In & out
The latest iteration of DSL shows that it’s better to give and receive.
Mitch Kahn, Excess Bandwidth

In business, small- and medium-scale companies are only now gaining access to half the capability required to use the Internet the way the experts say it should be used. And despite the escalating war of words in ad campaigns, the real benefits of the digital revolution continue to elude the mass market.

Based on the broadband pipes available today, principally ADSL (asymmetric digital subscriber line) and cable, it is no surprise that streaming audio and video are gaining in popularity among general consumers and, for applications such as distance learning, in corporate environments. These services are based on relatively passive models of information consumption. Surely, though, the picture of the future must entail more than just feeding content to digital couch potatoes.pp

Broadband is held back today by a basic weakness in current technologies, which becomes evident when you take a step back from the battle cable providers and telecommunications companies are fighting for consumers' dollars. The problem lies in the asymmetric nature of practically every available high-bandwidth technology. Granted, the data rates emanating from either cable or telco central office (CO) sites are orders of magnitude greater than the analog modem technology today's users grew up on. However, upstream rates, from the user site back to the CO, are not nearly as impressive.

Of course, asymmetric communication is fine if you are a cable company, since "providing" content to consumers is rolled up in the definition of your business model. But for local exchange carriers, including both the incumbent and competitive companies now fighting for market share across the US, the story is all about communications. And communications is a two-way street. The real broadband revolution begins when carriers can provision services to business and consumers using affordable and consistently reliable high-data-rate symmetric connections.

What do we gain when downstream and upstream data rates are equal? Quite simply, the ability to create a communications channel. The laundry list of Internet Protocol (IP) services that has been the subject of countless marketing promotions finally becomes deliverable. There's still work to do, but reliable, two-way, high-bandwidth connectivity changes the picture. In fact, it makes exchange of pictures, even real-time videoconferencing, practical.

For real

Experience in the field shows that many consumers end up keeping a separate POTS line to protect the basic voice service, which tends to weaken the economic case for DSL compared to cable. With symmetric DSL, the ability to provision VoDSL (voice over DSL) service becomes a powerful new weapon in a carrier's arsenal. In particular, businesses can now be offered a viable alternative source for voice phone services.

The basic hardware for VoDSL comprise three elements: an IAD (integrated access device) at the customer site and a DSLAM (DSL access multiplexer) and a voice gateway at the CO. As these technologies become available from multiple vendors, VoDSL is gaining momentum. Researchers at the Yankee Group forecast a $475 million market for business VoDSL in 2001, expanding to a little less than $3 billion in 2003.

Replacement of POTS is just the tip of the iceberg. Symmetric DSL's impact on the total communications market will be enormous, because the technology enables carriers to optimize the use of the existing copper-wire infrastructure.

Both ILEC and CLEC (incumbent and competitive local exchange carrier) companies have made enormous investments to create fiber-based backbone communications infrastructures. Today, much of the inter-city and major portions of the intra-city backbone infrastructure are digital. But connection of even the most advanced digital networks to most of the end-user community continues to take place over the existing copper wires of the PSTN (public switched telephone network).

This is the so-called "last mile" of copper (when applied to single-site businesses and residences) or the "last thousand feet" of copper (when applied to multi-tenant buildings connected to the local office by digital facilities). In all cases, it is this existing copper "loop" that is the prime-limiting factor in bringing broadband digital services to the consumer (Figure 1).

Help or hindrance

In any event, the mass of twisted pairs of copper wires that has served as the telephone infrastructure for the last 100 years can be seen as a tremendous handicap—or an asset of inestimable value. This network allows the relatively inexpensive provisioning of service between customer premises and central offices. However, learning to drive data quickly and accurately down a twisted-pair configuration that was intended for analog, not digital, communications has proven to be a very tough problem.

One important note about the last mile is that, in the US at least, it is often more like the last few miles. Carriers have been investing heavily to shrink local loops from lengths of 18,000 feet or more, to more functional ranges averaging 6000 feet. This is typically accomplished through installation of distribution area access nodes connected to a CO. This capital intensive, and thus lengthy, process will not be complete for many years. Thus, standards organizations defining new transmission technologies, including ANSI (the American National Standards Institute) in the US and the ITU (International Telecommunication Union) globally, often cite ranges of 8000 to 18,000 feet.

DSL technology is the industry's collective response to the last-mile challenge. Transmission techniques defined by standards bodies to deal with the local loops must address a wide range of issues. Chief among these is the fact that the loop between CO and business or residence connections is by far the "dirtiest" part of the network. In many cases it has seen decades of use. The quality of the wire itself, the number of dropouts, obsolete taps and termination points, and dozens of other variables may not even be fully documented. This unpredictability of conditions within the copper-wire infrastructure has made it very difficult to guarantee the performance of DSL.

As noted earlier, symmetry of data transmission rates is a key to unlocking DSL's potential. Symmetry extends to issues such as latency, since a fast duplex data rate has little value in two-way communications if processing overhead causes a perceivable lag on either end of the link. For carriers, the issues inherent in the dirty network of copper wire dictate that digital solutions also be evaluated for performance in two other areas: reach and robustness.

With all DSL technologies, there is a trade-off between data rate and the reach of the digital link. The maximum data rate is a function of the length of the loop between end user and CO. The longer the loop, the lower the maximum possible data rate, due to attenuation of signal strength and increased noise. This is why current DSL service—at any data rate—is only available within a 3-mile (approximately 15,000 feet) maximum radius from a CO site. Even relatively short-distance connections may not achieve expected data rates, due to the noise issues raised by the unpredictability of the copper wire in any given loop.

The ideal DSL technology would be able to "learn" the true line conditions and make adjustments to accommodate the varying conditions that are the rule, not the exception, in the local loop.

Robustness is a quality-of-service concern that is related to reach-vs-data-rate issues, but extends into many other areas. Spectral compatibility, for example, deals with the ability of a line technology to withstand interference from adjacent pairs. Twisted-loop pairs run from CO to subscriber distribution points in bundles of 25 or more pairs known as binder groups. Crosstalk—the engineer's term for when signals bleed from one pair into another—is practically unavoidable, and for many DSL technologies, an enormous hindrance.

Given these issues, it is a testimony to perseverance that the ILECs and CLECs have been able to provide the types of DSL service available today. Hundreds of thousands of homes and business locations have been wired with limited, asymmetrical DSL. And many point-to-point connections in the intra-city backbone, which is somewhat more predictable than the last mile, have been wired with implementations of symmetric DSL, as summarized below.

DSL today

One of the earliest DSL technologies was HDSL (high-bit-rate DSL), a transmission technique intended to be an all-digital replacement for analog T1, a 1.544-Mbit/sec service, and E1, a 2.048-Mbit/sec service. HDSL provides a symmetric 1.544- or 2.048-Mbit/sec data transmission rate both from the local office to the user and from the user to the local office. It achieves this rate using two pairs of wires (up to three pairs for the E1 rate).

Single-pair symmetric DSL (SDSL) is another non-standard approach for enterprise and branch-office applications. It uses HDSL technology over a single pair, but the data rates it provides are determined by the line conditions at any given time, and are thus not guaranteed by the service provider. Under the best of conditions, SDSL rates either don't approach those of T1/E1, or they provide T1/E1 speed only over short distances, requiring several repeater stages for a typical twisted-pair loop.

As noted earlier, the most common form of DSL being installed today is ADSL. ADSL data flows downstream from the local office to the user at a 1.5 to 9 Mbits/sec, but the upstream data rate has a limit of 640 kbits/sec. In practice, downstream rates are in the 768-kbit/sec to 1.5-Mbit/sec range. This is quite sufficient for the primary service of ADSL—delivery of fast downloads of information from the World Wide Web.

As mentioned earlier, another problematic area for ADSL is latency. A few milliseconds of added delay is unnoticeable for Web browsing and even streaming audio and video. However, for real-time two-way communications—such as digital telephony and video conferencing—latency produces an unacceptable amount of echo on the line, similar to a bad cellular phone connection.

In addition to the asymmetric constraints that limit the applications ADSL can support, early deployment of this technology has not been smooth, with issues of reach, crosstalk corruption, and poor performance often plaguing installations. The problems may escalate as DSL deployment grows and the number of twisted pairs within wire bundles carrying DSL traffic increases, resulting in even greater crosstalk. Furthermore, the reach of an ADSL link is limited by the fact that repeaters, or amplifiers, cannot be practically used in an asymmetric environment.

Enter next-gen DSL

Just how do telecommunications providers move from the current, asymmetric state of affairs to a symmetric future?

The ANSI T1E1.4 committee has defined a single-pair symmetric technology called HDSL-2 (HDSL second generation) as a T1/HDSL replacement. A related international standard being defined by the ITU, G.shdsl, is a more generalized version of the specification. G.shdsl is rate-adaptive service primarily intended for the consumer and SOHO markets now served by ADSL and SDSL. Business customers also are likely to be offered G.shdsl broadband connection technology, as it allows carriers to provide a very robust, high-speed symmetric service at very reasonable cost.

The G.shdsl standard provides for the transmission of data at distances exceeding 20,000 feet on a single twisted pair, with the data rate decreasing from 2.3 Mbit/sec to 192 kbits/sec as distance increases. This will allow service providers to offer high-speed, symmetric connectivity at predictable, reliable rates to a much broader percentage of customers. It can also immediately double (or triple, for a three-pair, HDSL-based E1 link in Europe) the number of available transmission lines, compared with HDSL. Even greater distances are achievable, as the symmetric nature of the data transmission supports the use of repeaters (Figure 2).

The true utility of symmetric DSL is not merely the promise of high-speed data, but the provisioning of data and multiple voice lines over a single twisted pair. While single-pair ADSL offers one baseband voice channel occupying the part of the frequency spectrum below the data, the symmetric bandwidth and low latency of G.shdsl/HDSL-2 supports multiple digital voice channels within the data stream. This is particularly valuable as the scarcity of twisted pairs worsens and more diverse services need to be deployed on fewer and fewer copper resources.

Designed for robustness

The specifications defined for G.shdsl/HDSL-2 fully address the robustness issues faced by ADSL. The standards bodies require that systems perform over a large set of copper-pair characteristics, with many different crosstalk scenarios. To ensure performance, a series of tests have been defined for specific test loops. The standard also defines a minimal test set, in which the crosstalk affecting those loops is 1 percent worst case—that is, 99 percent of the crosstalk scenarios will be more benign than this worst-case example.

It is clear that the ability to provide both data and multiple voice lines over just a single twisted pair will quickly become a potent weapon in carrier's arsenals. It allows service providers to establish both voice and data revenue streams, creating an optimum economic model for the delivery of digital broadband. With a solid economic model in place, usage will inevitably rise, creating more opportunities to develop and market new services based on the growing ubiquity of symmetric, broadband digital communications.

Author information

Mitch Kahn is the vice president of marketing for Excess Bandwidth, a Silicon Valley startup focusing on developing semiconductors for the symmetric DSL market.