1 T1E1.4/2002-194 [ COMMITTEE T1 TELECOMMUNICATIONS Working Group T1E1.4 (DSL Access) Westminster, CO; August l 8-23, 2002 T1E1.4/2002-194 CONTRIBUTION TITLE: SOURCE: PROJECT: STATUS: SHDSL enhancement: The performance of multiple G.SHDSL lines using identical and non-identical rates over each pair. Eran Gerson, Amir Leshem, Metalink Broadband Access. SHDSL enhancement. For discussion. ABSTRACT In this contribution we address the possibility of delivering data over multiple copper lines using SHDSL. We study the performance of a system with unequal bit rate on various pairs compared to a system with equal bit rate per pair. The performance was checked using 25 binder 26AWG cable. We provide both 1 percent worst case and bit rate cumulative distribution function. NOTICE This contribution has been prepared to assist Standards Committee T1-Telecommunications. This document is offered to the committee as a basis for discussion and is not a binding proposal. The requirements are subject to change in form and numerical value after more study. Specifically, the right to add to, or amend, the statements contained herein is reserved. Contact: Amir Leshem, Eran Gerson. Metalink Broadband Access, Yakum business park, 60972 Yakum, Israel. E-mail: lehem@metalink.co.il. Tel: 972-9-9605555 Fax: 972-9-9605544
2 T1E1.4/2002-194 I. INTRODUCTION Recently an increasing interest in symmetric DSL transmission at a rate of 10Mbps has emerged. This is largely due to the desire for transport of Ethernet data over the public network. Further interest in multipair SHDSL technology stems from the desire to achieve increased reach for T1 and E1 systems. In this contribution we address the performance of 4 pairs SHDSL system working with identical and non-identical bit rates among the pairs. Equal rate yield good results assuming the noise profile is the same for each pair. However real life noise profile is different for each pair due to different location within the binder as well as different twisting density. We demonstrate this variability in [1]. In this contribution we take a closer look at the effect of the noise variance on different pairs within a binder. We show that higher data rates can be achieved, or equivalently longer reach for a fixed rate service. In In this contribution a 25 pairs, 26 AWG cable data is used, the binder cable cross talk data was provided by Telecordia [4]. II. SIMULATED PERFORMANCE ON THE TELCORDIA CABLES In the simulation we have tested the capacity of a 4 pairs system using identical and non-identical bit rates. In the simulation we have performed 2000 Monte Carlo trials. At each trial we have randomly selected 4 pairs. All pairs were used to generate NEXT (24 self next for each pair). The SNR at each pair was calculated. The target SNR was 7 chosen to support the relevant modulation at bit error rate (BER) of 10 with 6db noise margin: 1. Identical rates The lowest SNR (among the 4 pairs) is used to determine the transmission rate for all four pairs. 2. Non-identical rates For each pair the transmission rate is calculated separately for each pair. In both cases we have used granularity of 16 Ksym/s for the rates on the various pairs. The maximum symbol rate allowed for each pair vary as a function of the reach according to the guidelines of T1.417 [5]. Note that simulation results differ from the worst case NEXT models used for single pair transmission [5]. This is largely due to the cable crosstalk characteristics. The simulation results show the aggregate bit rates for the equal and unequal rates. The graph shows the rate versus the probability. Figure 1 presents the simulations for cable length of 8kft. Figure 1: Cumulative distribution function of total rates with identical and non identical transmission rates. 26 AWG. L= 8Kft. p=4. 32-TCPAM.
3 T1E1.4/2002-194 Figure 2: Cumulative distribution function of total rates with identical and non identical transmission rates. 26 AWG. L= 17.5 Kft. p=4. 16-TCPAM. Figure 3: Cumulative distribution function of total rates with identical and non identical transmission rates. 26 AWG. L= 17.5 Kft. p=4. 32-TCPAM. III. ONE PERCENT WORST CASE PERFORMANCE We now test the one percent worst case bit rate in each scenario as above. Figure 4 demonstrates the result for aggregate bit rates of approximately 10Mbps. The simulation results shows a net reach gain of 300ft for the system with different rate per pair compared to a system with constant rate among the pairs.
4 T1E1.4/2002-194 Figure 4: Identical and non identical rates for 10Mbps systems using four pairs Figure 5 demonstrates the result for transport over 4 pairs with aggregate bit rate of approximately 2048Kbps. The simulation results show a net reach gain of 1Kft for the system with different rate per pair compared to a system with constant rate among the pairs. In this simulation we have used 16 TCPAM transmission. Figure 6 demonstrates the same scenario, but using 32 TCPAM. Figure 5: Identical and non identical rates for 2048 Kbps using 4 pairs and 16TCPAM
5 T1E1.4/2002-194 Figure 6: Identical rate and non identical rate for 2048Kbps systems, using 32TCPAM. IV. CONCLUSIONS In this contribution the capacity of a 4 pairs g.shdsl system was simulated with identical and non-identical bit rates. The simulation was based on a real crosstalk measurement provided by Telcordia. It was shown the nonidentical rates yield better performance than identical rates system. V. ACKNOWLEDGEMENT We would like to thank Ken Kerpez for providing the data that was the basis for this contribution. REFERENCES [1] ITU-T g.991.2: Symmetrical digital subscriber line transceivers. 2000. [2] A. Leshem and E. Gerson. Statisitical characterization of multipair noise Some real data. T1E1.4/2002-yyy. [3] E. Gerson. G.shdsl.bis: Extending the 4-Wire Mode to Multiple Pairs with different rate T1E1.4/2002-zzz. [4] K. Kerpz 26 gauge pair-to-pair NEXT coupling magnitude and phase measurements.http://net3.argreenhouse.com:8080/dsltest/help/helpterms.htm [5] T1.417 2001. Spectrum management for loop transmission systems. American national standard.