isen logo

ADSL: Fast Speeds, Phone Wires, no ISDN

by David S. Isenberg

I was one of the first fans of ADSL (asymmetric digital subscriber line). It sounded great, snaring bandwidth of 1.5 Mbps or even 6 Mbps over my existing telephone wires. I had visions of riffling through online catalogs, surfing 500 channels (did I say 500?) infinite libraries of movies, time-shifted TV shows, every rerun of M.A.S.H., Grateful Dead concerts from 1969. All this would be brought to me on the same telephone wires that I have now, on my settop box, on my TV-cum-joystick, in my living room. That was 1993.

Then, in 1994, there was a story about something called Mosaic on the front page of The New York Times. Internet providers, springing up in basements, closets, wherever you could wire a rack of modems, discovered that 20 hours of access for $20 a month worked better than a dollar an hour. Suddenly, everybody I knew and every business I dealt with had an email address. It was hard to find an activity involving people in which Internet did not come up.

In stark contrast, by 1995 it was hard to find an interactive TV (ITV) trial that could be called more than a qualified success. Cable companies, needing fast cash to feed their debt monster, cranked down the heat on ITV and turned their attention to services with more promise of shorter-term payoffs. They did well in the wireless spectrum (PCS) auctions and got busy building two-way hybrid fiber-coax (HFC) to support it.

Meanwhile, ADSL, which was a gleam in the marketing department's eye in 1993, was inching towards production. I decided to take another look.

ADSL applications, competing technologies

ADSL delivers data much faster in one direction (usually downstream, to the end user) than the other. The first products, introduced in 1994 and 1995, run at 1.5 Mbps downstream and at 16 kbps upstream. A second generation of products is now beginning production. These feature downstream speeds of up to 8 Mbps and upstream speeds as high as 640 kbps.

Most end-user digital applications are, in fact, asymmetrical, and could, potentially, be very suitable to ADSL. ITV, for example, requires a very high downstream bandwidth: 1.5 Mbps for VCR quality using MPEG-1 coding, and 2 or 3 Mbps for broadcast quality using MPEG-2. In the other direction, the upstream communication data rate is negligible; specifying which program or channel, the play, pause, rewind functions, and so on, require just a few bits of bandwidth. Browsing an electronic catalog, scanning an electronic publication, examining a remote X-ray or CAT scan make similar asymmetrical demands. LAN interconnect, with mostly typed user input but occasional file uploads, is a bit more demanding, but well within ADSL's upstream capabilities. Video conferencing becomes feasible as newer ADSL equipment achieves upstream speeds of 64 kbps and up.

But ITV and video-on-demand (VOD) just are not panning out to be the killer apps that they were thought to be in the early '90s, and this severely vexes the business planners of the major telcos. Virtually every VOD/ITV trial has been scaled back or discontinued. For instance, the original Time Warner business case for the Orlando Full Service Network trial reportedly projected $125 in spending per month per household. Apparently, average people won't pay that much for the applications that are being tested, at least not today (see Vol. 4, No. 8, p. 24).

Another problem for VOD and ITV over ADSL is that ADSL will not support normal analog cable TV, and most folks do not want another system connected to their television set. Ultimately, they will somehow have to pay for two terminal devices. In contrast, the cable operators have an easier time of it. They need to provide one, albeit upgraded, piece of terminal equipment. Also, if the channel switching is occurring back at the telco's central office, channel surfing is likely to be clumsier with a higher switching latency.

Today, Internet access seems to be the newly promised killer app. It is, at first glance, ideally suited for ADSL. Certainly, most of today's Internet applications fit ADSL's capabilities perfectly. But here, three new technologies look like competition: cable modems, wireless cable and ISDN.

Cable modems. When it comes to Internet, cable has a built-in advantage over the telcos (see Vol. 4, No. 4, p. 10). The data feeds (plus all the other cable channels) are always on, while the telcos still engineer their service on a dial-in, by-the-minute scarce-resources model. Internet providers and other online services have determined, through trial and error, that customers far prefer to pay for access by the month, on a subscription basis. The always-on model works better. Furthermore, the Internet's packet structure is more consistent with always-on than it is with dial-in.

Cable modems are asymmetrical, much as ADSL is, and they can handle downstream bursts of up to 10 Mbps. Upstream is still a problem today, and several flavors of cable modem use a telephone line for upstream transmission. Nevertheless, bursts of up to 128 kbps are well within the range of today's technology. I would keep scanning this front for rapid improvements.

One issue for cable modem architecture is that the bandwidth must be shared by all those logged on in, say, a 500-home neighborhood. If, hypothetically, ten percent of the 500 homes have cable modems, 20 percent of these are active in peak hours, and all ten users try to download a fat file at once, then performance would suffer. But if today's Internet consumer behavior is any indicator, customers will tolerate occasional peak-hour sluggishness.

The scenario above assumes deployment of an HFC architecture. While some proponents of ADSL argue that only 21 percent of homes will be included within HFC areas by the turn of the century, since the cable operators are not as burdened by universal access requirements as the telcos, this 21 percent is likely to be among the wealthiest 21 percent of American homes.

Big players are making big bets on cable modems. End-to-end prices are now around $1,000 per customer (about one-third of ADSL), and are expected to fall in accordance with Moore's Law. Zenith, Motorola and Intel are expected to release a new round of products for this market in 1996. Most cable operators have trials or nascent ventures, the biggest of these being @Home, a partnership of TCI, Microsoft and the venture capital firm Kleiner Perkins Caulfield & Byers (see Vol. 5, No. 7 & 8, p. 3). Conservative analyses would still see that cable modems will give ADSL some serious competition.

Wireless cable. Telephone companies are falling in love with this oxymoron (see Vol. 5, No. 6, p. 15). This is not uncharacteristic, given their short-term strategies for in-state long distance services, for example. Wireless cable provides a potential way to install full-service cable, plus a wide range of digital two-way services. Currently there are two forms of wireless cable: CellularVision and MMDS. CellularVision operates at about 29 GHz, and has been tested for several years in Brooklyn by a partnership of Bell Atlantic, the Hovnanian family, and others.

MMDS (multipoint multichannel distribution service) runs at around 2.5 GHz. Last summer, Pacific Telesis bought a 42,000-subscriber system in Southern California. Nynex and Bell Atlantic have taken substantial positions in MMDS equipment maker CAI Wireless Systems, based in Rochester, N.Y. While today's wireless cable systems are analog, the potential is to make them digital, and there is every reason to believe that in the near future they could become elegant, broadband, two-way symmetrical platforms for all kinds of digital services. The telcos are interested because of this, and because of the extremely low cost of cabling a neighborhood.

ISDN.Many of the RBOCs have already invested heavily in ISDN (integrated services digital network). PacTel has had significant success at selling 64- and 128-kbps Internet access in California; they also have a large third-party reseller system, much like cellular sales. The cost of this mature technology is around $500 per customer. Given the other current bottlenecks on the Internet, today it might be difficult to tell the difference between a last mile at 128 kbps and at 6 Mbps. However, as the Internet's infrastructure matures, ISDN will eventually be a drag. But when the telcos are faced with investing in new ADSL components and operations, they may opt, instead, to sell the right-now technology of ISDN.

Must-win, dustbin, or fit-in?

When ADSL manufacturers talk of alternative technologies, they usually list ISDN and optical fiber. They usually do not think of cable modems or wireless cable, yet these are viable technologies, as shown above. After all, ADSL is a telco technology.

Consider ISDN. It took some 20 years for it to get from invention to the meager penetration it has today. ADSL is about where ISDN was 15 to 18 years ago in the telco deployment process.

Now consider fiber. HFC is today's architecture of choice for cable operators, and the telcos have noticed. Nynex, for example, is building an HFC trial in Rhode Island and eastern Massachusetts (initially for traditional cable, but migrating to other digital services). In new construction, HFC is now significantly cheaper than twisted pair. Fiber to the home, the ultimate solution, is possible for another $200 to $400 per home. According to Ray Smith, CEO of Bell Atlantic, fiber plant could save up to $100 per year per line in maintenance alone over twisted pair.

It is likely that ADSL will be viable in certain niche applications. One of the most promising is work-at-home. The cable industry in the US is very balkanized, and an employer who wanted to LAN- interconnect its workers' homes would have to cut deals and coordinate service among multiple cable operators. In contrast, the legacy of the Bell system would allow an employer to deal with one, or possibly two telcos in a given area. A college offering at-home study would have a similar attraction to the contiguous territory of the telco, which would have the advantage of selective deployment of ADSL modems. The main issue here is whether ADSL volumes will allow it to hit its price points in a meaningful time frame.

A second sweet spot application of ADSL might be in developing countries, for remote education, telemedicine, and the like. Here, again, the issues are price points and competing technologies. The ADSL Forum points out that a complete conversion of the twisted pair plant to fiber could take 25 years, if you project current telco conversion rates. Selected homes in non-HFC, non-ISDN, non-MMDS areas may well want high-downstream-rate digital services, and the local telco could be in a good position to do this with ADSL.

Trials and conclusions

Bell Atlantic is perhaps the most visible US supporter of ADSL. The company's currently running a 1,000-home ADSL ITV/Home Shopping trial in Reston, Va. This was originally supposed to be 2,000 homes, growing to 20,000 homes with regulatory approval. But today, Ray Smith talks of this trial as a means to learn more about interactive services. Meanwhile, Bell Atlantic is aggressively repricing residential ISDN; aggressively investing in wireless cable; doing an HFC trial in Dover Township, N.J.; and doing Project Renaissance in Clifton and Passaic, N.J., using an all-digital fiber architecture to prove feasibility. One could conclude that Bell Atlantic is most visible in ADSL because it's actively exploring many new digital technologies. Perhaps the company's attitude was best summed up by vice chairman Lawrence Babbio, who said ADSL offers promise in niche markets or where switched digital video (a.k.a. fiber-to-the-curb) won't be available.

British Telecom is market-testing ADSL in 2,500 households in Colchester & Ipswich, England. This includes VOD (including TV, educational TV, movies & music), online shopping and banking, a magazine service, and a local community link and information service. We have no read on how this is going. Telecom Italia is testing ADSL in 1,000 households in Italy, according to Westell, an ADSL equipment manufacturer. Telstra, the second telephone company of Australia, has a 300 customer trial of ADSL in Melbourne with technology from NEC, Amati, CLI and DEC.

US West & Nynex are rumored to have been interested for some time, but I have found no independent confirmation of any trial activity. Telefónica de España tested ADSL, but will not pursue deployment, because it is viewed as an unnecessary middle step between copper and fiber.

ADSL technology has certainly advanced significantly; it now looks like speeds of 6 to 8 Mbps is a reality. But to date, virtually all market trials for interactive services have had universally thin results. The exception to this picture now appears to be Internet access. Yet numerous promising alternative technologies exist, each with their own set of advantages. There may be a place for ADSL in the mosaic of digital services, but if volume manufacturing does not materialize, will the cost be low enough to justify niche deployment? Watch for reports in Digital Media as new information rolls in.

David Isenberg is a technology analyst for AT&T Bell Labs. According to him, his expertise is in predicting the present.

Published in Digital Media 5(9)

Date last modified: 23 Dec 97