> online showcase > articles > The Last Mile Problem

September 1, 2002

The Last Mile Problem: Some Innovative Emerging Solutions


The last mile problem, the task of providing high-bandwidth Internet access to outlying areas, has still not been solved. The majority of potential Internet users are located beyond the reach of the Internetís fast optical-fiber infrastructure, and current links across this gap either have inadequate bandwidth or are limited in other ways. This paper describes four innovative emerging technologies that promise to close the gap with enough bandwidth to support present and future Internet applications: very high rate digital subscriber line (VDSL), non-line-of-sight fixed wireless systems, mesh-connected free-space optical links, and the Teledesic satellite network. While the four technologies have some design elements in common, each technology has unique objectives, advantages, and challenges.

1. Introduction

The last mile problem is the perennial technical, economic, and regulatory challenge of connecting outlying homes and businesses to the Internet with the ever increasing amount of bandwidth that is demanded. The Internetís multibillion dollar optical-fiber backbone certainly provides sufficient bandwidth for all of todayís Internet applications, as well as for the bandwidth-intensive applications that will be developed well into the future. This backbone, however, fails to reach a majority of the potential users of the Internet.[1] Although the length of the gap between an Internet user and the nearest optical-fiber pipeline varies widely, it is often only about a mile—hence the expression last mile problem that is used to describe the task of bridging the gap.

Because some sort of Internet connection is available to most homes and businesses in the industrialized nations, it might seem that the gap has already been closed. The problem, however, resides in the limitations of these current connections.[2] For one user, perhaps an adequate connection is too expensive (a possible problem with current "broadband" solutions).[3] For another user, maybe the only available connection requires dialing each time access is desired and is prone to being dropped (problems inherent in modem connections). For yet another user, perhaps a sufficiently fast connection has a high latency, making video conferencing and other real-time applications difficult (a problem with current satellite links). The most prevalent limitation with current Internet connections, however, is a lack of sufficient bandwidth. A modem connection clearly lacks adequate bandwidth for todayís Internet applications. But even current "broadband"[4] solutions such as ADSL and television cable might well become inadequate for future high-bandwidth applications such as digital video services.

A brute-force solution to last mile problem is to simply remove the last mile—that is, to extend optical fiber all the way to every home and business. This solution, however is obviously too expensive. Acampora (2002) states that optical-fiber extensions to individual users would cost $100,000 to $500,000 per mile, and Peterson and Davie (2000) point out that bringing optical fiber to the 100 million homes in the United States alone would cost $100 billion. Clearly, more innovative, economical, and practical solutions are required. This paper discusses four such solutions that are currently under development: very high rate digital subscriber line (VDSL), non-line-of-sight fixed wireless systems, mesh-connected free-space optical links, and the Teledesic satellite network. These four solutions were selected for coverage in this paper from the set of all solutions currently under development because they are particularly innovative, technically interesting, diverse, and typify the types of approaches that must be taken to solve the last mile problem.

2. Very High Rate Digital Subscriber Line (VDSL)

VDSL, which has been under development since around 1997, promises to bring Internet access to individual homes or businesses with tens of megabytes of bandwidth using the existing system of copper wires.[5] The acronym "VDSL" has various interpretations—for example, "very high rate digital subscriber line," "very-high-speed digital subscriber line," and "very high bit rate digital subscriber line."[6]

The primary downside of VDSL is that the distance over which copper wire can carry such high-bandwidth data is limited. Therefore, telephone companies need to bring Internet access over optical fiber to network units in each neighborhood, which then connect to individual customer sites using VDSL over existing copper wire. However, VDSL may provide a cost effective alternative to bringing optical fiber all the way to each home or business.

Because VDSL bandwidths are so high, it is regarded as a solution not only for providing voice communication and high speed Internet access to homes and businesses, but also for delivering digital video services such as digital television (DTV), high definition television (HDTV), and interactive video-on-demand with multiple simultaneous video streams.

2.1 VDSL Basics

Regarding the promise of VDSL, the first question is likely to be "how is it possible to transmit high-bandwidth data over telephone lines, which had supposedly reached their maximum capacity with 56 Kb/s modems?" The short answer is that the 56 Kb/s limit is not inherent in the copper twisted pair that carries the signal from the customerís site to the telephone companyís central office (CO). Rather, for a signal originating from an ordinary POTS modem, the limit is imposed by the narrow-bandwidth voice switches in the switched voice network over which the signal is sent from the CO. In contrast, for a signal coming from a VDSL system, the CO sends out the signal over a high-bandwidth optical-fiber network directly to the Internet service provider (and from there to the Internet), thereby eliminating the voice switch bottleneck.[7]

Although copper wires are intrinsically capable of transmitting high-bandwidth data, the signal strength does strongly attenuate with distance. To achieve a high bandwidth, a VDSL signal must be sent using a high frequency.[8] Unfortunately, higher frequency signals attenuate more rapidly over distance than lower frequency signals. Therefore, the tradeoff for VDSL is that a high bandwidth can be achieved, but it can be sent only over a relatively short distance. To overcome the distance limitation, VDSL systems use the two-part topology discussed in the next section. VDSL is thus optimized for bandwidth rather than distance. A typical downstream bandwidth achievable by VDSL is 26 Mb/s over a distance of up to 1 km. In contrast, other forms of DSL—such as ADSL—are optimized for distance. They use a lower frequency, lower bandwidth signal that can travel greater distances, normally from the CO all the way to the customerís site. (Typical ADSL downstream rates are 6 Mb/s over about 3.66 km or 1.5 Mb/s over about 5.5 km.)[9]

Another likely question about VDSL is "why wasn't it available previously, since copper wires have been around for a long time?" The reason is that copper wires typically have many impairments in addition to the attenuation factor just discussed.[10] Advances in microelectronics were needed to permit the extensive signal processing that is required to send data reliably in the face of these impairments. Of all forms of DSL, VDSL is the most difficult to implement over impaired lines.[11]

2.2 VDSL Topology

The limited transmission distance of VDSL requires using a two-part delivery system: The first part of the connection consists of optical fiber extending from the telephone companyís central office (CO) to an optical network unit (ONU) in each neighborhood. The ONU is close enough to the customer sites so that VDSL can be used to carry the signals the rest of the way over existing copper wires. An ONU typically supports 15 to 30 subscribers. Some telephone companies are including support for VDSL delivery in the new telephone systems they are installing. Others are planning to retrofit VDSL support to existing telephone systems in areas where they expect a sufficient number of subscribers.[12]

An ONU may be located in a street cabinet that serves a group of surrounding residences, in a local exchange that serves a collection of nearby businesses, or in a cabinet installed in the basement of a large industrial or apartment building.[13]

The VDSL connection between the ONU and the customer site can be either a point-to-point or a point-to-multipoint link. A point-to-point link is a dedicated connection between the ONU and a box at the customer site, which is known as a residential gateway. The residential gateway then distributes the VDSL service to the different Internet devices on the premises, such as personal computers and TVs. A point-to-multipoint link connects the ONU directly to each Internet device at the customer site, with no residential gateway.[14]

2.3 VDSL Bandwidths

VDSL service is offered with either asymmetric or symmetric upstream/downstream bandwidths. Asymmetric mode is designed primarily for residential customers, who need a large downstream bandwidth for receiving digital video services. Typical bandwidths for residential customers located within 1 km of an ONU are 13-26 Mb/s downstream and 2-3 Mb/s upstream, which are sufficient for the delivery of digital TV (DTV), high definition TV (HDTV), and very fast Internet access. Downstream bandwidths for customers within 300 m of an ONU can be as high as 52 Mb/s, which would allow delivery of several simultaneous DTV or HDTV channels.[15]

Symmetric mode is intended chiefly for business customers, who typically use the connection for sharing resources and exchanging data in both directions on the link. A typical symmetric bandwidth for customers located up to 1 km from an ONU is 13 Mb/s, and for customers within 300 m of an ONU, 26 Mb/s.

For businesses and campuses, an ideal use for VDSL is to interconnect LANs located in separate buildings, using single existing phone line connections. (This design is less expensive than the common alternative practice of using multiple 1.5 Mb/s non-VDSL phone lines to connect two LANs.) A single LAN within the complex could then be connected to the Internet via an optical-fiber cable.[16]

2.4 VDSL Transmission Methods

VDSL systems employ two primary strategies for generating reliable high-bandwidth transmissions using the high-frequency VDSL signals. First, to achieve the desired bandwidth, a VDSL system must reliably encode multiple bits within every change of state of the signal. In other words, the bit rate must be made significantly higher than the baud rate, without introducing errors. This is achieved by using sophisticated methods of modulation.[17]

Second, a VDSL system must separate the upstream and downstream signals to avoid crosstalk. Crosstalk is noise that is induced on a phone line by electromagnetic radiation emitted from other phone lines that are in close proximity. If upstream and downstream VDSL transmissions were generated using the same frequency band and at the same time, a type of crosstalk known as near-end crosstalk (NEXT) would occur. Crosstalk is especially severe at the high frequencies used for VDSL transmissions. Therefore, either the upstream and downstream signals must be placed in separate frequency bands (an approach known as frequency division duplexing, or FDD), or the signals must be generated at different times (a method known as time division duplexing, or TDD).[18]

Two distinct VDSL transmission methods have evolved to implement these strategies:

Each transmission method involves a modulation scheme for encoding digital information onto the signal, as well as a duplexing method for separating the upstream and downstream transmissions.

2.4.1 The CAP/QAM Transmission Method

The CAP/QAM transmission method uses either the CAP (carrierless amplitude/phase modulation) or the QAM (quadrature amplitude modulation) technique of modulation. For separating the upstream and downstream transmissions, it uses frequency division duplexing (FDD). The primary advantages of the CAP/QAM method are that it is less complex, less expensive, and uses significantly less power than the DMT method.[19]

The CAP/QAM method is backed by the VDSL Coalition (at http://www.vdslcoalition.net), which includes Infineon Technologies, Lucent Technologies, Broadcom Corporation, and other companies.[20]

The basic CAP/QAM transmission method employs a single carrier signal (hence the alternative term single carrier for this transmission method).[21] This signal is modulated using either CAP or QAM. A dual-mode receiver can detect which of these two modulation schemes is used and then properly process the signal.[22]

QAM is the more common of the two methods. (It is also used by conventional POTS modems and cable modems.)[23]

QAM generates a high-bandwidth signal by encoding more than one bit with every change of state of the signal. A somewhat simplified model that can help in understanding how QAM works is to consider the signal as consisting of one or more sine waves. First, consider a single wave. Within each wave cycle (wavelength), there is a peak at 90 degrees and a trough at 270 degrees. The amplitude of the wave can be modulated at 90 degrees or it can be modulated at 270 degrees (not at both positions). And, the amplitude can be modulated to one of two values (call them 1 and 2). Thus, the amplitude modulation within a wave cycle can occur in one of the following four ways:

The four different types of amplitude change (that is, state change) can be used to encode a 2-bit sequence. Thus, the bit rate is twice the baud rate.[25]

To increase the bit rate even more, what QAM does is to include an additional wave in the signal. The additional wave is phase shifted 90 degrees from the first. (If the first wave is considered a sine wave, the second would be a cosine wave.) The term quadrature derives from the fact that the signal contains a second wave that is phase shifted by 90 degrees, that is, by one quadrant. The second wave provides 2 more possible positions where one of two amplitude changes can occur, thus allowing one of 8 different types of amplitude change to occur in each cycle. This scheme permits a 3-bit sequence to be encoded for every state change.[26]

QAM allows additional phase shifted waves to be added to the signal to generate even more possible state changes per cycle. The collection of possible state changes for a particular form of QAM is known as a constellation. (When the state changes are plotted using polar coordinates, they form a symmetric constellation of points around the origin.)[27] According to Oksman and Werner (2000) the maximum QAM constellation size possible with VDSL is 256, which would encode 8 bits per state change.

The CAP method of modulation is derived from QAM and is almost identical to it.[28]

To implement frequency division duplexing (FDD), the CAP/QAM method uses distinct frequency bands for upstream and downstream signals. To support high performance for both symmetric and asymmetric services in same cable bundle, it is necessary to have more than one frequency band in each direction. As of the year 2000, the standard design was to have two frequency bands for upstream transmissions and two frequency bands for downstream transmissions, with unused guard bands safely separating the upstream and downstream bands:[29]

Last Mile Problem 1

2.4.2 The DMT Transmission Method

The DMT transmission method uses DMT (discrete multitone) modulation, and it separates the upstream and downstream transmissions using time-division duplexing (TDD). The primary advantage of DMT is that it can dynamically adapt to changing line conditions to provide a highly reliable high-bandwidth connection. Also, it can coexist with ADSL on the same wire.[30]

The DMT method is backed by the VDSL Alliance (at http://www.vdslalliance.com), which includes Alcatel, Texas Instruments, STMicroelectronics, Stanford University, and other organizations.[31]

The DMT modulation technique is an enhanced form of QAM. DMT uses the QAM technique to modulate its signals. However, rather than sending all of the data over just a few signals (as does the standard QAM technique employed by the CAP/QAM method), it divides the data among multiple—usually 256—separate QAM signals (hence the alternative term multiple carrier for this transmission method). Each signal—also called a tone—is sent over a separate frequency band, known as a subchannel. The subchannels are adjacent and have equal bandwidths. DMT thus evenly divides its total available frequency band into a large number of independent subchannels, each one modulating a separate signal that carries a portion of the data:[32]

Last Mile Problem 2

Each subchannel carries the maximum number of bits (per second) that is permitted by its current signal-to-noise ratio. Because signal attenuation is greater at higher frequencies, the higher frequency subchannels carry a steadily diminishing number of bits. Also—and here is the power of DMT—if noise occurs on a particular subchannel, the system responds by dynamically decreasing the number of bits sent over that channel. DMT thus always sends the maximum number of bits that can be transmitted without incurring an unacceptable probability of errors. (The dynamic adjustment is achieved as follows: The receiver measures the quality of each tone, and reports any required change in bit distribution to the sender, using a reliable low-speed channel. The sender then modifies the bit distribution accordingly.)[33]

To implement time-division duplexing (TDD), also known as ping-pong transmission, all adjacent lines are synchronized using the same network clock so that all subchannels in all lines "ping" (travel one way) at the same time and "pong" (travel the other way) at the same time. Thus, in contrast to the CAP/QAM method, each frequency band supports both upstream and downstream transmissions, although the transmissions occur at different times. To deliver symmetric upstream and downstream bandwidths, the DMT method makes the "ping" and "pong" time periods equal. The DMT method can create an asynchronous mode with any upstream/downstream bandwidth ratio by simply varying the ratio of upstream/downstream transmission times (these times are software programmable).[34]

The overall modulating and duplexing scheme used by the DMT method is known as synchronized discrete multitone (SDMT).[35]

2.5 Current Challenges to VDSL

According to Manners (2002)[36] VDLS is currently facing a combination of legal, standardization, and economic challenges:

Secker (2002) quotes several sources who believe that VDSL is not yet a proven technology and that it might take several years before it will be adopted on a broad commercial basis. However, Secker indicates that a number of providers are undertaking extensive VDSL trials, and that QWest has already fully implemented a 70,000 line VDSL service in Phoenix.

3. Non-Line-of-Sight Fixed Wireless Systems

With the fixed wireless system of Internet access, each customer communicates directly with the provider via a microwave signal transmitted between the customerís antenna and the providerís antenna. In outlying areas where neither DSL nor cable TV Internet access is available, fixed wireless would seem to be an ideal solution to the last mile problem. However, most current fixed wireless access systems require that the customerís antenna be within direct line-of-sight of the providerís antenna. These are known as line-of-sight (LOS) systems, the most common of which is MMDS (multichannel multipoint distribution service, operating in the 2.5 GHz band).[37]

Although capable of transmitting high-bandwidth data over relatively long distances (tens of kilometers), LOS systems have problems. In addition to the obvious problem that many potential customers in the service area might not be within direct line-of-sight of the providerís antenna, LOS systems have other limitations:[38]

As a result of these problems, some providers have either limited development of new LOS systems or have discontinued existing LOS service—especially service to residential customers, who tend to be price-sensitive.[44]

Fixed wireless systems that do not require direct line-of-sight between the customer and the provider are currently being developed and constitute an important emerging solution to the last mile problem. To overcome the problems inherent in LOS systems, a non-line-of-sight (NLOS) system must be able to exchange signals with a customer site that is hidden from direct view of the providerís antenna and is a reasonable distance away from the base station, and yet provide a signal-to-noise ratio that is high enough to carry high-bandwidth Internet access. The customerís antenna must not be highly directional and installation of the antenna must be easily accomplished by the customer, preferably indoors. And, the entire system must scale well as the number of customers increases.[45]

Two basic types of NLOS systems are being developed to meet these challenging requirements. One type of system, called a point-to-multipoint (or non-mesh) system, uses a smart antenna at the base station to achieve non-line-of-sight communication with customers in the service area with adequate signal strength. The other type of system, called a mesh system, uses a network of separate intercommunicating nodes to get around line-of-sight obstacles and to maintain adequate signal strength.[46]

Before exploring the details of these two types of systems, it is important to realize that a NLOS system typically cannot reach 100% of the customers who are out of direct line-of-sight. Kintzel (2001) quotes a representative from Iospan Wireless (their NLOS system is described later in this paper) as defining NLOS as the "ability to reach 90% to 95% of users within cell range, with transmission rates of 2 Mb/s and above." Kintzel also quotes a spokesperson from Cisco Systems who states that in order for a NLOS system to be viable, a provider must be able to reach 60% of the potential users in the area served.[47]

3.1 Point-to-multipoint NLOS Systems

The essential feature of a point-to-multipoint NLOS system is a smart antenna at the providerís base station, which communicates directly with the customers in the service area. There are no intermediate nodes between the smart antenna and a given customer for redirecting or boosting the signal.[48]

When the smart antenna transmits data to a particular customer, it directs the signal precisely to the customerís site, even if that customer is not within direct line-of-sight of the antenna, and it makes the received signal strong enough so that the customer can receive the signal using a simple indoor antenna.[49] According to Kintzel (2001), some of the signal comes through walls, but most enters through windows.

The basic strategy that the smart antenna uses to accomplish this feat is to analyze the signal received from the customer during a handshaking process, and then customize the signal it sends to the customer according to the analysis. This process is based on the two types of antennas that are used—the customer has a simple omnidirectional antenna (usually a small whip antenna), while the smart antenna at the base station consists of an array of separate radiating elements, each of which can transmit a signal in a separate direction and with a specific amplitude and phase shift (time delay).[50]

Because the customerís antenna is omnidirectional, the signal that the customer sends out radiates in all directions and reflects off various objects (such as mountains or buildings). The signal thus takes multiple paths to the base station, and when it reaches the base station it has broken apart into separate signals that vary in the direction from which they come, in phase, and in attenuation (the signals that take the longer paths will tend to be delayed and attenuated the most).[51] When the base-station subsequently transmits to the customer, it uses a set of complementary signals. As a simple example, assume that the base station receives one signal from the east, and 3 microseconds later it receives another signal from the south that is 5 dB weaker, as shown here:[52]

Last Mile Problem 3

In this case, to create the complementary signals needed to transmit to the customer, one of the array elements on the base-station antenna would send a strong beam to the south, and 3 microseconds later another array element would send a beam to the east that is 5 dB weaker.[53]

The complementary signals come together precisely at the customerís antenna, in phase, and with equal signal strengths. They therefore combine constructively and establish a single strong signal. Because the environment can change rapidly (a reflected signal might have been off a passing airplane), the smart antenna must constantly monitor the customerís transmissions and dynamically adjust the outgoing signal.[54]

3.1.1 Implementations of Point-to-multipoint NLOS Systems

A number of companies, most of them relatively new, have started testing or implementing point-to-multipoint NLOS systems. This section presents a few examples of current activity in this area.[55]

Navini Networks Inc. has incorporated point-to-multipoint NLOS technology in its Ripwave line of products. These products operate in both a licensed and an unlicensed band around 2.5 GHz, and the smart antenna at the base-station has 8 radiating elements in the array.[56]

Sprint PCS has recently been conducting tests in Huston and Montreal using equipment from Navini (as well as from IPWireless). Sprint claims that these are the largest NLOS tests performed so far by any United States Web service provider. The tests were scheduled to be completed in July, 2002.[57]

Iospan Wireless has enhanced the basic point-to-multipoint NLOS system described previously in this paper by using multiple transmitting and receiving antennas at each end of the link to multiply the overall bandwidth. This technology is known as MIMO (multi-input multi-output). With a customer 1.5 km away, Iospanís system has achieved a downstream bandwidth of over 13 Mb/s and an upstream bandwidth of 4 Mb/s.[58]

Iospanís antenna can be self-installed inside the customerís home (technicians are not needed to angle the antenna correctly). Iospan demonstrated the strong NLOS characteristics of their customer antenna by showing that a cookie sheet can be placed in front of the antenna without blocking the signal.[59]

On February 11, 2002 Iospan announced its first customer, ARM International, which is an Indian telecommunications equipment manufacturer whose customers include several major wireless carriers in India.[60]

3.1.2 Microcells

To reduce initial financial outlays, some companies are taking a scalable approach to implementing point-to-multipoint NLOS systems. That is, they are planning to start with small coverage areas, each around 4-12 miles in radius, which are known as microcells. They then plan to add more microcells as their customer bases grow. Because NLOS technologies use tightly targeted transmissions, they work well with this type of scalable approach.[61]

3.1.3 Spectrum Reuse

A provider of point-to-multipoint NLOS systems can send several simultaneous data streams, each using a separate frequency channel. However, because the provider has available only a narrow band of the microwave spectrum, the number of possible channels is strictly limited (4 to 6, for example).[62]

Fortunately, the provider can divide the service area into separate sectors (as many as 24 to a microcell) and use the full set of channels concurrently in each sector. That is, the company can use all of the channels in all of the sectors at the same time. Because the NLOS technologies create tightly targeted signals, a channel sent to one sector is unlikely to interfere with a simultaneous, same-frequency channel sent to a different sector. This approach maximizes the capacity of the limited available spectrum, and greatly increases the scalability of coverage.[63]

3.2 Mesh NLOS Systems

A second major approach to achieving NLOS delivery of high-bandwidth Internet access is to use a mesh topology consisting of a network of discrete intercommunicating nodes. The nodes are located at a few central locations and at every customerís site. Each node contains a wireless router that boosts the signal and (except for an end node) rebroadcasts the signal to one or more additional nodes. A node at a customerís site also connects the customer to the network. (See the figure in the next section, which illustrates the specific topology used by one company.)[64]

Because the signal is boosted each time it is relayed from one node to another, a mesh NLOS system solves the problem of signal attenuation that would tend to occur with a single long link. (A LOS system solves this problem by using a strong signal and requiring direct line-of-sight. A point-to-multipoint NLOS system solves the problem by sending out multiple signals that combine constructively at the customer site.)[65]

Although the individual links in a mesh network do require direct line-of-sight, the mesh topology serves to move the signal around any line-of-sight obstacles, thus solving the line-of-sight problem. And because each customer generally has several links to other nodes on the network, if a particular link gets blocked (for example, a neighbor adds a second story) the connection to the network will not be lost. (A point-to-multipoint NLOS system solves the line-of-sight problem by essentially using signal-reflecting environmental objects as nodes in a "mesh.")[66]

3.2.1 An Implementation of a Mesh NLOS System

Nokiaís Wireless Routing Group is one of the companies offering a wireless router that can be used for building a mesh NLOS system. (It is known as the R240 Wireless Router.) This router has a simple, multidirectional rod antenna that must be mounted outside (typically on a roof), but does not need to be aimed. Among the wireless Internet service providers that use this router is Vista Broadband Networks, which allows its customers to select one of two symmetric access bandwidths: 384 Kb/s or 1-Mb/s.

With this type of wireless router, a high router density is required to maintain an adequate signal-to-noise ratio for high-bandwidth communications. Specifically, a router needs to be mounted approximately every 50 m. If a provider needed to rent space for a base station for each router, the cost of implementing a mesh NLOS system would be prohibitively expensive. An important benefit of Nokiaís system is that the routers are included in the customersí equipment (which costs only about $800 per customer), eliminating the need to rent space for each node.

The following figure shows the topology of Nokiaís implementation of the mesh NLOS system. The mesh is organized as a hierarchy, with one central broadcasting aggregation point, a broadcasting AirHead in each neighborhood (a served neighborhood is known as an AirHood), and a wireless router at each customerís site. Like a customerís wireless router, an AirHead is simply roof mounted and does not need to be aimed or carefully sited. The places in this figure where no link is drawn between two adjoining routers represent situations where direct line-of-sight is blocked (for example, a tree stands between the two routers).[67]

Last Mile Problem 4

3.2.2 The Latency Problem

Having a series of signal-boosting routers in each data path maintains sufficient signal strength to permit a high-bandwidth data stream. However, these routers add latency to the Internet connection. Latency can be a problem for real-time interactive communications, such as voice or video conferencing. Nokia is working on an enhancement to their system that will allow them to offer a quality of service guarantee stating that if a customer is within 3 router hops of an AirHead, latency will be low enough to permit high quality interactive voice communication.[68]

3.3 Current Challenges to NLOS Systems

The following are among the challenges to implementing NLOS systems:

Despite these challenges, some NLOS equipment makers—such as Navini Networks and Iospan Wireless—are already beginning to find customers for their technology.[72]

4. Mesh-Connected Free-Space Optical Links

A network of free-space optical (FSO) links is capable of bringing Internet access to individual homes or businesses at bandwidths exceeding 1 Gb/s and at a cost of about one-tenth to one-third the cost of installing underground cable.[73] At each end of an individual link is both a transmitter and a receiver, allowing data to be sent in both directions. The transmitter sends a sequence of low-power infrared light pulses that encode the data using simple on/off modulation, and the receiver contains a sensitive photodetector that receives the signal. Acampora (2002) describes an FSO link as "fiber-optic communications without the fiber."[74]

The optical power of the pulses is relatively low (one reason for this is that the beams cut through inhabited neighborhoods and must therefore be eye-safe). Accordingly, the range of a single link is limited, with a maximum length of approximately 1 km—or significantly less in the presence of fog or another obstructing weather condition. Therefore, multiple short links are arranged in a mesh topology, where each node contains the transmitters and receivers needed to communicate with several nearby nodes. The mesh can reliably deliver data from a central dissemination center to an entire town, city, or region.[75]

Although the FSO system was originally invented in the 1970s, the growing demand for high-bandwidth Internet connectivity has resulted in a revival of interest in this technology in the last few years, and several small-scale FSO systems have recently been implemented.[76]

4.1 The Potential Bandwidth of an FSO System

Of the four Internet connectivity solutions described in this paper, the FSO system has the greatest potential bandwidth. According to Acampora (2002), with currently available FSO equipment the system can achieve bandwidths ranging from 10 Mb/s to 1.25 Gb/s. Furthermore, using current state-of-the art laser diodes (which have not yet been adapted for FSO applications) the FSO system could potentially achieve a bandwidth of 9.6 Gb/s. (With these diodes, each optical pulse lasts for only 100 picoseconds.) Acampora and Krishnamurthy (1999) make the more modest claim that an individual FSO link can easily operate at 622 Mb/s (matching the rate of an OC-12 line) and that in principle an FSO network can deliver this rate to "some number of clients."[77]

4.2 The Operation of an Individual FSO Link

An individual FSO link uses equipment and techniques that were originally developed for optical-fiber systems. At each end of the link is both a transmitter and a receiver (collectively known as a transceiver), which are mounted together on a window or roof and allow data to be transmitted in both directions across the link. The transmitter uses an infrared laser diode that emits a sequence of infrared optical pulses. These pulses are focused by a lens and are sent out as a collimated beam of light. A portion of the light[78] strikes the aperture lens of the receiver on the FSO node at the other end of the link. The receiver focuses the collected optical power onto a photodetector, which converts each pulse of light into a weak electrical signal. A sensitive electronic receiver (a signal regenerator) then amplifies and cleans up the signal. The data can then be sent either to the attached client or—as explained in the next section—on to another link (via another FSO transmitter at the node).[79]

The optical signal is modulated using simple on/off signaling, also known as keying. With this method, a light pulse represents a binary 1 and the absence of a light pulse represents a binary 0. To enhance transmission efficiency, the data is divided into separate packets, all of which have the same format and can be individually addressed and sent. To increase the number of signals that can be sent on a single optical path, an FSO link can support wavelength division multiplexing (WDM), which allows tens of separate signal channels to be sent on an optical path by using a slightly different wavelength for each one.[80]

4.2.1 Beam Divergence

The maximum length of a link depends upon the data rate, the optical power at which the beam is transmitted, the size of the receiving lens, the sensitivity of the optical receiver, and the divergence (that is, spread) of the beam over its path. Because of beam divergence, the beam forms a fairly large cone by the time it hits the receiver. Divergence causes the optical power striking the receiver's collecting lens to fall off rapidly with distance (the received optical power varies inversely with the square of the distance). Divergence can be minimized by increasing the diameter of the transmitting lens (divergence varies inversely with the transmitting lens diameter).[81]

Minimizing divergence, however, creates another problem: The resulting thin, pencil-like beam is difficult to aim, especially given the presence of normal building sway and thermal expansion and contraction of building materials. To successfully keep the beam properly aimed, an FSO system uses auto-tracking mechanisms at both ends of the link. These mechanisms keep the beam that is emerging from the transmitter pointed directly at the receiver and keep the receiver's aperture pointed directly at the transmitter. Feedback controls provide ongoing adjustments to maintain the beam aim.[82]

4.3 The Fog Problem

Fog and—to a much lesser extent—rain or snow seriously reduce the received optical power of a single FSO link, and the resulting attenuation factor increases exponentially with distance. Fog can thus limit the maximum range of an FSO link to well below the 1 km distance that is possible under ideal weather conditions. For example, in moderately dense fog, an optical signal might lose 90% of its strength every 50 m. This would mean that over a distance of 100 m 99% of the optical power would be lost, and over 150 m 99.9% would be lost.

An FSO link must therefore include a link margin—extra optical power that can be deployed in foggy conditions. One way to quantify the effect of the link margin is to use a metric known as link availability or uptime, which is the percent of total operating time that the link is up and functioning. For private enterprise networking, a link availability of 99.9% is considered acceptable. With this level of availability the link would  be down about 9 hours per year. For public carrier-class service (which a carrier provides to its prime business customers), link availability should be 99.999%. This level of link availability is known as the five nines benchmark and is the level provided by optical fiber. A link that provides this level of availability would be down only 5 minutes per year.

According to Acampora and Krishnamurthy (1999), data from the national climatic data center indicates that in many cities an FSO link could provide a 99.999% level of availability with a bit error rate of 10-12 and a bandwidth of 622 Mb/s (OC-12), provided that the link length is less than 200 m (under favorable climatic conditions the length can be greater). An FSO link can provide 99.99% availability in many more cities with a link length that is greater than or equal to 200 m.[83]

4.4 The FSO Mesh Topology

To overcome the distance limitations of an individual FSO link, multiple short two-way links can be arranged in a network with a mesh topology. The network can extend the Internet connection to residential or business customer sites that are too far from the optical-fiber backbone to be reached with a single link, providing service for a neighborhood, town, city, or an entire region.[84]

Each node in such a network consists of a multitransceiver optical node, which is a single roof or window mounted unit that has several pairs of transmitters and receivers. A root node is one that connects directly to the optical-fiber pipeline. A single network can have multiple root nodes in order to provide greater capacity to the network subscribers.[85]

Last Mile Problem 5

The intermediate (non-root) nodes serve as regenerative repeaters to maintain the signal strength. Each node also contains a switch that routes and multiplexes the packets of data. The switching software can sense a node failure and route data around it (the network designer must therefore provide some unallocated capacity on each link).[86] According to Acampora and Krishnamurthy (1999) each switch has maximum dimensionality of 5 x 5, which allows a maximum of four two-way links to other nodes (as shown in the figure above), plus one two-way "drop" to the equipment located in the building (not shown in the figure).

Thus, the mesh topology not only serves to strengthen signals by repeatedly regenerating them, but also it provides redundant routes from any given building back to the root node so that recovery from failure of an individual link can be rapid.[87]

4.5 Current Challenges FSO Systems

A significant obstacle to the commercial adoption of FSO systems is the fog problem described previously. Even if multiple links are arranged in a mesh-topology network, providing the 99.999% uptime demanded of public carrier-class service can be challenging. Ironically, one solution might come through a competing access technology that was described in section 3 of this paper: fixed wireless systems. According to Acampora (2002), the solution does not lie in simply using a fixed wireless system rather than an FSO system, because cost-effective, high-bandwidth fixed wireless systems are subject to severe signal attenuation in heavy rain. However, because heavy rain (which blocks an economical fixed wireless link) does not occur simultaneously with heavy fog (which blocks an FSO link), an economical fixed wireless system and an FSO system could be used as complementary technologies within the same area for providing highly reliable service.[88]

According to Acampora (2002), many experts think FSO technology—despite the technical and economic challenges—has the best overall chance of succeeding when compared to DSL, fixed wireless systems, and TV cable Internet access. The article lists ten companies that are already developing FSO technologies.

5. The Teledesic Satellite Network

Teledesic LLC is a privately held company with headquarters in Bellevue, Washington.[89] The primary mission of the company is to create an advanced two-way satellite communications network that will bring global, cost-effective, "fiber-like" telecommunications services to government, business, non-profit organizations, and individuals. The proposed telecommunications services will include:

Teledesic describes their proposed satellite network as the Internet-in-the-SkyTM, indicating not only that the network will extend Internet access globally but also that the topology and protocols of the network itself will be modeled on those of the Internet.[91]

5.1 The Design of the Teledesic System

The design of the Teledesic system has changed radically since the company was founded in 1990. The general trend has been toward fewer satellites, placed at a higher orbits. The original design specified 840 satellites in low-Earth orbit. The second design, adopted at the time of a major Boeing investment in 1997, called for 288 satellites, also in low-Earth orbit.[92]

The most recent design, announced in February of 2002 when Teledesic signed a contract with Alenia Spazio for building the first two satellites, features a total of 30 satellites in medium-Earth orbit. The following sections focus on the most recent design, except where otherwise noted.[93]

The goal of all Teledesicís designs has been to ensure seamless compatibility with optical-fiber networks. The satellite network has therefore been designed to have the same basic characteristics as fiber: high bandwidth, low latency, and low error rates.[94] The latest (30 satellite) plan, which Teledesic calls their "improved satellite network design," was conceived to achieve all of these goals while greatly reducing the overall cost of the network. This design promises to achieve upload bandwidths of 128 Kb/s to 100 Mb/s and download bandwidths up to 720 Mb/s.[95]

5.1.1 The Satellites

According to the current design, the Teledesic system will consist of a constellation of 30 interconnected satellites. Initially, Teledesic plans to deploy just twelve of these satellites, which will provide continuous coverage of selected areas of the earth. At a later date, they plan to deploy the remaining 18 satellites, which will permit global coverage.[96]

Each satellite will provide service to a relatively small portion of the earth. (The served area is known as the satelliteís footprint.)[97] According to Teledesic (n.d.) in describing the original 840 satellite design, the potential coverage areas of the satellites will overlap. This coverage redundancy, plus the use of in-orbit spare satellites, will provide high network reliability. This same article also states that the 840 satellite system would cover 95% of the earthís landmass and almost 100% of its population. Kohn (1996) points out that the relatively small footprint of each satellite (compared to traditional satellites, which are discussed in the next section) will allow a high level of frequency reuse and therefore an efficient use of spectrum resources (very much like the frequency reuse in microcells in point-to-multipoint non-line-of-sight systems, as explained in section 3.1.3).

In 1999 Teledesic signed a launch contract with Lockheed Martin, a major aerospace company and launch provider.[98] The company plans to use a mix of different launch systems to establish and to maintain the satellites in orbit.[99]

5.1.2 The Orbits

Each of the Teledesic network satellites will occupy a non-geostationary (NGSO), medium-Earth orbit. This means that the satellites will move with respect to the earth, and that they will occupy a lower orbit than traditional geostationary (GSO) satellites, which must be in a high-Earth orbit.[100] The relatively low orbits of the Teledesic satellites will eliminate the long latencies and high power requirements of the satellite-to-earth transmissions of traditional GSO high-Earth orbit satellites.[101]

The long latencies of traditional satellite systems make them unsuitable for latency-sensitive applications, such as real-time video conferencing and online games. Also, as explained in Kohn (1996), high latencies can restrict data throughput under the reliable protocols, such as TCP, that will be used across the network.[102] Furthermore, as Kohn goes on to point out, applications that do not require a reliable protocol, such as video conferencing, tend to be ones that are inherently latency sensitive.[103]

In contrast, the relatively low orbits of the Teledesic satellites will allow latencies that approach those of optical fiber (at least in the earlier Teledesic designs), minimizing latency problems. The need for "fiber-like" latencies exemplifies the importance of Teledesicís general goal of creating seamless compatibility with fiber.[104]

One of the implications of the high power requirements of traditional satellite systems is that upload bandwidths can be severely limited, because high data rates would demand dangerously high transmitting power from usersí antennas.[105] The relatively low orbit of Teledesic satellites will allow the system to send and receive high-bandwidth data using small, low-power user equipment. This equipment will be provided by various manufacturers, and will mount on a rooftop and connect to a computer or network inside the building.[106]

5.1.3 The Frequency Band

The Teledesic system will use the high-frequency Ka-band of the radio spectrum. Specifically, it will use the 28.6-29.1 GHz band for uploading data and the 18.8-19.3 GHz band for downloading data. These are the two 500 MHz bands—one for upload and one for download—that were designated internationally by the World Radiocommunication Conference and domestically by the FCC for use by NGSO fixed satellite systems such as the one proposed by Teledesic.[107]

The Teledesic system will need to use high frequencies, such as those in the Ka-band, in order to generate high-bandwidth data transmissions. A problem with these high frequency bands, however, is that a signal traveling at a low angle with respect to the horizon (the elevation angle) tends to be blocked by line-of-sight obstacles and to be attenuated by rain. Therefore, the Teledesic satellite that services a particular user terminal will need to be positioned at a relatively high elevation angle from the userís site. Fortunately, unlike a traditional GSO satellite, a NGSO Teledesic satelliteís orbit will not have to be above the equator—and therefore at a low elevation angle for users located well to the north or south of the equator.[108]

The direct line-of-sight requirement of the high-frequency Teledesic system will make it suitable only for fixed applications (where each user has a stationary, roof-mounted antenna), or for maritime or aviation applications where line-of-sight is not a problem (airplanes and ships are unlikely to move behind line-of-sight obstacles). The Teledesic system will not support mobile applications, such as mobile voice and paging services.[109]

Taken together, the medium-Earth orbit (necessary for low latency and power, as discussed in the previous section) and the high elevation angle (necessary to avoid signal blocking and attenuation) will dictate the required number and placement of the satellites needed to support global coverage, according to simple geometry. (The reduction in the number of proposed satellites from 288 to 30 necessitated raising the satellites from low-Earth to medium-Earth orbits.)[110]

5.1.4 The Network

In addition to the satellite-to-earth links that allow users to access the Teledesic network (discussed previously), each satellite will have a link to each of its neighboring satellites. The constellation of satellites will thus form a mesh-topology, non-hierarchical network similar to the Internet. As in the Internet, the mesh topology will provide redundant pathways that will render the network tolerant to faults and congestion in individual switches (described next).[111]

The description of the 840-satellite design given in Teledesic (n.d.) indicates that the Teledesic satellite network will use a form of fast packet switching that combines features of the ATM and IP protocols, where each satellite will serve as a node in the network and contain a switch. Like ATM, the packets will be short and have a fixed length.[112] The conversion to and from the packet format will take place at each user terminal, which will also encrypt and decrypt the data to provide security. Like IP, the network will use a connectionless protocol and a continuously active distributed adaptive routing algorithm.[113]

Each user terminal will connect directly with the Teledesic network via a gateway switch and communicate through the network to a destination user terminal. The destination terminal can be part of a private network, or—to connect to the Internet—it can belong to a local Internet service provider (ISP). A user will typically connect to an ISP located within the same country.[114]

5.2 The Teledesic Saga

This section provides a brief timeline of the major events in the Teledesic saga.

1990 Craig McCaw and Ed Tuck founded Teledesic, then named Calling Communications, and put together a team of engineers to begin initial research and development. (McCaw is currently Chairman and Co-Chief Executive Officer of the company.)

1994 The company name was changed to Teledesic. Teledesic completed the initial system design and filed an application with the U.S. Federal Communications Commission (FCC) for a license to build, launch, and operate a global high-bandwidth satellite communications network.

1997 Teledesic cleared its two major regulatory hurdles:

1998 Prince Alwaleed Bin Talal of Saudi Arabia invested $200 million in Teledesic.

1999 Teledesic signed a major launch contract with Lockheed Martin, a large aerospace company and launch provider. This same year Teledesic received a private equity investment of $121 million raised by the Abu Dhabi Investment Company (ADIC).

2002 Teledesic signed a contract with the Italian satellite manufacturer Alenia Spazio SpA to build the first two satellites for the Teledesic system.[117] Teledesic is currently negotiating with Alenia Spazio and other satellite manufacturers for the remaining satellites.

2005 This is Teledesicís current target year for beginning service.[118]

5.3 Current Challenges to the Teledesic System

Even though Teledesic passed its major regulatory hurdles with the FCC and the World Radiocommunication Conference, they still have to deal with literally every country in the world to secure "landing rights."[119]

As with the other systems discussed in this paper, economic issues present a significant challenge to the Teledesic system, especially considering the huge scale of the project and the fact that unlike the other systems it would be difficult to start small and expand gradually. Although Teledesicís current design will cost significantly less than their original plan, the construction phase alone for the 12 satellites needed in the first deployment phase will cost nearly $1 billion. According to Richards (1998), critics estimate a much higher cost.[120] Furthermore, the company anticipates that high bandwidth rates will drop dramatically between now and when the system is first deployed.[121] The Teledesic plan is certainly visionary, but according to Richards (1998) some financial, satellite, and telecommunications experts have felt it to be unrealistic.

Teledesic, however, remains optimistic and plans to be both competitive and profitable at the time of deployment. They recognize that optical fiber will continue to be the most cost effective high-bandwidth channel for the Internet backbones, where the high cost of fiber can essentially be amortized over a continuous high volume of data. But they feel that the Teledesic system, like the other systems discussed in this paper, will be much more economical than optical fiber for providing "last mile" connections—that is, for extending access to areas of low to medium user density or to low-bandwidth users (in other words, to most users).[122]

The Teledesic system has a unique economic advantage over the other systems. As stated in Teledesic (2002) it has the "ability to aggregate diffuse demand wherever in the world it exists at a cost independent of user density and independent of location." In other words, while each of the other systems discussed in this paper will need to wait for the demand to grow to a "critical mass" in a particular area before the system can be economically deployed in that area, the Teledesic system, by essentially tapping demand from the entire globe, might be able to become economically viable sooner. Teledesic feels that their potential market is so large that, in their own words, "Teledesic will succeed even if it serves only a small part of the overall broadband market." Of course many failed Internet start ups were fueled by this same logic, showing that even if demand for a product is drawn from a global market, it might not be sufficient. At the least, however, Teledesic is clearly the most viable system of those discussed in this paper for creating truly global Internet access.[123]

As pointed out by Kohn (1996), a uniquely appealing aspect of the Teledesic system (at least to egalitarians) is that the technology is inherently egalitarian. Because the satellites occupy a NGSO and therefore keep moving with respect to the earth, it is impossible to restrict coverage to selected wealthy areas. In Kohnís words, "It is a form of cross-subsidy from the advanced markets to the developing world, but one that does not have to be enforced by regulation but rather is inherent in the technology."

6. Conclusion

Although the solutions described in this paper are quite diverse, they exhibit a number of common design patterns. For example, three of the solutions—VDSL, mesh NLOS, and mesh-connected FSO links—have a two-part design in which optical fiber brings Internet access to a central location in each neighborhood and then a less expensive technology extends Internet access to each customer within that neighborhood. As another example, three of the solutions—mesh NLOS, mesh-connected FSO links, and the Teledesic satellite network—employ a mesh topology similar to that of the Internet itself.

This paper has demonstrated that literally all of the technology required to bring high-bandwidth Internet access to outlying areas has already been invented. Why then doesnít everyone have this access today? Although the regulatory hurdles are often nontrivial, the main general constraint seems to be economic. Developing these solutions is quite expensive.[124] Providers tend to be uncertain whether the current demand for high-bandwidth Internet access justifies the required investments,[125] especially in the face of competitive systems, the present economic downturn, and the lack of well-established companies involved in making the equipment (most suppliers of the newer systems are startups, still relying on funding). Perhaps compelling new high-bandwidth Internet services and applications will eventually stimulate demand enough to justify the investments.

Which of the solutions discussed in this paper is most likely to succeed? This is a difficult question because of the complex and rapidly changing economic and regulatory forces that affect these technologies. It is also difficult to compare the solutions because of the dissimilarities in the technologies and in the problems they are designed to solve. Each solution has unique compelling features. For example, a VDSL system can be implemented on existing copper wires and thus requires relatively little new equipment. Non-line-of-sight fixed wireless systems scale well and thus can easily be implemented in stages. Mesh-connected free-space optical links are capable of providing massive bandwidth. And the Teledesic satellite system promises to create truly global Internet access.

7. References

Acampora, A. (2002, July). Last mile by laser. Scientific American, 48-53.

Acampora, A. & Krishnamurthy, S. (1999, October). A broadband wireless access network based on mesh-connected free-space optical links. IEEE Personal Communications, 6(5), 62-65.

Charny, B. (2002, January 16). Seeing isn't believing for fixed wireless. CNET News.com. Retrieved July 22, 2002 from http://news.com.com/2100-1033-816598.html

Charny, B. (2002, February 11). Newfangled-wireless firm wins a client. CNET News.com. Retrieved July 22, 2002 from http://news.com.com/2100-1033-834669.html

Charny, B. (2002, May 7). Can your Net access travel through walls? CNET News.com. Retrieved July 22, 2002 from http://news.com.com/2100-1033-901554.html

Cioffi, J. M., Oksman, V., Werner, J., Pollet, T., Spruyt, P. M. P.,  Chow, J. S., & Jacobsen, K. S. (1999, April). Very-high-speed digital subscriber lines. IEEE Communications Magazine, 37(4), 72-79.

Kintzel, J. (2001, March 1). Outta sight. Wireless Review, 18(5), 24-28.

Kohn D., Teledesic Corporation (1996, June). The Teledesic network: using low-Earth-orbit satellites to provide broadband, wireless, real-time Internet access worldwide. INET96 Proceedings, Montreal, Canada. Retrieved August 20, 2002 from http://www.isoc.org/isoc/whatis/conferences/inet/96/proceedings/g1/g1_3.htm

Losowick, P. (1997, May 5). VDSL gains as technology barriers fall. Electronic News, 43, 52.

Manners, D. (2002, June 12). Big bandwidth VDSL networks held back as suppliers push competing standards. Electronics Weekly, 5.

Manners, D. (2002, June 26). The trials of VDSL. Electronics Weekly, 24.

Oksman, V. & Werner, J. (2000, May). Single-carrier modulation technology for very high-speed digital subscriber line. IEEE Communications Magazine, 38(5), 82-89.

Peterson, L. & Davie, B. (2000). Computer Networks, A Systems Approach. San Francisco, CA: Morgan Kaufmann.

Richards, K. (1998). Internet via satellite: the Teledesic model. Institute for Telecommunications Studies Projects Web site. Retrieved August 20, 2002 from http://www.tcomschool.ohiou.edu/its_pgs/teledesic.html

Rosner, M.C. (1997). Carrierless AM/PM. Worcester Polytechnic Institute Web site. Retrieved July 28, 2002 from http://www.ece.wpi.edu/courses/ee535/hwk97/hwk3cd97/mrosner/node4.html

Rubenstein, R. (1999, November 17). VDSL: The promise to come. Electronics Weekly, 18.

Schrick, B. & Rjezenman, M. (2002, June). Wireless broadband in a box. IEEE Spectrum, 39(6), 38-43.

Secker, M. (2002, June). A business model toddler. Telecommunications (International Edition), 36(6), 28-30.

Segal, N. (2001, October 25). Teledesic: a space-based Internet. Streaming Media World Web site. Retrieved August 20, 2002 from http://www.streamingmediaworld.com/yours/docs/teledesic/

Teledesic (2002). Corporate Web site. Accessed August, 2002, at http://www.teledesic.com

Teledesic (n.d.). Technical details of the Teledesic network. www.comlinks.com Web site. Retrieved August 21, 2002 from http://www.comlinks.com/sat/teled.htm

Some clipart used in creating the figures was supplied by Microsoft Office XP.

[1] According to Acampura (2002), nine out of 10 U.S. businesses with more than 100 workers are out of reach of an optical-fiber link. He also points out that as a result of the lack of high-speed connections to the optical-fiber backbone, only 2% to 5% of the backbone in the United States is being utilized.

[2] From a global perspective, however, the most prevalent problem is a total lack of Internet access. According to Teledesic (2002), the vast majority of the worldís population still lacks telephone service and therefore even basic modem access to the Internet.

[3] As discussed in the paperís conclusion, most Internet access customers are still quite price sensitive, and many consider current ďbroadbandĒ solutions too expensive.

[4] "Broadband" is quoted because it is an ill-defined term with a rapidly changing meaning.

[5] Losowick (1997)

[6] These interpretations come, respectively, from Rubenstein (1999); Cioffi, Oksman, Werner, Pollet, Spruyt, Chow, and Jacobsen (1999); and Secker (2002).

[7] Cioffi et al. (1999)

[8] According to Cioffi et al. (1999), on a copper twisted pair, POTS or ISDN signals are sent in the 0-120 KHz frequency band, while VDSL signals are sent in the 300 KHz-30 MHz band.

[9] Losowick (1997)

[10] Examples of additional impairments given by Cioffi et al. (1999) are crosstalk (noise on a phone line caused by electromagnetic radiation from other phone lines in close proximity), ingress (interference from radio frequency transmissions), and impulse noise (temporary, random noise caused by electronic and electromechanical devices). The section "2.4.2 The DMT Coding Scheme" briefly explains how DMT modulation minimizes the effects of these impairments.

[11] Cioffi et al. (1999)

[12] Losowick (1997)

[13] Oksman and Werner (2000)

[14] Losowick (1997)

[15] A more recent article, Secker (2002), quotes a source who claims that providers are abandoning the goal of delivering 52 Mb/s VDSL.

[16] Cioffi et al. (1999) (all of section 2.3, except previous footnote)

[17] Losowick (1997)

[18] Cioffi et al. (1999)

[19] Losowick (1997)

[20] Rubenstein (1999)

[21] More accurately, as explained later in this section, to support high performance for both symmetric and asymmetric services, current CAP/QAM systems normally use two signals for downstream transmissions and two for upstream transmissions, with each signal employing a separate frequency band.

[22] Oksman and Werner (2000)

[23] Losowick (1997)

[24] Rosner (1997) (from the beginning of the paragraph preceding the bulleted list)

[25] Rosner (1997)

[26] Rosner (1997)

[27] Rosner (1997)

[28] Losowick (1997)

[29] Oksman and Werner (2000)

[30] Cioffi et al. (1999). As explained later, DMT uses a broad range of frequencies. To coexist with ADSL, it merely needs to turn off its lower frequencies—that is, the frequencies below 1.104 MHz, which are used by ADSL. (Cioffi et al. 1999)

[31] Rubenstein (1999)

[32] Cioffi et al. (1999)

[33] Cioffi et al. (1999)

[34] Cioffi et al. (1999)

[35] Cioffi et al. (1999)

[36] Manners (2002, June 12) and Manners (2002, June 26)

[37] Schrick and Rjezenman (2002)

[38] Schrick and Rjezenman (2002)

[39] Schrick and Rjezenman (2002)

[40] Schrick and Rjezenman (2002)

[41] Schrick and Rjezenman (2002)

[42] Charny (2002, February 11)

[43] Kintzel (2001)

[44] Schrick and Rjezenman (2002). Among the LOS providers who either limited LOS service development or discontinued service to customers in 2001 are AT&T, Sprint, and WorldCom. Three LOS providers, Teligent, Winstar, and WorldCom have filed for bankruptcy. (Schrick and Rjezenman 2002)

[45] Schrick and Rjezenman (2002)

[46] Schrick and Rjezenman (2002)

[47] Schrick and Rjezenman (2002)

[48] Schrick and Rjezenman (2002)

[49] Schrick and Rjezenman (2002)

[50] Schrick and Rjezenman (2002)

[51] This phenomenon is known as multipath distortion. With radio communication, multipath distortion is a problem. However, as explained shortly, with point-to-multipoint NLOS transmission it is an essential feature.

[52] Schrick and Rjezenman (2002)

[53] Schrick and Rjezenman (2002)

[54] Schrick and Rjezenman (2002)

[55] Schrick and Rjezenman (2002)

[56] Schrick and Rjezenman (2002)

[57] Charny (2002, May 7)

[58] Charny (2002, May 7)

[59] Charny (2002, January 16)

[60] Charny (2002, February 11)

[61] Kintzel (2001)

[62] Kintzel (2001).

[63] Kintzel (2001).

[64] Schrick and Rjezenman (2002)

[65] Schrick and Rjezenman (2002)

[66]Schrick and Rjezenman (2002)

[67] Schrick and Rjezenman (2002) (all of section 3.2.1)

[68] Schrick and Rjezenman (2002)

[69] Schrick and Rjezenman (2002)

[70] Schrick and Rjezenman (2002)

[71] Charny (2002, May 7)

[72] Schrick and Rjezenman (2002)

[73] The set up time for an FSO system is also considerably less than that for underground optical fiber (a few days vs. six to twelve months).

[74] Acampora (2002)

[75] Acampora (2002)

[76] Acampora (2002)

[77] Acampora (2002). A VDSL system can achieve a maximum bandwidth of about 52 Mb/s, a NLOS fixed wireless system a maximum of about 13 Mb/s, and the current proposed Teledesic design a maximum of 720 Mb/s.

[78] Despite the initial focusing, the beam diverges and its power disperses with distance.

[79] Acampora (2002)

[80] Acampora (2002)

[81] Acampora (2002)

[82] Acampora (2002)

[83]Acampora (2002) (all of section 4.3)

[84] Acampora (2002)

[85] Acampora (2002). Acampora and Krishnamurthy (1999) define the network capacity as "the maximum number of QoS-guaranteed virtual connections that our broadband access network is always capable of delivering to an access point, independent of how the calls are distributed among clients (as long as no client generates/receives more than C calls, where C is the single-link capacity)." The article then presents and proves a formula for calculating capacity. However, it is difficult to apply the result to the network described in Acampora 2002, since the two articles seem to describe slightly different topologies.

[86] Acampora (2002)

[87] Acampora (2002)

[88] Acampora (2002)

[89] Teledesic has additional offices in cities around the world. Their web site is at www.teledesic.com.

[90] Teledesic (2002) (from the start of the paragraph preceding the bulleted list)

[91] Teledesic (2002)

[92] Richards (1998)

[93] Teledesic (2002). Because Teledesic has yet to publish the latest technical descriptions for many of the system features, this paper describes some of the features according to the published specifications for earlier designs. Of course, any of the features might change before the system is actually deployed (in 2005, according to current plans). (Teledesic 2002)

[94] Kohn (1996). According to Teledesic (n.d.), the 840 satellite system was designed to achieve a bit error rate of less than 10-10 and a link availability of 99.9% over most of the U.S.

[95] Teledesic (2002)

[96] Teledesic (2002)

[97] Kohn (1996)

[98] See "5.2 The Teledesic Saga" for a timeline that includes this event as well as other important milestones.

[99] Teledesic (2002)

[100] According to Kohn (1996), there is single possible GSO orbit, which is located over the equator and has an altitude of 36,000 km.

[101] Teledesic (2002). According to Teledesic (n.d.), in describing the 840-satellite design, the low orbit will also reduce the required size of the usersí antennas.

[102] As explained in Peterson and Davie (2000), TCP uses a sliding window protocol that restricts the number of bytes the sender can send per network round-trip-time (RTT) to 64KB or less (barring a TCP extension). The higher the RTT, the less throughput this restriction allows. For example, Kohn (1996) states that a traditional GSO satellite might have an RTT of 0.5 s, which would limit total throughput to 128 KB/s, or 1.024 Mb/s (only 68% of a T1 line).

[103] Kohn (1996)

[104] Kohn (1999)

[105] For example, according to Starband's Web site (http://www.starband.com), the upload bandwidth for their two-way traditional satellite Internet access system is limited to 33.6 Kb/s.

[106] Teledesic (2002)

[107] Teledesic (2002)

[108] Kohn (1996)

[109] Kohn (1996)

[110] Kohn (1996)

[111] Teledesic (n.d.)

[112] The packet size, however, is 512 bits vs. 424 bits for ATM.

[113] Teledesic (n.d.)

[114] Teledesic (n.d.). The Teledesic company, whose success depends upon acceptance in a large number of countries, seems to be sensitive to the importance of not bringing competition to local economies.

[115] of the International Telecommunication Union, a United Nations organization based in Geneva

[116] For a description of the spectrum, see "5.1.3 The Frequency Band."

[117] In 1991, Alenia Spazio became the first satellite manufacturer to build and launch a GSO Ka-band satellite, Italsat F1.

[118] Teledesic (2002) (all of section 5.2)

[119] Richards (1998)

[120] As of the end of 1999, Teledesic, still a privately held company, had raised a total of $1.5 billion.

[121] Teledesic (2002)

[122] Teledesic (2002)

[123] Teledesic (2002)

[124] For example, according to Charny (2002, May 7) a NLOS network such as the one being tested by Sprint PCS can cost up to $250 million to build.

[125] According to Schrick and Rjezenman (2002) a survey from the Strategis Group in Washington D.C. indicated that only 12% of online customers were willing to pay $40 per month for ďbroadbandĒ access, and only 30% were willing to pay $25 per month.

Valid HTML 4.01 Transitional