New high-bandwidth options may eventually compete against ISDN and coaxial cable

On the Internet, the quest for more bandwidth continues at an ever frenzied pace. What started out as a text-only medium has become graphics-intensive, with Java animations, sound and even Quicktime movies. There are blinking logos, scrolling text, and the prospect, at least, of thousands of other applets that will download to your computer and run in realtime. Previous articles in this Pacific Connection series have looked at some of the high bandwidth solutions that go beyond the 28.8Kbps speeds available through a standard dialup connection. This month, we take a look at a new technique with potential both here in the U.S. and in Japan: Asymmetric Digital Subscriber Line (ADSL). We'll also take a quick look at a technology a bit further over the horizon: satellite Internet connectivity.

Actually, ADSL is not new. The term was first coined in 1989 by Bell Communications Research (Bellcore), the research arm of the seven "Baby Bells" that were formed after the breakup of AT&T. The original conception for ADSL was to carry video- on-demand service to the home--a market that has so far remained dormant.. Meanwhile, ADSL has been given new life as a high-speed Internet carrier.

ADSL has at least two big advantages over ISDN: cost of implementation and speed of delivery. ISDN costs more to implement at the telephone central office--requiring new, expensive switching equipment. Not surprisingly, telephone companies don't want to make such an investment unless they are certain there is a demand. By contrast, ADSL deployment is relatively cheap.

ADSL--even at its slower speeds--is faster than ISDN for downstream communications--1.5Kbps versus 128Kbps. That speed has made early ADSL users happy. One ADSL pioneer, Howard Maher, vice president of marketing for ProTech Books, told the magazine Network World: "Our ADSL modems are so fast that many of our customers thought we were loading web pages from a huge hidden hard drive. It's incredible."

ADSL provides asymmetric "downstream" transmission of between 1.5 and 6Mbps of downstream transmission, depending on transmission lengths, with 9Mbps coming in the future. Notice the word "downstream." With ADSL, data flowing from a server to a client moves much faster than the "upstream" communications from the client back to the server--which travels only at 800 Kbps.

While at first glance that might appear to be a problem, the fact is that the Internet bottleneck is largely unidirectional: a lot of data heads from Web servers to client sites, but relatively little data goes from users back to servers. The one notable exception: video conferencing. Even the prospect of Java applets coming into widespread use wouldn't much change this scenario. Imagine, for example, a Java applet in the form of a spreadsheet being downloaded to your computer. The way Sun Microsystems and other Java companies envision it, you might manipulate data on this applet locally on your machine to figure how to refinance a car loan. Only after this local manipulation takes place would you send any data back to the site. The applet itself travels strictly downstream and "dies" on the client.

Unlike cable modem technology, ADSL takes advantage of the existing infrastructure of twisted pair phone lines running to homes and businesses. There are only about 12 million homes worldwide that can support cable modem, compared with 700 million homes already wired for telephone. On the surface, ADSL's speed of 1-6Mbps might appear slower than cable's downstream bandwidth of 30Mbps--except that bandwidth is shared among many users on the line. Indeed, as I found out in talking to several prospective cable modem providers, it is difficult to get anyone to commit or agree to what coaxial cable actually delivers to the home.

ADSL--what is it?

ADSL was initially developed to distribute video in real-time over ordinary telephone lines. But in 1995, it became clear that the more viable application--at least for the short term- -would be as a data communication technology. As such, ADSL has been speeded up--from 1.5Mbps to 6Mbps, with some predicting it will reach at least 9 Mbps.

"ADSL went from a fixed speed to a variable speed, allowing phone companies to reach more people without having to change a line," said Kim Maxwell, chairman of the ADSL Forum. Maxwell predicts that ADSL will eventually serve 80 percent of subscribers with the existing telephone infrastructure, with the remaining 20 percent served by running a fiber optic cable from the central office to a remote node. Such a node would provide 6Mb service up to 9,000 feet, or a circle of about nine square miles.

ADSL makes use of three information channels--a high-speed channel for downstream server-to-client communication, a medium-speed duplex channel for upstream communication, and a third channel for conventional voice transmission, known as POTS, or plain old telephone service.

The high-bandwidth channel speeds currently ranges from 1.5 to 6.1Mbps, with speeds of up to 9Mbps predicted. The actual speed depends in part on the wire gauge and part on the distance of transmission. For example, a signal traveling 5.5 kilometers over .5mm wire will yield up to 2Mpbs bandwidth. The same wire size at 12,000 feet will yield 6.1Mpbs. According to the ADSL Forum, 95 percent of the existing wired infrastructure will cover these capabilities.

One source of confusion for ADSL is that it supports two competing modulation techniques. In Carrier-less Amplitude/Phase (CAP) modulation, incoming data is modulated into a single carrier that is then transmitted via a telephone line. The technique is similar to Quadtrature Amplitude Modulation, which is used in V.34 modems. The other technique, Discrete Multi-Tone (DMT), collects and distributes multicarrier incoming data over a large number of small individual carriers. DMT creates these channels using a digital technique known as Discrete Fast-Fourier Transform. DMT is the basis of ANSI Standard T1.413.

"People say DMT is better and will offer higher speeds--but the equipment is much more expensive, said Lev Kaye, vice president of sales for InterAccess Inc., the first ISP to offer ADSL on a subscription basis. "When we see low cost DMT equipment shipping, that's probably when will begin to implement it. But right now we are implementing CAP because it is here today."

As a standard, ADSL is still being specified. A working group of the American National Standards Institute (ANSI) recently approved an ADSL standard at rates up to 6.1Mbps, while the European Technical Standards Institute (ETSI) contributed its own standards to reflect European requirements.

ADSL's biggest disadvantage compared with ISDN is that it probably won't be broadly available until at least the end of 1999--while ISDN service can now be purchased from almost anywhere in the U.S. Nevertheless, small ADSL trial programs are taking place both in the U.S. and Europe. The participants include virtually every mid-to-large size telephone company in the U.S.--including Nynex, Pacific Bell, Bell Atlantic and US West.

One hotbed of ADSL activity is the city of Chicago, where, within a single week last September, an Internet service provider and a Bell company both announced plans for ADSL access. The ISP, InterAccess, is probably the world's first provider to offer ADSL as a subscription service, while the telephone company, Ameritech Corp., is launching its first field trial. Both companies are using the same ADSL supplier- -Westell Technology Inc.

Ameritech started a six-month trial last October with 200 of its customers. InterAcess is a small ISP, and won't disclose how many of its customers have actually signed up for ADSL service--implying, at least, that the number is still unimpressive. To launch the service, the company staged a face- off pitting ADSL against ISDN and a standard 28.8Kbps modem. The goal: download a 190 KB image of the Mona Lisa. ISDN came in about five times slower.

InterAccess and other independent ISPs face one big obstacle: ADSL's limited transmission length. ADSL requires that equipment be co-located in or near every telephone central office--which in the U.S., requires the cooperation of the local phone company. "We either have to get their cooperation or find a way to get the door to the central office opened up, or we have to open up our own offices next door to the central offices," said Kaye. "That's going to be the tricky point over the next 12-18 months."

But InterAccess still views ADSL as the best high-bandwidth solution. "We see ISDN as an interim technology," Kaye said. "It's too complex to ever become widespread--and will be reserved for the techie elite. By contrast, ADSL is so easy to implement that it can become ubiquitous. And once the connection is set up, it always there--you are always connected to your ISP. If you wanted to, you could leave your computer running when you go to work, and when you come home you would just touch the keyboard to see your all your e-mail messages waiting for you. It's not a dial up connection. You are always on."

Wireless connections via satellite

For Japan, as well as the U.S., ADSL has the advantage of using the copper line telephone infrastructure already in place. Another emerging high-bandwidth option--satellites-- doesn't even bother with the wires. In the United States, satellite technology--which involves putting a small dish outside your home--is beginning to challenge cable as a provider of premium television service, called "direct broadcast service" or DBS. In some small European towns, such small satellite dishes are often mounted on almost every house.

One company, Hughes Network Systems, has begun to offer Internet service via a mounted satellite dish, as well. The advantage is that you can receive broadband from virtually anywhere with an unobstructed path between the satellite and your dish. The disadvantage is that, so far, satellite connectivity is--to put it politely--difficult. Satellite connectivity requires elaborate installation, a large up-front investment of $700, and hours of tinkering to get the dish pointed precisely. Users also found that early versions of Hughes' software crashed under Windows 95, and that you must establish two Internet connections, one for receiving via the satellite, the other for sending it through a standard analog phone connection.

That was last summer. By autumn, Hughes said it had revamped its software and made the whole process easier. Most surprisingly, the company even swung a deal with CompUSA--an American computer store chain, to offer do-it-yourself kits through its 20 California stores. The kits include a 21-inch dish, 100 feet of coaxial cable, an adapter card and Windows 95 software.

Clever marketing, but this is still a very costly service. In addition to the $700 up-front price and the time to set up the disk, users pay $9.96 a month access fee plus an additional monthly fee of from $40 for night and weekend access to $130 for unlimited around-the-clock access.

So with the cost and installation hurdles, who is a good candidate for Internet connectivity via satellite? In part, of course, it is those people in the U.S. who have replaced their television cable service with direct broadcast service (DBS) and have hence already installed the service. "Those people will be our customer set, and people in rural communities as well," said Paul Gaske, senior vice president, Satellite Networks Division, Hughes Network Systems. "Some small offices are also good candidates, depending on the price structure they are offered."

Gaske also sees possibilities in Japan. Last April, Hughes announced an alliance with Space Communications Corporation, a unit of Mitsubishi Group, for what it called the first digital satellite service in Japan. The service is in the test stages. "Japan has very wide ISDN capability, with not much coaxial cable hookups to speak of--so there may be some unique opportunities for us," said Gaske.

>From a business perspective, satellite technology has one big advantage over all other high-bandwidth technologies to date: the system can expand without additional costs. And those costs can be prohibitive. TCI, for example, the large U.S. cable television provider, has announced it will stop upgrading its cable system, relying instead on a new compression technology to cram more bits into the coaxial "pipe." In the process, it will stop installing cable modems until there is a demonstrated demand.

It's a tough choice to invest in infrastructure if you aren't at all certain whether you will get your money back. And so, pricing is a critical issue. "We're not only trialing the speed and convenience of ADSL, but also trying to determine what people will pay," said Ameritech spokesman Rick Aspan.

What will people pay? I don't know about you, but I'm on the Internet daily--and most of the time, I'm just interested in text. So like a lot of people, I uncheck the "load images" option on my browser so that downloads finish sooner, and most of the time, that approach works just fine. Maybe I'm a cheapskate or maybe as a writer I'm more oriented toward words. But until I see more compelling images on the Internet, I'd rather take the money I might have invested in high-bandwidth services and put it into something else. Those cross-country skis I just bought, for instance, are part of the broadest-band experience imaginable: real life.

An interview with Kim Maxwell, chairman of the ADSL Forum.

Kim Maxwell started Racal-Vadic, a major modem supplier until the early 1980s, and in 1990, he formed Amati Communications, which specializes in ADSL. He accepted the chairmanship of the ADSL Forum in 1994.

What are ADSL's prospects in Japan?

It's hard to say. I believe that NTT still wants to bring fiber to the home--which would deliver many gigabits, but is expensive--perhaps $600-$700 a home. But if NTT unbundles the local loop [the wiring from the telephone central office to the subscriber's telephone]--thereby giving copper pairs to alternative suppliers, then ADSL could be offered by other companies.

Why is ISDN so costly and will ADSL be any cheaper?

ISDN is expensive because it is a circuit-based system that requires telephone companies to upgrade the switch--even if you have a digital switch. And the software is expensive, as is the maintenance. The installation of ISDN is very complicated because of the way it was designed. There's millions of options and if you move it you have to change all the options. And it's got the same problems as ADSL in terms of transmission lengths.

ADSL doesn't go through a switch. It comes into a central office to a concentrator, which connects to a remote ATM switch. The typical configuration in the United States, which I think will be true elsewhere, is one big ATM switch for 40 or 50 central offices. They will be connected to the central offices over a fiber link, which is already in place. Each central office will have an ATM access concentrator--a simple device.

What does this work out to in terms of cost for the implementor?

Under $500 a line.

As opposed to ISDN?

I don't know what those costs are. If the tariffs reflect real costs, ISDN is about $800 to $1,000.

More important than cost, is the speed of implementation. An ADSL concentrator can support between 200 and 400 customers in about a two foot square footprint. An area the size of your desk will support 1000 to 1500 customers. The ADSL footprint is small. It's a completely separate network, and it's not that difficult to install. I believe that unlike ISDN, which covers 75 percent of the United States, the phone companies will have coverage for ADSL higher than that by the end of this decade.

So you are predicting ADSL will ultimately surpass ISDN?

No question.

When can we expect general service?

By 1998, you'll start seeing technical bugs worked out well enough that you'll see a major program underway for rapid distribution. The service will not be available from every central office until the end of 1999.

Is ADSL going to be restricted to the Bell companies that have access to a central office?

We don't know. If the FCC has its way, the phone companies would make access to copper relatively easy. In practice, it will probably vary by region. For example, Ameritech is not particularly unhappy about having non-phone company equipment on their premises, in which case it's not that difficult. If the phone company says you can have access to the wires but you can't put equipment on their precious territory, ADSL providers will have to do something more complicated, like run some copper wires from the main distribution frame over to another building.

However, that won't be a big impediment in the long run. I believe the structure of how the phone company deals with its enemies will all be worked out.

How do you compare with coaxial services, like TCI's @Home?

Fundamentally, while there are some differences, the appearance to the user of a coax modem and ADSL will be virtually identical. The coax modems deliver 30Mb downstream, and have some real problems upstream. But the 30Mb is shared among all users. It's a function of how many users are on the line at one time, and what they are doing.

Even so, will the speeds be comparable to the end user?

In terms of raw performance, after 1Mb you won't see any difference. The truth is that if you are four miles away and I can get 1Mbps to your house, you'll be happy. If the guy next to the central office gets 9Mbps, you'll each have equal performance for quite a while because of inherent delays in the Internet. That will eventually change, assuming there are a lot of proxy servers out there and people are spitting out video--then the difference between 1 and 6 and 9 Mbps will start making a lot of sense.