REAL TIME IMPLEMANTATION OF LMS BEAMFORMER FOR cdma2000 3G SYSTEM USING TI TMS320C6701 DSP

Size: px

Start display at page:

Download "REAL TIME IMPLEMANTATION OF LMS BEAMFORMER FOR cdma2000 3G SYSTEM USING TI TMS320C6701 DSP"

Gregory Brown
7 years ago
Views:

1 REAL TIME IMPLEMANTATION OF LMS BEAMFORMER FOR cdma2000 3G SYSTEM USING TI TMS320C6701 DSP Kerem Küçük, Mustafa Karakoç, and Adnan Kavak + Kocaei University + Kocaei University Eectronics and Computer Ed. Dept. Computer Eng. Dept. İzmit, Turkey İzmit, Turkey E-mais: kkucuk, mkarakoc, akavak@kou.edu.tr Key words: Beamforming, DSP, LMS, smart antenna, software radio 3G system ABSTRACT Digita Signa Processor (DSP) is an emerging technoogy that offers great opportunities for the deveopment of future wireess communications systems. In this paper, we demonstrate an impementation of rea time adaptive beamforming for cdma2000 reverse ink using TI C6701 DSP. We consider that the DSP executes east-mean-square (LMS) agorithm for the beamformer weight vector computation. The resuts show that the system can hande computation of optimum weight vector within reasonabe time for 3G appications. The computed weight vector is observed to steer the main beam of the antenna array towards the desired signa direction. Compexity and execution time of the agorithm increase with the number of antenna eements. I. INTRODUCTION Smart antennas [1] that compose of an antenna array and advanced signa processing modue in the baseband are being regarded by many as the key soution for increasing the spectra efficiency and improving the system performance in mobie communications. In principe, an antenna is smart ony when it can recognize and track the signa of a mobie user whie suppressing interfering signas. The antenna array empoys adaptive beamforming agorithms for recognize, track and suppress the interference. In this study, we investigate the feasibiity of impementing an adaptive beamforming agorithm, namey Least Mean Square (LMS) beamforming method, utiizing TI C6701 foating point DSP for rea time cdma2000 appications. The simuation system consists of a PC for data generation, Matab s TI code composer toobox for interfacing modue, and TI TMS320C6701 evauation modue (EVM) for the beamformer weight vector computation. Simuations are performed as the foowing steps: First, received data in cdma2000 reverse ink format [2] is generated using Matab and saved in a data fie of a PC. In the signa mode, wireess propagation channe and interference effects are considered by setting mutipe access interference (MAI), additive white Gaussian noise (AWGN), and channe parameters propery. Second, the adaptive agorithm is coded in ASM and compied finay to be oaded in a memory of DSP board (TMS320C6701). DSP board computed the weight vector that is based on LMS agorithm and beamformer generated the weight vector for muti-users in rea time. Third, the weight vector obtained in the beamforming modue is feedback to the PC via Matab s interfacing functions and code composer studio (CCS) to check the performance based on off-ine processing. II. SYSTEM MODELING Consider a reverse ink channe [2], assuming that there are N subscribers in a ce or a sector, the signa received by M eement antenna array at the base station can be expressed as, L X(t) = α s(t τ ) a(θ ) + I(t) + N(t) (1) = 1 where, s(t) is transmitted signa with compex path j attenuation α = β e φ and time deay τ for each mutipath signa, I(t) is the mutipe access interference (MAI) given by N 1 L q I (t) = α,s(t τ ). a(θ (2) q= 1 = 1 q q, q, ) This work is supported by The Scientific and Technica Research Counci of Turkey TUBITAK under grant EEEAG / 102E015

2 W 1 Desired user S(t) )) OCNS (mutipe access interference),i(t) AWGN X(t) W 2 W 3 y(t) Wireess Propagation Channe W n DSP (Adaptive Beamforming Agorithm) Figure 1 Received signa mode at an antenna array. and N(t) is the Mx1 compex-vaued additive white Gaussian noise vector, a (θ ) is the Mx1 array response vector for the mutipath arriving at direction of arriva (DOA) θ. At the antenna array output as shown in Figure 1, compex baseband signas are mutipied by compex weight vectors [ W W.... ] T =. (3) W 1 2 W M the weight vectors W is determined by some optimization (or beamforming) agorithm to yied optimum signa-tointerference (SINR) power ratio or some other objective function. The array output is mutipied by these compex weights to yied y(t) = W X(t). (4) The generated signa in (1) is fed to DSP for the computation of the optimum weight vector via LMS agorithm. The DSP processes this signa and cacuates a weight vector at each iteration, which converges to optimum weight vector [3]. In the LMS agorithm that is most widey used non-bind type adaptive agorithm introduced by Widrow & Hoff in 1959, antenna weights are recursivey obtained to minimize the mean square error using the foowing equations: w(t + 1) = w(t) + µ X(t) e (t) (5) e(t) = d(t) y(t) (6) y (t) = w H (t) X(t) (7) where w(t) is the antenna weight vector, X(t) is received signa vector at the antenna output, e (t) is the error signa between the desired response d (t) and weighted antenna output y(t). The superscripts and H, represent the compex conjugation and Hermitian operation, respectivey. µ is the step size parameter chosen as; λ max 0 < µ < (8) 10 where the λ is the eigenvaue of pre-correation matrix. The agorithm does not require measurements of the pertinent correation functions, nor does it require matrix inversion. The major benefit of the LMS agorithm is its simpicity compared to other adaptive agorithms [4,5]. III. SIMULATION SYSTEM SETUP The parameters in the signa modeing of eq (1) have been setup as summarized in Tabe 1. LMS Agorithm Conditions number of sampes 100 number of eements in the ULA desired DOA interference DOAs SINR 1, ,85,145 1,9934 2,0021 Tabe 1 Simuation vaues

3 Convergence criteria for the LMS agorithm is determined from the norm of error vector given by w = w( t + 1 ) w( t). (9) Simuation is performed according to the fow chart outined in Figure 2. This fow chart is reaized in the simuation with Matab Link for Code Composer Studio. This deveopment toos hep use Matab functions to communicate with Code Composer Studio and with information stored in memory and registers on a target DSP. With these inks, it is easy to transfer information to and from the memory and registers in the DSP. Figure 3 iustrates data fow in the simuation system. The signa received by the smart antenna array is modeed and generated in Matab in a PC. This noisy signa is fitered with the weight vector obtained via the adaptive agorithm executed in C6701 DSP.. Initiaization 1 w [n] = 0 Initiazation 2 y = 0 Acquisition read the input sampes: y X Computation 1 N = 1 w [i] X i= 0 Computation 2 e = d X µe = µ e [n], d[n] [i] PC MATLAB 6.5 signa generation performance computation CCS Target DSP EVM Board (TI C6701) Weight Computation Figure 3 Genera bock structure of simuation system IV. RESULTS AND DISCUSSION Tests for 3,5 and 7 antenna eements of uniform inear array (ULA) were performed. We assume that with this number of eements executes beamforming at the system. Tabe 2 beow tabuates the coected performance data. We consider one direct path for the desired user, and three interference signas. DSP Processing Time Antennas C6701 Cock(Cyce) C6711 Cock(Cyce) C6701 Time(msec) 0, , ,04909 C6711 Time(msec) 0, , ,19055 Sampe Computation 3 wn(k) = wn(k 1) + µex(n k) Update X (i) = X(i 1) Output Y Figure 2 Steps for impementing a LMS adaptive fiter Tabe 2 DSP beamformer processing times for various antenna numbers with various DSP (C6701 and C6711) The resuting signa and weight errors obtained with LMS agorithm are shown in Figures 4, 5 and 6 for 3, 5, and 7 ULA, respectivey. Spatia spectrums formed with the weight vector resuting from the agorithm outputs are shown in figures 7, 8, and 9. In these figures, we see that antenna array main beam direction is steered towards to DOA of the desired signa. This is because the desired signa was the strongest and the agorithm sense the signa that has the argest power as the desired signa. We note that beamwidth decreases as the number of antenna eements is increased.

4 Figure 4 Weight and signa error for m=3 antennas Figure 7 Spatia spectrum with 3 antennas Figure 5 Weight and signa error for m=5 antennas Figure 8 Spatia spectrum with 5 antennas Figure 6 Weight and signa error for m=7 antennas Figure 9 Spatia spectrum with 7 antennas

5 V. CONCLUSION In this paper we have demonstrated the use of a digita signa processor in the impementation of rea time adaptive beamforming agorithm for cdma2000 3G mobie communication systems. It seems possibe from our simuations that the proposed system (TI C6701 DSP and Matab functions) can hande optimum weight vector computation. We have found that agorithm computation time and accuracy of beam pattern direction are proportiona with the number of antenna eements. Athough, the resuts are preiminary they are vauabe in terms of finding new directions for optimizing the assemby coding in DSP and thus enabing software radio impementation of various beamformer agorithms. REFERENCES [1] T. S. Rappaport, J. C. Liberti, Smart Antennas for Wireess Communication, Prentice Ha, NJ, [2] TIA/EIA Interim Standard, Physica Layer Standard for cdma2000 spread spectrum systems, TIA/EIA/S [3] H. Im, S. Choi, and K. Lee, "Impementation of a smart antenna systems test-bed for wide-band CDMA WLL channe," Proc. VTC2000, / IEEE. [4] Naim Dahnoun, DSP Impementation using the TMS320C6000 DSP Patform, Prentice Ha, NJ, 2000 [5] Mohammad Darwish, Ceci Lau, A Software Radio Architecture for Smart Antennas, Spectrum Signa Processing,

Simultaneous Routing and Power Allocation in CDMA Wireless Data Networks

Simultaneous Routing and Power Allocation in CDMA Wireless Data Networks Simutaneous Routing and Power Aocation in CDMA Wireess Data Networks Mikae Johansson *,LinXiao and Stephen Boyd * Department of Signas, Sensors and Systems Roya Institute of Technoogy, SE 00 Stockhom,