit spectra is deined in terms o break requencies, as summarized in table 6 and 7. The associated values are constructed with straight lines between these break requencies, when plotted against a logarithmic requency scale and a linear dbm scale. The requency in this table reers to the sub-carrier spacing o the DMT tones o ADSL. Table 6 is to be used or modeling "guard band FDD modems", usually equipped with steep iltering or improving the separation between up and downstream signals. They leave 7 tones unused to enable this guard band. Table 7 is to be used or modeling "adjacent FDD modems", usually enhanced by echo cancellation or improving the separation between up and downstream signals. Because a guard band is not needed here, only 1 tone is let unused. The source impedance equals 100Ω. Guard band FDD (using ilters) ADSL.FDD over POTS Up ADSL.FDD over POTS Down DMT carriers [k 1:k 2] [7:30] DMT carriers [k 3:k 4] [38:255] [Hz] P [dbm/hz] [Hz] P [dbm/hz] k k k k ( 28.03) ( ) ( ) ( ) ( ) ( ) k ( ) M -100 x = <TBD> M M M M M M -112 = khz = khz Table 6. PSD template values at break requencies or modelling "ADSL.FDD over POTS", implemented as "guard band FDD" (with iltering). This PSD allocates 7 unused tones; Adjacent FDD (using echo cancellation) ADSL.FDD over POTS Up ADSL.FDD over POTS Down DMT carriers [k 1:k 2] [7:31] DMT carriers [k 3:k 4] [33:255] [Hz] P [dbm/hz] [Hz] P [dbm/hz] k k k k ( 28.03) ( 97.03) ( ) ( ) ( ) ( ) k ( ) M -100 x = <TBD> M M M M M M -112 = khz = khz Table 7. PSD template values at break requencies or modelling "ADSL.FDD over POTS", implemented as "adjacent FDD" (with echo canceling). This PSD allocates 1 unused tone, since a guard band is not required here. <POWER BACK-OFF HAS BEEN LEFT FOR FURTHER STUDY>

14 Transmitter signal model or "ADSL over ISDN" The PSD template or modeling the "ADSL over ISDN" transmit spectrum (EC variant) is deined in terms o break requencies, as summarized in table 8. The associated values are constructed with straight lines between these break requencies, when plotted against a logarithmic requency scale and a linear dbm scale. The requency in this table reers to the sub-carrier spacing o the DMT tones o ADSL. The source impedance equals 100Ω. ADSL over ISDN (EC) Up ADSL over ISDN (EC) Down DMT carriers [k 1:k 2] [33:63] DMT carriers [k 3:k 4] [33:255] [Hz] P [dbm/hz] [Hz] P [dbm/hz] k ( 97.03) ( 97.03) ( ) ( ) ( 273,84) ( ) ( ) -55 x = <TBD> ( ) M ( ) M M M M M M -112 = khz = khz Table 8. PSD template values at break requencies or modeling "ADSL over ISDN (EC)" <POWER BACK-OFF HAS BEEN LEFT FOR FURTHER STUDY> Transmitter signal model or "ADSL.FDD over ISDN" The PSD template or modeling "ADSL.FDD over ISDN" transmit spectra is deined in terms o break requencies, as summarized in table 9 and 10. The associated values are constructed with straight lines between these break requencies, when plotted against a logarithmic requency scale and a linear dbm scale. The requency in this table reers to the sub-carrier spacing o the DMT tones o ADSL. Table 9 is to be used or modeling "guard band FDD modems", usually enhanced by steep iltering or improving the separation between up and downstream signals. They leave 7 tones unused to enable this guard band. Table 10 is to be used or modeling "adjacent FDD modems", usually enhanced by echo cancellation or improving the separation between up and downstream signals. Because a guard band is not needed here, no tone is let unused. The source impedance equals 100Ω.

15 15 Guard band FDD (using ilters) ADSL.FDD over ISDN Up ADSL.FDD over ISDN Down DMT carriers [k 1:k 2] [33:56] DMT carriers [k 3:k 4] [64:255] [Hz] P [dbm/hz] [Hz] P [dbm/hz] c = c = c = c = c = c = c = c = x = <TBD> c = M c = M M M M M M -112 c = khz c = khz Table 9. PSD template values at break requencies or modeling "ADSL.FDD over ISDN", implemented as "guard band FDD" (with iltering). This PSD allocates 7 unused tones. adjacent FDD (using echo cancellation) ADSL.FDD over ISDN Up ADSL.FDD over ISDN Down DMT carriers [k 1:k 2] [33:63] DMT carriers [k 3:k 4] [64:255] [Hz] P [dbm/hz] [Hz] P [dbm/hz] c = c = c = c = c = c = c = c = x = <TBD> c = M c = M M M M M M -112 c = khz c = khz Table 10. PSD template values at break requencies or modeling "ADSL.FDD over ISDN", implemented as "adjacent FDD" (with echo canceling). This PSD has no guard band. <POWER BACK-OFF HAS BEEN LEFT FOR FURTHER STUDY> 4.5 Cluster 5 transmitter signal models Transmitter signal model or "VDSL" <or urther study>

16 16 5 Generic receiver perormance models or xdsl A receiver perormance model is capable o predicting up to what perormance a data stream can be recovered rom a noisy signal. In all cases it assumes that this recovery meets predeined quality criteria such as a maximum BER (Bit Error Ratio). Values like BER<10 7, during a time interval o several minutes, are not uncommon. The word perormance reers within this context to a variety o quantities, including noise margin, signal margin and max data rate. When the receiver is ideal (zero internal receiver noise, ininite echo cancellation, etc), quantities like noise margin and signal margin become equal. Perormance models are implementation and line code speciic. Perormance modeling becomes more convenient when broken down into a cascade o smaller sub models: A line code independent input (sub)model that evaluates the eective SNR rom received signal, received noise, and various receiver imperections. Details are described in clause 5.1. A line code dependent detection (sub)model that evaluates the perormance (e.g. the noise margin at speciied bit rate) rom the eective SNR. Details are described in clause 5.2. An echo coupling (sub)model that evaluates what portion o the transmitted signal lows into the receiver. Details are described in clause 5.3. This clause details all the above mentioned sub models, being used or evaluating the perormance o receivers under noise conditions. This clause 5 is dedicated to generic perormance models only. Clause 6 is dedicated to speciic models by assigning values to all parameters o a generic model. 5.1 Generic input models or eective SNR An input (sub) model describes how to evaluate the eective SNR, as intermediate result (see igure 1), rom various input quantities and imperections. received signal P RS received noise P RN input block Eective SNR detection block echo P RE transmitted signal echo coupling P TS Transmitter block (or opposite direction) xdsl transceiver Figure 1: Flow diagram o a transceiver model, build up rom individual sub models. On input, the input model or eective SNR requires values or signal, noise and echo. The low diagram in igure 1 illustrates this or an xdsl transceiver that is connected via a common wire pair to another transceiver (not shown).

17 17 The received signal power P RS carries the data that is to be recovered. This signal originates rom the transmitter at the other side o the wire pair, and its level is attenuated by cable loss. The received noise power P RN is all that is received when the transmitters at both sides o the link under study are silent. The origin o this noise is mainly cross talk rom internal disturbers connected to the same cable (cross talk noise), and partly rom external disturbers (ingress noise). The received echo power P RE is all that is received when the transmitter at the other end o the wire pair is silent, as well as all internal and external disturbers. It is a residue that will be received when a transmitter and a receiver are combined into a transceiver en co-connected via a hybrid to the same wire pairs. When the hybrid o that transceiver is unbalanced due to mismatched termination impedances (o the cable), then a portion (P RE ) o the transmitted signal (P TS ) will leak into the receiver and is identiied as echo. On output, the input model evaluates a quantity called eective SNR (Signal to noise Ratio) that indicates to what degree the received signal is deteriorated by noise, residual echo and all kinds o implementation imperections. Due to signal processing in the receiver, the input SNR (the ratio between signal power, and the power-sum o noise and echo) will change into the eective SNR at some virtual internal point at the receiver. The eective SNR can be better or worse then the input SNR. Receivers with build-in echo cancellation can take advantage o a-priori knowledge on the echo, and can suppress most o this echo and thus improving the eective SNR. On the other hand, all analog receiver electronics produce shot noise and thermal noise, the A/D-converter produces quantization noise, and the equalization has its limitations as well. The combination o all these individual imperections deteriorates the eective SNR. In principle all parameters o the eective SNR can be assumed as requency dependent, but this dependency has oten been omitted here or reasons o simplicity. In addition, external change o signal and noise levels will modiy the value o this eective SNR. Eective SNR, in oset ormat or margin evaluations To simpliy urther analysis o perormance quantities like noise margin and signal margin, the eective SNR is oten expressed in its oset ormat, characterized by an additional parameter m. The associated expression is deined or each model individually. With this parameter m the external noise level can be increased (or noise margin calculations) or the external signal level can be decreased (or signal margin calculations). The convention is that when m=1 (equals zero db) the eective oset SNR equals the eective SNR itsel. When the value o parameter m increases, the eective oset SNR decreases First order input model This input model is quite a simpliied model that assumes that the SNR o the input signal is internally modiied by internal receiver noise (P RN0 ). Most imperections o the receiver (such as imperect echo suppression, imperect equalization and quantization noise) are assumed to be concentrated in a single virtual internal noise source (P RN0 ). Figure 2 shows the low diagram o an xdsl transceiver model that incorporates a linear irst order model or eective SNR evaluation.

18 18 received signal P RS (First order) input model Receiver Eective received noise P RN SNR detection block P RN0 internal receiver noise transmitted signal P TS Transmitter block (or opposite direction) xdsl transceiver Figure 2: Flow diagram o a transceiver model that incorporates a linear irst order input model or the determination o the eective SNR. Expression 4 summarizes how to evaluate the eective SNR or this model, and it has been speciied in plain and oset ormats. Table 11 summarizes the involved parameters. Plain ormat: SNR() = Noise oset ormat: SNR os,n(m, ) = Signal oset ormat: SNR os,s(m, ) = P P RN RN P RS + P P RN 0 RS m + P PRS / m P + P Expression 4: Eective SNR, in various ormats, or a irst order input model RN RN 0 RN0 Input quantities linear In db remarks Received signal power P RS 10 log 10(P RS) Frequency dependent Received crosstalk noise P RN 10 log 10(P RN) External noise Model Parameters Receiver noise power P RN0 10 log 10(P RN0) Internal noise Output quantities Signal to noise ratio (eective) SNR 10 log 10(SNR) Frequency dependent Table 11: Involved parameters and quantities or a irst order input model Second order input model (with residual distortion) <or urther study> Second order input model (with residual echo) <or urther study; may be removed>

19 Third order input model <or urther study; may be removed> 5.2 Generic detection models This clause identiies several generic (sub) models or the detection block: one line code independent model derived rom the Shannon capacity limit, and various line code dependent models dedicated to PAM, CAP/QAM or DMT line coding. Table 12 summarizes the naming convention or input and output quantities. Input quantities linear In db remarks Signal to Noise Ratio SNR 10 log 10(SNR) Ratio o powers (requency dependent) Output quantities Noise margin m n 10 log 10(m n) Ratio o noise powers Signal margin m s 10 log 10(m s) Ratio o signal powers Table 12. Symbols used or input and output quantities o detection models On input, the detection block requires an eective SNR, as provided by the input block. This SNR is a unction o the requency. When the oset ormat is used or describing the SNR, it will also be a unction o the oset parameter m.this oset ormat is speciied individually or each model in clause 5.1. On output, the detection block evaluates a signal margin m n (or a noise margin m s when more appropriated). This margin parameter is a dominant measure or the transport quality that is achieved under noisy conditions. The Noise Margin m n indicates how much the received noise power can increase beore the transmission becomes unreliable. The Signal Margin m s indicates how much the received signal power can decrease beore the transmission becomes unreliable. Unless explicitly speciied otherwise, the word margin reers in this document to noise margin. NOTE From an xdsl deployment point o view, the analysis o noise margin is preerred over signal margin, since the (cross talk) noise is the quantity that may increase when more systems are connected to the same cable. Many xdsl implementations, however, do report margin numbers that are not exactly equal to this noise margin, since the detection circuitry cannot make a distinction between external noise (due to cross talk) and internal noise (due to imperect electronics). These margins are oten an estimate closer in value to the signal margin then the noise margin.

20 Generic Shited Shannon detection model The calculation o the margin m using the generic Shited Shannon detection model, is equivalent with solving the equation in expression 5. It has been derived rom Shannon's capacity theorem, by reducing the eective SNR ("shiting" on a db scale) by the SNR-gap Γ, to account or the imperections o practical detectors. The associated parameters are summarized in table 13. The eective SNR is to be evaluated by using one o the input models described in clause 5.1. Depending on what oset ormat is used or the SNR expression (see clause 5.1), the calculated margin m will represent the noise margin m n or the signal margin m s. b c + B / 2 = c B / 2 SNRos log 2 1+ Γ ( m, ) d Expression 5: Equation o the Shited Shannon detection model, or solving the margin m. Model Parameters linear In db remarks SNR gap Γ 10 log 10(Γ) Data rate d all payload bits that are transported in 1 sec Line rate b = DateRate + overhead bitrate Bandwidth B Width o most relevant spectrum Table 13. Parameters used or Shited Shannon detection models. The various parameters used within this generic detection model are summarized in table 13. The model can be made speciic by assigning values to all these model parameters. The SNR-gap (Γ) is a perormance parameter that indicates how close the detection approaches the Shannon capacity limit. The linerate is usually higher then the data rate (0 30%) to transport overhead bits or error correction, signaling and raming. The Bandwidth is a parameter that indicates what portion o the received spectrum is relevant or data transport. The model assumes that this portion passes the receive ilters Generic PAM detection model The calculation o the margin m using the generic PAM detection model is equivalent with solving the equation in expression 6. This model assumes ideal decision eedback equalizer (DFE) margin calculations. The associated parameters are summarized in table 14. The eective SNR is to be evaluated by using one o the input models described in clause 5.1. Depending on what oset ormat is used or the SNR expression (see clause 5.1), the calculated margin m will represent the noise margin m n or the signal margin m s. SNR req s N H 2 b = Γ ( ) = exp ln1 + SNRos ( m, + n s ) s 0 n= N L d Expression 6: Equation o the PAM-detection model, or solving the margin m.

21 21 The SNR gap Γ, being used in the above expression 6, is a combination o various eects. This Γ parameter is oten split-up into the ollowing three parts: Its theoretical value Γ PAM (in the order o 9.75 db, at BER=10 7 ) A theoretical coding gain Γ coding (usually in the order o 3-5 db), to indicate how much additional improvement is achieved by the chosen coding mechanism. An empirical implementation loss Γ impl (usually 1.6 db or more), indicating how much overall deterioration is caused by implementation dependent imperections in echo cancellation, equalization, etc, without identiying its true cause. When Γ is split-up into the above three parts, its value shall be evaluated as ollows: SNR gap (linear): Γ = Γ PAM / Γ coding Γ impl SNR gap (in db): Γ _db = Γ PAM_dB Γ coding_db + Γ impl_db Model Parameters linear In db remarks SNR gap (eective) Γ 10 log 10(Γ) = SNR req / (2 2 b 1) SNR gap in parts: Γ PAM 10 log 10(Γ PAM) Modulation gap or PAM Γ coding 10 log 10(Γ coding) Coding gain Γ impl 10 log 10(Γ impl) Implementation loss Required SNR SNR req 10 log 10(SNR req) = Γ (2 2 b 1) Data rate d all payload bits that are transported in 1 sec Line rate b = DateRate + overhead bitrate Symbol rate s = b / b Bits per symbol b = b / s (can be non-integer) Summation range N L, N H On deault: N L= 2 and N H=+1 Table 14. Parameters used or PAM detection models. The various parameters in table 14 used within this generic detection model have the ollowing meaning: The SNR-gap (Γ) and required SNR (SNR req ) are equivalent parameters and can be converted rom one to the other. The advantage o using Γ over SNR req is that Γ can be deined with similar meaning or all theoretical models in the requency domain (Shited Shannon, CAP, PAM, DMT). The advantage o using SNR req over Γ is that this quantity is closer related to the SNR observed at the decision point o the detection circuitry. The line rate is usually higher then the data rate (0 30%) to transport overhead bits or error correction, signaling and raming. The symbol rate is the line rate divided by the number o bits packed together in a single symbol. The summation range or n is rom N L to N H, and this range has to be deined to make this generic model speciic. Commonly used values or PAM, using over sampling, are N L = 2 and N H =+1. This correspond to T/3-spaced equalization. Wider ranges are not excluded.

22 Generic CAP/QAM detection model The calculation o the margin m using the generic CAP/QAM detection model is equivalent with solving the equation in expression 7. This model assumes ideal decision eedback equalizer (DFE) margin calculations. The associated parameters are summarized in table 15. The eective SNR is to be evaluated by using one o the input models described in clause 5.1. Depending on what oset ormat is used or the SNR expression (see clause 5.1), the calculated margin m will represent the noise margin m n or the signal margin m s. SNR req s N H b ( ) 1 Γ 2 1 = exp ln1+ SNRos ( m, + n s ) s 0 n= N L d Expression 7: Equation o the CAP/QAM-detection model, or solving the margin m. The (eective) SNR gap Γ, being used in the above expression 7, is a combination o various eects. This Γ parameter is oten split-up into the ollowing three parts: Its theoretical value Γ CAP (in the order o 9.8 db or BER=10-7 ) A theoretical coding gain Γ coding (usually in the order o 3-5 db), to indicate how much additional improvement is achieved by the chosen coding mechanism. An empirical implementation loss Γ impl (usually 1.6 db or more), indicating how much overall deterioration is caused by implementation dependent imperections in echo cancellation, equalization, etc, without identiying its true cause. When Γ is split-up into the above three parts, its value shall be evaluated as ollows: SNR gap (linear): Γ = Γ CAP / Γ coding Γ impl SNR gap (in db): Γ _db = Γ CAP_dB Γ coding_db + Γ impl_db Model Parameters linear In db remarks SNR gap (eective) Γ 10 log 10(Γ) = SNR req / (2 b 1) SNR gap in parts: Γ CAP 10 log 10(Γ PAM) Modulation gap or CAP/QAM Γ coding 10 log 10(Γ coding) Coding gain Γ impl 10 log 10(Γ impl) Implementation loss Required SNR SNR req 10 log 10(SNR req) = Γ (2 b 1) Data rate d all payload bits that are transported in 1 sec Line rate b = DateRate + overhead bitrate Symbol rate s = b / b Bits per symbol b = b / s (can be non-integer) Summation range N L, N H On deault: N L=0 and N H=+3 Table 15. Parameters used or CAP/QAM detection models. The various parameters in table 15 used within this generic detection model have the ollowing meaning: The SNR-gap (Γ) and required SNR (SNR req ) are equivalent parameters and can be converted rom one to the other. The advantage o using Γ over SNR req is that Γ can be deined with similar meaning or all theoretical models in the requency domain (Shannon, CAP, PAM, DMT). The advantage o using SNR req over Γ is that this quantity is closer related to the SNR observed at the decision point o the detection circuitry. The line rate is usually higher then the data rate (0..30%), to transport overhead bits or error correction, signaling and raming. The symbol rate is the line rate divided by the number o bits packed together in a single symbol.

23 23 The summation range or n is rom N L to N H, Commonly used values or CAP/QAM systems using over sampling are N L =0 and N H =+3. This holds when the carrier requency positions the spectrum low in the requency band (e.g. CAP-based HDSL). Other values may be more appropriated when the carrier requency moves the spectrum to higher requencies (e.g CAP based VDSL) Generic DMT detection model The calculation o the margin m using the generic DMT detection model is equivalent with solving the equations in expression 8, or a given line rate b (or given data line rate bd ). The associated parameters are summarized in table 16, and unction load is speciied by the chosen bit-loading algorithm. The eective SNR is to be evaluated by using one o the input models described in clause 5.1. Depending on what oset ormat SNR os (m, ) is used to express this eective SNR or margins other then m=1 (equals zero db), the solved margin m will result in the noise margin m n or the signal margin m s. b k bd b SNRos = log 2 1+ Γ = b = = bd sd + bs sd ( m, ) k tones k load( b k ) [ bit / tone / symbol] [ bit / s] [ bit / s] Expression 8: Equations o the DMT-detection model, or solving the margin m or a given data line rate bd, and a given data symbol rate sd The latter excludes all DMT symbols dedicated to synchronisation. Bit-loading algorithm The DMT sub-carriers are all positioned (centred) at a multiple o the sub-carrier requency spacing, and each subcarrier theoretically may carry any (ractional) number o bits per symbol. The way this bit space (bits per tone per symbol) is used to load each sub-carrier with bits is implementation dependent. Bit-loading algorithms do commonly use masking. Masking means skipping carriers or loading when their bit space b k is below some predeined minimum value b min, and limiting the bit-loading to some pre-deined maximum when the bit space b k exceeds some predeined maximum b max. This masking process is summarized in expression 9. b k < b min load(b k) 0 b min b k b max load(b k) b k b k > b max load(b k) b max Expression 9: The bit loading used in (ractional) bit-loading algorithms When the data transport is operating on its limits (margin m=1, or zero db), the ollowing bit-loading algorithms may apply, in addition to masking: Fractional bit-loading (FBL), sometimes reerred to as water-illing - is a pure theoretical approach enabling loading o any real number o bits per symbol in any sub-carrier k (including non-integer ractions). This maximizes the use o the available capacity, but is unpractical to implement. Truncated bit-loading (TBL) - is a more easible algorithm in practice, and loads on each sub-carrier k a number o bits equal to the largest non-negative integer below the bit space b k. Rounded bit-loading (RBL) - is also easible in practice, and loads each sub-carrier k a number o bits equal to the nearest non-negative integer o bit space b k.