The Wireless Last Mile

Entangled in their copper wires, phone companies may miss out on the broader broadband revolution

9 min read


OWENSBORO, A SMALL KENTUCKY CITY on the Ohio River, is famous for nothing, not even its International Bar-B-Q Festival, now in its 25th year. So it's not exactly the kind of place you'd expect to find technologies and business models that will either make or break some multibillion-dollar enterprises--namely, the U.S. regional Bell operating companies.

But find them you will. In October 2002, after a five-month pilot program, the local electricity and water provider, Owensboro Municipal Utilities (OMU), rolled out a high-speed broadband service to the city's 58 000 residents at US $25 a month, just $2 more than what many were paying for wire01.jpglow-speed dial-up access.

Soma Networks CEO Yatish Pathak [far left] cradles his company's all-in-one antenna and modem terminal. It communicates wirelessly with a faraway base station, connecting devices, such as COO Greg Caltabiano's laptop, to the Internet.

Owensboro isn't the only obscure town conducting an experiment with huge implications for broadband. Two months earlier and 3000 km away, in Klamath Falls, Ore., a small start-up company, Always On Network Inc. (Chiloquin, Ore.), began serving up broadband to 30 test customers, converting them into paying customers a few months later.

What makes these enterprises novel isn't the data rates, which aren't exceptional for broadband, at 250-1000 Kb/s. It's the way that the bits are delivered--wirelessly--at least for the critical last mile to the home. Bypassing the copper wires that connect a phone company's central offices to its customers, these wireless Internet service providers can deliver broadband more cheaply than digital subscriber lines (DSLs) and can reach out to rural homes and others not currently served at all except by dial-up.

In truth, OMU will never go beyond western Kentucky, and Always On, critically underfunded, will be fortunate to become a major regional provider in Oregon. Even so, the two providers are destined to be midwives to the next generation of broadband: wireless metropolitan-area networks (MANs). Propelled in part by a new standard, IEEE 802.16, wireless MANs are expected to do for neighborhoods, villages, and cities what IEEE 802.11, the standard for wireless local-area networks, is doing for homes, coffee shops, airports, and offices.

How quickly wireless MANs reach your own city may well depend on what happens in Owensboro, Klamath Falls, and a few other testbeds. The reason is that OMU and Always On represent opposing approaches to wireless MANs from two of the technology's top system vendors, Alvarion Ltd. (Tel Aviv, Israel, and Carlsbad, Calif.) and Soma Networks Inc. (San Francisco). Alvarion has almost half of what's already a $250 million-a-year market for wireless MANs. Soma, a well-funded start-up, is now in trials with one of the largest phone companies in the world, NTT Communications Corp. (Tokyo). The two companies, and the market itself, are at a crossroads. Alvarion has embraced the 802.16 standard and is a founding member of the WiMax Forum (San Jose, Calif.), an industry consortium created to commercialize it and a corresponding standard from the European Telecommunications Standards Institute known as HIPERMAN. The institute is based in Sophia Antipolis, France.

Alvarion's existing wireless last-mile products, and those of the other WiMax members, are designed for wireless Internet service providers (WISPs), making the connection between homes and the Internet backbone that lets end users bypass their telephone companies. For its part, Soma, reluctant to abandon or change a five-year odyssey of its wireless MAN technology development, and believing it to be superior to anything its competitors have, is ready to stand apart from the standard and go it alone. Its system, while eminently usable by WISPs, is chock full of quality-of-service features that help it transport voice-over-Internet-protocol packets. That means it's especially well positioned to be adopted by the phone companies themselves.

In effect, the two initiatives represent the dilemma the wireless last mile presents to those phone companies: buy into wireless MAN systems or risk being made increasingly irrelevant by them.

The most expensive mile

DSL could use a push. In the United States, it's mainly sold by the regional phone giants--Verizon, SBC Communications, BellSouth, and Qwest. But those companies are in trouble.

First they were damaged by the dot-com collapse and subsequent accounting scandals. Then the most avid Internet users, who had initially put in second or third phone lines for dial-up access, started turning them off after moving on to broadband. Finally, the rise of cellular phones has forever altered the economics of telephony, to the long-term detriment of the regional phone companies.

Andrew B. King of Web Site Optimization LLC (Ann Arbor, Mich.) predicts that 50 percent of all Internet access will be broadband by July 2004, and that this will climb to two-thirds just a year later. But if the future of telecommunications lies in broadband services, that only spells more trouble for phone companies. They've staked their broadband futures on DSL, which provides data rates 10-20 times as fast as dial-up on the same old copper phone lines.

Yet cable has twice as many customers as DSL in the United States [see ”Telcos in Trouble”], in part because it costs so much to provide DSL service to an area. A large part of the problem is that old--in some cases, ancient--copper network.

Switching from copper to fiber-optic cable improves data rates greatly but adds yet more cost. Forget the expensive fiber--anything as labor-intensive as digging up the streets or stringing cables from central offices to homes is phenomenally costly. Cable television companies have been able to justify upgrading their last-mile networks, which consist of coaxial cable, because doing so allows them to deliver more television channels, as well as other services like pay-per-view and video-on-demand.

The expenses, in other words, are spread between two types of services: access to movies and television shows on the one hand and Internet access on the other. Phone company business models have only one of those legs to stand on. As a consequence, widespread fiber to the home is going to take another 10 years--and as much as $40 billion per carrier [see ”U.S. Phone Companies Set Stage for Fiber to the Curb”].

In their competition with the cable guys, phone companies are also hampered by DSL's limited range, generally no more than 5000 meters for business customers and sometimes as low as 3600 meters to homeowners. In sparsely populated places, there may not be a sufficient number of customers near enough to a central office to justify more than a modest investment in DSL. The result: a self-reinforcing spiral of few customers, high per-user back-end costs, high prices, and little revenue with which to expand. It's no surprise, then, that Verizon and some of the other large telephone companies have been testing a wireless last mile for a while now.

In contrast, creating a wireless network is relatively simple. At its heart is a base station, which can be put on top of a building's roof, a cellular tower, or even a water tower. The base station is the bridge between the wired world of the Internet, on one end, and subscribers, with whom it is connected by radio waves, on the other. It largely takes the place of what in the DSL world is the DSL server in a phone company's central office [see ”How a Wireless DSL Network Works”]. With each station generally serving a 10- to 15-km radius, base stations can be put up where--and only where--they're economically justified.

Will o' the WISP

Klamath Falls is typical of the problem with wires. While businesses there are served by several DSL providers, with subscriptions starting at $105 per month, homeowners have only one supplier, Klamath Falls Internet (Medford, Ore.), at $55 per month. That's quite a bit more than the $35-$40 charged by the local cable provider, Charter Communications, for broadband and by the town's several start-up WISPs, of which cash-starved Always On is one. Dan Stanton, Always On's chief operating officer, says he needs $4.2 million to pay for a base station and enough home devices for the 6000 households it plans to sign up within a 500-km2 area centered on Klamath Falls.

Stanton says that the return on that investment would be $2.5 million per year. Even if that figure, and the 6000 subscribers, were to prove unrealistic, it is clear that the costs of serving a rural area are quite a bit lower than they are for wired DSL--lower by at least an entire order of magnitude, according to Greg Caltabiano, Soma's chief operating officer.

”A 200-square-kilometer service area costs a DSL provider over $11 million. The same area can be served wirelessly for about $450 000,” he insists. That covers the costs of a base station, permits, power, and antennas, but not of what the industry calls consumer premises equipment (CPE)--the antenna, modem, and other gear that has to be installed at the subscriber's house. Nor does it include customer support, administrative expenses, advertising, and so on. Soma's CPE is a single box that includes both the antenna and the modem. Most other wireless providers, including Alvarion, separate those two pieces of equipment, requiring a technician to install the antenna outside the house, at additional expense to the subscriber or the service provider.

”CPEs are key,” says Lindsay Schroth, an analyst of broadband access technologies for the Yankee Group (Boston). ”They can cost $400 or more until they are produced in high volumes. That's one of the biggest things holding back the broadband wireless industry--you're trying to compete with DSL and cable, where the CPE is under $50 now.”

Since CPE expenditure increases with the number of users, it can dwarf all the others, coming to be as much as half of all hardware costs. That's been the experience in Owensboro, where Alvarion's CPEs are about $500 each. Roughly, that's half the total system cost per user and leads to a payback period that will be three years for each OMU customer. (Obviously, charging only $25 per month, as OMU does, plays a big role in drawing out the payback as well.)

One thing that keeps expenses low for OMU is the use of unlicensed spectrum. Always On's costs, by comparison, do include licensed radio spectrum but can be modest, except in major cities. Using licensed spectrum makes sense for a Soma-based network, since the company sees the existing incumbent local exchange carriers (ILECs--the BellSouths and NTTs of the world) as key potential customers. In fact, Soma's ongoing trial with NTT, which is Japan's largest telephone company, began in mid-2002. Verizon's test project, which is in Fairfield, Va., also uses licensed spectrum.

That use doesn't surprise Schroth, who told IEEE Spectrum, ”No major ILEC would deploy an unlicensed solution.” She was similarly unsurprised to hear that a municipal utility, such as the one in Owensboro, was using unlicensed spectrum. ”It's going to be common in the MUSH market,” she says, using an industry term that refers to municipalities, universities/utilities, schools, and hospitals. The Alvarion-based network currently uses the same unlicensed frequencies that most Wi-Fi networks do, the 2.4-GHz band.

Soma designed its technology with a nearly blank slate, keeping some critical desiderata in mind, according to Caltabiano. In fact, those ”must items” are almost a wish list for wireless DSL systems. Not every system needs each and every feature, but the more one has, the better off one is.

Besides having a service that is non-line-of-sight and installable by the customers themselves, Soma wanted the largest possible coverage per base station, to reduce the number of stations needed in a service area. With the number of people that can be served in an area depending on the square of the radius, having twice the range means quadrupling the coverage.

Soma's networking software also includes some sophisticated quality-of-service features, such as adding prioritization markers to some data packets that place them ahead of other packets. That helps applications, like voice telephony and streaming music, in which latency--delays between when an information stream is sent and when it's received--is at least as important as overall throughput. Other applications, such as e-mail or ordinary Web browsing, use what are called best-effort methods, because a delay of a few more milliseconds doesn't matter much.

United we stand?

Soma's system uses proprietary devices that don't communicate with those of other vendors or adhere to any particular standard. That could be a problem, though not for the present. (Right now, all wireless last-mile systems have proprietary devices.) It usually takes a variety of vendors to create the critical mass needed to drive down component costs and get enough market share to compete with alternative ways of doing the same thing, in this case, last-mile broadband. Some companies are selling chipsets, others antennas and radios, and still others are putting the parts together in a complete package that a system operator can deploy.

That's what happened with wireless local-area networks (WLANs), and it's going to happen again with wireless metropolitan-area networks, says Jeff Orr of Proxim Inc. (Sunnyvale, Calif.), a company that makes products in both those areas.

Orr notes that wireless LANs didn't take off until the IEEE 802.11 standard was ratified and a nonprofit companion industry consortium was formed, now known as the Wi-Fi Alliance (Mountain View, Calif.). The vendor group is needed to ensure that products operate with one another, something that conformance to the standard doesn't guarantee. The analogous standard for wireless MANs, IEEE 802.16, was ratified last year, and there's a similar consortium, the WiMax Forum. [See ”IEEE 802.11 and 802.16: A Tale of Two Standards" or see a list of WiMax members who sell wireless last-mile systems.]

Several technical differences exist between Soma's system and one based on IEEE 802.16. While both are frequency-based methods, for example, they encode and channelize the spectrum differently: Soma uses code-division multiple access (CDMA) to divide up a given chunk of radio spectrum among simultaneous users, while WiMax uses orthogonal frequency-division multiplexing (OFDM). Other differences, though, may be more important, such as the range of each base station, which determines how many are needed. Soma claims a much greater range than WiMax.

Then there are those darned CPEs. WiMax-compliant units will soon be lower in price than proprietary ones--maybe much lower. Looking at the Wi-Fi experience, Orr says that IEEE 802.11 chipsets are ”$20 in volume today, driving toward the single digits. And there's nothing in 802.16 that wouldn't allow it to follow a similar pattern.” There are already three chipset manufacturers in WiMax: Fujitsu (Tokyo), Wavesat Wireless (St. Laurent, Que., Canada), and Intel (Santa Clara, Calif.), whose representative currently serves as the forum's president.

The proprietary versus standards-based battle won't flare up until at least the second half of 2004, which is the earliest date WiMax-compliant products might show up. A June market analysis by Global Information Inc. (Tokyo and West Hartford, Conn.) predicts that standards-based systems won't predominate until 2005 or 2006.

Meanwhile, carriers needn't wait to take their last miles wireless, and some aren't. Alvarion is a founding member of WiMax, but Carlton O'Neal, the company's vice president of marketing, expects it to sell about $100 million worth of its existing proprietary wireless last-mile products this year. Recently, Alvarion concluded a large agreement for its products with China United Telecommunications Corp. (Beijing), that country's No. 2 mobile carrier.

The wireless last mile doesn't compete with third-generation (3G) telephony, because it isn't mobile; there's no way of handing off a connection from one base station to another, or even to maintain a connection while traveling at high speed. Rather, it allows a cellular carrier like China United to start competing for the broadband portion of consumers' ever-growing telecom budgets.

Someday, we'll all be walking around with 2-Mb/s cellphone wristwatches capable of conducting videoconference calls while downloading e-mail and our favorite music in the background. Until then, broadband and cellular will be two different things. But broadband and wireless need not be.

This article is for IEEE members only. Join IEEE to access our full archive.

Join the world’s largest professional organization devoted to engineering and applied sciences and get access to all of Spectrum’s articles, podcasts, and special reports. Learn more →

If you're already an IEEE member, please sign in to continue reading.

Membership includes:

  • Get unlimited access to IEEE Spectrum content
  • Follow your favorite topics to create a personalized feed of IEEE Spectrum content
  • Save Spectrum articles to read later
  • Network with other technology professionals
  • Establish a professional profile
  • Create a group to share and collaborate on projects
  • Discover IEEE events and activities
  • Join and participate in discussions