When Will PC OEMs Bundle ADSL Modems?

As the Internet became more widely used in the late 1980s and early 1990s, modems started to become a commodity, although one not yet bundled with new PCs (Fig 1). During this time, more internal (ISA and then PCI) modems were available through retail channels and even service providers, and the price war began. The transition from 14,400 BPS to 28,800 BPS (and then from 28,800 BPS to 33,600 BPS) introduced another variable--up gradeability. A user could upgrade a modem with a single program memory chip (EPROM) that supported the new standard (V.34 and higher), although the upgrade procedure was highly technical and sensitive.

As the Internet reached the mainstream in the mid-90s, PC manufacturers realized they could generate more revenue by bundling modems with new PCs. They began with high-end computers, equipping them with the top-of-the-line sound cards, hard drives, and modems. Eventually, even low-end, entry-level computers came equipped with dial-up modems. The manufacturing decision was driven by price; modem bundling did not occur while the prices of external modems were around $300, and the prices of internal modems around $150. It was only when the price of a modem dropped below $35 that PC manufacturers decided to bundle modems into all new computers.

With the initiation of the modem "price wars" in the late 1990s, modem and modem chipset manufacturers started looking for ways to lower the component cost. Several chipset manufacturers introduced the "controller-less" modem that had some functionality (mainly protocol-level) implemented by the host CPU, thus lowering the size and complexity of the silicon chips in the modem--and therefore lowering the cost and the price of the modem.

Some modem chipset manufacturers took this idea to the next step by completely implementing all of processing functions through the host CPU, achieving a true "software modem." Although the price of these modems and chipsets were the lowest available, PC manufacturers were still reluctant to use them due to operating system instability and CPU requirements. While a hardware modem would not consume any CPU resources, a software modem would consume around 40 MHz from the host CPU. In 1994, when the top-of-the-line processor was the Pentium 75 MHz, consuming 40 MHz really affected the PC performance; thus, the idea of bundling the software modem was rejected by PC manufacturers. With the appearance of the Pentium 166 MHz, however, this problem was solved.

In 1999, 38 percent of the dial-up modems sold were full-hardware modems, 37 percent were controller-less and only 25 percent were full-software modems. In 2004, analysts project that only 3 percent of all modems will be full-hardware, 37 percent will be controllerless, and as many as 60 percent of all modems will be full-software modems (source: Cahners In-Stat, 2000).

Projected Technology Adoption Cycle for ADSL Modems

Today, the ADSL market is in its early adopter phase. ADSL provisioning is accomplished using external ADSL modems connected to PCs through either the Ethernet or USB cable. Service providers give their customers free or subsidized-cost ADSL modems in return for long-term service contracts. Subscribers see the benefit of ADSL service as being a high data-rate, always on Internet connection, and are willing to pay for it. However, there are many problems that prevent ADSL technology from mass deployment, almost all of which may be described under the title of infrastructure.

Although consumers have demonstrated a willingness to purchase DSL service, telephony infrastructure was never designed to support frequencies higher than the 4 KHz required for vocal transmission; ADSL communications require 1.1 MHz. This means that, due to the telephone line characteristics, the maximum distance between the central office and the subscriber is 18,000 feet. In the United States, only 60 percent of the population lives within this distance restriction. Additionally, 30 percent of U.S. households are connected to the telephone network through a Digital Loop Carrier (DLC), which does not support ADSL communications.

These problems can be identified prior to the installation of an ADSL modem by submitting the telephone number of the prospective subscriber to the Local Exchange Carriers (LEC) database; however, there is yet another problem. Bridge taps and load coils affect ADSL communications dramatically. While the loop lengths are documented by the LECs, bridge taps, load coils, and other impairments are not documented. This means that, even if a prospective subscriber pre-qualifies for ADSL service, there is still a 25-30 percent probability that the user will not be able to get ADSL service due to these loop impairments.

These are the major hurdles facing manufacturers who might be interested in bundling an ADSL modem with a new PC. if a consumer buys an ADSL modem through a retail outlet and cannot get service, he or she will simply return the modem. However, when a user buys a computer bundled with an ADSL modem and cannot get service, he or she will return the computer, which is a riskier proposition for the PC manufacturer.

Another indication of the early adoption state of the ADSL market is the cumbersome process of upgrading to DSL through a service provider. Expensive truck rolls are involved in order to test the line, solve problems, install, and configure the modem. ILECs (Incumbent LECs such as Verizon, Bell South, and Qwest) have complained that the overall cost of provisioning a single line exceeds $750. With a service fee averaging only $40 per month, it is easy to see why they require a long-term contract. The situation worsens for CLECs (Competitive LECs such as Covad, Rhythms, and NorthPoint--now defunct) that incur costs exceeding $1,200 per subscriber, yet still have to pay up to $20 per month to the ILEC for sharing its line. The current lag time of initiating service up to six weeks or even several months after it was ordered is another indication of an immature market.

These problems must be addressed before ADSL technology can move into the mainstream market. But, assuming these infrastructure problems will be resolved, and that the ADSL market will follow the technology adoption curve of the dial-up modem, it would be reasonable to then assume the market will develop as illustrated in Figure 2. Interoperability issues for external and internal ADSL modems will be resolved; service providers can then stop subsidizing the modems and force the consumers to purchase them in retail stores. This may occur later this year or in early 2002 (Fig 2).

At this time, modem branding will become an important issue as end users are willing to pay more for products with a name they recognize. Price wars will start driving down modem prices, while driving up the number of users buying internal modems instead of more expensive external modems. 3Com, Creative Labs, and similar companies with strong aftermarket brand recognition will have an advantage at this stage of the market; however, in order for an end user to purchase an ADSL modem, he or she must still be able to pre-qualify for service. Therefore, a kiosk must be installed in retail stores that enables users to pre-qualify for service at the point of sale.

The ADSL market will achieve its mainstream stage when: 1) the probability of being able to subscribe to ADSL service will be very close to 100 percent; 2) all infrastructure problems are solved; 3) the actual number of ADSL users are in the tens of millions; and 4) modems are priced below $25. At this point, PC manufacturers will decide that it is time to bundle ADSL modems with new PCs. Initially, just as the case with the introduction of dial-up modems, only the high-end computers at $1,500 or more will have an ADSL modem. As prices drop and adoption increases, all computers will include ADSL modems.

Lowering the cost of the ADSL modem will be achieved in many ways, but much like the dial-up modems, the lowest cost solution will be achieved by software-based ADSL modems. While software-based dial-up modems consume 40 MHz of the CPU resources, an ADSL modem operating at 1.5 MBPS may consume 200 MHz. Intel recently announced that it expects to deliver a 10 GHz processor within the next five years, so it is safe to assume that the CPU resource consumption by a "soft" ADSL modem will be negligible.

The advantages of a software modem offering the "future standard compatibility guarantee" will be emphasized, and, along with the lowest possible price, will constitute an order winner to the PC manufacturers.

PC OEM ADSL Modem Buying Criteria

When making a decision to buy, integrate, or use a CPE (Customer Premises Equipment) ADSL modem, a service provider, PC manufacturer, system integrator, or other must consider interoperability, total cost of installation loperation, performance, and price.

Lowering the price of a modem will not win orders if that modem does not prove to be at least as good as the competition in all other areas. In fact, a more expensive modem may be chosen if its interoperability, total cost of installation/operation, or performance are better. Only when the interoperability, total cost, and performance are similar (or at least exceed a minimum required level) will the price of the modem become the order winner.

Although the ADSL standards seem finalized (ITU standards G.992.1 and (G.992.2), DSLAM vendors continue to implement ADSL in slightly different manners. This was also true for dial-up modems when two 56 KBPS standards were introduced: Rockwell's (now Conexant) K56Flex and U.S. Robotics' (now 3Com) x2. Eventually, the two merged into the ITU standard known today as V.90.

Since ADSL implementation varies with DSLAM vendors, a CPE modem must be interoperable with as many DSLAMs as possible to be widely adopted. In the future, ADSL will become more standardized and interoperability will no longer be an issue; for now, a modem capable of interoperating with many different DSLAMs is a lower risk for the service provider or PC manufacturer. The interoperability issue emphasizes the advantages of a software-based modem, capable of offering a "future interoperability guarantee" simply by software driver upgrades.

The cost of supporting and installing an ADSL modem has several components. One is the cost of installation and provisioning. Today, installing an ADSL line involves expensive service technician "truck rolls" for (a) physically testing the quality of the copper loop between the central office and the user premises, (b) fixing the copper loop and the in-house telephone wiring, if necessary for qualifying the loop for ADSL provisioning, and (c) installing the CPE modem and configuring it. While the price of a CPE ADSL modem can be as high as $150, the cost of the truck rolls can easily exceed $300. Introducing modems that can be easily installed, self-configured, and that can possibly even simplify the loop qualification process through added diagnostics can lower the overall cost of installing and provisioning ADSL. "Auto-config," an activity by the OpenDSL and DSL Forum is helping to develop a standard by which ADSL CPE modems will self-configure upon connection with a DSLAM.

A second component of the cost of installation and operation is the support cost. A less-reliable modem will generate more technician calls to the service providers than a modem that is easier to operate with self-diagnostic capabilities. While lowering the price of a modem by $10 is significant, it does not have the same impact as a modem that generates three fewer service calls, potentially worth $45 in savings.

Modem performance is another important decision criterion can be defined as a modem's ability to deliver high-speed data to a distant end-user over real-world telephone lines. There are several factors to evaluating performance, including high-speed data rates and longer telephone lines.

Maintaining a comparable high-speed data rate to other moderns at the same distance from the central office is important especially to remotely located provision users. A modem capable of delivering 768 KBPS over a line on which another modem is only capable of delivering 256 KBPS is considered better, and allows the service providers to better serve the users within a certain reach of td e central office.

Being able to deliver the same data rate over a longer telephone line is important as it increases the customer base potential to a central office. A modem capable of delivering 384 KBPS over 18,000 feet (provided that this is the level of service guaranteed by the service provider to an end user) is better than a modem capable of delivering the same data rate to 15,000 feet. Increasing the reach of a modem by 20 percent, as in this example, will increase the target customer base by 44 percent!

The telephone line infrastructure was never, designed to carry the high frequencies (over 1.1 MHz) used by, the ADSL technology. It was designed to carry only 4 KHz used for voice conversations. While bridge taps, load coils, and other loop impairments will not affect the ability of a telephone line to carry voice, they dramatically decrease its ability to carry ADSL data, as the high frequencies are the first to be affected. A modem better suited to handling imperfect lines (i.e., delivering higher data rates over that line) allows the service provider to extend good service to more customers, especially those with impaired telephone lines.

It is worth mentioning that, while the ILECs and CLECs have databases that indicate, the loop length to their customers, the load coils and bridge taps on the loops were never properly documented. As a result, customers prequalifying for ADSL service using these records are doing so with inaccurate data; in fact, 25-30 percent of them will not receive service due to loop impairment.

MIPS Consumption. Hardware-based modems consume the least amount of CPU resources during operation. Controller-less modems consume some CPU resources to perform mainly protocol-level processing. Software-based modems consume the maximum amount of CPU resources, since the CPU is being used to perform Digital Signal Processing functions (sometimes referred to as Host Signal Processing or HSP in this context) as well as protocol-level functions. PC manufacturers and system integrators will probably care less about the CPU resource consumption of software-based modems since they will be bundled in new computers equipped with very fast processors. On the other hand, the aftermarket of service providers or retail chains will continue to be concerned with CPU requirements, since these modems will be installed in existing computers possibly equipped with a CPU incapable of supporting the requirements of a soft-ADSL modem.

Combination with V90 modems. Both the ADSL modem and the V.90 dial-up modem access the same telephone line. Combining them on a single board may reduce the overall price of the ADSL + dial-up modem by as much as 20 percent. Since existing computers already have dial-up modems installed, selling ADSL modems to the aftermarket will be difficult since they may not consider this feature an advantage. In fact, adding an additional V.90 dial-up modem may cause more harm than good. On the other hand, the PC manufacturers and system integrators consider this feature an advantage, since it lowers the combination modem purchase price by 20 percent and integrates the installation of two modems into one.

There is no doubt that ADSL is here to stay as the future method for residential broadband access. There is also no doubt that ADSL technology is still in its early adoption phase. PC manufacturers focusing on today's market with its less than perfect DSL provisioning process are being shortsighted. The ADSL modem market will follow the same development path as the dial-up mode, meaning that, as soon as the current problems are resolved, new PCs will ship with bundled ADSL modems. As this market moves into its mainstream phase, it is time for PC manufacturers to realize this.

Youram Solomon is vice president and general manager of the advanced communications business unit at PCTEL, Inc. (Milpitas, CA).

[Graph omitted]

[Graph omitted]