first three. Because UMTS W-CDMA and cdma2000 use direct sequence spread spectrum, only this access scheme will be explained in this section. In a direct-sequence spread spectrum system [6], [7], also known as the CDMA, each user is assigned a unique PN code. Its data stream is first spread out by that PN code, and then modulates the carrier frequency. The clock rate of the spreading code is known as the chip rate. This principle is illustrated in Figure 3-3. The chip rate of the PN code is usually much higher than the user data rate. The ratio of the chip rate to the data rate is called the spreading factor. In UMTS, the spreading factor varies from 4 to 256. CDMA Technology Direct-Spread CDMA Principles As will be seen later, PN codes have some unique properties. One of them is that any physical channel or user application, when spread by a PN code at the transmitter, can be uniquely identified at the receiver by multiplying the received baseband signal with a phase- coherent copy of that PN code. To illustrate how a CDMA receiver can detect the signal from a desired user in the presence of signals received from other users in a CDMA system, consider Figure 3-4 (a), which shows the block diagram of an overly simplified CDMA receiver. Suppose that the receiver wants to detect the data stream Chapter 3 60 Digital Modulator User Data X X Output Code Generator c f User's PN Code Figure 3-3 Principles of a DSSS 61 Principles of Wideband CDMA (W-CDMA) Demodulator X Integrator )(tr x )( 1 tC )( 1 ts )(tr )( 1 tr Decoder (a) Figure 3-4 A CDMA system: (a) A simplified CDMA receiver. The received signal at the input of the demodulator is composed of signals from multiple users. The data stream from user 1 is being detected in this figure. (b) Diagrams showing how the signal arriving at a receiver is composed of the transmitted signals from multiple users. (c) Diagrams illustrating how the signal from a desired user can be detected. The decoder reads the integrator output at the end of each symbol period, and if it is positive, it takes the data to be a binary zero. If it is negative, the data is decoded to be a binary one. Notice that after the decoder has read the integrator output, the integrator must (cont.) c T b T 1 0 0 User 1 Data User 1 PN Code )( 1 tC )( 1 ts )(*)( 11 tCts )( 2 ts )( 2 tC )(*)( 22 tCts )( 3 ts )( 3 tC )(*)( 33 tCts t t 1 0 Demodulator Output = stCt ii i () () = ⌺ 1 3 (b) from user 1. The received signal from multiple users is first demod- ulated. The output of the demodulator, which is a baseband signal, is multiplied by the PN code assigned to user 1. The resulting output is applied to the input of an integrator where it is integrated over each symbol period. The decoder reads the output of the integrator and Chapter 3 62 Figure 3-4 cont. be reset (that is, its output must be dumped) so that the process can recommence at the start of the next symbol period. 1 0 User 1 Decoded Data User 1 PN Code )( 1 tC t 1 Demodulator Output 1 2 3 -3 -2 -1 Demodulator Output * User 1 Code Integrator Output 2 4 6 12 -2 -4 -6 (c) decodes it into binary data, following certain rules. The result is the recovered data from user 1. To see that this indeed is the case, assume that the data stream from any user is represented by s i (t) and its associated PN code by C i (t). The output at the transmitter after spreading is n i (t) ϭ s i (t)*C i (t). Notice that in s i (t) or C i (t), the signal level is either ϩ1 or Ϫ1, with ϩ1 representing a binary 0 and Ϫ1 a binary 1. If the noise introduced by the channel is negligible, the demodulated signal at the baseband is given by where N is the number of users in the system. If r(t) is now multi- plied by a copy of the PN code C 1 (t) of user 1, the resulting output is given by Because the cross-correlation between C 1 (t) and C 2 (t) is very small, the second term appears as noise so that when it is integrated over a symbol period, the output of the integrator due to this term is virtually zero. The same is true of the third and following terms. However, the output of the integrator due to the first term, when averaged over a symbol period, is s 1 (t) because These ideas are illustrated in the time diagrams of Figure 3-4(b) and (c). Capacity of a CDMA System Consider a single cell CDMA system where a number of mobiles are simultaneously transmitting at the same frequency. Here, each mobile is assigned a unique PN code sequence. Let P ϭ carrier power, E b ϭ Energy per bit B c ϭ spread spectrum signal bandwidth C 1 1t2*C 1 1t2ϭ 1 ϭ s 1 1t2*C 1 1t2*C 1 1t2ϩ s 2 1t2*C 2 1t2*C 1 1t2ϩ s 3 1t2*C 3 1t2*C 1 1t2ϩ p r 1 1t2ϭ C 1 1t2*r1t2ϭ C 1 1t2* a N iϭ1 s i 1t2*C i 1t2 r1t2ϭ a N iϭ1 s i 1t2*C i 1t2 63 Principles of Wideband CDMA (W-CDMA) f data ϭ information bit rate I ϭ power due to interference N o ϭ noise power per bit Then, (3-1) So, Here G p is the RF bandwidth divided by the information bit rate. In the CDMA system being discussed here, the signal is Quadrature Phase Shift Key (QPSK)-modulated, where the RF bandwidth is approximately equal to the chip rate. In other words, if f chip is the chip rate, then the RF bandwidth B c ϭ f chip , and in that case G p ϭ f chip /f data is called the process gain. For a given bit error rate, E b /N 0 is fixed. Consequently, the greater the process gain, the larger the allowable interference (that is, I/P) for that bit error rate. If there are N transmitters, all transmitting at the same power and using the same chip rate, then So, using equation (3-1), Or, (3-2) for large values of N. N ϭ 1 ϩ G p E b >N o Ϸ G p E b >N o I>P ϭ 1N Ϫ 12P P ϭ N Ϫ 1 ϭ G p E b >N o I ϭ 1N Ϫ 12P E b >N o ϭ P I B c f data ϭ P I ϫ G p N 0 ϭ I>B c E b >N 0 ϭ P N 0 f data E b ϭ P>f data Chapter 3 64 TEAMFLY Team-Fly ® Notice that for a fixed bit error rate (that is, a fixed value of E b /N 0 ), the greater the process gain, the larger the capacity N of the system. Similarly, with a fixed process gain, the capacity increases if the value of E b /N 0 required to provide a satisfactory operation decreases. The capacity given by the previous equation is achieved only under ideal conditions. In actual practice, it may be significantly less for a number of reasons. For example, the capacity will decrease if the power control is not perfect. Similarly, in a multicell system, where each cell operates at the same frequency, transmissions in other cells may cause the interference to be increased by 60 — 85 percent. Because the system is interference limited, the capacity of the sys- tem can be increased by reducing the interference. There are a num- ber of ways of doing this. First, the interference due to other users can be reduced by replacing an omnidirectional antenna with a directional one. For example, a 3-sector antenna would increase the capacity by a factor of about 2 — 3. Second, human conversation is characterized by talk bursts fol- lowed by silence periods. If the transmitter is turned off during these silence periods, the interference to other transmitters will decrease, and consequently, the overall system capacity will increase. Thus, actual capacity may be given by (3-3) where a is the correction factor due to imperfect power control, b is the effect of co-channel interference from other cells in a multicell system, and n is the voice activity factor. Table 3-1 gives some typical values of these parameters. As an example, suppose that a ϭ 1 (that is, perfect power control), n ϭ 0.4, b ϭ 0.85 for a 3-sector cell, data rate ϭ 9.6 kb/s for an 8 kb/s vocoder, and chip rate ϭ 1.2288 Mc/s. The required E b /N 0 ϭ 7dB. So, the value of E b /N 0 ϭ 10 0.7 ϭ 5.01. G p ϭ 1.2288 ϫ 10 6 /9600 ϭ 128. So, N ϭ 1 ϩ (128/5.01)(1/1.85)(1/0.4) ϭ 35. This capacity is also known as the sectorized pole capacity. Notice that the capacity can be increased by simply reducing E b /N 0 , but that would result in increased bit error rates for all users. On the other hand, it is possible to minimize E b /N 0 without N ϭ 1 ϩ G p E b >N o a 11 ϩ b2v 65 Principles of Wideband CDMA (W-CDMA) necessarily running the risk of increasing the bit error rate. One way to do this is to select an appropriate modulation technique. For example, if the desired bit error rate is 10 Ϫ5 , the required E b /N 0 is 12.6 dB with Binary Frequency Shift Keying (BFSK), whereas it is only 9.6 dB for Binary Phase Shift Keying (BPSK) or QPSK using coherent detection [9]. Because the bit error rate increases as the signal-to-interference ratio is minimized, it is necessary to use an error-correcting code. The coding that is normally used in CDMA and W-CDMA systems is convolutional coding where it is possible to achieve a coding gain of 4 — 6 dB with hard decision sequential and soft decision Viterbi decod- ing. Thus, the capacity of a CDMA system can be increased by using channel coding. 3 It is interesting to know the minimum signal-to-noise ratio (SNR) that one can possibly use. The maximum attainable data rate R max on a channel with infinite bandwidth in the presence of Gaussian noise is given by Shannon’s channel capacity theorem: (3-4) Comparing equations (3-1) and (3-4), we see that the minimum SNR is ln 2 ϭ 0.693, that is, 10 log(0.693) or Ϫ1.6 dB. The maximum data rate is determined not only by this SNR but also the transmit- ter power. R max ϭ P N 0 ln2 Chapter 3 66 Parameter Average Values Power control correction factor, a 0.5–1.0 Voice activity factor, n 0.4–0.6 Effect of co-channel interference 0.5–0.9. A typical value for a 3-sector cell is from other cells in the system, b 0.85. For an omnidirectional antenna, it is 0.6. Table 3-1 Typical values of parameters that affect the system capacity 3 This is true even though the use of a channel code leads to a somewhat increased channel rate. 3G Radio Transmitter Functions To understand the technology used in the implementation of the physical layer functions of a typical W-CDMA system, consider Fig- ure 3-5, which shows a simplified block diagram of the transmit functions of a multicarrier cdma2000 base transceiver station. The incoming data stream (destined to each user) is encoded into a CRC code and passed through a convolutional encoder of 67 Principles of Wideband CDMA (W-CDMA) X X X X A W A W Q Complex Spreading Code CRC Encoder Convolutional Encoder o o Block Interleaver User 1 Data (U1) Pulse-Shaping Filter X X o 90 t A ω cos To Amplifier Symbol Mapper IS QS IQ QS IS IQ k = 9 r = 1/3 Symbol Repetition Demul- tiplexer Long Code Mask Long Code Generator Decimator + Repeated for Each User Power Control o o from Other Users MUX P/S I Q (Walsh Code for Carrier A) Channel Gain Channel Gain X X j I QI jSS Repeated for Each CDMA Carrier 1 A 2 A N A AU 1 1() A 1 Pulse-Shaping Filter ∑ ∑ + − + ⌺ Figure 3-5 A simplified block diagram of the transmit functions of a base transceiver station of a multicarrier cdma2000 system constraint length 9 and rate 1 / 3 . 4 Depending upon the data rate, the output of this coder may have to be repeated a few times in the block marked Symbol Repetition so that the resulting output matches the physical channel rate. The output of the symbol repetition block is applied to an inter- leaver that spreads out in time adjacent bits of the input stream to provide protection against burst errors. A long PN code that is unique for each user scrambles the output of the interleaver and applies the scrambled sequence to a demultiplexer where it is bro- ken into N subsequences, where N is the number of CDMA carriers. Each of these subsequences is transmitted over a separate CDMA carrier as shown in the lower half of the diagram. In other words, data is transmitted in parallel over multiple CDMA carriers. The chip rate used is N ϫ 1.2288 Mc/s. The value of N may be 1, 3, 6, 9, or 12. However, standards currently specify N ϭ 1 and 3 only. Subsequences (shown as A 1 , A 2 , . . .) are multiplexed with the power control bits, converted into a parallel form, and then split into I and Q streams corresponding to the in-phase (I) and quadrature (Q) component of the transmitted signal. Each bit of the I and Q stream is mapped into a BPSK symbol: a zero into ϩ1 and a one into Ϫ1. The symbols of the I and Q branches are multiplied by a gain fac- tor and spread by a Walsh code, say, W A , which is different for each CDMA carrier. As described later, these codes are constructed with the elements of a row of an orthogonal matrix, whose entries are either ϩ1 or Ϫ1 so that when channels are spread with different Walsh codes, they become mutually orthogonal. The I and Q symbols after Walsh spreading are added in quadra- ture to form complex symbols that are again spread (that is, multi- plied) by a complex PN code S I ϩ jS Q , where S I and S Q are, respectively, the cell-specific I-channel and Q-channel pilot PN sequence. The I and Q components of the output from complex spreading are passed through a pulse-shaping filter and modulate the desired CDMA carrier shown as v A in Figure 3-5. Chapter 3 68 4 In IS-95, the convolutional code used has a constraint length 9 and rate 1 / 2 . Basic ideas behind the system components, such as a channel encoder, an interleaver, PN code sequences, and so on, are presented in the following section. Speech Encoding Speech encoders used in different mobile communications systems are listed in Table 3-2. UMTS uses Adaptive Multirate (AMR) coding of speech based on the principles of Algebraic Code Excited Linear Prediction (ACELP) [43] — [46]. Eight encoded bit rates are supported: 12.2 (GSM enhanced full rate), 10.2, 7.95, 7.40, 6.70, 5.90, 5.15, and 4.75 kb/s. ACELP coders belong to the vocoder class of encoders that, unlike a waveform quantizer, model the vocal tract as a time-varying digital filter such that when it is excited with an appropriate input, the out- put is a desired speech signal. The filter coefficients are determined 69 Principles of Wideband CDMA (W-CDMA) Mobile Communications System Speech Coding Algorithm IS-54/IS-136 Vector-Sum Excited Linear Predictive Coding (VSELP). The bit rate is 7.95 kb/s. The coder operates on 20-ms frames. Its output consists of 159 bits per frame. European Telecommunications 13 kb/s Regular Pulse Excitation with Standards Institute (ETSI)/GSM Long Term Predictor (RPE-LTP). Every Standard 06.xx 20 ms, the speech encoder generates 260 bits. cdmaOne based on IS-95A Code Excited Linear Predictive Coding and IS-95B (CELP). The bit rate may be 9.6, 4.8, 2.4, or 1.2 kb/s. UMTS AMR based upon ACELP. There are eight possible bit rates varying from 4.75 kb/s to 12.2 kb/s. At 12.2 kb/s, the output of the coder is 244 bits for every 20-ms frame. Table 3-2 Speech encoders used in mobile communications systems [...]... the null-to-null bandwidth is Rb when using rectangular pulses and about 0.75Rb with a raised cosine filter 0 10 Figure 3-1 5 The bit error rate performance of BPSK with coherent detection in the presence of Gaussian noise -1 Bit Error Rate 10 -2 10 -3 10 -4 10 -1 0 7 -8 -6 -4 -2 0 2 Signal-to-Noise Ratio (dB) 4 6 8 In other words, the SNR required to achieve the same bit error rate is 2 dB more for the... 3-7 (b) The input is shifted k bits at a time As there are n output bits for each set of k bits at the input, the code is of rate k/n and constraint length mk In Figure 3-7 (a), k ϭ 1, n ϭ 2 and m ϭ 3 The two generating functions of the code of Figure 3-7 (a) are g1 1x2 ϭ 1 ϩ x ϩ x2, and g2 1x2 ϭ x ϩ x2 Principles of Wideband CDMA (W-CDMA) Figure 3-7 Convolutional encoders: (a) An encoder of rate 1/2 and. .. Odd Bits Data Serial-toParallel Converter Pulse-Shaping Filter X Σ Symbol Mapper Q Even Bits Pulse-Shaping Filter X Ac Sin ω c t Output Principles of Wideband CDMA (W-CDMA) 81 performance degrades to some extent For example, this degradation is about 2 dB for BPSK or QPSK.7 If the channel noise is Gaussian, coherent detection gives the best performance Under these circumstances, BPSK and QPSK have almost... the shift reg- C(j) Figure 3-1 7 The autocorrelation function of a maximal length sequence 15 -8 -6 -4 -2 2 4 6 8 Principles of Wideband CDMA (W-CDMA) 87 ister of Figure 3-1 6 However, this time, outputs of registers 2 and 0 are added modulo 2, and then applied to the input of the first stage as shown Figure 3-1 8 Assuming that the initial states of the register array are 1111, the subsequent states are... dB, whereas in W-CDMA with a bandwidth of 5 MHz, the signal experiences a Ϫ5.75 dB fade with the same probability Thus, fading is reduced with increased bandwidth When the signal bandwidth is sufficiently large, and multipath components are delayed by more than a chip period, they can be Principles of Wideband CDMA (W-CDMA) 95 10 9 8 Fade Depth (dB) Figure 3-2 4 The effect of channel bandwidth on fading... heavy lines, and those due to an incoming bit 1 by light lines Alongside each line is a 2-bit number that represents the output of the encoder Referring to Figure 3-9 , the initial state of the encoder at t ϭ 0 is 00 If now the input to the encoder is 0, the output is 00, and the new Principles of Wideband CDMA (W-CDMA) 75 Trellis Depth 3 Figure 3-9 The trellis diagram for the encoder of Figure 3-7 T 2T... Ϫ 1 Principles of Wideband CDMA (W-CDMA) 83 Thus, for example, H20 ϭ H1 ϭ 314 H21 ϭ H2 ϭ c 1 1 H22 ϭ H4 ϭ ≥ 1 1 1 0 1 0 1 1 1 1 0 0 1 d 0 1 0 ¥ 0 1 Channelization codes used in UMTS W-CDMA and cdma2000 are variable-length Walsh codes, also known as orthogonal variable spreading factor (OVSF) codes The spreading factors in UMTS may vary from 4 to 256 chips on uplink channels and from 4 to 512 chips... diagram of Figure 3-1 2(a) It consists of an interleaver and two encoders The two encoders, however, are identical and are built around a 3-bit shift register as Output 0 Figure 3-1 0 A convolutional encoder of rate 1/2 for UMTS + 7 6 5 4 3 2 1 0 + Output 1 Output 1 Figure 3-1 1 A convolutional encoder of rate 1/3 for UMTS + 7 6 5 3 4 2 + Output 2 + Output 3 1 0 Chapter 3 78 Figure 3-1 2 Turbo coder used... principle of PN codes, consider Figure 3-1 6 The shift register array consists of four single-bit shift registers and is clocked at the chip rate The outputs of registers 3 and 0 are added in a modulo-2 adder (that is, an exclusive-OR circuit), and then applied to the input of the array Assume that the initial states of all stages of the shift register are 1, 1, 1, and 1 Then at instant t ϭ 0, the output... are read into the encoder in 20 ms frames and are analyzed to derive the short-term coefficients of an LP filter of order 10.6 Because speech is encoded in public switched telephone networks (PSTNs) using the 8-bit A-law (or mlaw) companding schemes, the speech signal received from a PSTN is first converted into 13-bit uniform pulse code modulation (PCM), and then applied to the input of an ACELP encoder . 1 1 034 56 Figure 3- 1 0 A convolutional encoder of rate 1 / 2 for UMTS + + + 12 3 45 6 7 0 Output 1 Output 2 Output 3 Figure 3- 1 1 A convolutional encoder of rate 1 / 3 for UMTS shown in Figure 3- 1 2(b) input. 75 Principles of Wideband CDMA (W-CDMA) 00 01 10 11 t State T2T3T4T5T6T 10 01 10 11 11 00 00 00 00 00 10 10 00 01 01 Trellis Depth 3 Figure 3- 9 The trellis diagram for the encoder of Figure 3- 7 Decoding. Wideband CDMA (W-CDMA) Mobile Communications System Speech Coding Algorithm IS-54/IS- 136 Vector-Sum Excited Linear Predictive Coding (VSELP). The bit rate is 7.95 kb/s. The coder operates on 20-ms frames.