52 Point-to-Point Protocols and Links Chap.2 on the other hand,is that it asserts that with the use of error-correction coding,any rate less than C can be achieved with arbitrarily small error probability.The capacity of voice-grade telephone channels is generally estimated to be about 25,000 bps,indicating that the data rates of voice-grade modems are remarkably close to the theoretical limit. High-speed modems generally maintain signal timing,carrier phase,and adaptive equalization by slowly updating these quantities from the received signal.Thus.it is essential for these modems to operate synchronously;each T seconds,k bits must enter the modem,and the DLC unit must provide idle fill when there are no data to send.For low-speed modems,on the other hand,it is permissible for the physical channel to become idle in between frames or even between individual characters.What is regarded as high speed here depends on the bandwidth of the channel.For modems on 3 kHz voice-grade channels,for example,data rates of 2400 bps or more are usually synchronous,whereas lower rates might be intermittent synchronous or character asynchronous.For a coaxial cable,on the other hand,the bandwidth is so large that intermittent synchronous data rates of 10 megabits per second (Mbps)or more can be used.As mentioned before,this is essential for multiaccess systems such as Ethernet. 2.2.6 Frequency-and Time-Division Multiplexing In previous subsections,bandwidth constraints imposed by filtering on the physical chan- nel and by noise have been discussed.Often,however,a physical channel is shared by multiple signals each constrained to a different portion of the available bandwidth;this is called frequency-division multiplexing(FDM).The most common examples of this are AM,FM,and TV broadcasting,in which each station uses a different frequency band.As another common example,voice-grade channels are often frequency multiplexed together in the telephone network.In these examples.each multiplexed signal is constrained to its own allocated frequency band to avoid interference (sometimes called crosstalk)with the other signals.The modulation techniques of the preceding subsection are equally applicable whether a bandwidth constraint is imposed by FDM or by channel filtering. FDM can be viewed as a technique for splitting a big channel into many little channels.Suppose that a physical channel has a usable bandwidth of It hertz and we wish to split it into m equal FDM subchannels.Then,each subchannel has W/in hertz available,and thus W/m available quadrature samples per second for sending data (in practice,the bandwidth per subchannel must be somewhat smaller to provide guard bands between the subchannels,but that is ignored here for simplicity).In terms of Eq.(2.13). each subchannel is allotted 1/m of the overall available signal power and is subject to I/m of the overall noise,so each subchannel has 1/m of the total channel capacity. Time-division multiplexing (TDM)is an alternative technique for splitting a big channel into many little channels.Here,one modem would be used for the overall bandwidth W.Given m equal rate streams of binary data to be transmitted,the m bit streams would be multiplexed together into one bit stream.Typically,this is done by sending the data in successive frames.Each frame contains m slots,one for each bit stream to be multiplexed;a slot contains a fixed number of bits,sometimes 1,sometimes 8,and sometimes more.Typically,each frame also contains extra bits to help the
52 Point-to-Point Protocols and Links Chap. 2 on the other hand, is that it asserts that with the use of error-correction coding, any rate less than C can be achieved with arbitrarily small error probability. The capacity of voice-grade telephone channels is generally estimated to be about 25,000 bps, indicating that the data rates of voice-grade modems are remarkably close to the theoretical limit. High-speed modems generally maintain signal timing, carrier phase, and adaptive equalization by slowly updating these quantities from the received signal. Thus, it is essential for these modems to operate synchronously; each T seconds, k bits must enter the modem, and the DLC unit must provide idle fill when there are no data to send. For low-speed modems, on the other hand, it is permissible for the physical channel to become idle in between frames or even between individual characters. What is regarded as high speed here depends on the bandwidth of the channel. For modems on 3 kHz voice-grade channels, for example, data rates of 2400 bps or more are usually synchronous, whereas lower rates might be intermittent synchronous or character asynchronous. For a coaxial cable, on the other hand, the bandwidth is so large that intermittent synchronous data rates of 10 megabits per second (Mbps) or more can be used. As mentioned before, this is essential for multiaccess systems such as Ethernet. 2.2.6 Frequency- and Time-Division MUltiplexing In previous subsections, bandwidth constraints imposed by filtering on the physical channel and by noise have been discussed. Often, however, a physical channel is shared by multiple signals each constrained to a different portion of the available bandwidth; this is called frequency-division multiplexing (FDM). The most common examples of this are AM, FM, and TV broadcasting, in which each station uses a different frequency band. As another common example, voice-grade channels are often frequency multiplexed together in the telephone network. In these examples, each multiplexed signal is constrained to its own allocated frequency band to avoid interference (sometimes called crosstalk) with the other signals. The modulation techniques of the preceding subsection are equally applicable whether a bandwidth constraint is imposed by FDM or by channel filtering. FDM can be viewed as a technique for splitting a big channel into many little channels. Suppose that a physical channel has a usable bandwidth of W hertz and we wish to split it into m equal FDM subchannels. Then, each subchannel has VV 1m hertz available, and thus VV 1m available quadrature samples per second for sending data (in practice, the bandwidth per subchannel must be somewhat smaller to provide guard bands between the subchannels, but that is ignored here for simplicity). In terms of Eq. (2.13), each subchannel is allotted I/m of the overall available signal power and is subject to 11m of the overall noise, so each subchannel has 11m of the total channel capacity. Time-division multiplexing (TDM) is an alternative technique for splitting a big channel into many little channels. Here. one modem would be used for the overall bandwidth VV. Given m equal rate streams of binary data to be transmitted, the m bit streams would be multiplexed together into one bit stream. Typically, this is done by sending the data in successive frames. Each frame contains m slots, one for each bit stream to be multiplexed; a slot contains a fixed number of bits, sometimes I, sometimes 8, and sometimes more. Typically, each frame also contains extra bits to help the
Sec.2.2 The Physical Layer:Channels and Modems 53 receiver maintain frame synchronization.For example,TI carrier,which is in widespread telephone network use in the United States and Japan,multiplexes 24 data streams into slots of eight bits each with one extra bit per frame for synchronization.The overall data rate is 1.544 Mbps,with 64,000 bps for each of the multiplexed streams.In Europe, there is a similar system with 32 data streams and an overall data rate of 2.048 Mbps. One way to look at FDM and TDM is that in each case one selects I quadrature samples per second (or more generally,2W individual samples)for transmission;in one case,the samples are distributed in frequency and in the other they are distributed in time. 2.2.7 Other Channel Impairments The effects of filtering and noise have been discussed in preceding subsections.There are a number of other possible impairments on physical channels.Sometimes there are multiple stages of modulation and demodulation internal to the physical channel.These can cause small amounts of phase jitter and carrier frequency offset in the received waveform.Amplification of the signals,within the physical channel and in the modems, can cause nonnegligible nonlinear distortion.Impulses of noise can occur due to lightning and switching effects.Repair personnel can short out or disconnect the channel for periods of time.Finally,there can be crosstalk from other frequency bands and from nearby wires. To a certain extent,these effects can all be considered as extra noise sources. However,these noise sources have different statistical properties than the noise assumed by Shannon's theorem.(Technically,that noise is known as additive white Gaussian noise.)The principal difference is that the errors caused by these extra noise sources tend to occur in bursts of arbitrarily long length.As a result,one cannot,in practice, achieve the arbitrarily low error probabilities promised by Shannon's theorem.One must also use error detection and retransmission at the DLC layer;this is treated in Sections 2.3and2.4. 2.2.8 Digital Channels In many cases,physical channels are designed to carry digital data directly,and the DLC unit can interface almost directly to the digital channel rather than to a modem mapping digital data into analog signals.To a certain extent,this is simply the question of whether the channel supplier or the channel user supplies the modem.This is an oversimplified view,however.A channel designed to carry digital data directly (such as the TI carrier mentioned earlier)is often capable of achieving higher data rates at lower error probabilities than one carrying analog signals. A major reason for this improved performance is the type of repeater used for digital channels.Repeaters are basically amplifiers inserted at various points in the propagation path to overcome attenuation.For a channel designed for analog signals, both the signal and the noise must be amplified at each repeater.Thus,the noise at the final receiver is an accumulation of noise over each stage of the path.For a digital
Sec. 2.2 The Physical Layer: Channels and Modems 53 receiver maintain frame synchronization. For example, TI carrier, which is in widespread telephone network use in the United States and Japan, multiplexes 24 data streams into slots of eight bits each with one extra bit per frame for synchronization. The overall data rate is 1.544 Mbps, with 64,000 bps for each of the multiplexed streams. In Europe, there is a similar system with 32 data streams and an overall data rate of 2.048 Mbps. One way to look at FOM and TOM is that in each case one selects W quadrature samples per second (or more generally, 2HT individual samples) for transmission; in one case, the samples are distributed in frequency and in the other they are distributed in time. 2.2.7 Other Channel Impairments The effects of filtering and noise have been discussed in preceding subsections. There are a number of other possible impairments on physical channels. Sometimes there are multiple stages of modulation and demodulation internal to the physical channel. These can cause small amounts of phase jitter and carrier frequency offset in the received waveform. Amplification of the signals, within the physical channel and in the modems, can cause nonnegligible nonlinear distortion. Impulses of noise can occur due to lightning and switching effects. Repair personnel can short out or disconnect the channel for periods of time. Finally, there can be crosstalk from other frequency bands and from nearby wires. To a certain extent, these effects can all be considered as extra noise sources. However, these noise sources have different statistical properties than the noise assumed by Shannon's theorem. (Technically, that noise is known as additive white Gaussian noise.) The principal difference is that the errors caused by these extra noise sources tend to occur in bursts of arbitrarily long length. As a result, one cannot, in practice, achieve the arbitrarily low error probabilities promised by Shannon's theorem. One must also use error detection and retransmission at the OLC layer; this is treated in Sections 2.3 and 2.4. 2.2.8 Digital Channels In many cases, physical channels are designed to carry digital data directly, and the OLC unit can interface almost directly to the digital channel rather than to a modem mapping digital data into analog signals. To a certain extent, this is simply the question of whether the channel supplier or the channel user supplies the modem. This is an oversimplified view, however. A channel designed to carry digital data directly (such as the Tl carrier mentioned earlier) is often capable of achieving higher data rates at lower error probabilities than one carrying analog signals. A major reason for this improved performance is the type of repeater used for digital channels. Repeaters are basically amplifiers inserted at various points in the propagation path to overcome attenuation. For a channel designed for analog signals, both the signal and the noise must be amplified at each repeater. Thus, the noise at the final receiver is an accumulation of noise over each stage of the path. For a digital
54 Point-to-Point Protocols and Links Chap.2 channel,on the other hand,the digital signal can be recovered at each repeater.This means that the noise does not accumulate from stage to stage,and an error occurs only if the noise in a single stage is sufficient to cause an error.In effect,the noise is largely suppressed at each stage.Because of this noise suppression,and also because of the low cost of digital processing,it is increasingly common to use digital channels (such as the TI carrier system)for the transmission of analog signals such as voice.Analog signals are sampled(typically at 8000 samples per second for voice)and then quantized (typically at eight bits per sample)to provide a digital input for a digital channel. The telephone network can be separated into two parts-the local loops,which go from subscribers to local offices,and the internal network connecting local offices, central offices,and toll switches.Over the years,the internal network has become mostly digital,using TDM systems such as the T1 carrier just described and a higher-speed carrier system called T3.at 44.736 Mbps,that will multiplex 28 T1 signals. There are two consequences to this increasing digitization.First,the telephone companies are anxious to digitize the local loops,thus allowing 64 kilobit per second (kbps)digital service over a single voice circuit.An added benefit of digitizing the local loops is that more than a single voice circuit could be provided to each subscriber. Such an extension of the voice network in which both voice and data are simultaneously available in an integrated way is called an integrated services digital network (ISDN) The other consequence is that data network designers can now lease TI lines (called DSI service)and T3 lines (called DS3 service)at modest cost and thus construct networks with much higher link capacities than are available on older networks.These higher-speed links are making it possible for wide area networks to send data at speeds comparable to local area nets. With the increasing use of optical fiber,still higher link speeds are being stan- dardized.SONET (Synchronous Optical Network)is the name for a standard family of interfaces for high-speed optical links.These start at 51.84 Mbps (called STS-1), and have various higher speeds of the form n times 51.84 Mbps (called STS-n)for n =1,3,9,12,18.24,36,48.Like the TI and T3 line speeds,each of these speeds has a 125 us frame structure which can be broken down into a very large number of 64 kbps voice circuits,each with one byte per frame.What is ingenious and interesting about these standards is the way that the links carrying these frames can be joined together, in the presence of slight clock drifts,without losing or gaining bits for any of the voice circuits.What is relevant for our purposes,however,is that 1.5 and 45 Mbps link speeds are now economically available,and much higher speeds will be economically available in the future.The other relevant fact is that optical fibers tend to be almost noise-free, with error rates of less than 10-10. ISDN As outlined briefly above,an integrated services digital network(ISDN)is a telephone network in which both the internal network and local loops are digital.The "integrated service"part of the name refers to the inherent capability of such a network to carry both voice and data in an integrated fashion.A telephone network with analog voice lines for local loops also allows both data and voice to be carried,but not easily at the same time and certainly not with much speed or convenience.There has been
54 Point-to-Point Protocols and Links Chap. 2 f channel, on the other hand, the digital signal can be recovered at each repeater. This means that the noise does not accumulate from stage to stage, and an error occurs only if the noise in a single stage is sufficient to cause an error. In effect, the noise is largely suppressed at each stage. Because of this noise suppression, and also because of the low cost of digital processing, it is increasingly common to use digital channels (such as the Tl carrier system) for the transmission of analog signals such as voice. Analog signals are sampled (typically at 8000 samples per second for voice) and then quantized (typically at eight bits per sample) to provide a digital input for a digital channel. The telephone network can be separated into two parts-the local loops, which go from subscribers to local offices, and the internal network connecting local offices, central offices, and toll switches. Over the years, the internal network has become mostly digital, using TDM systems such as the Tl carrier just described and a higher-speed carrier system called T3, at 44.736 Mbps, that will multiplex 28 Tl signals. There are two consequences to this increasing digitization. First, the telephone companies are anxious to digitize the local loops, thus allowing 64 kilobit per second (kbps) digital service over a single voice circuit. An added benefit of digitizing the local loops is that more than a single voice circuit could be provided to each subscriber. Such an extension of the voice network in which both voice and data are simultaneously available in an integrated way is called an integrated services digital network (ISDN). The other consequence is that data network designers can now lease T1 lines (called DS 1 service) and T3 lines (called DS3 service) at modest cost and thus construct networks with much higher link capacities than are available on older networks. These higher-speed links are making it possible for wide area networks to send data at speeds comparable to local area nets. With the increasing use of optical fiber, still higher link speeds are being standardized. SONET (Synchronous Optical Network) is the name for a standard family of interfaces for high-speed optical links. These start at 51.84 Mbps (called STS-l), and have various higher speeds of the form n times 51.84 Mbps (called STS-n) for n = 1,3,9,12, 18,24,36,48. Like the Tl and T3 line speeds, each of these speeds has a 125 p,s frame structure which can be broken down into a very large number of 64 kbps voice circuits, each with one byte per frame. What is ingenious and interesting about these standards is the way that the links carrying these frames can be joined together, in the presence of slight clock drifts, without losing or gaining bits for any of the voice circuits. What is relevant for our purposes, however, is that 1.5 and 45 Mbps link speeds are now economically available, and much higher speeds will be economically available in the future. The other relevant fact is that optical fibers tend to be almost noise-free, with error rates of less than 10- 10• ISDN As outlined briefly above, an integrated services digital network (ISDN) is a telephone network in which both the internal network and local loops are digital. The "integrated service" part of the name refers to the inherent capability of such a network to carry both voice and data in an integrated fashion. A telephone network with analog voice lines for local loops also allows both data and voice to be carried, but not easily at the same time and certainly not with much speed or convenience. There has been I
Sec.2.2 The Physical Layer:Channels and Modems 55 considerable effort to standardize the service provided by ISDN.Two standard types of service are available.The first,called basic service,provides two 64 kbps channels plus one 16 kbps channel to the user.Each 64 kbps channel can be used as an ordinary voice channel or as a point-to-point data channel.The 16 kbps channel is connected to the internal signaling network of the telephone system(i.e.,the network used internally for setting up and disconnecting calls,managing the network,etc.).Thus,the 16 kbps channel can be used to set up and control the two 64 kbps channels,but it could also be used for various low-data-rate services such as alarm and security systems. In ISDN jargon,the 64 kbps channels are called B channels and the 16 kbps channel is called a D channel;the overall basic service is thus referred to as 2B+D.With these facilities,a subscriber could conduct two telephone conversations simultaneously,or have one data session (at 64 kbps)plus one voice conversation.The latter capability appears to be rather useful to someone who works at home with a terminal but does not want to be cut off from incoming telephone calls.Most terminal users,of course,would be quite happy,after getting used to 300 bps terminals,to have the data capabilities of the 16 kbps D channel,and it is likely that subscribers will be offered a cheaper service consisting of just one B and one D channel. The 64 kbps channels can be used in two ways:to set up a direct connection to some destination (thus using ISDN as a circuit switched network),and as an access line into a node of a packet switching network.On the one hand,a 64 kbps access into a packet switched network appears almost extravagant to one used to conventional networks.On the other hand,if one is trying to transfer a high-resolution graphics image of 109 bits,one must wait for over 4 hours using a 64 kbps access line.This illustrates a rather peculiar phenomenon.There are a very large number of applications of data networks that can be accomplished comfortably at very low data rates,but there are others (usually involving images)that require very high data rates. Optical fiber will gradually be introduced into the local loops of the telephone net- work,and this will allow ISDN to operate at much higher data rates than those described above.Broadband ISDN(BISDN)is the name given to such high-speed ISDN networks. There has already been considerable standardization on broadband ISDN,including a standard user access rate of 155 Mbps (i.e.,the SONET STS-3 rate).These high data rates will allow for high-resolution TV as well as for fast image transmission,high-speed interconnection of supercomputers,and video conferencing.There are many interesting questions about how to build data networks handling such high data rates;one such strategy,called asynchronous transfer mode (ATM),is discussed briefly in Section 2.10. The 2B+D basic service above is appropriate (or even lavish,depending on one's point of view)for the home or a very small office,but is inappropriate for larger offices used to using a PBX with a number of outgoing telephone lines.What ISDN (as proposed)offers here is something called primary service as opposed to basic service. This consists of 24 channels at 64 kbps each (in the United States and Japan)or 31 channels at 64 kbps each (in Europe).One of these channels is designated as the D channel and the others as B channels.The D channel is the one used for signaling and call setup and has a higher data rate here than in the basic service to handle the higher traffic levels.Subscribers can also obtain higher rate channels than the 64 kbps
Sec. 2.2 The Physical Layer: Channels and Modems 55 considerable effort to standardize the service provided by ISDN. Two standard types of service are available. The first, called basic service, provides two 64 kbps channels plus one 16 kbps channel to the user. Each 64 kbps channel can be used as an ordinary voice channel or as a point-to-point data channel. The 16 kbps channel is connected to the internal signaling network of the telephone system (i.e., the network used internally for setting up and disconnecting calls, managing the network, etc.). Thus, the 16 kbps channel can be used to set up and control the two 64 kbps channels, but it could also be used for various low-data-rate services such as alarm and security systems. In ISDN jargon, the 64 kbps channels are called B channels and the 16 kbps channel is called a D channel; the overall basic service is thus referred to as 2B + D. With these facilities, a subscriber could conduct two telephone conversations simultaneously, or have one data session (at 64 kbps) plus one voice conversation. The latter capability appears to be rather useful to someone who works at home with a terminal but does not want to be cut off from incoming telephone calls. Most terminal users, of course, would be quite happy, after getting used to 300 bps terminals, to have the data capabilities of the 16 kbps D channel, and it is likely that subscribers will be offered a cheaper service consisting of just one B and one D channel. The 64 kbps channels can be used in two ways: to set up a direct connection to some destination (thus using ISDN as a circuit switched network), and as an access line into a node of a packet switching network. On the one hand, a 64 kbps access into a packet switched network appears almost extravagant to one used to conventional networks. On the other hand, if one is trying to transfer a high-resolution graphics image of 109 bits, one must wait for over 4 hours using a 64 kbps access line. This illustrates a rather peculiar phenomenon. There are a very large number of applications of data networks that can be accomplished comfortably at very low data rates, but there are others (usually involving images) that require very high data rates. Optical fiber will gradually be introduced into the local loops of the telephone network, and this will allow ISDN to operate at much higher data rates than those described above. Broadband ISDN (BISDN) is the name given to such high-speed ISDN networks. There has already been considerable standardization on broadband ISDN, including a standard user access rate of 155 Mbps (i.e., the SONET STS-3 rate). These high data rates will allow for high-resolution TV as well as for fast image transmission, high-speed interconnection of supercomputers, and video conferencing. There are many interesting questions about how to build data networks handling such high data rates; one such strategy, called asynchronous transfer mode (ATM), is discussed briefly in Section 2.10. The 2B + D basic service above is appropriate (or even lavish, depending on one's point of view) for the home or a very small office, but is inappropriate for larger offices used to using a PBX with a number of outgoing telephone lines. What ISDN (as proposed) offers here is something called primary service as opposed to basic service. This consists of 24 channels at 64 kbps each (in the United States and Japan) or 31 channels at 64 kbps each (in Europe). One of these channels is designated as the D channel and the others as B channels. The D channel is the one used for signaling and call setup and has a higher data rate here than in the basic service to handle the higher traffic levels. Subscribers can also obtain higher rate channels than the 64 kbps
56 Point-to-Point Protocols and Links Chap.2 B channels above.These higher rate channels are called H channels and come in 384, 1536,and 1920 kbps flavors.Such higher rate channels partially satisfy the need for the high data rate file transfers discussed above. One of the interesting technical issues with ISDN is how to provide the required data rates over local loops consisting of a twisted wire pair.The approach being taken,for basic service,is to use time-division multiplexing of the B and D channels together with extra bits for synchronization and frequency shaping,thus avoiding a dc component.The most interesting part of this is that data must travel over the local loop in both directions, and that a node trying to receive the attenuated signal from the other end is frustrated by the higher-level signal being generated at the local end.For voice circuits,this problem has been solved in the past by a type of circuit called a hybrid which isolates the signal going out on the line from that coming in.At these high data rates,however,this circuit is not adequate and adaptive echo cancelers are required to get rid of the remaining echo from the local transmitted signal at the receiver. 2.2.9 Propagation Media for Physical Channels The most common media for physical channels are twisted pair (i.e.,two wires twisted around each other so as to partially cancel out the effects of electromagnetic radiation from other sources).coaxial cable,optical fiber,radio,microwave,and satellite.For the first three,the propagated signal power decays exponentially with distance (i.e.,the attenuation in dB is linear with distance).Because of the attenuation,repeaters are used every few kilometers or so.The rate of attenuation varies with frequency,and thus as repeaters are spaced more closely,the useful frequency band increases,yielding a trade-off between data rate and the cost of repeaters.Despite this trade-off,it is helpful to have a ballpark estimate of typical data rates for channels using these media. Twisted pair is widely used in the telephone network between subscribers and local stations and is increasingly used for data.One Mbps is a typical data rate for paths on the order of I km or less.Coaxial cable is widely used for local area networks,cable TV, and high-speed point-to-point links.Typical data rates are from 10 to several hundred Mbps.For optical fiber,data rates of 1000 Mbps or more are possible.Optical fiber is growing rapidly in importance,and the major problems lie in the generation,reception, amplification.and switching of such massive amounts of data. Radio,microwave,and satellite channels use electromagnetic propagation in open space.The attenuation with distance is typically much slower than with wire channels, so repeaters can either be eliminated or spaced much farther apart than for wire lines. Frequencies below 1000 MHz are usually referred to as radio frequencies,and higher frequencies are referred to as microwave. Radio frequencies are further divided at 30 MHz.Above 30 MHz,the ionosphere is transparent to electromagnetic waves,whereas below 30 MHz,the waves can be re- flected by the ionosphere.Thus,above 30 MHz,propagation is on line-of-sight paths. The antennas for such propagation are frequently placed on towers or hills to increase the length of these line-of-sight paths,but the length of a path is still somewhat limited and repeaters are often necessary.This frequency range,from 30 to 1000 MHz,is used
56 Point-to-Point Protocols and Links Chap. 2 r B channels above. These higher rate channels are called H channels and come in 384, 1536, and 1920 kbps flavors. Such higher rate channels partially satisfy the need for the high data rate file transfers discussed above. One of the interesting technical issues with ISDN is how to provide the required data rates over local loops consisting of a twisted wire pair. The approach being taken, for basic service, is to use time-division multiplexing of the Band D channels together with extra bits for synchronization and frequency shaping, thus avoiding a de component. The most interesting part of this is that data must travel over the local loop in both directions, and that a node trying to receive the attenuated signal from the other end is frustrated by the higher-level signal being generated at the local end. For voice circuits, this problem has been solved in the past by a type of circuit called a hybrid which isolates the signal going out on the line from that coming in. At these high data rates, however, this circuit is not adequate and adaptive echo cancelers are required to get rid of the remaining echo from the local transmitted signal at the receiver. 2.2.9 Propagation Media for Physical Channels The most common media for physical channels are twisted pair (i.e., two wires twisted around each other so as to partially cancel out the effects of electromagnetic radiation from other sources), coaxial cable, optical fiber, radio, microwave, and satellite. For the first three, the propagated signal power decays exponentially with distance (i.e., the attenuation in dB is linear with distance). Because of the attenuation, repeaters are used every few kilometers or so. The rate of attenuation varies with frequency, and thus as repeaters are spaced more closely, the useful frequency band increases, yielding a trade-off between data rate and the cost of repeaters. Despite this trade-off, it is helpful to have a ballpark estimate of typical data rates for channels using these media. Twisted pair is widely used in the telephone network between subscribers and local stations and is increasingly used for data. One Mbps is a typical data rate for paths on the order of I km or less. Coaxial cable is widely used for local area networks, cable TV, and high-speed point-to-point links. Typical data rates are from 10 to several hundred Mbps. For optical fiber, data rates of 1000 Mbps or more are possible. Optical fiber is growing rapidly in importance, and the major problems lie in the generation, reception, amplification, and switching of such massive amounts of data. Radio, microwave, and satellite channels use electromagnetic propagation in open space. The attenuation with distance is typically much slower than with wire channels, so repeaters can either be eliminated or spaced much farther apart than for wire lines. Frequencies below 1000 MHz are usually referred to as radio frequencies, and higher frequencies are referred to as microwave. Radio frequencies are further divided at 30 MHz. Above 30 MHz, the ionosphere is transparent to electromagnetic waves, whereas below 30 MHz, the waves can be reflected by the ionosphere. Thus, above 30 MHz, propagation is on line-of-sight paths. The antennas for such propagation are frequently placed on towers or hills to increase the length of these line-of-sight paths, but the length of a path is still somewhat limited and repeaters are often necessary. This frequency range, from 30 to 1000 MHz, is used