Unlimited bandwidth for a pittance. If that seems like an unrealistic dream, compare telecommunications networks to the computer. In 1980, who would have conceived of having on their desk more computing power than IBM's largest mainframe for less than $1000? Everyone knows how rapidly computers have advanced in the last two decades, but few realize that the cost of sending a million bits of information across the continent is just a fraction of the cost of a decade ago. That doesn't necessarily mean the price of service has dropped, but the service providers' cost of the bandwidth itself is so low that the pricing model has changed significantly in the last few years, with more changes yet to come. Give credit to the optical network -- that phenomenon that launches bits of information across tiny glass strands by encoding it into infrared pulses.

Back when the nation was laced with microwave radios and the oceans were bridged with copper cables, the cost of the transmission medium was a major part of the cost of delivering telecommunications service. With optical networking today, the cost of the medium is small, and the ultimate capacity of a single fiber-optic pair has scarcely been utilized. The manufacturers regularly leapfrog one another with new records for the amount of information a fiber pair can transmit. If the original fiber, which operated at 45 megabits per second (Mb/s), were the equivalent of rocketing to the moon, today we'd be exploring the outer reaches of the galaxy.

The importance of fiber optics to the world economy cannot be overstated. The demand for information transport is insatiable and highly price-elastic. As the price of bandwidth drops, applications will expand to absorb it, and the impact is incalculable. Today's computer networks are distributed to bring information closer to the user, but with cheap bandwidth it may be better to centralize it. The growth of telecommuting, video conferencing, distance learning, and a host of other applications is restrained only by cost, which will drop as optical networking expands. Today, fiber is being installed at a mind-boggling pace.

This book explains optical technology and how it is applied in telecommunications networks ranging from in-building networks to campuses, metropolitan areas, and worldwide long-haul networks. Optical networking is about much more than just fiber optics. Fiber is the medium used to guide lightwaves, which carry the information in the form of coded pulses. Optical networking includes other components that are discussed here: lightwave amplifiers, wavelength division multiplexing equipment, and optical switching or cross-connects.

Lightwave's Astounding Growth

Looking back over the past 150 years of telecommunications, it's easy to identify a handful of technologies that have driven the remarkable progress the industry has made. In the middle of the nineteenth century the telegraph was the primary means of long-distance communications. Then, in short order came the telephone, automatic switching, radio, and the vacuum tube, which relegated the telegraph to the background of telecommunications.

Technology advanced steadily, but unspectacularly, until the next two dramatic developments -- the transistor in 1948 and fiber optics in the mid-1970s. Together, these two technologies have revolutionized telecommunications and the computer industry. As telecommunications technologies go, fiber optics is a relative newcomer, although the idea of communicating with light has been around for a long time. In 1880 Alexander Graham Bell constructed a device he called the "Photophone," over which he could talk if conditions were right, which they seldom were. Scientists experimented with light transmission for years until 1970, when Corning researchers finally discovered the secret of making glass so transparent that it could guide lightwaves encoded for information transport.

The first commercial use of lightwave transmission was in a Bell Laboratories field trial in 1977. That system operated at 45 Mb/s, and was capable of carrying 672 voice channels. The original fiber and most metropolitan systems today carry lightwaves at baseband -- that is, a single stream of pulses occupies the entire bandwidth of the fiber-optic medium. Baseband fiber still has many uses, but its capacity is limited compared to fiber that is expanded with wavelength division multiplexing (WDM). WDM divides the fiber-optic transmission medium into different "colors," or wavelengths. The spectrum used is below that of visible light, but the color analogy is accurate and each different wavelength can carry as much information as the baseband. The Greek letter lambda (l) is used to denote a wavelength. As many as 160 lambdas can be multiplexed on a fiber at today's level of technology, but no one is claiming that this is the limit. To put that in perspective, 160 lambdas, each carrying 10 gigabitsper second (Gb/s), provides 1.6 terabits of capacity. In terms of telephone calls, that amount of bandwidth would carry more than 20 million simultaneous conversations -- and 10 gigabits per lambda is not even the limit.

As Bandwidth Grows, Applications Follow

Humans have an insatiable desire to communicate. The decades of the 1980s and 1990s saw intense activity to lace the globe with fiber optics. Not only the traditional carriers but also gas pipeline, electric transmission, and railway companies became major developers of fiber-optic capacity, thanks to the rights-of-way they own. All of this happened in time to meet the explosive growth of the Internet, which began consuming bandwidth at an unprecedented rate. From the telecommunications service providers' point of view, the Internet sparked an intense peak in demand, but at the same time, it eliminated the predictability that had always characterized conventional voice and data networks. The prodigious growth of the Internet means that bandwidth is needed in abnormal quantities and with lead times measured in days rather than the months and years that were characteristic of the past.

Without fiber optics, the Internet today would resemble its ARPANET predecessor -- a specialized and limited network designed to enable a select class of users to exchange email and files and log onto remote computers. The World Wide Web, when it was introduced, not only insulated users from the arcane commands of the UNIX operating system, it also swelled the size of files by several orders of magnitude. The Web would have quickly overwhelmed the microwave radio and coaxial cable networks of the past. Not only is the bandwidth of these older technologies insufficient, their error rate cannot match the purity of an optical fiber system.

Optical Networking Technologies

Today, the world is beginning to deploy new optical technologies that will multiply the capacity of existing fiber and lower the cost of bandwidth to a fraction of its historical levels. Where the original fibers carried information at 45 megabits per second, the current standard is 10 Gb/s with 40 Gb/s waiting in the wings. In addition, by multiplexing different wavelengths using dense wavelength division multiplexing (DWDM), a fiber can support multiple channels, each capable of carrying 10 Gb/s or more of bandwidth. This leads to a demand for other technologies to amplify and switch the light without converting it from an optical to an electrical signal and back to an optical signal. The equipment that accomplishes this falls under the blanket term of optical networking.

Optical networking includes several elements. At the physical level is the fiber-optic cable itself. An equipment layer known as synchronous optical network (SONET) is used to multiplex multiple services on the fiber. SONET offers a standard hierarchy of bandwidths at the electrical level. It is a subset of the international synchronous digital hierarchy (SDH), and is discussed further in Chapter 3. Binary electrical pulses from the multiplexing equipment are converted to optical pulses and carried across a lightwave span. SONET/SDH equipment brings several important functions to the optical network. Not only is it a standard that permits interoperability among equipment of different manufacturers, it also enables the service providers to add and drop bandwidth at the electrical level, a function that is known as add-drop multiplexing (ADM).

A second function at the electrical level is known as a digital cross-connect system (DCS). A DCS is, in effect, a digital switch capable of forming high-bandwidth connections between input and output ports. A DCS replaces physical connections with logical connections that are set up under software control, making rearrangements much faster and easier. DCSs are available in a wide range of sizes, and are used by both common carriers and end users. Carriers require huge DCSs at main cross-connect points. These become a major source of capital expenditure and operational costs. Optical networking's major objective is to perform the cross-connections at an optical level, avoiding the need to convert optical signals to electrical signals just to switch and route them. Every optical-to-electrical-to-optical (OEO) conversion is expensive in terms of hardware, floor space, electrical power, and administration. The optical network aims to keep bandwidth in the optical domain to the maximum degree, converting it tothe electrical domain only where it is delivered to the customer.

As demand increases and carriers begin to exhaust their capacity, DWDM technology becomes a relatively inexpensive way of multiplying the carrying capacity of a fiber. Although 160 lambdas can be carried on a single fiber, a more practical limit at the present state of technology is 40 lambdas. If each lambda must undergo an OEO conversion at each major branching point, the amount of SONET and lightwave equipment needed is enormous. To avoid this, the industry is working toward optical cross-connect switches to switch lambdas and add, drop, and shift wavelengths in the same way that SONET multiplexers function on the electrical signal. Optical switches exist today, but those that are available do not scale to handle the thousands of lambdas that will eventually need to be routed through major hubs.

Optical and digital cross-connects are also used in the provisioning process. In telecommunications vernacular, provisioning is the lengthy process of accepting a service order, assigning it to facilities, adding capacity if spare capacity is not available, and making the necessary physical connections to create an end-to-end facility. Digital and optical cross-connects not only are used in the provisioning process but also provide fault management and protection switching for service restoration at the optical level.

Structure of the Optical Network

When digital transmission first became practical in the early 1960s, AT&T's Bell Laboratories defined the multiplexing hierarchy in North America using a basic building block of twenty-four 64 kb/s voice channels, which is known as a DS-1. The DS hierarchy defines the digital signal structure, which is used to provide services known as T1 and T3. Customers can employ these services to carry information at bandwidths of 1.5 and 45 Mb/s, respectively. The Consultative Committee on International Telephone and Telegraph (CCITT), which has since evolved into International Telecommunications Union-Telephone (ITU-T), promulgated European standards. Instead of the 24-channel structure of T1, European standards use a 32-channel building block known as E1. Thirty channels are used for information transport and two for signaling.

At the time T1/E1 was developed, the lack of compatibility between North American and European standards was not a major problem because the ability to exchange intercontinental digital signals did not exist and no technology was in sight to change that. The world then relied heavily on analog microwave radio and coaxial cable for transcontinental communications and satellite and analog copper cable for intercontinental service. It was recognized that these would be insufficient for the future, but the technology for making glass of sufficient transparency for lightwave communications did not exist. Bell Laboratories was developing precisely aligned underground microwave waveguides as the next-generation transmission system. They were also planning to convert analog coaxial cable to digital, and accordingly defined a 274 Mb/s T4 signal as the top level in the multiplexing hierarchy.

Meanwhile, scientists were working toward the objective of making glass that was sufficiently transparent to carry an optical signal. A loss of 20 decibels per kilometer (20 dB/km) was at the time considered to be the greatest loss that could be tolerated. The decibel is a logarithmic scale. Each 3 dB of loss cuts the signal power in half, so a reduction of 20 dB would mean that after a distance of 1 km only about 1 percent of the original signal would remain. In 1970, Corning Glass Works reached the 20-dB/km threshold, and focused the attention of the world on fiber optics as the next communications technology. Today, the technology has advanced to the point that it is possible to purchase fiber with loss as little as 0.2 dB/km.

The motivations for using lightwave transmission were many. By confining the transmission to a thin strand of glass, the signal does not radiate into the atmosphere. This eliminates the drawbacks of microwave: interference, fading, and noise. The glass is impervious to induction from nearby electrical equipment and it does not deteriorate with exposure to the elements as copper cable does. Most compelling is its enormous bandwidth. One strand of fiber offers more bandwidth than the entire radio spectrum.

The earliest optical networks consisted of a basic link of fiber-optic cable and lightwave terminating equipment at each end of the link. The lightwave equipment was a laser to convert electrical pulses to optical pulses and a photo diode to convert optical back to electrical pulses. Links were connected back-to-back through regenerators, which restored the shape of the electrical pulses. The industry has seen significant improvements in the technology, but the basic principle of the optical link terminated in electrical regenerators still remains. The span of the link has been expanded by the use of lightwave amplifiers that boost the light signal without regeneration of the pulses.

Fiber-Optic Cable

Fiber is a hair-thin waveguide that closely resembles a monofilament fishing line. A fiber consists of three parts: inner core, outer cladding, and a protective coating around the cladding. The outside diameter is 125 mm, or millionths of a meter, also known as a micron. The inner core is highly transparent glass that guides light pulses within the bounds of the cladding, which is glass with a different refractive index than the core. As light travels down the core, some of it strikes the cladding where it is either refracted back into the core or it escapes. The angle at which light must enter the fiber to be contained within the core is known as the angle of acceptance.

Although the glass used in the manufacture of the fiber is exceptionally pure, imperfections limit the length of a fiber span. As light pulses travel along the length of the fiber, some of the signal is attenuated by various properties of the glass. This attenuation is referred to as loss and it is measured in decibels (dB). Loss is inherent in the fiber, and it can be increased by improper installation. For example, sharp bends reduce the diameter of the core and can cause some light to escape through the cladding.

Light propagates through the core in modes, which are the different rays or paths a lightwave can follow. Fiber is manufactured as either multimode or single-mode, depending on the core diameter. Multimode has a core diameter of 50 to 200 mm, with 50 and 62.5 mm the most common. Single-mode fiber has a core of about 8 to 9 mm. In multimode fiber, some lightwaves are focused exactly parallel to the core and travel directly to the receiver. Some waves aren't so precisely focused and are launched at a greater angle. These are reflected from the cladding and therefore take a longer path from the transmitter to the receiver. These variations in path length cause the pulses to spread out. This pulse-spreading phenomenon is known as modal dispersion. If a light pulse spreads so far that the trailing edge of one pulse merges with the leading edge of the next, bit errors result and the link is useless. The greater the core diameter, the greater the amount of dispersion, which can be overcome only by regenerating thesignal. Single-mode fiber propagates only one mode of light, and therefore does not suffer from modal dispersion.

Single-mode fiber and some low-cost multimode fiber have a uniform refractive index throughout the core. Such fiber is known as step index fiber. The effects of modal dispersion are reduced by using a more complex type of multimode fiber known as graded index. In this type of fiber, the characteristics of the core are altered throughout its diameter by using a slightly different glass composition at each layer of the core. The refractive index is higher at the core and reduces uniformly nearer the cladding. The lightwaves propagate at slightly lower speeds near the core than nearer the cladding, which diminishes the pulse rounding. This reduces the amount of dispersion compared to step index fiber. Figure 1-1 shows the dispersion profiles of types of fiber. The bandwidth of multimode fiber is expressed as a product of speed and length. A fiber with a bandwidth of 1000 MHz at 1 km of length will have a bandwidth of 500 MHz at 2 km, 250 MHz at 4 km, and so on.

Both single-and multimode fibers are subject to another form of dispersion called chromatic dispersion, which results from multiple wavelengths being produced at the source and propagated through the core. Chromatic dispersion is the phenomenon that occurs when light passes through a prism, which breaks the fundamental color into a rainbow. Although single-mode fiber is immune to modal dispersion because its core propagates only one path, it is affected by chromatic dispersion. Chromatic dispersion is controlled by the quality of the glass and of the laser used in the transmitter. High-quality lasers emit a narrower band of wavelengths, resulting in less dispersion.

A third type of dispersion is known as polarization mode dispersion (PMD). PMD is caused by small variations in the shape of the fiber core. When light travels down a fiber it has polarization modes that are at right angles to each other. If the core is not perfectly symmetrical, one mode travels faster than the other, resulting in pulse spreading. Both chromatic and polarization dispersion increase as the square of the bit rate. This means the effects are four times as great at 40 Gb/s as at the current 10 Gb/s standard.

Fiber waveguides are not capable of carrying all wavelengths of light with equal efficiency. As we will discuss in Chapter 4, the manufacturing process leaves the fiber with three wavelength "windows" of sufficient transparency for optical communications. All three windows are lower in frequency (i.e., they have longer wavelengths) than the light spectrum that is visible to the human eye. These infrared windows center about 850, 1300, and 1550 nanometers (nm). A nanometer is one-billionth of a meter, or 1 3 10 meters. In general, the longer wavelengths have lower loss, but the lightwave equipment is more expensive. The 850-nm wavelength is used primarily in customer premise applications, and the 1550-nm wavelength is used almost exclusively in long-haul carrier networks.

Lightwave Transmission Equipment

The 0's and 1's of a digital signal are used to drive a fiber-optic transmitter, which emits a light pulse for a 1 and turns it off for a 0. Other modulation methods such as intensity modulation are used, but on-off triggering is the most common. Fibers are generally deployed in pairs, with separate fibers used for transmitting and receiving. Either a laser or a light-emitting diode (LED) is used for the electrical-to-optical conversion. Lasers are required in long-haul networks, but they are generally too expensive for short-range private systems, which often use LEDs driving multimode fiber. At the receiving end, an optical receiver detects the light and converts it back to an electrical pulse. Depending on the quality of the cable, regenerators using back-to-back receivers and transmitters are placed at intervals of 50 km or so. To protect against link failure, a spare link is usually waiting in hot standby to assume the load if a link fails. The service provider determines how many active linksare protected by a single standby link. Figure 1-2 shows the layout of a basic lightwave transmission system.

Around 1980 when the expansion of lightwave networks was just beginning, lightwave systems were proprietary. The multiplexing method was based on some multiple of the North American T3 at 45 Mb/s or the European E3 at 34 Mb/s. Manufacturers were vying to increase the carrying capacity of their fiber-optic systems, and selected some multiple of the T3/E3 hierarchy. An operational lightwave system requires nonpayload or overhead bits between terminals for a variety of purposes such as alarms and order wires. Overhead bits are also required for evaluating the quality of the link and initiating a switch to protection in case of link failure. Since each manufacturer developed its own method of assigning various overheads, interoperability between systems was usually not possible. Also, the asynchronous nature of the T3/E3 signal made it difficult to access the underlying T1/E1 channels. The answer to these problems was the development of SONET/SDH standards, which we will discuss in Chapter 3.

By the time the Internet explosion hit, the carriers had fiber in place and expanded it to meet the demand. The first step in capacity increase is to increase the speed of the multiplexing equipment, which can often be done by changing the optical equipment and cards in the multiplexers. Today's systems operate routinely at either 2.5 Gb/s (SONET OC-48) or 10 Gb/s (OC-192), with 40 Gb/s (OC-768) equipment soon to reach the market. When this kind of bandwidth is converted to voice signals, it appears that the capacity of a fiber would never be exhausted. One pair of fibers running at 40 Gb/s could handle more than a half-million voice sessions, which is more bandwidth than most large metropolitan areas could possibly consume -- if the traffic were voice-based.

Optical Amplifiers

As lightwaves travel along the glass fiber, losses gradually diminish the signal strength to the point that it must be amplified. The OEO conversions of conventional fiber are costly, both in terms of floor space and power requirements as well as the equipment itself. An important element of the optical network is the optical amplifier. These devices, which operate in the 1550-nm window, typically use erbium-doped fiber amplifiers (EDFAs) to amplify the light signal. Today, EDFAs can be spaced at about 80 km, or even farther apart with the high-purity fiber used in submarine cables. Approximately every five amplifier points, the signal must be regenerated with an OEO conversion. This still represents a large number of OEO conversions in a transcontinental link. Toward that end, several manufacturers are beginning to produce ultra-long-haul equipment that can span as much as 4000 km without an OEO conversion.

Wavelength Division Multiplexing

Most fiber facilities have been designed so that when the base fiber is exhausted, its capacity can be multiplied by a factor of four or more with WDM. Several years ago, dense WDM (DWDM) equipment began to enter the market with much closer wavelength spacing than had been thought possible. DWDM facilities use EDFA amplifiers spaced at about 80 km. Figure 1-3 shows a typical DWDM section. The long-haul carrier backbone consists of multiple sections, most of which today are deployed as SONET rings for service protection. The OEO conversion takes place in a central office facility if possible because it is expensive to construct a space solely for the purpose of regenerating signals. As a result, in many cases the full span of a DWDM section cannot be met because of the geography of the serving area.

Optical Switching

DWDM equipment enables the carriers to increase their capacity dramatically, but it raises additional issues that are not so easily resolved. The cost per megabit of the physical facility becomes insignificant by itself, but the cost of operations, administration, maintenance, and provisioning (OAM&P) becomes more complex with the thousands -- perhaps millions -- of cross-connections that must be managed.

Today, only the largest customers of the fiber-optic carriers subscribe to dedicated fiber or even dedicated lambdas. Most customers need much narrower bandwidths and, in fact, most carriers don't offer so-called dark fiber, which is fiber without lightwave equipment. As we will discuss later, the bandwidth is separated into building blocks that may be routed to different destinations. In the earliest digital systems, circuits were routed to different destinations by making back-to-back physical connections between apparatus. Later, DCS made it possible to administer these connections under computer control. DCS systems are widely deployed today as a convenient way of routing bandwidth where it is needed. Where SONET/SDH multiplexers are used, an even simpler method is using an ADM. Both ADM and DCS work only in the electrical domain, however. Where electrical access is not needed, the bandwidth can be routed in the optical domain. This is the function of the optical cross-connect (OCX), or optical switchas it is sometimes called.

The terms cross-connect and switching are often ambiguous because their function differs with the class of service in which they are being employed. Telephone calls use a form of switching known as circuit switching in which two or more voice-grade circuits are connected end-to-end for the duration of a session. When the session is completed, the connection is taken down and the circuits are available for use by another session.

This form of switching is much different than the switching done in a DCS or an OCX. The most common use of these types of system is for service provisioning, which means the bandwidth is connected by a service order and remains connected until a subsequent order disconnects it. This is the kind of service needed by Internet Service Providers (ISPs), Interexchange Carriers (IXCs), and other such service providers. Both the DCS and the OCX can also be used for dynamic bandwidth allocation. This is bandwidth that is temporarily assigned under software control to fulfill a temporary demand. The demand might be caused by an overload, a failure, or a scheduled activity such as backup of a large server. In this situation, the cross-connect system takes the place of a manual cross-connect, which would involve connecting bandwidth through patch cords and jacks.

Another concept that should be understood in considering switching is the method of control. One form of switching requires external control to set up and take down connections. This is the method the telephone system and digital and optical cross-connect systems use. By contrast, devices such as local area network (LAN) switches control the connection by reading the destination address inside the protocol data unit (PDU) and as such make real-time switching decisions. Digital and optical cross-connect systems today are unable to make decisions based on the content of the PDU,and the technology for doing that appears to be a long way off.

Demand Drivers

Shortly after fiber was proved practical, long distance became fully competitive in the United States as a result of the consent decree that broke up the Bell System. This generated a flood of new construction as AT&T's traditional competitors, MCI and Sprint, were joined by other companies that began to take advantage of their right-of-way resources to build fiber facilities. As other countries joined the United States in deregulating their telecommunications services, similar activity occurred internationally. Undersea cables were placed in major routes, displacing satellite communications to all but the least-developed countries.

In just a few years, fiber optics replaced microwave radio and coaxial cable, which up to that point had been the mainstay of long-haul communications. Through the 1980s and early 1990s, the volume of voice communications outweighed data by a large margin. By the mid-1990s, however, the Internet changed all that. The growth in demand had been easy to forecast when voice was the principal application, but the Internet was anything but predictable.