Abstract The adaptive antenna array is one of the advanced techniques which could be implemented in the IMT-2 mobile telecommunications systems to achieve high system capacity. In this paper, an integrated adaptive antenna array system for W-CDMA, which consists of both the uplink and the downlink beam-formers in the baseband, is described. The uplink beam-former is based on the concept of finger beam-former in which each dominant component of the multi-path signal is allocated to a dedicated beam, and the normalised least mean square algorithm (NLMS) is used as the beam-forming algorithm. In the meanwhile, a low complexity but high performance algorithm known as the iterative beam steering (IBS) is applied to the uplink signal to form a steering beam for the downlink. Simulation results show that, compared with using the conventional sectorial antenna, about four times as much system capacity can be achieved by employing a four-element adaptive antenna array. In addition, the effectiveness of the antenna array is demonstrated by experimental results. Y. J. Guo He has been with Fujitsu Europe Telecom R & D Centre, U. K., working on advanced BTS techniques for IMT-2. Marío A. Bedoya-Martinez He joined Fujitsu Europe Telecom R&D Centre (UK), where he has been working on R&D of Second-and Third-Generation mobile communication systems. Currently he is a Principal Engineer in Advanced Wireless Communications Section. Jahangir E. Austin He has now been with Advanced Wireless Communications Section of Fujitsu Europe Telecom R & D Centre Ltd. Principal Engineer responsible for the hardware development of Secondand Third-Generation mobile telecommunication systems. 66 FUJITSU.51, 1, pp.66-72 (1,2)
1. Introduction Currently, the mobile telecommunications industry is working on the 3 rd generation mobile telecommunications systems known as IMT-2 (International Mobile Telecommunications 2). IMT-2 is designed to provide wireless access to the global telecommunication infrastructure through both satellite and terrestrial systems, serving fixed and mobile users in public and private networks. One of the most promising members of the IMT-2 family is based on the wideband code division multiple access (W-CDMA) technique jointly developed in Japan and Europe. 1) Compared with the 2 nd generation mobile communications systems like GSM or PDC, the W-CDMA is expected to offer much higher system capacity and to accommodate larger number of high data rate (HDR) applications such as mobile multimedia and mobile computing. In a W-CDMA system, different users are allocated with different codes and all the active user signals are present all the time. For any given active user, the interference caused by other active users can be effectively rejected by virtue of the so-called processing gain. When the number of active users in a cell is very large, however, the total interference caused by many users to any particular user, which is known as multiple access interference (MAI), becomes so severe that the signal quality at the receiver degrades to an unacceptable level. A conventional solution to this problem is the sectorisation technique, which is to divide each cell into several sectors (normally three or six) and group the active users accordingly. Because different sectorial antennas are used in different sectors and each antenna beam is concentrated in its own sector, users in different sectors do not interfere with each other strongly and the total MAI experienced by each active user is significantly reduced. The problem with the conventional sectorisation techniques is that each sectorial beam is designed to support all the users in the sector so the beam may not be optimal for any individual one. When there is a HDR user in a sector, which is equivalent to many co-located speech users, the number of channels available to other users will be greatly decreased. Instead of having all the users in a sector share the same antenna beam, the adaptive antenna array is aimed to generate an optimal antenna beam pattern for each user. By creating a narrow beam pointed at the wanted user and a null in the direction of the HDR user, the total MAI experienced by the former can be effectively reduced. As a result, the system can accommodate a much larger number of HDR and common users. In addition, employing the adaptive antenna array can also reduce the outage at the cell edge and within buildings, and extend cell range. When used in the uplink (from mobiles to a base station), it can reduce the required mobile terminal transmit power and hence increase the battery life in terms of both talk time and standby time. It is envisaged that many W-CDMA systems will be based on the six-sector site configuration in order to guarantee high initial capacity, and adaptive antenna arrays will be introduced in the second phase of the system deployment when traffic congestion caused by HDR users becomes severe. To this end, a four-element linear adaptive antenna array with an inter-element spacing of one wavelength of the operating frequency is studied for W-CDMA base stations using six sectors in macro-cellular environments. The NLMS and IBS beam-forming algorithms are used for the uplink and the downlink (from a base station to mobiles) respectively, but the same wide-band antenna elements are employed for the two FDD (frequency division duplex) sub-bands. The two algorithms have four salient features: fast convergence, low complexity, stability and ease of implementation. Compared with using one sectorial antenna for each sector, about four times of as much system capacity can be achieved using the proposed technique. 2. System configuration The proposed adaptive antenna array system 67
consists of both the uplink and the downlink beamformers in the base band. The uplink beam-former employs the NLMS algorithm with fading compensation. The downlink beam is synthesised based on the angle of arrival (AoA) information and the downlink data rates required by mobile users, with the former being derived from the uplink signal using a new algorithm called iterative beam steering (IBS). The combination of the two beam-forming algorithms which are suited for two links provides an effective means of improving the system capacity of an IMT-2 system. 2.1 Uplink beam-forming The uplink beam-former employs the socalled finger beam-former configuration, in which a set of finger beams are formed to track different signal components of each active mobile user. 2) The number of fingers allocated to a user is determined by the number of signal components caused by the multipath in the channel, and the beam of each finger is locked to each path. A dynamic finger beam-former allocation scheme will be employed in the proposed system. In other words, a base station has a number of finger beamformers as a shared resource and up to four finger beam-formers will be activated by each active user. Compared with the common beam-former which employs only one common beam for all the multipath components, the finger beam-former configuration offers much better performance when the angular spread of the mobile signal is large, as a narrower finger beam can be produced to point at the AoA of each path of the wanted user signal. Another advantage of the finger beam-former configuration is that it can be operated at the symbol rate instead of the chip rate, so the power consumption can be reduced and the algorithm can be implemented using reprogrammable signal processing devices, which offers flexibility in fine-tuning and updating the algorithm at a later stage. A generic architecture of the proposed uplink beam-former using NLMS is shown in Figure 1, where there are a number of finger beamformers, each equipped with a dedicated NLMS algorithm and a channel estimator, and the blocks refer to correlators. Also, a combination of the pilot signal and tentative decisions is taken as the common reference signal. It is noted that signals at the output of the correlators are normalised to the mean signal intensity of each individual path before being fed into the beamformer. Furthermore, the reference signal is phase compensated before being subtracted from the output of the beam-former. These two functions are used to separate the fading compensation from the angular tracking of mobiles, thus improving the performance of the beam-former. 2.2 Downlink beam-forming Downlink beam-forming is inherently different from the uplink beam-forming. For the uplink, an optimum beam can be formed for each individual mobile to maximise the SINR experienced by the corresponding receiver at the base station. For the downlink, however, the quality of the signal received by each mobile depends on not only the beam patterns formed for the mobile and the associated transmit power, but also the beam patterns formed for other mobiles and the associated powers. In theory, therefore, all the beams should be jointly optimised in association with the transmit powers in order to achieve maximum SINR at each mobile and minimum interference NLMS Figure 1 NLMS based uplink beam-former. + - Ch Est MRC Combiner Tentative Decision Pilot 68
to other users. Unfortunately, such a global optimisation approach is not feasible in practice due to both the excessive signal processing complexity and the inaccuracy of downlink AoA estimation. Therefore, the following practical beam-forming approach is proposed. First, based on the signal received from the uplink, the IBS algorithm is used to estimate the AoA s of all the users. Second, based on the AoA s of each wanted user, a common steering beam is synthesised for it. It should be pointed out that although the propagation mechanism for the uplink is similar to that for the downlink, the angular positions of the dominant paths are generally different in the two cases because of the frequency difference in the FDD scheme. Therefore, a dedicated common beam for each user is required in the downlink to cover the spatial spread of the multi-path signals. As an illustration to the concept, a generic diagram of the downlink beam-former is shown in Figure 2. Compared with the uplink beam-former shown in Figure 1, the configuration of the downlink beam-former appears to be much simpler, which is due to the fact that major signal processing is performed by the AoA estimator. 3. Simulation results In order to obtain a quantitative capacity measure of the W-CDMA system employing adaptive antennas, a large number of multi-user simulations have been conducted using the parameters specified in the 3GPP document. 3) In the simulations, it is assumed that all the users are randomly located according to an uniform distribution in each sector, and the multi-path components of each user signal are spatially distributed according to a Gaussian distribution with 2.5º standard deviation. The power control mechanism is modelled as a Gaussian process with a 2.3 db standard deviation. The ITU Vehicular B model is used for all the radio channels and the Doppler frequency is assumed to be 8 Hz. The processing gain is chosen as 128, which corresponds to a 32 kb/s raw data rate, and a convolutional coding scheme with R = 1/3 and K = 9 is used. Previous study showed that the inter-element spacing is an important parameter in the design of an adaptive antenna array. 4) In the proposed system, this parameter is chosen as one wavelength of the operating frequency to achieve high capacity. Figure 3 shows the required E b /N o per antenna branch for different number of simultaneous speech channels which the adaptive antenna array can support in one sector to achieve the targeted bit error ratio (BER) of 1-3, where a practical antenna pattern with -3 db tapering at the sector boundaries is used. To account for discon- Data Code generator Weights Beam synthesis AoA AoA estimator No of speech channels per sector 35 3 25 2 15 1 5 Adaptive antenna 2 Branch diversity One antenna -2 2 4 6 8 1 12 14 16 Figure 2 Illustration of downlink beam-former. Uplink signal Required receive E b /N o per branch Figure 3 Capacity comparison of using one antenna, two branch diversity and adaptive antenna in uplink. 69
Tx power saving (db) 18 16 14 12 1 8 6 4 2 1 2 3 4 5 6 7 8 Sector load (Number of speech channels) Figure 4 Mobile transmit power saving in different loading conditions when using adaptive antenna array at base station compared with using single antenna. Signal intensity (db) -5-1 -15-2 -25 beam pattern wanted user HDR user -3-4 -3-2 -1 1 2 3 4 Angle of arrival (deg.) Figure 5 Formed beam pattern. tinuous transmission of voice signals, a voice activity factor of.5 has been assumed. For comparison, simulation results obtained using one antenna and two-branch diversity are also shown in the figure. It is seen that when only one antenna is used, the maximum number of speech channels which the system can support is about 82. The system capacity is increased to about 157 when two-branch diversity is employed. Using the adaptive antenna array, the maximum number of simultaneous speech channels the system can support reaches about 314. When the inter-cell and inter-sector interference is considered, it is found that the average system capacity becomes smaller than that shown in Figure 3, but the capacity achieved by using the adaptive antenna is still approximately four times of that achieved using one antenna. Figure 3 also reveals the advantage of using adaptive antennas for saving mobile transmit power. It is observed that in a system loaded with 8 speech users, using single antenna at the base station requires 15.8 db E b /N o to achieve a BER of 1-3, whereas with the adaptive antenna array only db E b /N o per branch is required to achieve the same BER. This implies that the transmit power of the mobile terminals can be reduced by 15.8 db, thus resulting in many fold increase in battery life. Alternatively, the communication range of the uplink could be extended by about 2.5 times under the same load. The mobile transmit power saving achieved in different loading conditions when using the adaptive antenna at the base station is shown in Figure 4. Another major advantage of using adaptive antennas is to accommodate HDR users. 4) The adaptive antenna can not only support much greater number of simultaneous HDR users by virtue of a narrow beam, but also reduce the strong interference caused by HDR users by nulling, thus increasing the overall system capacity. To illustrate this point, the following scenario has been studied. Assume that there are a.5 Mb/s HDR user located at -1º and a speech user located at 1º with 32 kb/s data rate. To the speech user, the HDR user is equivalent to a group of 3 co-located speech users interfering with it. When the adaptive antenna array is used, an optimum beam pattern can be so formed that the interference from the HDR user can be reduced to a negligible level. The beam pattern formed for the speech user is shown in Figure 5 and it is observed that a deep null is formed in the direction of the HDR user and the main beam is also shaped. Figure 6 shows the BER for the speech user achieved by using the adaptive antenna and also included is the BER achievable for a single user when there is no multi-user interference. It can 7
.1 1 User + 1 High data Rate interferer 1 user BER.1.1-4 -3-2 -1 E b /N o Figure 6 BER for speech user shown in Figure 5. Figure 8 Prototype model of adaptive antenna array. No of speech channels per sector 5 1 Antenna element 45 Adaptive antenna 4 35 3 25 2 15 1 5 4 6 8 1 12 14 16 18 2 22 Required E b /N o at receiver Figure 7 Capacity comparison of using one antenna and adaptive antenna array in downlink. be seen that, because of the use of the adaptive antenna array, the E b /N o required for the speech user to achieve the targeted BER of 1-3 in the presence of a HDR user is more or less the same as that for the single user case. In practice, this means that more active users can be accommodated in the sector and/or the power required for the speech user can be reduced. Figure 7 shows the capacity improvement when the proposed adaptive antenna array is employed in the downlink. It is observed again that employing the four-element antenna array improves the system capacity by about four times. Compared with Figure 3, it is seen that the downlink capacity for an isolated sector is greater than that of the uplink, which is due to the orthogonality of the downlink channel codes. 1) When inter-cell interference and the spectral occupancy of the common control channels are considered, 3) however, it is expected that the downlink capacity will be actually smaller than that of the uplink. 4. Prototype model Figure 8 shows a photograph of an experimental base station model where the proposed 4-element adaptive antenna array is implemented. Amongst various experiments, the generated antenna beam pattern as well as improvement of BER performance is being evaluated. Figure 9 shows generated directivity characteristics of the antenna array and Figure 1 shows the improvement of BER performance by the antenna array in presence of an equal power interferer. 5. Conclusions The architecture and beam-forming algorithms of an adaptive antenna array are described. The adaptive antenna consists of an uplink beamformer and a downlink beam-former, and is designed to operate in an existing six-sector site 71
BER Relative signal power (db) -1-2 -3-5 -4-3 -2-1 1 2 3 4 5 Direction of arrival (deg.) 1.E+ 1.E-1 1.E-2 1.E-3 1.E-4 1.E-5 1.E-6-125 1 path rayleigh fading 8 Hz 1 path static channel without FEC Desired user AoA = deg. -12-115 -11-15 -1-95 Received power of desired signal (dbm) in one antenna element Figure 1 BER performance improvement. Measured Theoratical Figure 9 Directivity characteristics of 4-element antenna array. Single element, equal power interferer Beamformer, equal power interferer AoA = deg. Beamformer, equal power interferer AoA = deg. Beamformer without interferer to improve system capacity. The uplink employs the finger beam-former configuration and the NLMS algorithm is used. The IBS algorithm is operated on the uplink signal to form a steering beam for the downlink. Simulation and experimental results demonstrate the effectiveness of using the adaptive antenna array to increase system capacity by a factor of four and to support HDR users in W-CDMA. References 1) F. Adachi, M. Sawahashi, and H. Suda: Wideband DS-CDMA for Next Generation Mobile Communications Systems. IEEE Communications Magazine., pp.56-69, Sept. 1998. 2) A. J. Paulraj and C. B. Papadias: Space-Time Processing for Wireless Communications. IEEE Signal Processing Magazine., pp.49-83, Nov. 1997. 3) Technical Specification, V2, Working Group 1 (WG1), Technical Specification Group (TSG), Radio Access Network (RAN), 3 rd Generation Partnership Project (3GPP), September 1999. 4) Y. J. Guo, M. Davies. M. Zarri, S. Vadgama, and E. Fukuda: Improving the System Capacity of UMTS Using Digital Beamformer. Proceedings of European Wireless 99, Munich, Germany, Oct. 1999. 72