rod -$: PriniPles HR served Spedtvun, 11.3 SPREAD-SPECTRUM COMMUNICATION SYSTEMS In our treatment of signal design for digital communi c ation over an AWGN channe}, the major objective has been the efficient utilization of transmitter power and channe| bandwidth. As shown in Chapter 13, channel coding allows us to reduce the transmit. ter power by increasing the transmitted signal bandwidth through code redundancy, this allows us to trade off transmitter power with channel bandwidth. To be spe. cific, let R denote the information (bit) rate at the input to the transmitter and fe, W denote the channel bandwidth. We also define the ratio W/R as the bandwidth expansion factor, denoted as B,, of the channel coded transmitted signal. The factor B, represents the amount of redundancy introduced through channel coding. Thus, by increasing B,, we can reduce the power in the transmitted signal that is required to achieve a specific level of performance. In most practical communication systems, B, is in the range of 2 < B, < 5. This is the basic methodology for the design of digital communication systems for AWGN channels. In practice, other factors influence the design of an efficient digital communi- cation system. For example, in multiple-access communication when two or more transmitters use the same common channel to transmit information, the interference created by the users of the channel limits the performance achieved by the system. The existence of such interference must be considered in the design of a reliable digital communication system. Even in this complex design problem, the basic system design parameters are transmitter power and channel bandwidth. To overcome the degradation in perfor- mance caused by interference, we may further increase the bandwidth of the transmit- ted signal so that the bandwidth expansion factor B, = W/R is much greater than unity. This is one characteristic of a spread-spectrum signal. A second important characteristic is that the information signal at the modulator is spread in bandwidth aero coer ia is independent of the information sequence. This code signal. This second characteris ree " Knowledge of the code to demodulng te system from the conventional Ic distinguishes a spread-spectrum communication signal bandwidth by means of channel code redheahee ye eanas the transite spectrum communication system, ement in the design of an efficient sprea Spread-spectrum signals for digital communi ‘ations were originally developed and used for military communications either (1) to paving ney develop Provide resistance to jammingi 653 ide the ade it difficult for an unintendeq ligeee by transmitting it at low power, which pility of intercept). However, Sprea 7 to d 7 letect its i ‘i 1 arene oe Presence in noise (lo ity ivili ru . cations in a variety of civilian applications ine ‘ ignals now provide reliable communi- cordless telephones, and interoffice wireless ‘uding digital cellular communications, In this section, we present the b; he a ommunications. and assess their performance in te )w proba- ic characteristi °* TMs of : amely, by direct ‘opping. Both methods tequire the use of eration is also described. whose gen "i meee Several applica- tions of spread-spectrum Signals 11.3.1 Model of a Spread-Spectrum Digital Communication System The basic elements of a spread-spectrum di in Figure 11.15. The channel encoder and decoder and the modulator and demod- ulator are the basic elements of a conventional digital communication system. In addition, a spread-spectrum system employs two identical pseudorandom sequence generators, one that interfaces with the modulator at the transmitting end and one that interfaces with the demodulator at the receiving end. These two generators pro- duce a pseudorandom or pseudonoise (PN) binary-valued sequence, which is used to spread the transmitted signal at the modulator and to despread the received signal at the demodulator. Time synchronization of the PN sequence generated at the receiver with the PN sequence contained in the received signal is required to properly despread the received spread-spectrum signal. In a practical system, synchronization is established prior to the transmission of information; this is achieved by transmitting a fixed PN bit pattern, which is designed so that the receiver will detect it with high probability in the presence of interference. After time synchronization of the PN sequence gen- erators is established, the transmission of information commences. In the data mode, the communication system usually tracks the timing of the incoming received signal and keeps the PN sequence generator in synchronism. : Interference is introduced in the transmission of the spread-spectrum signal through the channel. The characteristics of the interference depend toa large extent on its origin. The’ interference may be generally categorized as either broadband igital communication system are illustrated a ; Output formation Channel |_data a —s tence, | Channel Modulator ->| Channel | Demodulator decoder encoder t Pseudorandom Pseudorandom pattern pattern generator generator Figure 11.15 Model of a spread-spectrum digital communication system.654 Selected Topics in Digital Communications Chapter yy or narrowband (partial band) relative to the bandwidth of the informations signal and either continuous in time or discontinuous (pulsed) in tine, an interfering signal may consist of a high-power sinusoid in the bandwidth o¢ lpg by the information-bearing signal. Suc nal is narrowband, As a second CRAMP the interference generated by other users in a multiple-access channel depends gf the type of spread-spectrum signals that are used to transmit information, If al] Users employ broadband signals, the interference may be characterized as an equivaleny broadband noise. If the users employ frequency hopping to ¢ le spre Spectrum, signals, the interference from other users may be characterized as narrowband, Our discussion will focus on the performance of spread-spectrum signals for digital communication in the presence of narrowband and broadband interference, Two types of digital modulation are considered, namely, PSK and PSK. PSK modu. lation is appropriate for applications where phase coherence between the transmitted signal and the received signal can be maintained over a time interval that spans several symbol (or bit) intervals. On the other hand, FSK modulation is Appropri- ate in applications where phase coherence of the carrier cannot be maintained due to time variations in the transmission cha of the communications channel, For example, this may be the case in a communications link between two high-speed aircraft or between a high-speed aircraft and a ground-based terminal. The PN sequence generated at the modulator i ‘din conjunction with the PSK modulation to shift the phase of the PSK signal pseudorandomly at a rate that is an integer multiple of the bit rate. The resulting modulated signal is called a direct sequence (DS) Spread-spectrum signal. When used in conjunction with binary or M-ary (M > 2) FSK, the PN sequence is used to select the frequency of the transmitted signal pscudorandomly. The resulting signal is called a frequency-hopped (FH) spread-spectrum signal. Although other types of Spread-spectrum signals can be generated, our treatment will emphasize DS and FH Spread-spectrum communication systems, which are generally used in practice, Net Vor exany le, 11.3.2 Direct Sequence Spread-Spectrum Systems Consider the transmission of a binary information sequence by means of binary PSK. The information rate is R bits per second, and the bit interval is Ti, = 1/R seconds. The available channel bandwidth is B. Hz, where B. > R. At the modulator, the bandwidth of the information signal is expanded to W = B. Hz by shifting the phase of the carrier pseudorandomly at a rate of W times per second according to the pattern of the PN generator. The basic method for accomplishing the spreading is shown in Figure 11.16, The information-bearing baseband signal is denoted as v(t) and is expressed as VD = SO agree —nty), 3.) n==00 Where (ay = 1. —oo YY Blam) ECnvm)- a n=0 m=0 But E(Cn¢m) = bn. Therefore, 2%, : ElyPTp)) = ae 26 = 1.60%), (113.21) TyExample 11.3.1 The SNR required at the detector to achieve reliable communication in a DS spread- Spectrum communication system is 13 dB. If the interference-to-signal power at the receiver is 20 dB, determine the processing gain Tequired to achieve reliable communication, Solution We are given (P;/Ps)yg = 20 dB or, equivalently, P)/Ps = 100. We are also given (SNR)p = 13 dB, or equivalently, (SNR) = 20. The relation in Equation (11.3.29) may be used to solve for L,. Thus, 1(P; 3 (#) (SNR)p = 1000. Therefore, the processing gain required is 1000 or, equivalently, 30 dB.section 113 Spread-Spectrum Communication System: 663 8 The Interference Margin. . Equation (11.3.32) as $y We may express & i in the Q-function in — Ps/R _ W/R Io 1/W ~ PW > Pippy (1.3.33) Also, suppose we specify a required &, / Jy to achieve a desi e ; ‘Then, using a logarithmic scale, we may express Eat a ae Pr w ‘ 10 loss = 10 log — 10 log (#) o #) a (% Sp, (#z a ap - (2), (113.34) The ratio (P;/Ps)ap is called the interference margin. This is the relative power advantage that an interference may have without disrupting the communication system. Example 11.3.2 Suppose we require an (€4/Jo)an = 10 dB to achieve reliable communication. What is the processing gain that is necessary to provide an interference margin of 20 dB? Solution Clearly, if W/R = 1000, then (W/R)ap = 30 dB and the interference margin is (P)/Ps)ap = 20 dB. This means that the average interference power at the receiver may be 100 times the power Ps of the desired signal and we can still maintain reliable communication. . Performance of Coded Spread-Spectrum Signals. As shown in Chapter 13, when the transmitted information is coded by a binary linear (block or convolutional) code, the SNR at the output of a soft-decision decoder, at large SNR, is increased by the coding gain, defined as coding gain = R.di¥.,, (1.3.35) where R; is the code rate and @., is the minimum Hamming distance of the code. Therefore, the effect of coding is to increase the interference margin by the coding gain. Thus, Equation (11.3.34).may be modified as (#) -(%) + (Cova - (3) . (1.3.36) Ps) ap Ras lo] ap where (CG)gp denotes the coding gain in dB. 11.3.3 Some Applications of DS Spread-Spectrum Signals In this section, we briefly describe the use of DS spread-spectrum signals in four applications, Fire. we consider an application in which the signal is transmitted at very low power, co that a listener trying to detect the presence of the signal would encounter great difficulty in doing so?. A.second application is multiple access radio664 Selected Topics in Digital Communications Chapter 11 communications. A third application involves the use of a DS spread-spectrum signa} to resolve the multipath in a time-dispersive radio channel. The fourth application ig the use of DS spread-spectrum signals in wireless LAN’s. Low-Detectability Signal Transmission. In this application, the infor. mation-bearing signal is transmitted at a very low power level relative to the back. ground channel noise and thermal noise that is generated in the front end of a receiver. If the DS spread-spectrum signal occupies a bandwidth W and the power spectral density of the additive noise is No W/Hz, the average noise power in the bandwidth W is Py = WNo. The average received signal power at the intended receiver is Pr. If we wish to hide the presence of the signal from receivers that are in the vicinity of the intended receiver, the signal is transmitted at a power level such that Pr/Py < 1. The intended receiver can recover the weak information-bearing signal from the back- ground noise with the aid of the. processing gain and the coding gain. However, any other receiver that has no knowledge of the PN code sequence is unable to take advantage of the processing gain and the coding gain. Consequently, the presence of the information-bearing signal is difficult to detect. We say that the transmit- ted signal has a low probability of being intercepted (LPI), and it is called an LPI signal. The probability of error given in Section 11.3.2 also applies to the demodulation and decoding of LPI signals at the intended receiver. Example 11.3.3 ADS spread-spectrum signal is designed so that the power ratio Pg/ Py at the intended receiver is 10-, If the desired €4/No = 10 for acceptable performance, determine the minimum value of the processing gain. Solution We may write €4/No as Pele Pabske (Fe), Pr) No No No \WNo) *~ 4) - Since €,/No = 10 and Pr/Py = 10%, it follows that the necessary processing gain is Le = 1000. . Code Division Multiple Access. The enhancement in performance obtained from a DS spread-spectrum signal through the processing gain and the coding gain can enable many DS spread-spectrum signals to occupy the same channel bandwidth Provided that each signal has its own pseudorandom (signature) sequence. Thus, it is Possible to have several users transmit messages simultaneously over the same chan- nel bandwidth. This type of digital communication, in which each transmitter-receiver user pair has its own distinct signature code for transmitting over a common channel bandwidth, is called code division multiple access (CDMA). In the demodulation of each DS spread-spectrum signal, the signals from the other simultaneous users of the channel appear as additive interference. The level of interference varies as a function of the number of users of the channel at any41.3.4 Generation of PN Sequences ‘A pseudorandom or pseudonoise (PN) sequence is a code sequence of 1’s and 0's whose autocorrelation has properties similar to th it i i i we briefly describe the consiruction of some PN sesusnees aed thei oconelation and cross correlation properties. quences and their autocorrelation __ The most widely known binary PN code sequences are the maximum-length shift-register sequences. A maximum-length shift-register sequence, or m-sequence for short, has the length L = 2” —1 bits and is generated by an m-stage shift register with linear feedback, as illustrated in Figure 11.20. The sequence is periodic with period L. Each period contains 2"-! ones and 2-1 — 1 zeros. Table 11.2 lists shift register connections for generating maximum-length sequences. In DS spread spectrum applications, the binary sequence with elements (0, 1) is mapped into a corresponding binary sequence with elements {—1, 1}. ‘We shall call the equivalent sequence {c,} with elements {—1, 1} a bipolar sequence. ‘An important characteristic of a periodic PN sequence is its autocorrelation function, which is usually defined in terms of the bipolar sequences {cx} as L Rem) = > Cntntm OSmSL—A, (1.3.38) n=l where L is the period of the sequence. Since the sequence {cn} is periodic with period L, the autocorrelation sequence {Re(m)) is also periodic with period L. ttc Output et Dela |i Figure 11.20 General m-stage shift register with linear feedback.TABLE 11.2. SHIFT-REGISTER CONNECTIONS FOR GENERATING MAXIMUM-LENGTH SEQUENCES a j 668 Selected Topics in Digital Communications — Chapter 4, | | / | Stages connected Stages connected Stages connected m to modulo-2-adder =m to modulo-2-adder mt to modulo-2 adder 2 1, 13 1,10, 11,13 24 1, 18, 23, 24 3 1,3 4 1,5,9,14 501/23 4 4 15 115 26 1, 21, 25, 26 5 14 16 1,5, 14, 16 27° 1,23, 26;27 1 6 1,6 Ii 1S 2% 1,26 7 Wy? 18 1,12 29 1,28 8 15,67 191, 15, 18, 19 30° 1,8, 29,30 9 1,6 2 1,18 31 1,29 le 108) Bie 20) 2 eee aisle 3) Ho 410 2 1,2 3300121 12 1,7,9,12 B 119 341, 8,33, 34 Ideally, a PN sequence should have an autocorrelation function that has corre- lation properties similar to white noise. That is, the ideal autocorrelation sequence for {cn} is Re(0) = L and Re(m) = 0 for | < m < L—1. In the case of m-sequences, the autocorrelation sequence is L, m=0 Re(m) = (113.39) -l, l VL. Hence, ound, and they are optimal. On the other = J2L. Hence, they are slightly (1.3.49) which, for large values of L. Kasami sequences satisfy the lower bi hand, Gold sequences with m odd have an Rmax suboptimal. 11.3.5 Frequency-Hopped Spread Spectrum In frequency-hopped (FH) spread spectrum, the available channel bandwidth W is subdivided into a large number of non-overlapping frequency slots. In any signaling interval, the transmitted signal occupies one or more of the available frequency slots. ‘The selection of the frequency slot (s) in each signal interval is made pseudorandomly according to the output from a PN generator. ‘A block diagram of the transmitter and receiver for an FH spread spectrum system is shown in Figure 11.22. The modulation is either binary or M-ary FSK (MFSK). For example, if binary FSK is employed, the modulator selects one of two frequencies, such as fo or fi, corresponding to the transmission of a 0 or a 1. The resulting binary FSK signal is translated in frequency by an amount determined by the output sequence from a PN generator, which is used to select a frequency fe that is synthesized by the frequency synthesizer. This frequency is mixed with the output of the FSK modulator and the resultant signal is transmitted over the channel. For example, by taking m bits from the PN generator, we may specify 2” — 1 possible carrier frequencies. Figure 11.23 illustrates an FH signal pattern. Information carmen = ostput SH} Encoder FO} incttstor FO) Miter YChannel | Mixer cme SN >] PeeOder lemodulatoy PN fn t f f sequence P| Frequency Frequency Time sequence ihe synthesizer syne, PN sequence generator Figure 11.22 Block diagram of an FH spread-spectrum system.672 Selected Topics in Digital Communications Chapter 11 ‘hronization must be established to within a fraction of the other hand, in an FH system, the chip interval 7, gnal in a particular frequency slot of bandwidth ately 1/B, which is much larger than 1/W, FH system are not as stringent as in a DS a DS system, timing and syncl achip interval T. = 1/W. On the o is the time spent in transmitting a Si B < W. But this interval is approxim: Hence, the timing requirements in an system. Next, we shall evaluate the performance of FH spread spectrum systems under the condition that the system is slow hopping. Slow Frequency-Hopping Systems. Consider a slow frequency-hop- ping system in which the hop rate Ry, = 1 hop per bit. We assume that the inter. ference on the channel is broadband and is characterized as AWGN with power spectral density /9. Under these conditions, the probability of error for the detection of noncoherently demodulated binary FSK is P= 1 ml, (113.41) 2 where py = €p/Io is the SNR per bit. ‘As in the case of a DS spread-spectrum system, we observe that 6p, the energy per bit, can be expressed as € = P57} = Ps/R, where Ps is the average transmitted power and R is the bit rate. Similarly, fo = P:/W, where Py is the average power of the broadband interference and W is the available channel bandwidth. Therefore, the SNR pp can be expressed as bp W/R Po= T= Bayes (1.3.42) where W/R is the processing gain and P;/Ps is the interference margin for the FH spread-spectrum signal. Note that the relationship in Equation (11.3.42) for the FH spread-spectrum signal is identical to that given by Equation (11.3.33) for the DS spread-spectrum signal. Therefore, frequency hopping provides basically the same benefits as direct sequence spreading. Applications of FH Spread Spectrum. FH spread spectrum is a viable alternative to DS spread spectrum for protection against narrowband and broadband interference that is encountered in CDMA. In CDMA systems based on frequency hopping, each transmitter-receiver pair is assigned its own pseudorandom frequency hopping pattern. Aside from this distinguishing feature, the transmitters and receivers of all users may be identical, i.e., they have identical encoders, decoders, modulators, and demodulators. : : ; CDMA systems based on FH spread-spectrum signals are particularly attractive jor mobile (j.e., land, air, and sea) users because timing (synchronization) require- ments are not as stringent as in a DS spread-spectrum system. In addition, new fre- quency synthesis techniques and associated hardware make it possible to frequency- hop over bandwidths that are significantly larger, by one or more orders of magnitude,11.4.2 CDMA System Based on IS-95 As described in Section 11.3, the enhancement in performance obtained from a DS spread-spectrum signal through the processing gain and coding gain can enable many DS spread-spectrum signals to simultaneously occupy the same channel band- width provided that each signal has its own distinct pseudorandom sequence. Direct sequence CDMA has been adopted as one multiple-access method for digital cel- lular voice communications in North America. This first generation digital cellular (CDMA) communication system was developed by Qualcomm, and it has been stan- dardized and designated as IS-95 by the Telecommunications Industry Association (TIA) for use in the 800 MHz and the 1900 MHz frequency bands. A major advan- tage of CDMA over other multiple access methods is that the entire frequency band is available at each base station, i.e., the frequency reuse factor V = 1. "The nominal bandwidth used for transmission from a base station to the mobile Teceivers (forward link) is 1.25 MHz. A separate channel, also with a bandwidth of 1.25 MHz, is used for signal transmission from mobile receivers to a base station (reverse link). The signals transmitted in both the forward and the reverse links are678 Selected Topics in Digital Communications Chapter 14 PN code Pilot channel generator and other trafic Tchannel channels in the same cell Baseband Hadamard crane fier ate LL =| eat z 7 19.2 kbps Data | convotuion |_| Block 9.6 kbps encoder Sea transmitter, wee | ah SERRE | sepatiion 12 kbps Baseband shaping fitter | #% Pilot channel Mask | Long.code PN code and other traffic SEA Y generator LY Decimator generator channels in the 1.2288 Meps| 19.2 keps Qchannel same cell Figure 11.28 Block diagram of a IS-95 forward link. DS spread spectrum signal and they have a chip rate of 1.2288 x 10® chips per second (1.2288 Mchips/s). Forward Link. A block diagram of the modulator for the signals transmitted from a base station to the mobile receivers is shown in Figure 11.28. The speech coder is a code-excited linear predictive (CELP) coder that generates data at the variable rates of 9600, 4800, 2400, and 1200 bits/s, where the data rate is a function of the user’s speech activity in frame intervals of 20 ms. The data from the speech coder is encoded by a rate 1/2, constraint length L = 9 convolutional code. For lower speech activity, where the data rates are 4800, 2400, or 1200 bits/s, the output symbols from the convolutional encoder are repeated either twice, four times, or eight times to maintain a constant bit rate of 9600 bits/s. At the lower speech activity rates, the transmitter power is reduced by either 3, 6, or 9 dB, so that the transmitted energy per bit remains constant for all speech rates: Thus, a lower speech activity results in a lower transmitter power and, hence, a lower level of interference to other users. The encoded bits for each frame are passed through a block interleaver, which overcomes the effects of burst errors that may occur in transmission through the channel. The data bits at the output of the block interleaver, which occur at a rat of 19.2 kbits/s, are scrambled by multiplication with the output of a long code (period N = 2—1) generator running at the chip rate of 1.2288 Mchips/s, but whose, output is decimated by a factor of 64 to 19.2 kchips/s. The long code uniquely identifies # call of a mobile station on the forward and reverse links. Each channel user is assigned a Hadamard (also called a Walsh) sequence of length 64. There are 64 orthogonal Hadamard sequences assigned to each base station; thus, there are 64 channels available. One Hadamard sequence (the all-zere sequence) is used to transmit a pilot signal, which serves as a means for measuring | 1section 11.4 Digital Cellular Communication Sit as ems 1 characteristics, i i : i parameters are ee net the signal strength and the carrier phase offset. Thevhar Hadamard sequence is used ny na jpeform™ Phase coherent demodulation. Am possibly more if necessary, is used’ for qt’ synchronization. One channel for allocation to different users, paging. That leaves up to 61 channels Each user, using the assi; he assigned Hadamard s igned Hadamard sequence, multiplies the data sequence by quence. Thus, each encoded data bit is multiplied b: mard sequence of | aa Coe a ec oaa: the Hada 4 length 64, The resulting bin: i d Itiplication with two PN ig binary sequence is now sprea\ ie ean Y Sequences of length N = 2'5; this creates in-phase (J) and cea rature (Q) signal components. Thus, the binary data signal is converted to a four-phase signal and both the / and Q components are filtered by baseband spectral-shaping filters. Different base stations are identified by different offsets of these PN sequences. The signals for all 64 channels are transmitted synchronously so that, in the absence of channel multipath distortion, other signals received at any mobile receiver do not interfere because of the orthogonality of the Hadamard sequences. At the receiver, a RAKE demodulator resolves the major multipath signal com- ponents. These components are then phase-aligned and weighted according to their signal strength, using the estimates of phase and signal strength derived from the pilot signal. These components are combined and passed to the Viterbi soft-decision decoder. Reverse link. The reverse link modulator from a mobile transmitter to a base station is different from the forward link modulator. A block diagram of the modulator is shown in Figure 11.29. An important consideration in the design of the modulator is that signals transmitted from the various mobile transmitters to the base station are asynchronous; hence, there is significantly more interference among users. In addition, the mobile transmitters are usually battery operated; consequently, these transmissions are power limited. To compensate for these major limitations, a rate 1/3, K = 9 convolutional code is used in the reverse link. This code has essentially the same coding gain in an AWGN channel as the rate 1/2 code used in the forward link. However, it has a much higher coding gain in a fading channel. Again, for lower speech activity, output bits from the convolutional encoder are repeated either two, four, or eight times. However, the coded bit rate is 28.8 kbits/s. For each 20-ms frame, the 576 encoded bits are block-interleaved and passed to the modulator, The data is modulated using an M = 64 orthogonal signal set of Hadamard sequences each of length 64. Thus, a 6-bit block of data is mapped into one of the 64 Hadamard sequences. The result is a bit (or chip) rate of 307.2 kbits/s at the output of the modulator. Note that 64-ary orthogonal modulation at an error probability of 107° requires approximately 3.5 dB less SNR per bit than binary antiy i ing i WGN channel. 2 ay signaling in an to other users, the time position of the transmitted code symbol repetitions is randomized; thus, af the lower speech activity, ine utive bursts are not evenly spaced in time. Following the randomizer, the signal is Spread by the output of the long code PN generator, which is running at a rate ofReverse link data 9.6 kbps ae 48kbps Rate z.L=9} 24 kbps —>eonvolutional 12kbps {encoder and Block 64-ary [> }interleaver} —»prthogonal {modulator 288 sys Mask for __ [ Long code user i generator yur 9819900 S6-S1 aM jo wesiep OIL 6T'TT aan T2288 Mchipsls PN code generator for Ichannel 1.2288 Mehipsis [Basebana shaping filter 1.2288 Mchipsis PN code generator lor Q channel|