Media

Communications media fall into two categories, wireless and wired, with a fair number of iterations in each.

Wireless

The various frequency bands used for wireless transmission belong to what is known as the electromagnetic spectrum. This topic, together with descriptions of how waves carry information, is covered in Chapter 10.

Radio

Radio waves, roughly 1 KHz to 1 GHz in the frequency spectrum, carry radio, television, cellular telephone calls, and communications from ships, planes, drug dealers, cops, good buddies, garage door openers, and lots more. Altogether, a heck of a lot of stuff is carried by radio waves. Still, relatively little of that is data. The problem with using radio frequencies for data is that this is both a very crowded and a rather noisy part of the spectrum. The most reliable way of modulating a wave, the method least vulnerable to interference, is frequency modulation (see Chapter 10). But this is also the least efficient, since using two sets of frequencies for the binary numbers uses more bandwidth than amplitude or phase modulation.

Given all the competition for the use of radio waves, conserving bandwidth is a real issue. On the other hand, all that stuff being broadcast creates lots of interference—mostly through waves that are reflected by earth, sky, or buildings, and bounce back at the wrong frequency and/or the wrong phase. Also, as we've already noted in Chapter 10, interference with a signal's amplitude is very common. There are some radio-based local area networks, used mostly where cables just don't work, for example, where people are moving around a hospital with laptops or PDAs. There is also increasing use of radio for campus-type networks (as noted, campus in this case means a cluster of buildings—more than a LAN but less than a WAN). However, both types of use are usually for short distances and low data rates, for example, e-mail or similar messages.

The future will bring a sharp increase in digital connections in the radio band; the first change has already begun with the introduction of personal communications systems (PCS, digital cell phones) that provide both voice and pager-type functions. Next, digital television (DTV) will appear, and with it, some related information services. DTV broadcasts will have huge information carrying capacity, but it will be only one-way (television sets are receivers, not transmitters, and even if they could send data, there are no "return channels" for televison). Finally, the advent of the third generation of cellular networks, known simply as 3G cellular, will offer two-way data transmission at rates up to about 2 Mbps. The radio bands will also carry some of the very short range links—pico-nets—that will allow small devices like PDAs and home computers to communicate with each other. Still, the congestion and noise present in the radio bands limits their potential to take on a lot more capacity. That chore will go to microwaves.

Microwave

Generally speaking, the higher the frequency of a wave, the more likely it is to lose strength over distance (attenuation); a logical corollary of this weakness is that higher frequency waves are more easily blocked by physical objects (until you get to really high frequencies like X and gamma rays, which go through almost everything—for more information, consult your physicist). Microwaves, in roughly the 1 GHz to 60 GHz frequency band, are therefore, obviously, more prone to attenuation than radio waves. On the other hand, their higher frequency means they can be modulated to a greater bandwidth, and this in turn makes microwaves very attractive as broadband carriers of information. One easy way to make microwaves useful, of course, would be to give them a lot of power at the antenna. Unfortunately, this might leave a bounty of cooked pigeons around broadcast sites. A safer and more effective way is to focus the beam, using horn-type antennas at both ends, an approach that helps to keep power below dangerous levels. Because they are of relatively low power, and because they are highly focused, microwave broadcasts are limited to line of sight "connections" from antenna to receiver.

Unlike the situation with radio waves, an obstruction can easily block the signal (above 8 GHz or so, even rain can attenuate microwave transmissions), and a receiver not aimed at the focused beam will get a severely degraded signal, or none at all. Microwave communications were used in a limited way during World War II and became popular shortly thereafter. As we'll discuss later, until the advent of fiber optics, telephone companies made extensive use of microwave systems for long distance networks. Some independent users, for example, large corporations, also own and operate microwave links that they employ for data, for voice, or for both. Satellite systems, which use roughly the 3 GHz to 20 GHz portion of the microwave band, will be discussed in the section on wide area networks.

Infrared

Just below the portion of the frequency band that comprises visible light are infrared waves, around 400 GHz to 100 THz. Best known for their role in carrying information from couch potato to TV, infrared waves have very short range and are easily blocked by any solid object (e.g., the body, but not the head, of another couch potato). They can bounce off of surfaces such as white walls, but lose considerable power in doing so. Still, infrared waves can be useful in a limited number of short (up to 2 miles, max) line of sight, high bandwidth links or, more often, in special low bandwidth local area networks for which normal cables are infeasible (perhaps because the computers are mobile, as in forklifts in a warehouse). An example of an infrared LAN would be one in which wires connect computers to a ceiling or other high-mounted antenna, which then carries information via light beams to another receiver across the room or down the hall. The good news about infrared communications is that, unless Shaquille O'Neal is visiting or the staff has taken to keeping pet bats, you aren't likely to experience much interference—infrared waves are immune to radio, micro, or even other light waves. As noted earlier, infrared connections are also employed to link laptop to desktop computers for file transfers. It's not clear, though, that infrared communications have much future beyond these types of connections.

Wired

While wireless media are increasingly popular for carrying data, the real volume, now and in the future, is carried on wires or cables of one kind or another.

Twisted-pair

The most common kind of cable is what is known as twisted-pair, so called because a pair of wires is twisted around each other (each wire is covered with a light plastic coating to prevent short circuits). The twisting has the effect of shielding the cable from interference; the more twists, the better the shield. The two pairs are then covered with a plastic or similar coating that helps protect them from fraying or other damage but has no electrical benefit. Old style telephone wire (known as voice grade, or Category 1) has few twists and is a poor choice for data. More modern telephone cable, known as Category 3, will handle fairly high digital data rates, including local area networks, such as 10 Mbps Ethernet. With careful installation and the use of two or more pairs for one connection, Category 3 wiring can handle much higher data rates, though it is not commonly used in this way. The next level up, Category 5, gets greater capacity from tighter twists, though it needs to be carefully installed to ensure that the twists continue all the way to the connections. Category 5, using two to four pairs, can handle at least 100 Mbps Ethernet and is being tested for data rates as high as 1 Gbps. Some categories higher than 5 are now specified, but are not in wide use.

Drop It! A drop, in communications jargon, is the wire that connects a network node to the backbone, or hub. The term comes from the telephone world, where a drop wire is the one that drops from the telephone pole to the residence or business.

Twisted pair comes in various sizes, measured in the same gauges as for electrical cable such as lamp cord. The thicker the wire, the greater the capacity, and therefore, the lower the signal loss over distance. But it's not common to make the wire thicker than the standard 24-gauge (higher numbers mean smaller wire). The reason is that twisted-pair's greatest advantage is its comparatively small size and ability to twist and bend. This means that it's relatively easy to install twisted-pair in existing walls and conduits. This is true even for the standard data carrying version, which has four pairs in one sheath. If it were thicker, most of those advantages would disappear. The need for wire that is small and flexible also explains why standard twisted pair doesn't have any extra shielding, i.e., an external metal cover. Some types of twisted-pair do have such shielding; IBM favored a shielded twisted-pair system for a while, but it hasn't been popular because its expense, bulk, and inflexibility are not a good tradeoff compared to the standard unshielded options.

By the way, standard twisted-pair is often known as UTP, for unshielded twisted-pair, though given the dominance of the unshielded variety, this is a little like saying "rubber-tired vehicle." One final point about twisted pair—don't accept the idea that existing telephone cable in buildings can be easily converted to handle high-speed data. As noted, most older phone cable has few twists and is very vulnerable to interference. The newer Category 3 may or may not handle data—it all depends on the care with which it was installed. A really bad idea is to use some of the pairs in old-style 25-pair office phone system for data. The wires in these bundles are very thin and obviously quite close together. The lack of shielding works both ways. Not only are the wires poorly protected from outside interference, when operating at high data rates they emit a lot of their own. The result is that the high frequency transmissions from the data lines leak over into the voice wires—crosstalk in telephone parlance. A friend who worked in an office that had used wires in this way observed that he could hear his e-mail coming in when he was talking on the phone. More on the problem of crosstalk when leased lines are discussed in Chapter 12.

Coaxial

The other major kind of copper cable is known as coaxial for the fact that there are two layers of metal that share the same axis. The first metal layer, the part that carries the signal, is a solid piece of copper in the interior. The next layer is an insulating substance (usually a foam). Next is the second metal layer, a mesh of thin wire that serves as an electrical ground and that has the effect of shielding the core from interference. Outside is a plastic or similar cover. Coax is best known as the kind of cable used for cable television. The combination of a comparatively thick core and an external metal shield means that it is less vulnerable to interference and can therefore handle much higher data rates for longer distances than twisted-pair. Before its role in the entertainment business, coax was used by telephone companies for trunks—lines, such as long distance, that carry many circuits. In the data world, coax has had a variety of uses.

IBM long favored a pair of coaxial cables (twinaxial) for its mainframe systems. Big Blue's more modern systems are likely to use fibre channel for central connections (CPU to storage) and twisted-pair for desktop machines, but there is a lot of twinax still out there connecting IBM's venerable 3270 terminals and other peripherals. Coax's principal use is for the most common local area network, Ethernet. The original Ethernet used a comparatively thick coaxial cable. As the system became more popular, this heavy and expensive cable was relegated to backbone (i.e., server to server) connections and replaced with thinner coax (similar to standard cable TV) for normal server to desktop links. Virtually all of these connections are at 10 Mbps. Coax can handle lots more bits per second than Category 5 twisted-pair, but it isn't used even for the higher speed (100 Mbps) Ethernet. The reason for coax's loss of popularity is that twisted-pair is cheaper and easier to work with in the low to middle speed ranges, and fiber optic has totally staked out the high end. While comparatively little coax is now being installed, the stuff that's already there—whether in office LANs or cable company connections to homes—can still carry a lot of data and will still be very useful in the digital world.

Fiber Optic

The potential of light to carry sequential information at high speed attracted computer designers from the earliest days. The problem in using light waves was that they don't do well in the air—they spread out, are blocked by solid objects, and can be interfered with by other light sources of the same wavelength. People realized early on that an optical cable could solve these problems, but for a long time it simply wasn't possible to make glass strands of the necessary purity at a reasonable price. The problem was solved by Corning in the early 1970s. The original fibers were exclusively glass, but a plastic version is now sometimes used for very short hauls. Fiber optic cable uses a highly conductive (extraordinarily pure) center core to carry the light waves and an exterior cladding that causes the waves to be refracted in a way that keeps them in the core. To illustrate the purity of their product, purveyors of fiber optic cable say that if you had a window of this type of glass three miles thick, you could easily see through it. Like copper wire, there is also an external plastic cover—necessarily opaque in this case. Most fiber cable, and especially that used outside, will also have some metal strands inside the plastic cover to make it resistant to cuts and stretching. Since fiber optic cable carries no electricity, it can do things copper can't—for example, many long distance fiber optic links are carried inside natural gas pipelines.

Sending Information with Light · Pulses of light are fired into the core from the end; the pulses are produced either by a very small laser (required for greatest speed and distance) or a light emitting diode (used for shorter connections such as LANs). At the other end of the cable, a pulse or lack of a pulse is then interpreted by the receiver as either a one or a zero. Nearly all of the fiber optic cable in use is unidirectional—two way communication uses two cables. Bidirectional fiber is new and relatively much more expensive, though that may change in the near future. Fiber optics has overwhelming advantages over copper or other metal cable. One is capacity. Light waves are much shorter than electrical pulses; since they are shorter, there are more of them in a given space, which means more information carrying capacity in a unit of time.

The bandwidth potential of a single strand of optical fiber goes into terabits—trillions of bits per second—on the order of 10,000 times the maximum speed of copper. Part of fiber's advantage is its immunity from interference. Once the easy task of keeping stray light out is accomplished, exterior interference is nonexistent. This makes it very attractive for electrically noisy environments like factories. It also explains why electric power utilities have found it so easy to add fiber cables to their high voltage transit lines. Finally, light in pure glass or plastic maintains its integrity over distance much better than electricity in copper; laser-driven single mode fiber optic cable (described later in this section) needs amplifiers (repeaters) only every 80 or so kilometers vs. about 600 meters for copper, despite its much lower speed.

A major problem in early fiber optics was amplification—restoring a fading, hard-to-read signal to full strength. After six or seven kilometers, system designers had to build devices that first used optical sensors to receive the signal, then converted it to electricity for amplification, then converted it back to light for retransmission. This expensive and inefficient approach has been supplanted by the use of chemical lasers that are extremely efficient, as well as effective, in regenerating optical pulses to their original shape. As you would expect, the development of optical amplification was particularly welcome in making fiber optic transoceanic cables. Another early problem with fiber optics was the relatively mundane issue of cable splicing; joining two cables requires that their optical axes be aligned to a high degree of precision (even tiny amounts of dirt can be a major issue as well). Experts initially believed that fiber optics were so complicated that they would never take over telephone systems. Splicing, it was said, required "sending a Ph.D. into the manhole." However, advanced manufacturing has allowed suppliers to offer splicing techniques (e.g. the cool sounding fusion splice) and other connecting equipment that can be reliably used by people with only Master's degrees. Just kidding; any well-trained person can install current fiber optic systems.

Kinds of Fiber Optic Cable · There are two basic kinds of fiber optic cable: single mode and multimode. Multimode fiber, the original kind, has a thick core. As light travels through this medium, it disperses and follows a variety of paths; some rays go straight down the center, but others hit the cladding and are reflected back at varying angles. The receiver, therefore, has to account for the differing speeds of pulses that have taken different paths; the resulting complexity limits the effective distance before amplification is needed. Single mode fiber, as the name suggests, has a very narrow core and allows light to take only one path. Single mode fibers can go much farther without the need to regenerate the signal and can handle higher data rates. Because the core must be manufactured to very high tolerances, single mode fiber is more expensive. As often happens, though, improvements in manufacturing have rapidly narrowed the price gap such that single mode is now much more widely used. Advanced networks now use a special form of single mode, known as non-zero dispersion shifted fiber. You can tell that the marketing people haven't gotten their hands on this bit of technical slang—how about SuperTruLink?

New Technologies and Networks · One hot new development in fiber optics is the advent of multiple wavelength systems (the approach is usually called wave division multiplexing; see Chapter 12). Early lasers and LEDs sent out just a few wavelengths of light, but the fiber can handle a whole range at once. Light sources that can send multiple wavelengths offer the potential for vast increases in the capacity of a medium whose existing ability is as yet far from fully tapped. We have mentioned that fiber optics carry a signal further than copper without the need for amplification, but we should also mention that it is at the same time superior in data integrity—a fiber optic cable is only likely to produce an error in about 1 of 100 trillion bits (100,000,000,000,000). The good news doesn't end there. While fiber optic cable is normally installed in pairs—one for each direction—individual fibers are very thin and it's a rare cable that doesn't have lots of these pairs. Typical fiber optic bundles include 24, 72, or 96 fibers; the maximum is about 200. And, despite the huge growth in both data and voice traffic that has occurred since the cable was laid, some of these installed fibers are still unused. Called dark or unlit fiber, this reserve was, until recently, thought to provide an enormous network resource for future development. Observers believed that the nation's cable backbone, installed for the most part in the ground along railroad and pipeline rights of way, had capacity, both in the ability of the fiber to handle ever greater data rates and of the unlit fiber to take up the slack when that approach is exhausted, to meet voice and data needs for years, perhaps decades. The phenomenal growth of data networking has changed that assumption, though, and new fiber is now being installed at a frantic pace.

To illustrate the potential of fiber optic cable, consider the case of Qwest Communications (Qwest is not a typo; maybe Elmer Fudd is chairman of the board). Qwest has built a 125-city fiber network in the U.S., using mostly railroad right of way. There are two fiber cables, each with 96 fibers (48 pairs). Using an early form of wave division multiplexing, each of those fibers carries 16 different frequencies (colors) of light, and each frequency (according to the limit of the electronics at the time the system was built) handles about 10 Gbps. Given that the second cable is in reserve as a spare, and that current versions of WDM can carry more on the order of a hundred times that of the original, there is a lot of room for growth without additional digging. Qwest began operation strictly as a carrier's carrier, selling bandwidth to phone companies as well as to purely data systems like Internet backbones, though it now also has its own long distance service. In a break with telephone company tradition, it also makes dark fiber available to private companies as end-to-end leased lines. Qwest has taken many steps to enhance reliability: the use of railroad rights of way limits the threat from the cable-sensing backhoes that plague more public routes; the cable is buried extra deep (about four feet); and the topology is in the form of a self-healing ring (see the section on topology, below).

Although Qwest's network is impressive, it is only the first of the megaprojects. Subsequent efforts include that of a company called Level 3, which is conceptually similar to Qwest's network. Outside the U.S., there is Project Oxygen, which intends to literally circle the globe with undersea fiber links. There will be more by the time you read this.

The biggest barrier to achieving the potential of fiber optics is not in the fiber or in its optical components, but in the attached electronics. When fiber optics systems are transmitting in the Gbps range, the fiber is loafing but the electronics are struggling. It is difficult and expensive to build electronic systems that can feed bits into the lasers or LEDs, much less read their output, at the speeds now being used. There will need to be major increases in the capability of the electronics, including especially the use of high performance materials such as Gallium Arsenide (GaAs) and Silicon Germanium (SiGe) as a replacement for silicon in the substrates, if the bandwidth of fiber networks is to continue to grow to accommodate terabit levels.

One area of intensive research is in passive optical networks (PONs). These would do away with electronics entirely and use optical switching throughout. The first stage of this approach will be optical cross connects that are able to switch entire streams of data, rather than individual packets. Full switching may follow. The technological hurdles here are formidable, however, and we may see bandwidth expansion slowed, or perhaps made a great deal more expensive, by the cost of electronics before optical switching comes to the rescue.