You need fiber optic cable to your house for telecommuting, Internet access, and all those new interactive video services you see at the trade shows. You can't have it. Bellcore figures that it will be at least 2005 before fiber installation economics converge with the cost of operating, provisioning, and maintaining the existing twisted pair copper loop that delivers voice service to your home now. So what do you do? Maybe, just maybe, the new Asymmetric Digital Subscriber Line (ADSL) technology will be the thing that users need to inch their way toward broadband residential access speeds.

ADSL can allow users to download digital information at multimegabit speeds while still happily chatting on their analog phone, and all over the same twisted pair copper wire local loop that they used before. Service providers may not need to run any new outside plant wiring at all to offer the new service. Outside plant costs have historically been the highest component of service deployment and traditionally have the longest payback periods. Any way to deploy new services that people want on existing cable plant is attractive to users, vendors, and carriers alike.

It won't be an easy task, even for as promising a technology as ADSL. ADSL is really an umbrella term for many different schemes methods, all of which are represented by products and even commercially available services. This article will look at the many variations of the ADSL technology and examine some of the hoops of fire that ADSL must jump through before multimegabit access speeds come to your block. But rest assured, come they will, and you need not wait until 2005.

The very name ADSL is mildly controversial. The Digital Subscriber Line concept is as old as ISDN, so the sending and receiving of digital bits instead of analog signals with analog telephones or modems over the twisted pair local loop is not the controversial part. But the term "asymmetric" naturally implies that bit rate transfer speeds downstream, or inbound to your home, are not equal to (and actually much greater than) the transfer speeds upstream, or outbound from your home.

The problem is that not all DSL solutions are asymmetric. Some of them offer exactly the same speeds inbound and outbound. Some have argued that these symmetric variations mean that the term xDSL is much more accurate than ADSL. The "x" in xDSL can stand for any one of a number of competing or complementary technologies that describe a raft of other speeds and DSL arrangements. The latest xDSL menagerie is shown in Table 1.

In spite of all the variants, the term ADSL has been accepted by the networking community as the umbrella term for all new DSL technologies. The ADSL Forum (www.adsl.com) has much information about VDSL. From a marketing standpoint, this makes perfect sense. Consumers have a had a hard enough time grappling with such obvious technological distinctions as Beta versus VHS, or even analog voice versus "hear a pin drop" digital. The subtleties of HDSL versus ADSL will evade all but the most sophisticated consumers. So the marketing of any xDSL technology as ADSL is the safest and best approach for a manufacturer, reseller, or service provider to take. This will be especially true if HDSL and/or VDSL eventually follow and even subplant pure ADSL in the future.

It is important to point out that each of the speeds listed as characteristic of a given xDSL technology are not hard and fast, or even achievable in many cases. This may come as a surprise to those who are used to the fact that a T-1 always runs at 1.5 Mbps, or that an ISDN BRI (Basic Rate Interface) always has a aggregate bit rate of 144 Kbps for one and all. This is true no matter where nor exactly how the T-1 or ISDN BRI is offered for subscriber use.

Although xDSL is as digital a technology as T-1 or ISDN, xDSL shares an unfortunate characteristic with the analog modems everyone has come to love and hate. It is not unusual for a 28.8 Kbps or 33.6 Kbps analog modem to connect with a remote modem at speeds well below the standard, rated maximum, and often only half as fast. This is due to the need to use existing, and therefore unpredictable in terms of noise and signal loss, loops and lines whenever a modem connection is made. T-1 leased lines can be routed where quality supports the standard. ISDN is deployed according to the ISDN wiring standard. But ADSL standards control only the end equipment capabilities, not the wire quality in between. Many ADSL studies and trials have shown a distressing tendency to limit throughput well below the theoretical maximum.

Usually, when ADSL functions below the rated, standard speed, the decision is made at service provision time. That is, one neighbor may be able to get a full 1.5 Mbps inbound, while a home nearby may have to make due with only 768 Kbps. Regardless, the speeds should be consistent. However, the attraction of RADSL (Rate-adaptive DSL) equipment is that the speeds can be adjusted on the fly, in use, as line quality deteriorates or improves with melting ice, heavy rains, or the number of gnawing squirrels.

Digital Subscriber Lines were deployed in order to support ISDN services years ago. ISDN has been positioned and marketed as a service for high-speed Internet access, telecommuting, and so forth. It seems like the goals of ISDN are the same as ADSL. If this is true, then just what is the relationship between ADSL and ISDN? Is ISDN obsolete? Is ADSL superfluous? Will they coexist? Must one disappear?

Simply put, the pure ISDN BRI rate of 144 Kbps into and out of the home, is just too slow for hooking today's home-based clients up to today's network servers. And if the possibility of home-based servers and LANs is added to the equation, even an ISDN PRI (Primary Rate Interface) running at a full 1.5 Mbps in and out may not be enough. Of course, the most basic ADSL offers only 64 Kbps outbound, but this is exactly the same as an ISDN bearer channel (B channel). For inbound service, the simplest ADSL offers at least 1.5 Mbps, but this is a fully useful bandwidth for data packets. B channels on both the ISDN BRI and PRI can be bonded to yield higher aggregate bit rates, but this practice can be awkward for users and sometimes not even supported by the service provider. With ADSL, there are no real channels for users to be concerned with.

ISDN requires new users to swap out their existing analog phones for digital models or invest in some conversion equipment. Of course, nothing need be done if the ISDN line is a second line used for purely data applications. In that case, the former analog phone line is retained unchanged (and still paid for, of course). ADSL, on the other hand, typically handles analog voice as an analog baseband signal on the same wire pair that transfer the ADSL service digits. There are exceptions, but usually adding ADSL to a local loop does not require any extra equipment for the user's continued analog voice services, nor another pair of wires to provide the new digital services.

Moreover, ADSL offers a permanent, packet-switched connection for all services, while ISDN supports most of the things ADSL is used for only on traditional circuits (the B channels), which dedicates bandwidth and switching resources to a specific connection for the duration of the connection. This is fine for most videoconferencing applications, but can be an enormous waste for data applications that are bursty in nature, such as Web browsing or remote corporate database access.

Most users at home will at least consider ISDN BRI for higher speed network access. This is priced reasonably enough nationwide, as low as about $100 per month in many cases. But ADSL speeds are more often and properly compared not to BRI speeds of 144 Kbps, but ISDN PRI speeds of 1.5 Mbps. The ISDN PRI is much more pricey, averaging about $1300 per month nationwide. However, PRI will still be the choice for businesses, due to the better match with PBXs and routers, for the foreseeable future.

This does not mean that ISDN BRI will take a back seat to ADSL. Close to half a million BRI lines have been deployed, and until ADSL line conversion charges fall into the $500 per local loop range, where carriers want them, BRI will remain a popular high-speed access choice for many users.

There is another factor to consider when comparing and forecasting ISDN and ADSL. ISDN, both BRI and PRI, requires extensive and expensive software upgrades or wholesale replacement of central office equipment. The phone companies have tended to wait until a appreciable number of users in a given area seem like good candidates for ISDN marketing before spending these dollars. This has often meant a lack of ISDN availability in many rural areas and even whole states.

ADSL, on the other hand, can be deployed almost on a "one off" basis. A local loop converted to ADSL still funnels analog voice through the same voice switch (typically), but hands off digital traffic to a corporate private line for telecommuters, an Internet router for Web browsing, or a video server for more advanced offerings. Only the price of the ADSL modems and data communications equipment need be recovered to make the service viable and attractive. Users can be added incrementally just by adding more ADSL modems at the central office.

So it seems that corporate users will be happy with ISDN PRI for the time being. But ISDN BRI will find itself as more and more of a niche offering as ADSL lines become more common. There is no doubt that users will migrate from BRI to ADSL in droves once the added speeds available of ADSL lines become not only desirable, but absolutely necessary. One early former bonded B-channel BRI Web surfer reportedly opened up the ADSL modem device because of the conviction that a hidden hard drive was loading the Web pages faster than the user believed possible. Now, that's technology that will sell!

Whenever something sounds too good to be true, it's usually because it is. Sadly, ADSL may fall into that category, along with cheap, safe nuclear power, job security, and long, hot showers in hotel rooms. Because any ADSL method can only control the endpoints of the two-wire local loop, wide variations in speeds and performance have been reported in many early trials. And none have consistently supported the architected speeds. For example, a GTE trial using HDSL, which should have supported 8 Mbps into the residence, will end up supporting only speeds from 128 Kbps to 1.5 Mbps.

Why such a wide variation? It is one thing to say that local loops vary in quality and electrical characteristics. But just what is it about local loops that seems to drive digital schemes like ADSL crazy?

Figure 1 shows a telephone central office switch with all local loops divided into four major categories. The categories are admittedly somewhat arbitrary, but not totally subjective. Most carriers would agree to the broad terms of classification used in the figure. First, there are copper pairs that extend less than 18,000 feet (18Kft) from the office. Next, there are copper loops that are extended through the use of loading coils, which are nothing more sophisticated than big iron doughnuts that add inductance to the twisted pair to counteract the effects of the added distance on signal loss. These loops may extend to about 30 Kft or so. There are also local loops which have been extended by some other active equipment to reach up to about 25 Kft, known generically as line extenders. Finally, there are loops which consist of copper pairs that branch off into sub-loops with what are known as bridge taps. This is typically done to avoid the need to run new twisted pair all the way back to the central office. A new pair can be "bridged" onto an used pair already run on another block. Sub-loops typically use a finer gauge of wire than the "main" pairs, since the distance is less and this works just fine with analog voice signals.

The problem is that only one of the types of loops described is appropriate for any current ADSL technology. These are the plain vanilla cooper pairs running less than 18Kft from the central office (a little more than three miles). All others are unsuitable for ADSL. Loading coils add too much induction for digital signals. Line extenders cannot function in a digital environment. And bridge taps attenuate digital signals far too much. And if that were not enough, mixed wire gauges reflect digital signals and result in more disruption.

There are other, environmental factors at work on the local loops as well. These factors are most severe in the Northeast, where years of blizzards and blazing hot sun beating down on aerial wires (most northeast suburban wire is still on poles) have taken their toll. And the wire has had plenty of time to suffer in many cases. Some neighborhoods are still festooned with copper wire from the 1930s (this is not a typo!). Even for regular analog telephone service, these wires often hum with noise, frequently crackle and pop, and are prone to sudden, catastrophic failures. Digital techniques like ADSL or even ISDN are offered only at high risk for the carrier. As a resident of the Northeast for his entire life, the author can personally attest to each of these conditions. But the factors are also well documented.

So the cost of deploying ADSL, which is attractive to many carriers as an incremental cost and a means of digitizing the local loop, must be balanced by the possibility that vast areas of customer pools may remain beyond the reach of ADSL, especially in the customer-rich Northeast. The alternative is to eliminate the problem at its source. This means re-engineering the local loops into "tail ends" serviced by coaxial or (more often) fiber carrier systems (which is slowly and agonizingly being done anyway). In fact, at least one ADSL method, VDSL, relies on the presence of fiber carrier systems only 1,000 feet from a customer's doorstep. Most other ADSL technologies that offer 6 Mbps are intended for customers up to 12Kft away.

Of course, in this case the incentive provided by ADSL to preserve the existing copper pairs in considerably diluted. The re-engineering cost for ADSL must now be balanced against the cost of deploying other digital technologies to the home, such as ISDN BRI. Even CATV Hybrid Fiber-Coax (HFC) systems, which are all exclusively still analog, may be a possibility. At least one local carrier has seen HFC as the only viable long-term solution to the residential bandwidth squeeze.

The current limitations that the existing copper physical plant poses to ADSL technologies has not stopped numerous vendors from developing and marketing a whole line of ADSL products. All are called ADSL modems. Although often still called "modems", HDSL products are technically just another form of digital DSU/CSU (Digital Service Unit/Channel Service Unit). The term modem seems have become a generic term for any device that links a computer to a network, be it analog or digital. These products have been used in a number of carrier trials throughout the United States.

Although it may still be too early to handicap the big winners and losers in the ADSL device derby, some early leaders are emerging. However, this is all based on choices made by carriers and early returns, so all impressions are still liable to change. Some 20 companies currently make ADSL equipment or components. A full list of ADSL vendors with links to their own Web sites (where relevant) is available at the ADSL Forum Web site (http://www.adsl.com/adsl_vendors.html).

The most successful ADSL vendor to date has been Westell. They make it all and sell it all, from customer to central office gear, and in a range of ADSL technologies. Close behind are companies such as Performance Telecom, Orckit, and Amati, as well as such well-known firms as ADC Telecom and AT&T Paradyne (which licenses ADSL from GlobeSpan, another AT&T piece). The usual modem gang is involved as well, with Motorola, U.S. Robotics, Rockwell, Pairgain, and Alcatel all weighing in with their own products lines. The rest of the ADSL market is scattered with relative newcomers with names like Aware, Diamond Lane, Teltrend, Adtran, Copper Mountain, Netspeed, Analog Devices, and Cayman Systems.

Westell's FlexCAP ADSL "modem" has been the most common choice for ADSL carrier trials, usually running at 1.5 or 2 Mbps downstream. GTE's widely publicized ADSL trial with Microsoft in Seattle uses the Amati Overture 8 (8 Mbps downstream). Amati licenses its ADSL technology to Texas Instruments, NEC, Nortel, and Motorola. The Amati-Motorola relationship is interesting. Motorola is developing its own "single-chip" ADSL product called CopperGold. CopperGold will in turn be licensed to Amati, along with other interested companies such as Ericsson, Cabletron, and even Westell.

Space limits this quick look at ADSL vendors. Look for the ADSL market to become very crowded very quickly. The shakeout will occur shortly thereafter.

Until recently, ADSL service offerings could have been described as "escaped lab demos". All ADSL equipment so far has tended to be pricey. Costs of about $1500 per line have not been uncommon. But this compares with the cost of $2500 only a year or so ago. In order to charge an anticipated $40 to $100 per month for ADSL service, the carriers would like to get this cost down to about $500 per line. Certainly, heated competition and more volume production will help a lot.

All of the former pieces of the Bell System (the RBOCs like US West and BellSouth), GTE, and several smaller local companies, have ADSL trials in progress. Most tend to be rather small. GTE's trial with Microsoft is the most publicized, but is not the largest. Both GTE trials, one in Seattle and the other in Irving, Texas, involve only slightly more than 100 lines. The most ambitious is the Bell Atlantic market trial in Fairfax, Virginia, which involves some 1000 or so homes.

Most trials have no pricing associated with them at all. Those that do will typically try out a variety of rate and price plans as the service provider seeks to find the right combination of installation and monthly costs that maximize revenues and/or the number of customers. However, $100 per month seems to be the acknowledged ceiling for ADSL service. This compares quite favorably with ISDN pricing, of course.

Trials are fine as far as they go. However, there are even commercially available ADSL services around the country, with structured pricing, stable service, and even repair crews. The most publicized has been Chicago ISP InterAccess, which offers ADSL at 1.5Mbps/64Kbps for a modest $200 per month for home use, and $1,000 per month for a business's LAN. The Aspen Internet Exchange in Colorado offers ADSL pricing along the same lines from $100 to $1500 per month. CADvision in Calagary, Canada, offers an astonishing 2Mbps/500Kbps for only $53 per month.

The Anchorage Telephone Utility in Alaska has a variety of speeds available (depending on loop conditions) at about $200 per month. And in Massachusetts, HarvardNet will soon offer Internet service for LANs in Boston and Cambridge from 128 Kbps ($299/month) to 384 Kbps ($500/month) to 768 Kbps ($750/month). Ironically, most of the commercial ADSL services are squarely aimed at business customers.

What kind of future will ADSL have in a world that also contains ISDN, cable modems, direct broadcast satellite, and even other alternate methods for allowing residential users to enjoy the benefits of broadband access in their homes? After all, most technologies can flourish for a time in a niche, but eventually they must move into the mainstream or disappear.

While ADSL shows considerable promise, especially for some flavors of the technology, it remains to be seen if local loop performance limitations will eventually force carriers to abandon ADSL in favor of something else, something that perhaps has not even left the lab yet. However, all other current ADSL alternatives suffer from their own limitations. ISDN also has distance limits, basically the same as ADSL. Cable modems suffer from lack of two-way CATV systems, and are still analog. Satellite systems are digital, but lack wireless two-way capability.

ADSL has a window of opportunity, to be sure. But most carriers and vendors will not put all of their eggs in the ADSL basket just yet.