
- •Contents
- •Preface
- •1 Spread spectrum signals and systems
- •1.1 Basic definition
- •1.2 Historical sketch
- •2 Classical reception problems and signal design
- •2.1 Gaussian channel, general reception problem and optimal decision rules
- •2.2 Binary data transmission (deterministic signals)
- •2.3 M-ary data transmission: deterministic signals
- •2.4 Complex envelope of a bandpass signal
- •2.5 M-ary data transmission: noncoherent signals
- •2.6 Trade-off between orthogonal-coding gain and bandwidth
- •2.7 Examples of orthogonal signal sets
- •2.7.1 Time-shift coding
- •2.7.2 Frequency-shift coding
- •2.7.3 Spread spectrum orthogonal coding
- •2.8 Signal parameter estimation
- •2.8.1 Problem statement and estimation rule
- •2.8.2 Estimation accuracy
- •2.9 Amplitude estimation
- •2.10 Phase estimation
- •2.11 Autocorrelation function and matched filter response
- •2.12 Estimation of the bandpass signal time delay
- •2.12.1 Estimation algorithm
- •2.12.2 Estimation accuracy
- •2.13 Estimation of carrier frequency
- •2.14 Simultaneous estimation of time delay and frequency
- •2.15 Signal resolution
- •2.16 Summary
- •Problems
- •Matlab-based problems
- •3 Merits of spread spectrum
- •3.1 Jamming immunity
- •3.1.1 Narrowband jammer
- •3.1.2 Barrage jammer
- •3.2 Low probability of detection
- •3.3 Signal structure secrecy
- •3.4 Electromagnetic compatibility
- •3.5 Propagation effects in wireless systems
- •3.5.1 Free-space propagation
- •3.5.2 Shadowing
- •3.5.3 Multipath fading
- •3.5.4 Performance analysis
- •3.6 Diversity
- •3.6.1 Combining modes
- •3.6.2 Arranging diversity branches
- •3.7 Multipath diversity and RAKE receiver
- •Problems
- •Matlab-based problems
- •4 Multiuser environment: code division multiple access
- •4.1 Multiuser systems and the multiple access problem
- •4.2 Frequency division multiple access
- •4.3 Time division multiple access
- •4.4 Synchronous code division multiple access
- •4.5 Asynchronous CDMA
- •4.6 Asynchronous CDMA in the cellular networks
- •4.6.1 The resource reuse problem and cellular systems
- •4.6.2 Number of users per cell in asynchronous CDMA
- •Problems
- •Matlab-based problems
- •5 Discrete spread spectrum signals
- •5.1 Spread spectrum modulation
- •5.2 General model and categorization of discrete signals
- •5.3 Correlation functions of APSK signals
- •5.4 Calculating correlation functions of code sequences
- •5.5 Correlation functions of FSK signals
- •5.6 Processing gain of discrete signals
- •Problems
- •Matlab-based problems
- •6 Spread spectrum signals for time measurement, synchronization and time-resolution
- •6.1 Demands on ACF: revisited
- •6.2 Signals with continuous frequency modulation
- •6.3 Criterion of good aperiodic ACF of APSK signals
- •6.4 Optimization of aperiodic PSK signals
- •6.5 Perfect periodic ACF: minimax binary sequences
- •6.6 Initial knowledge on finite fields and linear sequences
- •6.6.1 Definition of a finite field
- •6.6.2 Linear sequences over finite fields
- •6.6.3 m-sequences
- •6.7 Periodic ACF of m-sequences
- •6.8 More about finite fields
- •6.9 Legendre sequences
- •6.10 Binary codes with good aperiodic ACF: revisited
- •6.11 Sequences with perfect periodic ACF
- •6.11.1 Binary non-antipodal sequences
- •6.11.2 Polyphase codes
- •6.11.3 Ternary sequences
- •6.12 Suppression of sidelobes along the delay axis
- •6.12.1 Sidelobe suppression filter
- •6.12.2 SNR loss calculation
- •6.13 FSK signals with optimal aperiodic ACF
- •Problems
- •Matlab-based problems
- •7 Spread spectrum signature ensembles for CDMA applications
- •7.1 Data transmission via spread spectrum
- •7.1.1 Direct sequence spreading: BPSK data modulation and binary signatures
- •7.1.2 DS spreading: general case
- •7.1.3 Frequency hopping spreading
- •7.2 Designing signature ensembles for synchronous DS CDMA
- •7.2.1 Problem formulation
- •7.2.2 Optimizing signature sets in minimum distance
- •7.2.3 Welch-bound sequences
- •7.3 Approaches to designing signature ensembles for asynchronous DS CDMA
- •7.4 Time-offset signatures for asynchronous CDMA
- •7.5 Examples of minimax signature ensembles
- •7.5.1 Frequency-offset binary m-sequences
- •7.5.2 Gold sets
- •7.5.3 Kasami sets and their extensions
- •7.5.4 Kamaletdinov ensembles
- •Problems
- •Matlab-based problems
- •8 DS spread spectrum signal acquisition and tracking
- •8.1 Acquisition and tracking procedures
- •8.2 Serial search
- •8.2.1 Algorithm model
- •8.2.2 Probability of correct acquisition and average number of steps
- •8.2.3 Minimizing average acquisition time
- •8.3 Acquisition acceleration techniques
- •8.3.1 Problem statement
- •8.3.2 Sequential cell examining
- •8.3.3 Serial-parallel search
- •8.3.4 Rapid acquisition sequences
- •8.4 Code tracking
- •8.4.1 Delay estimation by tracking
- •8.4.2 Early–late DLL discriminators
- •8.4.3 DLL noise performance
- •Problems
- •Matlab-based problems
- •9 Channel coding in spread spectrum systems
- •9.1 Preliminary notes and terminology
- •9.2 Error-detecting block codes
- •9.2.1 Binary block codes and detection capability
- •9.2.2 Linear codes and their polynomial representation
- •9.2.3 Syndrome calculation and error detection
- •9.2.4 Choice of generator polynomials for CRC
- •9.3 Convolutional codes
- •9.3.1 Convolutional encoder
- •9.3.2 Trellis diagram, free distance and asymptotic coding gain
- •9.3.3 The Viterbi decoding algorithm
- •9.3.4 Applications
- •9.4 Turbo codes
- •9.4.1 Turbo encoders
- •9.4.2 Iterative decoding
- •9.4.3 Performance
- •9.4.4 Applications
- •9.5 Channel interleaving
- •Problems
- •Matlab-based problems
- •10 Some advancements in spread spectrum systems development
- •10.1 Multiuser reception and suppressing MAI
- •10.1.1 Optimal (ML) multiuser rule for synchronous CDMA
- •10.1.2 Decorrelating algorithm
- •10.1.3 Minimum mean-square error detection
- •10.1.4 Blind MMSE detector
- •10.1.5 Interference cancellation
- •10.1.6 Asynchronous multiuser detectors
- •10.2 Multicarrier modulation and OFDM
- •10.2.1 Multicarrier DS CDMA
- •10.2.2 Conventional MC transmission and OFDM
- •10.2.3 Multicarrier CDMA
- •10.2.4 Applications
- •10.3 Transmit diversity and space–time coding in CDMA systems
- •10.3.1 Transmit diversity and the space–time coding problem
- •10.3.2 Efficiency of transmit diversity
- •10.3.3 Time-switched space–time code
- •10.3.4 Alamouti space–time code
- •10.3.5 Transmit diversity in spread spectrum applications
- •Problems
- •Matlab-based problems
- •11 Examples of operational wireless spread spectrum systems
- •11.1 Preliminary remarks
- •11.2 Global positioning system
- •11.2.1 General system principles and architecture
- •11.2.2 GPS ranging signals
- •11.2.3 Signal processing
- •11.2.4 Accuracy
- •11.2.5 GLONASS and GNSS
- •11.2.6 Applications
- •11.3 Air interfaces cdmaOne (IS-95) and cdma2000
- •11.3.1 Introductory remarks
- •11.3.2 Spreading codes of IS-95
- •11.3.3 Forward link channels of IS-95
- •11.3.3.1 Pilot channel
- •11.3.3.2 Synchronization channel
- •11.3.3.3 Paging channels
- •11.3.3.4 Traffic channels
- •11.3.3.5 Forward link modulation
- •11.3.3.6 MS processing of forward link signal
- •11.3.4 Reverse link of IS-95
- •11.3.4.1 Reverse link traffic channel
- •11.3.4.2 Access channel
- •11.3.4.3 Reverse link modulation
- •11.3.5 Evolution of air interface cdmaOne to cdma2000
- •11.4 Air interface UMTS
- •11.4.1 Preliminaries
- •11.4.2 Types of UMTS channels
- •11.4.3 Dedicated physical uplink channels
- •11.4.4 Common physical uplink channels
- •11.4.5 Uplink channelization codes
- •11.4.6 Uplink scrambling
- •11.4.7 Mapping downlink transport channels to physical channels
- •11.4.8 Downlink physical channels format
- •11.4.9 Downlink channelization codes
- •11.4.10 Downlink scrambling codes
- •11.4.11 Synchronization channel
- •11.4.11.1 General structure
- •11.4.11.2 Primary synchronization code
- •11.4.11.3 Secondary synchronization code
- •References
- •Index

300 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Spread Spectrum and CDMA |
||||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||
|
|
q1n(bi) |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
Deinterleaver |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
q2n(bi) |
|
|
|
|
|
|
|
|
|||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||
|
y2 |
|
|
|
First code |
|
|
Interleaver |
|
|
|
|
|
|
Second code |
|
|
|
Λ |
> |
1 |
bi |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||||||||
|
|
|
|
|
decoder |
|
|
|
|
|
|
|
|
decoder |
|
|
|
i < |
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|||
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
y1
Interleaver
y3
Figure 9.13 Iterative turbo decoder
and check symbols of this code. As initial information it uses a uniform a priori distribution q11(bi) ¼ 1/2, bi ¼ 0, 1, since setting all data bit patterns a priori equiprobable is natural. After this, the decoder of the second code computing a posteriori probabilities p(bijy1, y3) may rest not only on appropriate observations y1, y3 but also on the information delivered by the first decoder, using its a posteriori distribution p1(bijy1, y2) as the a priori one: q21(bi) ¼ p1(bijy1, y2). The result is the first approximation p1(bijy) of a posteriori distribution p(bijy). Since at this step the first decoder was not supported by the information of the second one, this is done at the second iteration, where the first decoder again decodes the first component code but using a priori distribution q12(bi) ¼ p1(bijy). Going on in this way, after the nth step the next approximation pn(bijy) of p(bijy) is formed by the second decoder and used by the first one as the next a priori distribution q1nþ1(bi) to output pnþ1(bijy1, y2). This latter in its turn is used by the second decoder as the next a priori distribution q2nþ1(bi) to produce the next approximation of the desired a posteriori probabilities pnþ1(bijy), etc. Since the interleaver permutes data bits before inputting the second encoder the interleavers permute in the same way observations y1 and a priori distribution q2n(bi) ¼ pn(bijy1, y2) entering the second decoder. Similarly, the deinterleaver restores the original order of bits in a feedback passing on pn(bijy) from the second decoder output to the first decoder input. With these rearrangements, all data processed by both decoders are aligned properly.
A vast simulation has confirmed the convergence of these iterations experimentally, although theoretical justification remains a matter of question.
9.4.3 Performance
As mentioned before, turbo codes were the first regular codes to provide reliable data transmission over the band-limited channel at near-capacity rates and low energies per bit. To illustrate it by examples, let us first examine some fundamental limits on BPSK data transmission. On the strength of the sampling theorem any bandpass signal of bandwidth W is a vector of dimension 2WT (see Section 2.3). In the case of BPSK each component of such a vector may take on only two values, implying that within the time–frequency resource WT the number M of BPSK signals obeys the bound M 22WT , or equivalently, no more than 2WT data bits may be transmitted. This
restricts the transmission rate achievable with |
BPSK within |
the |
bandwidth |
W as |
R ¼ ( log M)/T 2W, imposing, in its turn, |
the bound on |
the |
rate per |
Hertz: |
Channel coding in spread spectrum systems |
301 |
|
|
R/W 2 bps/Hz. Now consider a binary code with rate Rc ¼ 1/2, meaning that only every second component of a signal vector carries data, the rest being assigned to check symbols (two signal samples are spent per data bit), so that the ratio between data rate and bandwidth R/W ¼ 1 bps/Hz. Turning to Shannon’s bound (1.2), we may see that the minimum bit energy normalized to noise Eb/N0 (half of power bit SNR) necessary to provide errorless transmission over an AWGN channel at such a rate equals 0 dB. We are, however, now dealing not with an arbitrary Gaussian channel but with one whose input symbols are limited to the BPSK alphabet (Gaussian channel with binary input). This restriction increases the minimal Eb/N0 corresponding to R/W ¼ 1 bps/Hz up to 0.19 dB [97]. The turbo code of constraint length 5 and block length 65 536 proposed in [95,96] secures bit error probability Pb < 10 5 (this figure is often referred to as the practical criterion of wireless errorless operation) at Eb/N0 0:7 dB, i.e. yielding only about 0.5 dB to Shannon’s limit. At the moment of their publication these results seemed fantastic, since a long unsuccessful history had made many experts believe that finding deterministic coding rules allowing operation near Shannon’s bound was a hopeless task. Following the original works [95,96], other efficient turbo as well as serial concatenation codes have been found (see bibliography in [97]).
It should be noted that the asymptotic (Eb/N0 ! 1) behaviour of turbo codes is not better than that of convolutional codes of the same rate and memory, since they do not possess any advantage in minimum distance. Turning to (2.23), we may see that asymptotically (with growth of SNR) the effect of multiplicity nmin, i.e. the number of signals with minimum Euclidean distance dmin from a transmitted one, plays a secondary role against dmin itself due to the exponential drop of the Q-function with SNR (take the logarithm of Pe to make sure of it). For this reason dependence of Pe on Eb/N0 sooner or later achieves a ‘floor’ character determined by dmin and analogous to the one typical of other codes with the same minimum distance. This, however, occurs at SNR values securing very small bit error probabilities, falling far beyond the range of practical needs. The explanation of why turbo codes guarantee such excellent operational quality at low bit SNR is not in their large minimum distance, but rather in the relatively small number of words lying from each other at small distances, in particular the small multiplicity nmin in (2.23). This redistribution of distances towards a greater number of bigger ones versus convolutional codes happens due to the pseudorandom interleaving of data bits encoded by the second component encoder. If the data bit pattern is unlucky to generate a small-weight word of the first component code, its permutation may appear different enough to produce the second component word of a remarkably higher weight.
9.4.4 Applications
Despite their short history, turbo codes are now in widespread use and enter the specifications of many systems. The most interesting in our context is their involvement in 3G mobile radio standards. The UMTS specification includes turbo codes of rate 1/3 based on two component convolutional codes of constraint length 4 and interleaver of variable length in the range from 40 to 5114 [92,97]. The cdma2000 standard also contains two-component turbo codes of constraint length 4 with interleaver size ranging from 250 to 4090. Appropriate puncturing allows rates of 1/2, 1/3, 1/4 or 1/5 to be obtained [69,97].