The Promise of ADSL

Steve Shepard
March 1998

This paper originaly appeared in the
InFocus column in Telephony Magazine, March 16, 1998.

Beginning with the advent of the T-carrier in 1962, the telephone network has evolved from an all-analog environment to a virtually all-digital network, starting with the interoffice trunks, continuing through the switches, and working outward toward the customer.

To date, the trunks and switches have been 100% converted, while the local loops leading to the customer remain largely analog, with the exception of those environments where ISDN has managed to gain a foothold. Unfortunately, modern applications such as videoconferencing and Internet access demand more bandwidth than ISDN is able to provide. The applications have outpaced the capabilities of the loop, and show no signs of slowing down. Something better is clearly needed.

The telephone network is designed to transport voice and data within the confines of the 3-4 KHz voiceband. It is also designed around the known behavior of typical telephony users (both people and applications), that exhibit predictable and well-understood calling patterns and hold times. The unparalleled growth in Internet access has had a dramatic and all too often negative effect on the ability of the local switching infrastructure to handle the requested load, since most users log on and stay connected much longer than the typical telephone call.

An added wrinkle is the ongoing evolution of the user community. It is estimated that well over 10 million people now telecommute in the U.S., while an additional 8 million or so operate small businesses out of their homes--the so-called SOHO (Small Office/Home Office) model. These people need to connect to corporate network resources at substantially higher data rates than the traditional modem allows. Together with Net surfers, they pose a substantial challenge to existing network resources.

Studies show that the traffic characteristics of business users differ substantially from those of Web surfers. The most significant difference between the two is the degree of symmetry of the data exchange. Web surfers tend to generate traffic that is highly asymmetric in nature: that is to say, more bandwidth is required in one direction than the other.

This picture makes perfect sense. When a user requests information from a Web site, the number of bits in the upstream request is minuscule compared to the volume of data that flows downhill toward the user. Business users, which send a receive large amounts of data, generate traffic that is far more symmetrical than that of a Web surfer. Unfortunately, these high-bandwidth, low-bandwidth, symmetrical and non-symmetrical traffic flows all impinge on telephone company local switches, which have begun to smoke under the onslaught.

Several solutions have been proposed. They run the gamut from massive switch capacity retrofits and construction of overlay networks for Internet traffic to schemes designed to persuade customers to limit their connect time through adverse pricing mechanisms. The most promising scheme, however, is a technological one, and involves the use of digital subscriber line (DSL) technologies in the local loop. One of these technologies, the asymmetric digital subscriber line (ADSL), has begun to show promise as the next generation local loop.

ADSL offers enormous bandwidth (as much as 9 Mb/s downstream and 384 kb/s upstream, hence the name "asymmetric"), requires a minimal equipment retrofit, can be installed very quickly and easily on a customer-by-customer basis, and is a cost-effective solution for the growing bandwidth requirements of modern applications.

It offers an additional advantage: It allows voice traffic and data traffic to be separated from one another at the point where they enter the network, thus allowing the data to be routed around the local switch to an alternate transport mechanism, freeing the local switch to do what it was originally designed to do--carry voice. The data usually goes to a router on its way to the Internet or some other data network. ADSL, Internet traffic--the main cause of switch meltdown--never touches them.

How It Works

The local loop is typically viewed as a 4 kHz diameter pipe. Frankly, this is incorrect. The loop is bandwidth-limited because it is engineered that way, not because the medium cannot deliver greater bandwidth. The typical local loop can, in fact, provide 1.1 MHz of bandwidth if engineered properly. Using sophisticated coding schemes, ADSL and the other digital subscriber line technologies can achieve line speeds over the local loop that are well into the multimegabit range.

ADSL uses the existing local loop to deliver high-bandwidth services, but that is where the similarity to existing architectures ends. At each end of the circuit, modems are installed, which make possible the promise of ADSL. These modems create a high-bandwidth downstream channel, a smaller upstream channel and a basic telephone service channel for voice. The actual bandwidth provided is dependent on the length of the local loop, as shown in the table below.

Data Rate	Distance
1.544 Mb/s (T-1)	18 kft.
2.048 Mb/s (E-1)	16 kft.
6.312 Mb/s (DS-2)	12 kft.
8.448 Mb/s	9 kft.

In a typical ADSL installation, service modules (set-top boxes, routers, PC interface devices) attach at the customer premises to the premises distribution network (PDN). The PDN is the premises wiring scheme that interconnects customer premises equipment to the local loop.

The PDN is attached to a remote ADSL transmission unit (ATU-R), which in turn is connected to the local loop using a splitter. The splitter performs the "Stuff Separation." It accomplishes the logical separation of voice and data traffic.

At the network side of the circuit, the loop terminates at another voice-data splitter, which in turn is connected to a central ADSL transmission unit (ATU-C). The ATU-C is connected to the access node, which is the aggregation point for broadband and narrowband data sources delivered from a DSL access multiplexer, or DSLAM. This device allows TV signals, interactive video, Internet access and a wide variety of other data types to share access to the ADSL-equipped local loop.

As a result of this configuration, a service provider can provision a high-bandwidth, multiservice network.

ADSL relies on frequency division multiplexing to create the independent basic telephone service, upstream and downstream channels. It establishes a channel at the low end of the spectrum for voice; a medium frequency band for the upstream channel; and a higher frequency band for the high-bandwidth downstream channel. In some cases, the channels may overlap. This technique is called partially overlapped echo-canceled transmission, or POET.

In some cases, the downstream channel may be further subdivided into smaller channels using time division multiplexing; the upstream channel may also be subdivided as required. This channelization process is key to ADSL’s success in the telco domain. Because the voice and data traffic are transported on logically separate paths, the data can be routed around the local switch, thus providing load relief without massive hardware and software upgrades.

Two signal modulation techniques have been developed for use in ADSL implementations to achieve the very high bit rates that the service promises. The first, called carrierless amplitude phase modulation, or CAP, is similar to quadrature amplitude modulation (QAM), a technique that has been in existence for quite some time. QAM is simple and well-understood--and chipsets for the process are readily available. The downside of CAP is that it is only offered by AT&T Paradyne.

The second technique used in ADSL systems is called discrete multitone, or DMT. In DMT, the 1.1 MHz "channel" is broken into 256 4-kHz sub-channels, hence the term, "multitone." Each sub-channel has its own carrier, and the signal-to-noise ratio is constantly monitored by the DMT system to determine how many bits-per-signal can be carried in each sub-channel.

The system dynamically adjusts each channel accordingly, resulting in a technique that is by its very nature dynamically rate adaptive. If certain frequency ranges in the "spectrum of sub-channels" are noisy, they are not used. Developed by Amati Communications (which is now part of Texas Instruments), DMT is the broadly accepted coding standard for ADSL, but it is significantly more complex than CAP.

Other DSL Services

ADSL is one member of a family of digital subscriber line services, all of which provide high bit rate solutions at a reasonable cost to the customer. High bit-rate digital subscriber line, or HDSL, provides 1.544 Mb/s service over two pair, while HDSL-2, an emerging standard, will provide full T-1 capability over a single pair -- a service offering sure to be attractive to both competitive local exchange carriers (CLECs) and Internet service providers (ISPs).

Rate adaptive digital subscriber line, or RADSL, is able to dynamically select the most appropriate data rate given changing line conditions, and may well replace ADSL at some point in the future. It is attractive to service providers since the rate adaptation capability eliminates the need to dispatch a technician in many cases.

Very high bit rate digital subscriber line, or VDSL, provides as much as 54 Mb/s downstream and T-1 speeds upstream, although at those rates the loop length can only be about 900 feet. However, a 900-foot copper tail circuit attached to a switched digital video or fiber-to-the-curb network could prove to be an ideal solution for high-bandwidth services delivered to residential customers in the future.

In recent months, a new contender has emerged from the digital sidelines. Known as the consumer digital subscriber line (CDSL) introduced by Rockwell and Nortel, it provides 1 Mb/s of bandwidth downstream and 128 kb/s upstream.

CDSL is similar to RADSL in that it can save the service provider a significant amount of installation expense. Whereas other DSL technologies require the installation of additional hardware and a fairly complex wiring scheme, CDSL just replaces the existing line card.

And while it doesn’t offer the bandwidth that ADSL provides, its T-1 service may well be enough for a substantial portion of the market. Nortel recently signed an agreement with Rockwell to create products using Rockwell’s CDSL chipsets and Nortel’s 1-Meg modem technology. It should also be noted that standards work is underway for CDSL, which is not expected to be compatible with the Rockwell/Nortel scheme. Either way, this is a technology to watch.

ADSL vs. ISDN vs. Cable Modems

ADSL is often seen as a significant competitor for both ISDN and the cable modem market. While ADSL will undoubtedly have a significant impact on the installation of basic rate ISDN, it will have a minimal impact on primary rate ISDN installations. Basic rate ISDN doesn’t offer the bandwidth that ADSL provides, and is a fairly expensive installation because of the need to install updated switch generics. ADSL has no such software requirement. Furthermore, ADSL offloads data from the local switch, while ISDN does not. In this regard, ADSL offers a clear advantage.

ADSL is often viewed as the telco solution to cable modems. In theory, cable modems certainly offer access to greater bandwidth than ADSL. In reality, most cable plant today cannot provide access via cable modems to the full bandwidth of the coax pipe.

And while cable systems offer an upstream channel, the channel is shared among a large number of users, resulting in contention issues. The deployment of hybrid/fiber coax (HFC) architectures will greatly alleviate these concerns, but these systems are costly and it will be some time before they are widely deployed. ADSL, therefore, remains the most viable near-term solution.

Who’s Using It?

ADSL has now been tested by most providers, and many are planning service rollouts. All of the Bell regional holding companies have tested DSL technologies, and will have ADSL commercially available on a large scale by mid-1998. GTE intends to make the service available in first quarter. Other providers will undoubtedly follow suit.

As far as ADSL manufacturers are concerned, you can round up the usual suspects: Westell, Orckit Communications, ADC Telecommunications and AT&T Paradyne are all in the game, as are Motorola, U.S. Robotics, Rockwell International, PairGain Systems and Alcatel Telecom. The complete list of players, of course, is much longer. For additional information about service providers, standards bodies and manufacturers, visit the ADSL Forum’s Web page at www.adsl.com.

ADSL offers distinct advantages to the consumer, but its value to service providers must not be underestimated. In addition to load relief for an overtaxed network, ADSL provides carriers with new business opportunities. Consider, for example, the situation in which ADSL is used to route data away from the local switch, dropping it instead into a router.

In this scenario, the router is undoubtedly connected to the Internet, in which case the traditional local carrier instantly becomes an Internet service provider, allowing it to compete with incumbent ISPs. Regulatory concerns would have to be overcome, though, since traditional telephone service providers can only enter the ISP business through the creation of fully separate subsidiaries.

But there is a quid pro quo at work here: ADSL also allows ISPs to compete with ILECs for the role of access provider. Under the right circumstances, may even allow them to provide voice services as a limited competitive local exchange carrier.

Sound the starting gun. The race is on.

Steve Shepard is a Senior Member of the Technical Staff at Hill Associates in Colchester, VT. He can be reached at shepard@hill.com.