Tài liệu Điện thoại di động băng thông rộng không dây P9 doc

5.3.3 Space-time Equaliser Using Adaptive Antennas Space-time equalisers using the adaptive antennas with equalisers have also recently inves-tigated a more powerful technique than only

† Least-mean square (LMS) algorithm

† Recursive least squares (RLS) algorithm

† Sample matrix inversion (SMI) algorithm

The SMI algorithm, which is also known as direct matrix inversion (DMI) algorithm, has recently been used for 3G systems and beyond, because the fast convergence property makes

it suitable for use with high data rate transmissions [143,150] However the complexity grows three-orders exponentially with the number of the weights (M3) Recursive equations for the inverse of the correlation matrix thus had been used for the implementation on digital signal processors

5.3.3 Space-time Equaliser Using Adaptive Antennas

Space-time equalisers using the adaptive antennas with equalisers have also recently inves-tigated a more powerful technique than only using the adaptive antennas [13,36–39,46,63, 104,143,144,150] They are also called smart antennas, or intelligent antennas

Nonlinear adaptive equalisers such as decision-feedback equaliser (DFE) and maximum-likelihood sequence estimator (MLSE) had been investigated and implemented on commer-cial systems to compensate for the ISI However, DFE cancels undesired delayed paths by subtracting replica from received signal and thus cannot obtain path diversity gain

MLSE, which is well-known optimum equaliser and can be implemented by Viterbi

algo-Figure 5.22 Frame structures: (a) W-CDMA and (b) EDGE

rithm (VA), but instead can obtain path diversity gain by exploiting the delayed path infor-mation MLSE thus can have higher efficiency than DFE under the multipath-rich environ-ment However, the longer the span of the multipath, the more complicated the hardware implementation of the VA with the exponential behaviour in complexity as a function of the span of the ISI Adaptive antennas on the other hand can suppress the relatively longer-delayed paths without the hardware overhead, though adaptive antennas cannot obtain path diversity gain in the same way as the DFE The joint signal processing of the adaptive antennas and the equalisers can thus mutually compensate their drawbacks and provide higher transmission quality and capacity

Figure 5.23 shows a block diagram of the space-time equaliser [35,142–144] The scheme proposed in [36,37] consists of a couple of adaptive antenna array processors and the branch-metric combining maximum-likelihood sequence estimator (MLSE) Here the first arrival and one-symbol-delayed path components are treated as desirable Other longer delayed path components are suppressed as undesirable Each array-processor combines four space-diver-sity branches to maximise the signal-to-interference-plus-noise ratio (SINR) of the first arrival and the one-symbol-delayed path components One array processor combines space-diversity branches to pass the one-symbol-delayed path component into the array output with constrained first arrival path component while suppressing other longer delayed path components

Likewise, the other array combines space-diversity branches to pass the first arrival path component into the array output with constrained one-symbol-delayed path component while suppressing the other longer delayed path components Consequently, each array processor extracts both first arrival and one-symbol-delayed path components, whose SINRs in both diversity branches are improved Mean-square-error between the array outputs and the repli-cas are weighted with branch-metric-combining coefficients then combined and input to MLSE The adjustable weights in antenna array and one-symbol-delayed tap-coefficients

in the array-output-replica generator are estimated in adaptive weight controller using constrained-MMSE-criterion-based algorithm The branch-metric-combining coefficients

Figure 5.23 Space-Time equaliser

can be estimated based on the quality of each diversity branch By using the pair of array processors and the branch-metric-combining method for MLSE, a sufficient path diversity effect can be obtained when the phase differences between first arrival paths to antennas are significantly different from those on the one-symbol-delayed paths to antennas

5.3.4 Implementation of the Space-time Equaliser

The recent boom of hardware implementations of adaptive antennas and space-time equali-sers may be caused by recent advances of reconfigurable hardware such as central processor units (CPUs), digital-signal-processors (DSPs), field-programmable gate arrays (FPGAs) Adaptive antennas using digital array processing is thus also called software antennas, because the digital array processing can be implemented on those programmable devises

by software such as binary pattern We had developed an experimental system using CPUs, DSPs and FPGAs, and then evaluated the performances of the adaptive antenna and the space-time equaliser [35,142,143] Figure 5.24 shows a photograph of the experimental system and Table 5.2 describes the main specifications of the system

A lot of time and effort are still required for the development of the experimental systems for adaptive antennas and space-time equalisers, though the recent advance in the digital signal processors We therefore developed a real-time operating system (RTOS) embedded fully programmable system for easy implementations of various space and time processing and also to carry them out simultaneously for comparison in real time

Figures 5.25 and 2.56 show the experimental results of the adaptive antenna and the space-time equaliser Figure 5.25 illustrates bit error rate (BER) performances under

frequency-Figure 5.24 Photograph of the experimental system

Figure 5.25 Experimental result of the adaptive antenna (AA) and space-time equaliser (AA 1 BMC-MLSE) under frequency-selective fading channels

Figure 5.26 Delay time difference characteristics (3-path model)

selective fading channels, in which the number of arrival paths ranges from one to five and the path delays are fixed multiples of the symbol period (0, 1Ts, 2Ts, 3Ts, 4Ts) The average power

is equal along all paths of each antenna The received signal power represents the total power arriving along all paths, per antenna Therefore, the average desired power on each path is 1/L (L: the number of arrival paths) of the total signal power received at each antenna For the one-path model, that is, a flat fading channel, the measured BERs of both schemes are almost equal to a theoretical BER of the four-antenna maximal-ratio-combining (MRC) For the two-path model, the space-time equaliser has an improved BER because both space-and two-path diversity effects are obtained from signals on first and one-symbol-delayed paths The BER of the adaptive antenna instead fell as one of its degrees of freedom is consumed in suppressing the signal on the one-symbol-delayed path For the three-, four-, and five-path models, the pace-time equaliser has a significantly lower BER than that of the adaptive antenna The space and path diversity effects are especially true for the five-path model because all the degrees of freedom of the adaptive antenna are used up

Figure 5.26 shows the delay time difference characteristics in the three-path model, where the delay time for the second path is set to one symbol and the delay time for the third path is varied from zero to 6Ts The BER of the adaptive antenna increases as the difference in delay times increases When the space-time equaliser is used, however, the BER keeps low in the range from zero to one-symbol delay because of the space and path diversity effects from signals on both first arrival and one-symbol-delayed paths A one-symbol-delayed path does not always exist in real channels However, the one-symbol-delayed path may be able to be produced by a delay-diversity technique [151]

5.3.5 CDMA Adaptive Array Antennas

Application of adaptive array antennas is now under consideration for CDMA systems [2,5, 12,48,49,72,88–90,99,139] Using adaptive antennas at the base station, we can reduce co-channel interference, and increase the capacity of CDMA systems Furthermore, terminals in

Table 5.2 Specifications of the experimental system

Radio channel

Carrier frequency RF/IF 3.35 GHz/245 MHz

Modulation method QPSK

Transmission rate 4.096 Mb/s

Pulse shaping Root Nyquist filter (a ¼ 0.5)

Array signal processing

Number of antennas 4

DSP SHARC ADSP2106 (129 MFLOPS) £ 8 (Max 40) Real-time operating system VxWorks 3.5.1

Viterbi equaliser (VA)

Number of VA states 4 states

VA path memory length 10 symbols

different angular positions can be served on the same channel with little interference if they have sufficient angular separation

Many investigations have been performed on antenna arrays of CDMA systems, including capacity evaluation, call admission control, and signal processing techniques Most recent investigations have focused on space-time processing executed by means of antenna arrays and a RAKE receiver In the literature [90], the spatial matched filter is performed before the despreading process and the filter outputs are despread and coherently combined by a RAKE combiner In [99], joint space-time auxiliary-vector filtering is employed in the presence of multiple-access interference In literatures [2,5,12,139], whole space-time processing is performed after the despreading process The signals in different antennas are despread using the sequence of the desired terminal where the despread signal is composed of multiple delay paths The spatial signal processing is performed for each delay path and the outputs of the spatial processors are combined by the RAKE combiner In the spatial processors, the optimum weight vectors are given by the Wiener-Hopf solution In [2,139], the optimal weight vectors are obtained by the normalised least mean square (LMS) algorithm with pilot symbol-assisted decision-directed coherent adaptive array diversity (PSA-CAAD) Recently, NTT DoCoMo, Japan, has carried out field experiments and laboratory experiments

of PSA-CAAD with 1.990.5-MHz carrier frequency, 32-kbps information bit rate, 4.096-Mcps chip rate, and Rayleigh fading environments

A recent example of technology in other literatures includes multi-user adaptive arrays with a common correlation matrix (CCM) [49], in which one common correlation matrix is used to calculate the optimal weight vectors for multiple users Multi-user adaptive arrays with CCM can significantly decrease the computational complexity of a base station serving a number of active terminals

Another topic of CDMA systems with adaptive antennas is call admission control (CAC) With CAC, a new call is admitted if there is an available channel; otherwise the call is blocked Since the beam pattern of an adaptive array differs terminal by terminal, a new terminal may suffer from co-channel interference even if another new terminal with a differ-ent direction does not Therefore, the direction of the terminal must be considered in CAC In [48], the CAC procedure is carried out by estimating new terminal’s signal-to-interference-plus-noise ratio (SINR) at the output of adaptive array The admission of new terminal is determined based on the estimated SINR

CDMA systems with base-station adaptive arrays are expected to achieve a capacity about 20–30% greater than that of systems with antenna diversity More precise capacity evaluation will be required in future research

5.3.6 SDMA (Spatial Division Multiple Access)

The basic concept of spatial division multiple access (SDMA) [19,26,27,29–33,47,103,116, 147,158] is channel reuse within a cell With the use of adaptive arrays at the base station, terminals in different angular positions can share the same time slot reducing the power of other terminals’ signals Therefore, the SDMA system is an attractive scheme to increase the capacity

of mobile communication systems So far, the RACE TSUNAMI [147] project had field trial demonstration of both receive and transmit digital beamforming supporting SDMA systems In addition, many literatures described beamforming methods, assignment algorithms, and power control in SDMA systems Let us address some topics of these investigations

Figure 5.27 shows an example of a base station structure with L-branch adaptive antennas for an SDMA/TDMA system which N communication time slots Spatially separated Kn (, L) terminals within a cell share the same time slot n (n ¼ 1,2, ,N) as shown in Figure 5.28 The base station has a channel situation list (Figure 5.29), which stores the covariance matrix (Rn), the number of active terminals (Kn), and the received signal vector of each active terminal (Unk) for each time slot n The covariance matrix Rncan be obtained by calculating the autocorrelation coefficients of total received signals The matrix Rnincludes the inter-ference from the outer cell as well as the active terminal’s signal The parameters in the channel situation list are updated at a specific time intervals

In uplink, we can use the optimal weight vector for each user, i.e., Wiener-Hopf solution,

wnk¼ Rn21Unk In the weight vector calculation process, the channel situation list is referred to get the information of Rn21and Unk In contrast, the downlink optimal weight vector is difficult

to solve because the problem includes a nonlinear constrained optimisation problem Farsakh [29] and Zetterberg [158] demonstrated feasible downlink beamforming methods to reduce interference for higher frequency reuse

When the base station receives a new call request signal, it searches for an available time slot to assign If there is no available time slot, the new terminal is blocked Careful time slot assignment can minimise the blocking probability and allow the SDMA system greater capacity From such a point of view, a number of algorithms have been proposed for channel

or time slot assignment in an SDMA system Farsakh [28] described assignment algorithms based on spatial correlation coefficients Piolini [103] studied an assignment scheme with cost coefficients Shad [116] and Chen [19] provided assignment algorithms based on SINR In these investigations, algorithms based on SINR have advantages in managing new terminal’s signal quality easily because SINR is closely related to signal quality or bit error rate (BER)

Figure 5.27 SDMA/TDMA system with adaptive antennas

Recently, a time slot assignment algorithm based on estimated SINR has also been proposed [47] This algorithm estimates output SINRs of adaptive arrays for new terminal and for active terminals on the assumption that the new terminal is assigned to a specific time slot The estimated SINR of a new terminal for time slot n is represented by:

gn0 ¼ UH

0 R21n U0

Figure 5.28 Base station structure

Figure 5.29 Data structure of channel situation list

Here, U0represents a modified signal vector of a new call request signal and H denotes transpose conjugate The estimated SINR of active terminal k in time slot n is given by:

gnk ¼ UH

0 ðRn 1 U0 UH0 - Unk UnkHÞ21 U0

By using estimated SINRs, the proposed algorithm attempts a new call request signal, it calculates the signal vector of the new terminal and estimates SINRsgn0 to search for an available time slot considering not only the signal quality of the new terminal, but also the signal quality of active terminals Figure 5.30 shows a flowchart of highest SINR algorithm The time slots are sorted in order of magnitude of the estimated SINRs of the new terminal The base station begins to examine whether the time slot with the largest estimated SINR

of the new terminal is available If all the estimated SINRs of active terminals are above the required SINRgreq, the new terminal is assigned to the time slot Otherwise, the assignment process continues to the next time slot according to the time slot ordering until an available time slot is found If no time slot is available, the new terminal is blocked

In this algorithm, the active terminals are always guaranteed to have a suitable SINR after the time slot assignment process Therefore, the required signal quality is always maintained not only for the new terminal, but also for the active terminals Furthermore, performance

Figure 5.30 Flow chart of highest SINR slot assignment algorithm

evaluation shows that these time slot assignment algorithms have significantly better perfor-mance than sectored systems

Thus, both uplink time slot assignment and downlink slot allocation are important More consideration will need to be given to SDMA, including multimedia data transmission, in future research

5.3.7 Summary

This section briefly described basic concepts of adaptive antennas, and also introduced a space-time equaliser Furthermore, we present the recent investigations of CDMA adaptive array antennas and SDMA systems Adaptive antennas can be one of the key technologies in 3G wireless and beyond, and be put into practical use in several years time

5.4 Multiple Access Schemes

Studies on the concept of the 4G system (beyond IMT-2000), which will be the next genera-tion of mobile communicagenera-tion, are advanced now and the key technology has been examined Because more users will need transmission with a high bit rate and large capacity in such mobile communication systems in the future, the selection of a multiple access scheme is important as well as modulation and demodulation Moreover, the maximum transmission bit rate will be 20–100 Mbit/s and the transmission bit rate in the reverse link will be higher than

in the forward link In addition, the importance of transmitting IP packets has been shown by the development of recent Internet technology The system construction must be compatible with these technologies

The application of a powerful error correcting code and a multi-level modulation technol-ogy which increases the amount of information transmitted per symbol, is being studied to reduce multipath fading degradation, which becomes a problem in high bit rate transmission

in mobile communication To increase the transmission capacity, it is necessary to use parallel transmission, which allows simultaneous access by several users, and high efficiency modulation

In this section, we describe a recent study on transmission technology that focuses on multiple access Because it efficiently accommodates a lot of users, the multiple access method is important Code division multiple access (CDMA) is used in IMT-2000 For time division multiple access (TDMA), a lot of research on technology to counter fading has been studied It also has the advantage of making the system configuration comparatively easily Orthogonal frequency division multiplexing (OFDM) is used for digital broadcasting and is being researched actively TDMA and its combination with packet transmission and multi-carrier CDMA are being examined, though it is not possible to achieve multiple access with the OFDM unit It is being paid attention because the modulation method such a multi-carrier technique offers excellent bit rate and frequency availability

The transmission capacities of the forward and reverse links are expected to become asym-metrical because transmission in the reverse link is increasing Time division duplex (TDD), which changes the occupation time, and frequency division duplex (FDD), which changes the frequency band, are typical methods for dealing with this They give the system different timings for the start times for the forward and reverse links, and there are problems such as greater influence of interference from other cells, though TDD is more flexible than FDD