Chapter 13. Telecommunications
Until the invention of the telegraph, no reliable communications traveled long distance faster than via a human letter carrier. That didn’t stop people from trying, so in the pre-telegraph days a lot of exciting technologies promised to revolutionize communications—smoke signals, semaphore flags, and carrier pigeons were all attempts to break the bond between messenger and message.
The telegraph was able to let the messenger rest his feet and put pigeons back where they belong, atop the heads of dignitaries’ statues in the park. With the advent of the telegraph, messages finally were able to travel at the ultimate speed limit of the universe—the speed of light.
Well, not quite. The telegraph only changed the speed of the messenger. It let the pony express rider exchange his steed for Pegasus, but the content of those messages never reached the speed of light. They were also slowed by the interface technology—the telegraph key in the hands of a human being. With fast fingers and a good ear, a 19th-Century telegrapher might squeeze ten words a minute down the wire, equivalent to a bit rate of less than ten per second. At that rate, sending a snapshot home from summer camp might take a month.
Since the advent of the telegraph, technology has concentrated on this other side of the speed issue, making the message move through the data channel faster. Using essentially the same wires as Samuel Morse, you can move data roughly 100,000 times faster through the telephone system and ten million times faster through your home network.
Getting there isn’t always easy. The analog nature of your telephone line challenges data signals and imposes strict limits on how fast information can move through the system. The telephone line is an arcane world of strange technologies, all meant to get more speed from the simple telephone wire. It has been a long, hard battle, but one that has paid off. Today’s modems now race ahead of the theoretical speed of which information can move through a telephone line. (They can do that because they’re no longer really modems, but we’re getting ahead of ourselves.)
You can do better. Digital services promise to deliver data 20 times faster than the best modem. Although people aren’t switching over to the new, high-speed connections anywhere near as fast as once predicted (that’s one of the big reasons the telecommunications industry crashed in 2002), you can be sure someday you will have an all-digital connection.
A modem is standard equipment in every new computer—and with good reason. It’s what most people use to connect to the Internet. Consider a computer without a modem, and you might as well buy a vacuum cleaner.
Today’s modem is a world apart from those of only a few years ago, and the difference is speed. Nearly every new modem sold operates at a top speed of 56,000bps—fast enough that most people refuse to pay the higher charges demanded for faster-still all-digital services. Today’s modems are fast enough for e-mail and most surfing. They falter only when you want an edge in playing online games, want to download streaming video, or want to exchange photos with relatives and friends.
A true modem is a necessary evil in today’s world of telecommunications because we still suffer from a telephone system that labors under standards devised even before electronics were invented, at a time when solid-state digital circuitry lay undreamed, almost a hundred years off. The telephone system was designed to handle analog signals only because that’s all that speaking into a microphone creates. Over the years, the telephone system has evolved into an elaborate international network capable of handling millions of these analog signals simultaneously and switching them from one telephone set to another, anywhere in the world.
In the last couple decades, telephone companies have shifted nearly all their circuits to digital. Most central office circuitry is digital. Nearly every long distance call is sent between cities and countries digitally. In fact, the only analog part of most telephone connections is the local loop, the wires that reach out from the telephone exchange to your home or office (and likewise extend from a distant exchange to the telephone of whomever you’re calling).
The chief reason any analog circuitry remains in the telephone system is that there are hundreds of millions of plain-old telephone sets (which the technologically astute call simply POTS) dangling on the ends of telephone wires across the country. As long as people depend on POTS, they need an analog network to connect to.
Of all modern communications systems, the one with the most severe limits on bandwidth is the telephone channel. Instead of the full frequency range of a good-quality stereo system (from 20 to 20,000Hz), a telephone channel only allows frequencies between 300 and 3000Hz to freely pass. This very narrow bandwidth works well for telephones because frequencies below 300Hz contain most of the power of the human voice but little of its intelligibility. Frequencies above 3000Hz increase the crispness of the sound but don’t add appreciably to intelligibility.
What works well for voice is horrible for data. Although intelligibility is the primary concern with voice communications (most of the time), data transfer is principally oriented to bandwidth. The comparatively narrow bandwidth of the standard telephone channel limits the bandwidth of the modulated signal it can carry, which in turn limits the amount of digital information that can be squeezed down the phone line by a modem. The result is that dial-up telephone lines are poor data channels—but too often they are the only ones we have.
Engineers can use a variety of technologies to squeeze more data into a narrow-bandwidth channel, but they still face an ultimate limit on the amount of data they can squeeze through a tight channel such as an analog telephone line. This ultimate limit combines the effects of the bandwidth of the channel and the noise level in the channel. The greater the noise, the more likely it will be confused with the information that has to compete with it. This theoretical maximum data rate for a communication channel is called Shannon’s Limit. This fundamental law of data communications, discovered by Claude L. Shannon working at Bell Labs in 1948, states that the maximum number of digital bits that can be transmitted over a given communication path in one second can be determined from the bandwidth (W) and signal-to-noise ratio (S/N, expressed in decibels) by using the following formula:
In telephone circuits, the analog-to-digital converters used in telephone company central offices contribute the noise that most limits modem bandwidth. In creating the digital pulse-coded modulation (PCM) signal, they create quantization distortion, which produces an effective signal-to-noise ratio of about 36 dB. Quantization distortion results from the inability of the digital system with a discrete number of voltage steps (256 in the case of telephone company A/D converters) to exactly represent an analog signal that has an infinite number of levels. At this noise level, Shannon’s Limit for analog data on telephone lines is about 33,600 bits per second. Modern modems for dial-up data communications can be faster only by sidestepping some of the bandwidth issues.
Communications are supposed to be a two-way street. Information is supposed to flow in both directions. You should learn something from everyone you talk to, and everyone should learn from you. Even if you disregard the potential for success of such two-way communication, one effect is undeniable: It cuts the usable bandwidth of a data communication channel in one direction in half because the data going the other way requires its own share of the bandwidth.
With modems, such a two-way exchange of information is called duplex communications. Often it is redundantly called full-duplex. A full-duplex modem is able to simultaneously handle two signals, usually (but not necessarily) going in opposite directions, so it can send and receive information at the same time. Duplex modems use two carriers to simultaneously transmit and receive data, each of which has half the bandwidth available to it and its modulation.
The alternative to duplex communications is half-duplex. In half-duplex transmissions, only one signal is used. To carry on a two-way conversation, a modem must alternately send and receive signals. Half-duplex transmissions allow more of the channel bandwidth to be put to use but slow data communications because often a modem must switch between sending and receiving modes after every block of data crawls through the channel.
The term duplex is often mistakenly used by some communications programs for computers to describe echoplex operation. In echoplex mode, a modem sends a character down the phone line, and the distant modem returns the same character, echoing it. The echoed character is then displayed on the originating terminal as confirmation that the character was sent correctly. Without echoplex, the host computer usually writes the transmitted character directly to its monitor screen. Although a duplex modem generates echoplex signals most easily, the two terms are not interchangeable.
With early communications programs, echoplex was a critical setup parameter. Some terminal programs relied on modem echoplex to display your typing on the screen. If you had echoplex off, you wouldn’t see what you typed. Other terminal programs, however, displayed every character that went through the modem, so switching echoplex on would display two of every letter you typed, lliikkee tthhiiss. Web browsers don’t bother you with the need to select this feature. Most, however, work without echoplex.
To push more signal through a telephone line, some modems attempt to mimic full-duplex operation while actually running in half-duplex mode. Switching modems are half-duplex modems that reverse the direction of the signal at each end of the line in response to the need to send data. This kind of operation can masquerade as full-duplex because most of the time communications go only in one direction. You enter commands into a remote access system, and only after the commands are received does the remote system respond with the information you seek. Although one end is sending, the other end is more than likely to be completely idle.
On the positive side, switching modems are able to achieve a doubling of the data rate without adding any complexity to their modulation. However, the switching process itself is time-consuming and inevitably involves a delay because the modems must let each other know that they are switching. Because transmission delays across long-distance circuits are often a substantial fraction of a second (most connections take at least one trip up to a satellite and back down, a 50,000 mile journey that takes about a quarter of a second even at the speed of light), the process of switching can eat huge holes into transmission time.
Most software modem protocols require a confirmation for each block of data sent, meaning the modem must switch twice for each block. The smaller the block, the more often the switch must occur. Just one trip to a satellite would limit a switching modem with an infinitely fast data rate, using the 128-byte blocks of some early modem protocols, to 1024 bits per second at the two-switches-per-second rate.
Because of this weakness of switching modems, asymmetrical modems cut the waiting by maintaining a semblance of two-way duplex communications while optimizing speed in one direction only. These modems shoehorn in a lower speed channel in addition to a higher speed one, splitting the total bandwidth of the modem channel unequally.
Early asymmetrical modems were able to flip-flop the direction of the high-speed communications, relying on algorithms to determine which way is the best way. The modern asymmetrical technologies have a much simpler algorithm. Designed for Internet communications, they assume you need a greater data rate downstream (to you) than upstream (back to the server). This design is effective because most people download blocks of data from the Internet (typically Web pages rife with graphics) while sending only a few commands back to the Web server.
The latest V.90 modems operate asymmetrically at their highest speed. Cable modems and satellite connections to the Internet also use a variation on asymmetrical modem technology. These systems typically provide you with a wide bandwidth downlink from a satellite or cable system to permit you to quickly browse pages but rely on a narrow-channel telephone link—a conventional modem link—to relay your commands back to the network.
Getting the most from your modem requires making the best match between it and the connection it makes to the distant modem with which you want to communicate. Although you have no control over the routing your local phone company and long distance carrier give to a given call (or even whether the connection remains consistent during a given call), a modem can make the best of what it gets. Using line compensation, it can ameliorate some problems with the connection. Fallback helps the modem get the most from a substandard connection or one that loses quality during the modem link-up. Data compression helps the modem move more data through any connection, and error correction compensates for transitory problems that would result in minor mistakes in transmissions.
Although a long-distance telephone connection may sound unchanging to your ear, its electrical characteristics vary by the moment. Everything, from a wire swaying in the Wichita wind to the phone company’s automatic rerouting of the call through Bangkok when the direct circuits fill up, can change the amplitude, frequency, and phase response of the circuit. The modem then faces two challenges: not to interpret such changes as data and to maintain the quality of the line to a high-enough standard to support its use for high-speed transmission.
Under modern communications standards, modems compensate for variations in telephone lines by equalizing these lines. That is, two modems exchange tones at different frequencies and observe how signal strength and phase shift with frequency changes. The modems then change their signals to behave in the exact opposite way to cancel out the variations in the phone line. The modems compensate for deficiencies in the phone line to make signals behave the way they would have in absence of the problems. If, for example, the modems observe that high frequencies are too weak on the phone line, they will compensate by boosting high frequencies before sending them.
Modern modems also use echo cancellation to eliminate the return of their own signals from the distant end of the telephone line. To achieve this, a modem sends out a tone and listens for its return. Once it determines how long the delay is before the return signal occurs and how strong the return is, the modem can compensate by generating the opposite signal and mixing it into the incoming data stream.
When satellite rather than fiber optic technology dominated the long-distance telephone market, the half-second delay imposed by the long distance the signals traveled from earth to satellite and back created annoying “echoes” on the line. You spoke, and a fraction of a second later, you heard your own voice, delayed by the satellite hop, as it looped through the entire connection.
Switching to fiber optic long-distance connections minimizes the problem, at least to your ears. But modems can be confused even with the resulting short-delay echoes. A delayed pulse or phase-shift can sound like good data to a modem. To prevent such problems, modern modems use echo cancellation, which automatically subtracts the original signal from what the modem hears after the echo delay, thus canceling out the echo. To properly cancel the echo, the modem must be trained—during the initial handshake with a distant modem, it checks for echo and measures the delay. It later uses the discovered values for its echo cancellation.
Most modems use at most two carriers for duplex communications. These carriers are usually modulated to fill the available bandwidth. Sometimes, however, the quality of the telephone line is not sufficient to allow reliable communications over the full bandwidth expected by the modem, even with line compensation. In such cases, most high-speed modems incorporate fallback capabilities. When the top speed does not work, they attempt to communicate at lower speeds that are less critical of telephone line quality. A pair of modems might first try 56,000bps and be unsuccessful. They next might try 53,000 or switch from high-speed to conventional modem technology with a fallback to 33,600bps.
Most modems fall back and stick with the slower speed that proves itself reliable. Some modems, however, constantly check the condition of the telephone connection to sense for any deterioration or improvement. If the line improves, these modems can shift back to a higher speed.
Most modems rely on a relatively complex form of modulation on one or two carriers to achieve high speed. However, one clever idea (now relegated to a historical footnote by the latest modem standards) is the multiple-carrier modem, which uses relatively simple modulation on several simultaneous carrier signals. One of the chief advantages of this system comes into play when the quality of the telephone connection deteriorates. Instead of dropping down to the next incremental communications rate, thus generally cutting data speed in half, the multiple-carrier modems just stop using the carriers in the doubtful regions of the bandwidth. The communication rate may fall off just a small percentage in the adjustment. (Of course, it could dip by as much as a normal fallback modem as well.)
Although there’s no way of increasing the number of bits that can cross a telephone line beyond the capacity of the channel, the information-handling capability of the modem circuit can be increased by making each bit more meaningful. Many of the bits that are sent through the telecommunications channel are meaningless or redundant—they convey no additional information. By eliminating those worthless bits, the information content of the data stream is more intense, and each bit is more meaningful. The process of paring the bits is called data compression.
The effectiveness of compression varies with the type of data that’s being transmitted. One of the most prevalent data-compression schemes encodes repetitive data. Eight recurrences of the same byte value might be coded as two bytes, one signifying the value and the second the number of repetitions. This form of compression is most effective on graphics, which often have many blocks of repeating text. Other compression methods may strip out start, stop, and parity bits.
At one time, many modem manufacturers had their own methods of compressing data so that you needed two matched modems to take advantage of the potential throughput increases. Today, however, most modems follow international compression standards so that any two modems using the same standards can communicate with one another at compressed-data speeds. The most efficient of these international standards is called V.44.
These advanced modems perform the data compression on the fly in their own circuitry as you transmit your data. Alternately, you can precompress your data before sending it to your modem. Sort of like dehydrating soup, precompression (also known as file compression) removes the unnecessary or redundant parts of a file, yet allows the vital contents to be easily stored and reconstituted when needed. This gives you two advantages: The files you send and receive require less storage space because they are compressed, and your serial port operates at a lower speed for a given data throughput.
Note that once a file is compressed, it usually cannot be further compressed. Therefore, modems that use on-the-fly compression standards cannot increase the throughput of precompressed files. In fact, using one on-the-fly modem data-compression system (MNP5) actually can increase the transmission time for compressed files as compared to not using modem data compression.
Error Checking and Error Correction
Because all high-speed modems operate closer to the limits of the telephone channel, they are naturally more prone to data errors. To better cope with such problems, nearly all high-speed modems have their own built-in error-checking methods (which detect only transmission errors) and error-correction methods (which detect data errors and correct the mistakes before they get passed along to your computer). These error-checking and error-correction systems work like communications protocols, grouping bytes into blocks and sending cyclical redundancy checking information. They differ from the protocols used by communications software in that they are implemented in the hardware instead of your computer’s software. That means they don’t load down your computer when it’s straining at the limits of its serial ports.
It can also mean that software communications protocols are redundant and a waste of time. As mentioned before, in the case of switching modems, using a software-based communications protocol can be counterproductive with many high-speed modems, slowing the transfer rate to a crawl. Most makers of modems using built-in error-checking advise against using such software protocols.
All modem error-detection and error-correction systems require that both ends of the connection use the same error-handling protocol. In order that modems can talk to one another, a number of standards have been developed. Today, the most popular are MNP4 and V.42. You may also see the abbreviations LAPB and LAPM describing error-handling methods.
LAPB stands for Link Access Procedure, Balanced, an error-correction protocol designed for X.25 packet-switched services such as Telebit and Tymnet. Some high-speed modem makers adapted this standard to their dial-up modem products before the V.42 standard (described later) was agreed on. For example, the Hayes Smartmodem 9600 from Hayes Microcomputer Products included LAPB error-control capabilities.
Combining Voice and Data
Having but a single telephone line can be a problem when you need to talk as well as send data. In the old days, the solution was to switch. You’d type a message to the person at the other end of the connection, such as “Go voice,” pick up the telephone handset, and tell your modem to switch back to command mode so you could talk without its constant squeal.
In the early 1990s, several manufacturers developed the means of squeezing both data and voice down a single telephone line at the same time using special modem hardware. Three technologies—VoiceView, VoiceSpan, and DSVD—vied for market dominance.
Internet technology has made all three irrelevant. Instead of combining data and voice in modem hardware, the modern alternative is to combine them using software inside your computer. Your computer captures your voice with a microphone and digitizes it. Web software packages the voice information into packets that get sent to an ordinary modem exactly like data packets. At the other end of the connection, the receiving computer converts the voice packets back to audio to play through the computer’s speakers. The modem connection doesn’t care—or even know—whether its passing along packets of data or digitized audio.
Neither men nor modems are islands. Above all, they must communicate and share their ideas with others. One modem would do the world no good. It would just send data out into the vast analog unknown, never to be seen (or heard) again.
But having two modems isn’t automatically enough. Like people, modems must speak the same language for the utterances of one to be understood by the other. Modulation is part of the modem language. In addition, modems must be able to understand the error-correction features and data-compression routines used by one another. Unlike most human beings, who speak any of a million languages and dialects, each somewhat ill-defined, modems are much more precise in the languages they use. They have their own equivalent of the French Academy: standards organizations.
In the United States, the first standards were set long ago by the most powerful force in the telecommunications industry, which was the telephone company. More specifically, the American Telephone and Telegraph Company was the telephone company prior to is breakup announced on January 8, 1982, which resulted in seven local operating companies (in addition to AT&T). Before then, the Bell System created nearly all U.S. telephone standards, including two of the historically most important modem standards, Bell 103 and Bell 212A.
With the globalization of business and technology, communication standards have become international, and the onus to set new standards has moved to an international standards organization that’s part of the United Nations, the International Telecommunications Union (ITU) Telecommunications Standards Sector, which was formerly the Comite Consultatif International Telegraphique et Telephonique (in English, that’s International Telegraph and Telephone Consultative Committee). All the current standards for modems fall under the aegis of the ITU. You can purchase copies of the ITU standards from the organization’s Web site at www.itu.org.
Standards are important when buying a modem because they are your best assurance that a given modem can successfully connect with any other modem in the world. In addition, the standards you choose will determine how fast your modem can transfer data and how reliably it will work. The kind of communications you want to carry out will determine what kind of modem you need. If you’re just going to send files electronically between offices, you can buy two nonstandard modems and get more speed for your investment. But if you want to communicate with the rest of the world, you will want to get a modem that meets the international standards. Table 13.1 summarizes major modem speed standards.
Nearly all new modems now sold operate at a top speed set by the V.92 modem standard, the international standard for modem communications at 56,000 bits per second (or 56Kbps) across dial-up telephone lines. It is the highest speed modem standard in use today and possibly the highest speed true modems will ever achieve. Note that this standard is essentially the same as ITU V.90. The new designation does not indicate an increase in speed. The chief changes between V.90 and V.92 include improved connection setup and handshaking, so it takes a modem less time to set up a V.92 connection. The V.92 standard includes (and is compatible with) the V.90 standard, and modems matching the two standards will interconnect using V.90 technology.
Strictly speaking, V.92 is not a modem standard because, at its top speeds, it involves no modulation or demodulation. When line conditions are not favorable, however, it shifts back to analog technology to cope.
It is an asymmetrical standard, with a maximum upstream data rate of 48,000bps and a maximum downstream data rate of about 56,000bps. It uses switching technology and cannot send and receive simultaneously. In theory, in many connections, its downstream falls short of its maximum rate because of line conditions. In addition, telephone regulations at one time prohibited the V.92 top operating speed because the last bit of speed pushed the power level on the telephone line above the allowed standards. Most sources listed a top practical speed of 53,000 bits per second, but your actual connection speed is now determined by line conditions and not law.
V.92 takes advantage of the digital technology already in use to shift voice calls across long-distance digital lines. The telephone standard for voice is to sample analog signals 8000 times a second using an eight-bit digital code. V.92 translates the digital code into voltage levels on the telephone line from your local phone company’s digital switch to your home or office. Encoding digital information as voltage levels is a technology called pulse amplitude modulation (PAM). Because the V.92 standard requires modems to operate at frequencies in excess of normal voice circuits, its range is limited—V.92 connections require your modem to have no more than three miles of telephone wire between it and your telephone company’s switch.
It should take 256 voltage levels, which are technically called quantization levels, to encode the eight-bit data stream. However, noise and line conditions can often mask changes of 1/256th of the voltage on an ordinary telephone line. To avoid problems with noise, the V.92 system uses only 128 quantization levels to encode data, allowing a seven-bit digital code with an 8000Hz sampling rate or a 56,000bps data rate.
As originally developed as the V.90 standard, modems used the PAM system only for downstream communications (that is, from the telephone company’s switch to your modem). Upstream data (from your modem back to the phone company) relied on conventional modem technology under the V.34 standard, although limited to 31,200bps. The underlying assumption was that the phone company was better able to control digital signals and keep them within the required limits. Moreover, only the telephone company had direct access to the digital connection with long-distance trunk lines.
With the advent of the V.92 standard, this situation changed. Under V.92, modems use a technology called V.PCM upstream to let you use PAM to send data to other modems and services at a rate of 48,000 bits per second. The only difference between upstream and downstream is that the upstream signal is limited to only 64 quantization levels, thus allowing a six-bit code. The fewer voltage levels means that modem manufacturing inconsistencies and installation differences are less likely to cause signals exceeding the levels allowed by law.
To reduce the time needed for modems to connect, V.92 uses a technology called QuickConnect that takes advantage of the digital nature of the telephone system from the central office onward. The only part of the connection that needs line compensation is the local phone loop between where you use your modem and the telephone company’s central office. Unless you’re traveling with a notebook computer, this connection doesn’t often change (and even when you’re traveling, you’re likely to use the same hotel phone for several connections at a time). Consequently, QuickConnect checks and remembers the settings for the local telephone loop and tries to reuse them if possible.
In truth, QuickConnect does not reduce the time required to determine the quality of a connection and compensate for it. Rather, it remembers the quality of your last connection under the assumption that you’ll use the same telephone line for consecutive calls. The QuickConnect system stores in nonvolatile memory the equalizer and echo-cancellation settings as well as the digital characteristics of the line. When you place a subsequent call, your modem first examines the tone from the distant modem and compares it to the setting in memory. In case of a match, the modem starts with the stored settings to make a fast connection. If the modem detects a substantial change (such as you’re using a notebook computer in a different hotel room), it walks through the standard V.90 handshaking procedure. The first time you use a V.92 modem in a given location, it will use the standard V.90 handshaking, requiring 20 to 30 seconds to adjust for the local phone line. Subsequent QuickConnect handshaking usually takes about half as long.
Recognizing how modems are used in normal telephone systems, V.92 explicitly recognizes the need to be able to pause modem communications so you can take another call. A feature called Modem-on-Hold lets you suspend data transfer without looking at your modem connection. In effect, you can put your modem on hold, take another call, and then return to your modem connection.
The V.92 standard is regarded as a refinement of V.90, which evolved out of two competing 56Kbps systems. K56flex is a proprietary technology that was independently developed and initially marketed as two different and incompatible systems by Rockwell and Lucent Technologies. In November, 1996, the two companies agreed to combine their work into a single standard. x2 is a proprietary technology developed by U.S. Robotics. Both K56flex and x2 used the same PAM digital encoding as was adopted for V.90 and V.92. They differed from each other (and the standard) only in the handshaking used to set up the connection.
In addition to speed standards, several other standards have been used for data compression and error correction. Before the widespread adoption of international standards, many U.S. companies used technologies developed by Microcom Corporation and formalized as Microcom Networking Protocol Standards Levels 1 through 9. Currently, however, the dominant modem compression and error-control standards are those set by the ITU. These include V.42, a world-wide error-correction standard (which also incorporates MNP4 as an “alternative” protocol), V.42bis (data compression that can yield compression factors up to four, potentially quadrupling the speed of modem transmissions), and V.44, an improved compression system.
A modem is a signal converter that mediates the communications between a computer and the telephone network. In function, a modern computer modem has five elements: interface circuitry for linking with the host computer; circuits to prepare data for transmission by adding the proper start, stop, and parity bits; modulator circuitry that makes the modem compatible with the telephone line; a user interface that gives you command of the modem’s operation; and the package that gives the modem its physical embodiment.
For a modem to work with your computer, the modem needs a means to connect to your computer’s logic circuits. At one time, all modems used a standard or enhanced serial port to link to the computer. However, because the standard serial port tops out at a data rate that’s too slow to handle today’s fastest modems—the serial limit is 115,200 bits per second, whereas some modems accept data at double that rate—modem-makers have developed parallel-interfaced modems.
All modems, whether installed outside your computer, in one of its expansion slots, or in a PCMCIA slot, make use of a serial or parallel communications port. In the case of an internal computer modem, the port is embedded in the circuitry of the modem, and the expansion bus of the computer itself becomes the interface.
With an external modem, this need for an interface (and the use of a port) is obvious because you fill the port’s jack with the plug of a cable running off to your modem. With an internal modem, the loss is less obvious. You may not even detect it until something doesn’t work because both your modem and your mouse (or some other peripheral) try to use the same port at the same time.
In the case of serial modems, this interface converts the parallel data of your computer into a serial form suitable for transmission down a telephone line. Modern modems operate so fast that the choice of serial port circuitry (particularly the UART) becomes critical to achieving the best possible performance.
The serial and parallel ports built into internal modems are just like dedicated ports of the same type. They need an input/output address and an interrupt to operate properly. The Plug-and-Play system assigns these values to the modem during its configuration process. (Older modems required you to select these values with jumpers or switches.)
Ordinarily you don’t need to know or bother with these values. Some communications software, however, may not mesh perfectly with the Windows system. It may demand you tell it the port used by your modem. The modem’s properties sheet lists this value. You can check it under Windows by clicking the modem icon in Control Panel and then clicking the Properties tab.
Modern modem communications require that the data you want to send be properly prepared for transmission. This pre-transmission preparation helps your modem deliver the highest possible data throughput while preventing errors from creeping in.
Most modem standards change the code used by the serial stream of data from the computer interface into code that’s more efficient (for example, stripping out data-framing information for quicker synchronous transfers). The incoming code stream may also be analyzed and compressed to strip out redundant information. The modem may also add error-detection or error-correction codes to the data stream.
At the receiving end, the modem must debrief the data stream and undo the compression and coding of the transmitting modem. A micro controller inside the modem performs these functions based on the communications standard you choose to use. If you select a modem by the communications standards it uses, you don’t have to worry about the details of what this micro controller does.
The heart of the modem is the circuitry that actually converts the digital information from your computer into analog-compatible form. Because this circuitry produces a modulated signal, it is called a modulator.
The fourth element in the modem is what you see and hear. Most modems give you some way of monitoring what they do either audibly with a speaker or visually through a light display. These features don’t affect the speed of the modem or how it works but can make one modem easier to use than another. Indicator lights are particularly helpful when you want to troubleshoot communication problems.
Finally, the modem needs circuitry to connect with the telephone system. This line interface circuitry (in telephone terminology, a data access arrangement) boosts the strength of the modem’s internal logic-level signals to a level matching that of normal telephone service. At the same time, the line interface circuitry protects your modem and computer from dangerous anomalies on the telephone line (say, a nearby lightning strike), and it protects the telephone company from odd things that may originate from your computer and modem (say, a pulse from your computer in its death throes).
From your perspective, the line interface of the modem is the telephone jack on its back panel. Some modems have two jacks so that you can loop through a standard telephone. By convention, the jack marked “Line” connects with your telephone line; the jack marked “Phone” connects to your telephone.
Over the years, this basic five-part modem design has changed little. But the circuits themselves, the signal-processing techniques that they use, and the standards they follow have all evolved to the point that modern modems can move data as fast as the theoretical limits that telephone transmission lines allow.
Internal modems plug into an expansion slot in your computer. The connector in the slot provides all the electrical connections necessary to link to your computer. To make the modem work, you only need to plug in a telephone line. The internal modem draws power from your computer, so it needs no power supply of its own. Nor does it need a case. Consequently, the internal modem is usually the least expensive at a given speed. Because internal modems plug into a computer’s expansion bus, a given modem is compatible only with computers using the bus for which it was designed. You cannot put a computer’s internal modem in a Macintosh or workstation.
External modems are self-contained peripherals that accept signals from your computer through a serial or parallel port and also plug into your telephone line. Most need an external source of power, typically a small transformer that plugs into a wall outlet and—through a short, thin cable—into the modem. At a minimum, then, you need a tangle of three cables to make the modem work. You have two incentives to put up with the cable snarl. External modems can work with computers that use any architecture as long as the computer has the right kind of port. In addition, external modems usually give you a full array of indicators that can facilitate troubleshooting.
Pocket modems are compact external modems designed for use with notebook computers. They are usually designed to plug directly into a port connector on your computer, eliminating one interface cable. Many eliminate the need for a power supply and cable by running from battery power or drawing power from your computer or the telephone line.
PC Card modems plug into that PCMCIA slots that are typically found in notebooks. They combine the advantage of cable-free simplicity of internal modems with the interchangeability of external modems (the PCMCIA interface was designed to work with a variety of computer architectures). The confines of the PCMCIA slot also force PC Card modems to be even more compact than pocket modems. This miniaturization takes its toll in higher prices, however, although the ability to quickly move one modem between your desktop and portable computer can compensate for the extra cost.
The confines of a PCMCIA slot preclude manufacturers from putting a full-size modular telephone jack on PC Card modems. Modem-makers use one of two workarounds for this problem. Most PC Card modems use short adapter cables with thin connectors on one end to plug into the modem and a standard modular jack on the other. Other PC Card modems use the X-Jack design, developed and patented by Megahertz Corporation (now part of 3Com). The X-Jack pops out of the modem to provide a skeletal phone connector into which you can plug a modular telephone cable. The X-Jack design is more convenient because you don’t have to carry a separate adapter with you when you travel. On the other hand, the X-Jack makes the modem more vulnerable to carelessness. Yank on the phone cable, and it can break the X-Jack and render the modem useless. Yanking on an adapter cable will more likely pull the cable out or damage only the cable. On the other hand, the connectors in the adapter cables are also prone to invisible damage that can lead to unreliable connections.
The principal functional difference between external (including pocket) and internal (including PC Card) modems is that the former have indicator lights that allow you to monitor the operation of the modem and the progress of a given call. Internal modems, being locked inside your computer, cannot offer such displays. Some software lets you simulate the lights on your monitor, and Windows will even put a tiny display of two of these indicators on your task bar. These indicators can be useful in troubleshooting modem communications, so many computer people prefer to have them available (hence, they prefer external modems).
The number and function of these indicators on external modems vary with the particular product and the philosophy of the modem-maker. Typically you’ll find from four to eight indicators on the front panel of a modem, as shown in Figure 13.1.
The most active and useful of these indicators are Send Data and Receive Data. These lights flash whenever the modem sends out or receives in data from the telephone line. They let you know what’s going on during a communications session. For example, if the lights keep flashing away but nothing appears on your monitor screen, you know you are suffering a local problem, either in your computer, its software, or the hardware connection with your modem. If the Send Data light flashes but the Receive Data light does not flicker in response, you know that the distant host is not responding.
Carrier Detect indicates that your modem is linked to another modem across the telephone connection. It allows you to rule out line trouble if your modem does not seem to be getting a response. This light glows throughout the period your modem is connected.
Off-Hook glows whenever your modem opens a connection on your telephone line. It lights when your modem starts to make a connection and continues to glow through dialing, negotiations, and the entire connection.
Terminal Ready glows when the modem senses that your computer is ready to communicate with it. When this light is lit, it assures you that you’ve connected your modem to your computer and that your computer’s communications software has properly taken control of your serial port.
Table 13.2 summarizes the mnemonics commonly used for modem indicators and their functions.
Today, all of the international telephone network is digital, with the exception of one wire—the one running from the telephone company central office to your home. This local loop, sometimes called the last mile, is the same stuff that Alexander Graham Bell experimented with and uses the same technology he developed—a pair of copper wires designed to carry analog voice signals. Eventually, this local loop will be replaced with a newer connection technology, such as coaxial cable or fiber optics carrying only digital data.
Businesses already have this option available to them. Modern cable television systems offer digital data communications capabilities as well, although dial-up voice circuits remain the province of the telephone company (mostly for legal reasons). Even so, there’s little doubt you’ll eventually shift from analog to digital services for all of your telecommunications needs.
You can already get all-digital circuits for your data. The only question is who will provide the connection. Three technologies can provide you with a high-speed all-digital link—telephone, cable, and satellite.
All three work. All three deliver speeds that make modem connections seem like they are antiquated. In fact, the most important limiting factor is availability. Only satellite services can promise a link in any part of the United States (and most of the world). The others depend on your local telephone company or cable provider upgrading its facilities to handle digital subscriber services.
One key player in the supply of digital telecommunications services is quite familiar—the telephone company. Beyond traditional POTS, telephone companies have developed a number of all-digital communication services. Some of these have been around for a while, aimed at business users with heavy data needs. Several new, all-digital services are aimed directly at you as an individual consumer.
The range of digital services supplied by telephone companies is wide and spans a range of data rates. Table 13.3 lists many of these and their maximum data rates.
You will still talk on the telephone for ages to come (if your other family members give you a chance, of course), but the nature of the connection may finally change. Eventually, digital technology will take over your local telephone connection. In fact, in many parts of America and the rest of the world, you can already order a special digital line from your local telephone company and access all-digital switched systems. You get the equivalent of a telephone line, one that allows you to choose any conversation mate who’s connected to the telephone network (with the capability of handling your digital data, of course), as easily as dialing a telephone.
The basic high-speed service provided by the telephone company is called T1, and its roots go back to the first days of digital telephony in the early 1960s. The first systems developed by Bell Labs selected the now-familiar 8KHz rate to sample analog signals and translate them into eight-bit digital values. The result was a 64Kbps digital data stream. To multiplex these digital signals on a single connection, Bell’s engineers combined 24 of these voice channels together to create a data frame 193 bits long, the extra bit length defining the beginning of the frame. The result was a data stream with a bit rate of 1.544Mbps. Bell engineers called the resulting 24-line structure DS1. AT&T used this basic structure throughout its system to multiply the voice capacity of its telephone system, primarily as trunk lines between exchanges.
As telephone demand and private business exchanges (PBXs) became popular with larger businesses, the telephone company began to offer T1 service directly to businesses. As digital applications grew, T1 became the standard digital business interconnect. Many Web servers tie into the network with a T1 line.
A key feature of the DS1 format was that it was compatible with standard copper telephone lines, although requiring repeaters (booster amplifiers) about every mile. The signal itself is quite unlike normal analog telephone connections, however, and that creates a problem. Its signal transmission method is called Alternate Mark Inversion (AMI), a formatting code for T1 transmissions over twisted-pair copper cable. T1 transmissions are in bipolar form. AMI represents a zero (or space) by the absence of a voltage; a one (or mark) is represented by a positive or negative pulse, depending on whether the preceding one was negative or positive (that is, marks are inverted on an alternating basis). This encoding system generates a signal with a bandwidth about equivalent to its data rate, 1.5MHz. This high-speed signal creates a great deal of interference, so much that two T1 lines cannot safely cohabitate in the 50-pair cables used to route normal telephone services to homes.
Outside of the United States, the equivalent of T1 services is called E1. Although based on the same technology as T1, E1 combines 30 voice channels with 64Kbps bandwidth to create a 2.048Mbps digital channel.
Serious Web surfers dream of having a dedicated T1 line. The cost, however, is prohibitive. Installation is often thousands of dollars and monthly charges may be a thousand dollars, sometime more. Typically, your Internet Service Provider has an T1 (or better) connection and divides it up, giving each customer a single modem-slice.
The primary problem with T1 is the interference-causing modulation system it uses, one based on 1960s technology. Using the latest modulation techniques, the telecommunications industry has developed a service called High data rate Digital Subscriber Line (HDSL) that features the same data rate as T1 or E1 but requires a much narrower bandwidth, from 80 to 240KHz. One basic trick to the bandwidth-reduction technique is splitting the signal across multiple phone lines. For T1 data rate, the service uses two lines; for E1, three. Besides reducing interference, the lower data rate allows longer links without repeaters, as much as 12,000 feet.
HDSL delivers high-speed data networking, up to 1.544Mbps over two copper pairs and up to 2.048Mbps over three pairs, at a maximum range of 20,000 feet (about 3.8 miles or 6.1 km) from a central office. It is similar to Symmetrical Digital Subscriber Line (discussed later) and has symmetrical transmission capabilities. Most T1 lines installed today utilize this technology.
Unfortunately the “subscriber” in the name of the standard was not meant to correspond to you as an individual. It fits into the phone company scheme of things in the same place as T1—linking businesses and telephone company facilities.
The service you’re most likely to buy from your telephone company goes under the name Digital Subscriber Line. It uses your ordinary telephone wires to carry high-speed digital signals. In fact, today’s technologies let you use both the analog phone service and the digital capacity of the same wires at the same time. The high frequencies of the digital signal are easily split from the low frequencies of the analog signal, so one wire can do double duty. When telephone companies first introduced this piggyback service, they called it G.Lite, but it has become the standard for residential DSL.
G.Lite is a special case of a more generalized service called Asymmetrical Digital Subscriber Line or ADSL. It is asymmetrical because it offers a higher downstream data rate from the server compared to its upstream rates, from you back to the server.
When telephone companies first offered DSL services, they needed to send out a technician on each job to install the splitters required to separate the analog and digital signals. The need for a technician added hundreds of dollars to the cost of setting up an ADSL connection. To eliminate the need for the splitter and create a common consumer standard for ADSL, several manufacturers in the telecommunications industry banded together to create the Universal ADSL Working Group, the organization that defined the G.Lite standard under the formal designation G.992.2.
ADSL can move downstream data at speeds up to 8Mbps. Upstream, however, the maximum rate is about 640Kbps to 1Mbps. ADSL doesn’t operate at a single rate as does T1. Its speed is limited by distance, with longer distances imposing greater constraints. It can push data downstream at the T1 rate for up to about 18,000 feet from the central office. At half that distance, its downstream speed potential approaches 8.5Mbps. To distinguish the more general form of ADSL from G.Lite, this full-fledged ADSL is sometimes called ADSL Full Rate or G.dmt. Its official designation is G.992.1.
G.Lite is one variety of ADSL. G.Lite allows for a bandwidth downstream of up to 1.544Mbps. Upstream the asymmetrical system allows for a bandwidth of up to 512Kbps. The maximum length of a G.Lite connection stretching between the central office and your home or business is 18,000 feet.
Table 13.4 summarizes the downstream speeds and maximum distances possible with ADSL technology.
Another kind of DSL offers the same speed in both the upstream and downstream directions. Termed Symmetrical Digital Subscriber Line (SDSL), this kind of service is best suited for companies offering Web services from their own servers. The data speed of the SDSL system ranges from 160Kbps up to 1.544Mbps. The maximum usable rate depends on the distance of the subscriber from the central office. The lowest speed occurs at a maximum range of the SDSL system, which is 24,000 feet (about 4.5 miles or 7.2 km).
The acronym SDSL has been used in the past to describe another service called Single-line Digital Subscriber Line, which was an alternative to multiline data services for high-speed operation.
The modulation system used by all DSL services operates at frequencies above the baseband used by the ordinary telephone service, which is why a single line can carry high-speed digital signals and ordinary telephone signals simultaneously. In typical DSL implementations, the DSL signals start at about 50KHz, leaving the lower frequencies for carrying conventional voice signals.
The splitter that divides the signal between voice and data on the subscriber’s premises is the equivalent of a stereo speaker’s crossover; the splitter combines a high-pass filter to extract a data-only signal and a low-pass filter to extract the voice-only signal. Using the G.Lite system, you install the required splitters (sometimes called filters) by simply plugging them in using modular plugs and jacks like ordinary telephone equipment.
The next step above ADSL is the Very-high-data-rate Digital Subscriber Line (VDSL). A proposal only, the service is designed to initially operate asymmetrically at speeds higher than ADSL but for shorter distances, potentially as high as 51.84Mbps downstream for distances shorter than about 1000 feet, falling to one-quarter that at about four times the distance (12.86Mbps at 4500 feet). Proposed upstream rates range from 1.6Mbps to 2.3Mbps. In the long term, developers hope to make the service symmetrical. VDSL is designed to work exclusively in an ATM network architecture. As with ADSL, VDSL can share a pair of wires with an ordinary telephone connection or even ISDN service.
Switched Data Services 56 (sometimes shortened to Switched-56) is an archaic connection system that yielded a single digital channel capable of a 56Kbps data rate—the same as with a modem but with true digital signals. The Switched-56 signals traveled through conventional copper twisted-pair wiring (the same old stuff that carries your telephone conversations). For most telephone companies, it was an interim service to bridge the gap between POTS and ISDN service areas.
With Switched-56 you needed special head-end equipment—the equivalent of a modem—to link the wire to your computer. To take advantage of the connection, you also needed to communicate with someone who also had SDS 56 service.
In some locales, SDS 56 was no more expensive than an ordinary business telephone line. Installation costs, however, could be substantially higher (PacBell, for example, at one time charged $500 for installation), and some telephone companies added extra monthly maintenance charges in addition to the normal dial-up costs. With modern modems promising the same speed with no extra charges, little wonder Switched-56 gets discussed in the past tense.
The initials stand for Integrated Services Digital Network, a first attempt at bringing true digital communications to the home through existing telephone lines. Although the service is still available, it’s essentially irrelevant because DSL offers more speed at about the same cost.
ISDN predates DSL. Its start came in November 1992 when AT&T, MCI, and Sprint embraced a standard they called ISDN-1. Today, two versions of ISDN are generally available. The simplest is the Basic Rate Interface (BRI), which takes advantage of the copper twisted-pair wiring that’s already in place, linking homes and offices to telephone exchanges. Instead of a single analog signal, an ISDN line uses what is called “2B1Q line coding” to carry three digital channels: two B (for Bearer) channels that can carry any kind of data (digitally encoded voice, fax, text, and numbers) at 64,000bps, and a D (or Delta) channel, operating at 16,000bps, that can carry control signals and serve as a third data channel. The three channels can be independently routed to different destinations through the ISDN system.
The maximum distance an ISDN line can stretch from the central office is 18,000 feet (about 3.4 miles or 5.5 km). To accommodate longer runs, this distance can be doubled by adding a repeater in the middle of the line. A repeater is an amplifier that regenerates the digital signals, erasing the signal distortion that arises on long lines.
A single BRI wire enables you to transfer uncompressed data bidirectionally at the 64,000bps rate, exactly like a duplex modem today but with higher speed and error-free transmission, thanks to its all-digital nature. Even during such high-speed dual-direction connections, the D channel would still be available for other functions.
The more elaborate form of ISDN service is called the Primary Rate Interface (PRI). This service delivers 23 B channels (each operating at 64,000 bits per second) and one D channel (at 16,000 bits per second). As with normal telephone service, ISDN service is billed by time in use, not the amount of data transmitted or received.
The strength of BRI service is that it makes do with today’s ordinary twisted-pair telephone wiring. Neither you nor the various telephone companies need to invest the billions of dollars required to rewire the nation for digital service. Instead, only the central office switches that route calls between telephones (which today are mostly plug-in printed circuit boards) need to be upgraded.
The chief performance limit on telephone service is the twisted-pair wire that runs from the central office to your home or business. Breaking through its performance limits would require stringing an entirely new set of wires throughout the telephone system. Considering the billions of dollars invested in existing twisted-pair telephone wiring, the likelihood of the telephone company moving into a new connection system tomorrow is remote.
Over the past two decades, however, other organizations have been hanging wires from poles and pulling them underground to connect between a third to a half of the homes in the United States—cable companies. The coaxial cables used by most such services have bandwidths a hundred or more times wider than twisted pair. They regularly deliver microwave signals to homes many miles from their distribution center.
Tapping that bandwidth has intrigued cable operators for years, and the explosive growth of the Internet has set them salivating. The advent of digital cable has made Web connections an option for most cable subscribers.
The cable television system differs from the international telephone system in several ways. Most cable television systems are local. They are meant to cover only a limited geographic range. Cable systems do not interconnect. Each cable operator plucks the signals it needs from the air, either from distant broadcast stations or from satellite. There is no great cable web that allows you to directly link with another cable user anywhere in the world.
Moreover, cable systems are designed differently. Telephone systems are point to point, caller to caller, one on one. Cable systems are designed for broadcast in a one-to-many fashion. Cable systems send the same signals to each of their subscribers. The wiring for telephone systems resembles a star with a center hub (the central office) and individual nodes, each connected directly to the hub with its own (albeit low bandwidth) wire. The cable system is a spine—one big, wide bandwidth wire carrying signals for all subscribers, each of which taps into the same spine for the same signals.
The essence of the cable design is that all users share the same bandwidth. The cable is piled with as many as 500 television channels—that’s a full 3GB of bandwidth-ignoring guard bands. When operators put Internet signals on the cable system, users have to share that bandwidth, too. How wide a slice of bandwidth each user gets depends primarily on how many users are sharing. If you’re the only one to log on to the Internet on your cable system, you can get speeds that T1 users would envy. Log on when Microsoft offers free downloads of a new version of Internet Explorer or a beer company runs an online bikini competition, and you’re apt to get more free beer than bandwidth through the cable (that is, about zero). With a high-speed DSL telephone line, you are guaranteed bandwidth. With cable, you share and take your chances. You could do better—or much, much worse. In technical terms, the telephone system guarantees Quality of Service (QoS). The cable system may not (although some operators are making QoS guarantees to lure you over to coaxial cable).
Cable systems allow for individual addressing of subscribers’ equipment. Each cable box has an electronic serial number that the cable operator’s equipment can address individually (for example, to alter the services you are authorized to receive). But this individual addressing does not provide an individual channel to your home. The signals to control your cable box are broadcast to all subscribers on the cable system. Only your box with the correct serial number responds to commands sent to it.
During the first few years of cable-based data services, the one element lacking was standardization. Each cable operator used its own equipment designs to adapt data signals to the cable medium. The adapters used by different cable operators were incompatible, as were the data signals from the head-end, the cable company’s equivalent of the telephone company’s central office. Although such proprietary designs gave cable companies a measure of security—they helped thwart widespread hacking—they also made equipment more expensive and employee training more difficult.
Several efforts at developing cable standards started in the mid–1990s. The first major effort started at the IEEE, which formed a new working group called the Cable TV Media Access Control and Physical Protocol Working Group in May 1994, to define a standard for cable modems. The IEEE group members had difficulty agreeing to any single standard, however, and the group missed its original target of December 1995, for publishing a specification.
An impatient cable industry got tired of waiting for engineers to come up with a standard, so a group of cable television operators, including such major players as Comcast, Cox, MSOs, TCI, and Time Warner, started their own standardization effort. The companies formed an independent limited partnership called Multimedia Cable Network System Partners, Ltd. (usually shortened to MCNS), which Continental Cablevision and Rogers Cablesystems then joined. In little more than a year, the partnership sifted through 12 proposals to produce a draft standard that it published as the Data Over Cable Service Interface Specification (DOCSIS) in March, 1997. The industry research organization CableLabs took over the administration of the specification and developed a compliance program. By March 1998, the group had developed an interoperability certification program and, in March 1998, the ITU endorsed the DOCSIS standard as an official standard ITU J.112.
A revised version of the DOCSIS specification, version 1.1, was released in April 1999. The new standard better defined signal parameters to guarantee the bandwidth of the signal and minimize delays. In addition, the revision added standards for new services, including voice-over-Internet telephones and constant bit-rate services. A European version of DOCSIS, called EuroDOCSIS, is used with the different television standards prevailing there. Another update, to DOCSIS 2.0, is in the works. The complete DOCSIS specification is available online at www.cablemodem.com/specification.
DOCSIS fits data onto cable by taking over a single television channel for downstream data and a second channel for upstream data. Each television channel has a bandwidth of 6MB. In general, the system can use any channel in the VHF or UHF ranges (from 50 to 864MHz) for downstream data, but upstream data is restricted to lower frequencies (5 to 42MHz), which makes the adapters less expensive to manufacturer.
The DOCSIS standard allows cable operators flexibility in choosing either 64- or 256-state quadrate amplitude modulation for the downstream data signals they provide. Typically cable operators use 64-state QAM, with which a single 6MHz downstream television channel can carry data at a rate of about 27Mbps. Upstream, most cable operators have a choice of 16-state QAM or quadrature phase-shift keying, a more robust but slower modulation scheme most choose to ride over the higher noise levels prevalent at lower frequencies on cable. In this asymmetrical system, cable operators often limit upstream bandwidth from individual users, often to as little as 320Kbps, although the standard allows upstream rates as high as 10Mbps. At low-demand times, you might score downstream bandwidth approaching the full 27Mbps rate. Typically, however, cable services deliver downstream data at 1 to 3Mbps.
To allow multiple users to share the same bandwidth, DOCSIS uses time division multiple access (TDMA) technology. That is, each user gets a fraction of the total bandwidth in which to transmit and receive data. To gain access to the cable, DOCSIS uses the Ethernet MAC (Media Access Control) layer.
DOCSIS includes provisions for each user to have a 14-bit subscriber ID, which allows cable operators to individually tailor service much as they do premium cable television channels. For example, individual users may be assigned different bandwidths.
Before the wide adoption of DOCSIS, some cable companies deployed telco-return modems, which used the cable company’s high-speed coaxial cable for downstream data but made the upstream link through a conventional phone line (with all its bandwidth constraints). With the move to DOCSIS, most cable modems now use only a cable connection.
Because all signals ride across the same coaxial cable, all your neighbors have access to the packets of data you send and receive. To maintain privacy, your cable modem automatically encrypts everything you send and receive using the Data Encryption Standard (DES) algorithm.
The same technology used by direct-broadcast satellite television, through which you can grab viewable video directly from orbiting satellites, also works for data. Using a small parabolic antenna pointed at a geosynchronous satellite orbiting about 24,000 miles away, you can tap into the Internet at speeds well beyond the capabilities of dial-up telephone connections. Instead of television signals, the satellite simply beams a stream of data down to earth.
The leading satellite service, DirecPC from Hughes Electronics, initiated service in 1997. It bills itself as the fastest Internet service available nationwide. Although slower than either DSL or cable modems, it holds the advantage of availability—anywhere you can see the southern sky, you can make a connection to DirecPC.
Satellite systems are inherently asymmetrical. You don’t transmit your needs to the satellite—doing so would require an uplink and a much larger antenna. Instead, you use the satellite connection only for a downlink, to receive data. To send data, your computer connects to the Web through a conventional dial-up modem. The satellite-based downlink operates at 400Kbps while your phone-based uplink struggles along at modem speed, 14.4 to 56Kbps.
Satellites best fit a broadcast model. That is, they dump out their signals across wide areas for consumption by the multitudes rather than directly targeting individuals. By limiting the downlink bandwidth to 400Kbps, they maximize the number of subscribers that can share the system. In addition, DirecPC attempts to maximize the speed and usefulness of its product using push technology. The system pushes out selected Web and newsgroup information, and your computer captures it as it is sent out, spooling the data to disk. When you want to access one of the pushed Web sites or newsgroups, you can read it almost instantly from the cache. So your computer won’t clog up by trying to cache the entire Internet, the system allows you to choose which sites and groups to cache locally.
DirecPC uses a 21-inch elliptical antennae designed for roof mounting. The DirecPC antenna is a single-purpose device and can be used only for data, not satellite television reception. The system also requires a receiver (sometimes called a modem) that may be installed as an expansion board inside your computer or as a standalone external peripheral.