With 200 million users worldwide, digital subscriber line (DSL) is thepredominant means for service providers to deliver broadband servicesto the mass market. With DSL, service providers are able to increasetheir return on investment in copper infrastructure by offeringrevenue-enhancing triple play services.
But these services present a challenge: they require superiorperformance, greater dedicated bandwidth, enhanced quality of service(QoS) and “always-on” access to the network.
Rate/reach performance for multimedia services can be severelyimpacted by noise on VDSL2 lines that is a result of crosstalk couplingfrom electrical signals on adjacent pairs of copper wires or otherimpairments. Loop noise characteristics can change significantly as aresult of on/off switching of DSL lines.
While traditional seamless rate adaptation (SRA) and bitswaptechniques can change the data rate to adapt to time-varying loopconditions, they are often incapable of sustaining the integrity of theline and the associated overhead channel at all times, particularly inthe presence of a severely degraded (and sometimes negative) noisemargin.
What service providers need is an intelligent mechanism that allowsrapid and dynamic adjustment to the data rate in the presence of suddenwideband noise changes. This type of Rapid Rate Adaptation (RRA)strategy is critical for a robust deployment of premium and enhancedservices, such as Internet protocol television (IPTV) and real-timetriple play applications.
This article will discuss:
* Common causes of performance degradation on DSL lines such ascrosstalk, non-stationary radio interference disturbers, impulse noiseand temperature changes;
* Techniques to minimize the impacts of noise to ensure that enhancedservices are “always on”
Service Providers' Access Challenges
Service providers worldwide are introducing triple play services overan end-to-end IP-based xDSL access infrastructure. Some of the keyservice providers' business rollout strategies include introduction ofcost-effective, high-density, power-efficient and highly reliableVDSL2/ADSLx combo cards offering ubiquitous, on-demand and highlyintegrated rich services for any content, anywhere, anytime, to anyuser.
Cost-performance optimization for an xDSL system must be consideredat all levels of hierarchy in order to maximize bandwidth, minimizeerror rates, enhance robustness and improve operational efficiencies.An optimized DSL system, coupled with guaranteed service reliabilityand stability, would allow service providers to rapidly introduceinnovative offerings, increased average revenue per user, and reducedoperational and capital expenses.
Consumers want superior QoS, always-on connectivity to the network,exceptional reliability, and instantly available content from a varietyof service providers. Operators must avoid service interruptions andline retrains at any cost.
To meet the needs of triple play applications, DSL technologycontinues to evolve. The ITU G.993.2 VDSL2 standard was approved inNovember 2006 . Compared to earlier generations of xDSL, VDSL2adopts a number of improvements to better tackle the various bandwidthneeds of different operators.
For example, VDSL2 supports improved impulse noise protection (INP),impulse noise monitoring (INM), SRA, U0 PSD (power spectral density)shaping, and new service and initialization policies for enhancedtriple play stability. Other advanced technologies such as emergencyrate reduction (SOS), dynamic rate repartitioning (DRR), and crosstalkchannel estimation and mitigation techniques are also being proposed inthe standard.
VDSL2 also supports a wide range of deployment scenarios by using aprofile concept. Each profile is linked to a specific bandwidth (from 8MHz up to 30 MHz) and to different power and power spectral density(PSD) mask constraints. And each profile is optimized for a specificreach and rate. As such, VDSL2 extends the reach of VDSL1, pushing themaximum achievable bit rates above all other forms of DSL.
Future bandwidth and coverage improvements with DSL can potentiallytake help from sophisticated technologies such as dynamic spectrummanagement (DSM), which is the mitigation of crosstalk picked up by DSLpairs sharing the same binder.
Because higher levels of DSM require some form of coordination amongall users sharing a binder, these techniques are far from trivial.Non-technical issues, such as loop unbundling — where different twistedpairs are operated by different companies, are the true DSM bottleneck.As a result, these issues must be resolved before telcos can exploitthe full potential of DSM.
Common Causes of Performance Degradation
It is often assumed that the corrupting signal is an additive randomsource with Gaussian distribution. For example, bit loading algorithmsare typically designed based on the assumption of additive Gaussiannoise. With such algorithms, the impacts of other types of interferenceare underestimated, which results in an excessive error rate.
Furthermore, channel estimation procedures are also usually designedto optimize performance in the presence of stationary impairments.However these algorithms will not perform well at estimatingnon-stationary, quasi-stationary or time-varying interferences.
These DSL training procedures are usually well suited only tooptimizing loop performance-and robustness in the presence ofstationary and/or additive Gaussian impairments. This would result inleaving the modem receivers blind to other impairments such as impulsenoise.
Major sources of xDSL performance degradation can be classified inthe following categories:
Internal noisesources. Examples of internally generated impairments includeever-existing thermal noise, quantization noise, componentnonlinearities and distortions, echo leakage or trans-hybrid loss,dispersion caused by band-limiting filters (ISI) and clocking noise,such as phase noise and clock jitter. Poor designs, precision andperformance limitations in algorithms and signal processing canadversely impact the xDSL performance.
External(semi-) stationary noise. Commonly referred to as crosstalk,this noise includes interference generated from other services suchasT1 and xDSL alien and/or self-noise sources such as far-end crosstalk(FEXT) and near-end crosstalk (NEXT).
Although the crosstalk interference generated by the continuoustransmission of adjacent DSL lines is essentially stationary, thecrosstalk profile may change over time due to changes in customerwiring or temperature variations. Crosstalk is often the dominantsource of performance degradation in DSL systems.
Externalnon-stationary noise. The non-stationary noise affecting theperformance of xDSL systems can be as a result of various sources. Themain causes of degraded performance are:
1) Impulse noise fromvoltage spikes generated on the line. Common causes include lightningstrikes, transformer surges and turning home appliances and lightswitches on or off. Repetitive impulse noise (REIN) is also anon-stationary noise from repeated voltage spikes or surges of thenoise on the line, usually at the operating frequency of the powersupply (50 Hz or 60 Hz). Impulse noise causes cyclical redundancy check(CRC) errors and may result a DSL line retrain, thus interruptingservice.
2) Radio frequencyinterference (RFI) is usually constrained to a relatively narrow partof the spectrum and is due to similar sources as impulse noise signalsas well as HAM radio interference coupling into the signal. Spreadingof this sinusoidal-like interferer results in a potential data ratereduction across a large number of tones.
Although RFI typically affects a small area of a symbol in discretemulti-tone (DMT)-based DSL systems, its duration is much longer thanthat of impulse noise and many DMT symbols are impacted. This leads toincreased interleaving and/or Reed-Solomon (RS) parity requirements,which is not suitable for low-latency applications (such as VoIP, whereexcessive delay can cause voice quality to suffer).
3) Plain old telephoneservice (POTS) signaling generated by on/off hook transitions, dialpulses, ringing and ring trips.
4) Micro-cuts in the copperline lead to random micro-interruptions of the DSL signal. In order toachieve reliable transmission, a combination of filtering and signalprocessing is used to reduce generated CRCs.
5) Neighboring xDSLservices moving in and out of operation. This phenomenon can causesudden and large changes in the stationary noise levels when the modemsin the binder are switched on and off. Depending on the level ofcrosstalk and the available noise margin, this may result in serviceinterruptions and line retrains (resynchronization and training).
Figure 1 below shows a mixed central office (CO)/remoteterminal (RT) deployment scenario with various interference signals onthe downstream of a victim receiver.
|Figure1. A typical deployment scenario illustrating the impact of differentloop impairments, including FEXT and NEXT, RFI, and impulse noise onthe DS received signal.|
Noise Mitigation Strategies for DSL Networks
There are a number of practical strategies to mitigate variousimpairments in DSL systems.
Internally generated noise sources are mostly influenced by thetransceivers' design of CO and customer premises equipment (CPE) andgenerally cannot be mitigated through deployment practices. In thiscase, it is advisable to evaluate and use systems from reputedmanufacturers.
Robustness to loop impairments and internal noise sources can beimproved through system design guidelines and advanced DSL transmissiontechnologies, such as adaptive TEQ, adaptive echo-canceller, adaptivehybrid and programmable digital/analog filters.
External stationary noise limits, such as crosstalk, are oftenaccounted for in network design. Spectral compatibility is ensured byimposing restrictions on the transmit power PSD masks. Operatorscontrol parameters related to transmit PSD in order to minimizecrosstalk to neighboring lines. Advanced upstream and downstream powerback-off techniques (UPBO/DPBO) are now commonly used and have beenintroduced in G.993.2 VDSL2 standard.
Dynamic Spectrum Management
Dynamic spectrum management (DSM) comprises a set of techniques formulti-user power allocationand/or detection in DSL networks to ensurespectral compatibility under crosstalk assumptions.
With DSM, crosstalk is either reduced by shaping the spectra of thetransmit signals, or is (partially) cancelled within the binder. Thesetechniques are very effective for deployment scenarios where crosstalkis the dominant source of impairment. There are four levels of DSMcoordination:
1) DSM Level 0 correspondsto static spectrum management (SSM) maximizing individual DSL lineperformance without considering the performance of neighboring lines.
2) DSM Level 1 deals with anautonomous power allocation management used for crosstalk avoidance.
3) DSM Level 2 is acoordinated power allocation management between neighboring lines toavoid crosstalk.
4) DSM Level 3 is used forcrosstalk mitigation. It can be used only when either transmittersand/or receivers are collocated 
Virtual noise is another technique introduced in the G.993.2 (VDSL2)standard to increase line stability. Virtual noise is added to the lineover the tones that are expected to be impacted by crosstalk from aneighboring line switching on.
As shown in Figure 2 below , noise margin will adapt toaccommodate virtual noise. Noise margin can remain low at the tonesthat are not impacted by crosstalk for maximum bandwidth availability.Virtual noise is very effective if apriori knowledge of crosstalk isknown. In many cases, the accurate knowledge of the interference is notavailable apriori. Virtual noise mitigation technique is not dynamicand often results in a conservative bit loading.
|Figure2. Virtual noise is added to accommodate expected crosstalk.|
Interleaving and Reed-Solomon Coding
There are a variety of techniques to manage non-stationary impairments.Impulse noise protection (INP) error correction strategy involves acombination of Reed-Solomon coding and interleaving whose framingparameters can be selected among ranges of values specified by thestandard.
Data rate is often reduced for increased noise margin. In essence,the receiver operates with sufficient noise margin to maintain theerror rate within acceptable limits. This results in running a majorityof frames with excess margin to ensure integrity of the data while onlya small fraction of frames are directly impacted by impulse noise.
Other drawbacks, such as an increased interleaver memory, may not beappropriate for low-latency and interactive applications. Theinterleaving delay of DSL systems can be significantly reduced byperforming erasure decoding of the RS codewords at the receiver.
Seamless Rate Adaptation
The VDSL2 standard specifies Seamless Rate Adaptation (SRA) as atechnique to be used when the noise profile on the line changes. SRAensures that the line bandwidth is reduced to ensure that there isadequate noise margin on the line. When the noise on the lineincreases, SRA is used to reduce the bandwidth on the line, withoutdropping the line, and therefore, without interrupting the service.
For slow to moderately varying noise cases, one or any combinationof the following adaptive techniques including tone reordering,bitswap, and/or SRA are commonly used.
In case of sudden and severe noise conditions, none of the abovestrategies are effective. In this situation, a rapid rate adaptation(RRA) solution, known as SOS in the ITU-T standard, is the mostpromising mitigation strategy to sustain the link and prevent the DSLmodem to retrain.
Rapid Rate Adaptation (RRA) Technology
RRA (Rapid Rate Adaptation) is an automated, intelligent solutionallowing service providers to enhance link robustness, reliability andavailability under severe and time-varying noise environments.
RRA is an advanced technology that enables rapid and dynamicadjustment of the data rate, without dropping the line or retraining,in response to sudden or large wideband noise changes such as anincrease in crosstalk. As a result, the quality of triple play servicesis optimized and the broadband connection is greatly enhanced to managevolatile and evolving loop conditions.
RRA enables the transceivers to monitor line conditions anddynamically adapt the data rate without interrupting service forretraining. RRA helps maintain link integrity even if the noiseenvironment changes beyond its tolerable noise margin. RRA is highlyresilient to dynamic crosstalks in the binder from on/off switching ofxDSL lines.
RRA outperforms other standard mechanisms, such as seamless rateadaptation (SRA), in terms of robustness and speed for sudden, largewideband noise changes. Specifically, RRA can maintain link integrityin environments where SRA is no longer effective.
The total time for RRA to stabilize the link by changing to a newrate will be much faster than SRA. RRA has a faster response time todynamic noise changes. Table 1 below shows a comparison of RRAand SRA highlighting the effectiveness of these techniques in presenceof sudden and strong noise.
|Table1. RRA and SRA functionalities and their effectiveness in strong noiseenvironment.|
The triple play target market has extremely high expectations forservice quality and reliability. For traffic-sensitive applicationssuch as video services, to avoid tiling (pixelization), apriority-based algorithm can be implemented to ensure that the datarates of the right bearer channel is changed based on the QoScapabilities available in the product.
By doing so, it is guaranteed that if data rates become lower, thehighest priority traffic will maintain its bandwidth (depending on lineconditions) and that the lowest-priority traffic bandwidth is reduced.RRA coupled with QoS provides an unmatched application-aware resiliencyin severely degraded noise margin environments for operators andservice providers.
Enhanced robustness is crucial for successful deployment of triple playservices. Techniques such as Virtual Noise, Bitswap, SRA and RRA willboost the reliability of the system and enable carriers to reach morecustomers with the same product. These features are expected to be usedextensively by carriers in the future.
Manouchehr Rafie, Ph.D., is director, StrategicProducts/Technologies, at IkanosCommunications, Inc.
 ITU-T RecommendationG.993.2, Very high speed digital subscriber line transceivers 2,November 2006.
 J. Verlinden, T. Bostoen,G. Ysebaert: “Dynamic Spectrum Management for Digital SubscriberLines”, 2nd Edition, to be published 3rd Quarter 2005 onwww.alcatel.com.
 Ikanos, “G.vdsl: High-levelrequirements for SOS mechanism,” ITU-T SG15/Q4 Contribution CD-078,Denver, U.S.A., Sept. 2006.
 Ikanos Communications,”Proposal for SOS protocol”, ITU-T SG15/Q4 Contribution SD-074, Jan.2007
 Huawei, Ikanos, Centillium,Conexant, Infineon, SEI, NEC, “Proposal for SOS technical details”,ITU-T SG15/Q4 Contribution NC-090 Napa U.S.A., Apr.2007
 Ministry of InformationIndustry, Huawei Technologies Co., Ltd “VDSL2: Key issues to emergencyrate reduction (SOS)”, ITU-T SG15/Q4 Contribution CD-149, U.S.A., Oct.2006.