Measurement Based Capacity of Distributed MIMO Antenna System in Urban Microcellular Environment at 5.25 GHz

Transcription

1 Measurement Based Capacity of Distributed MIMO Antenna System in Urban Microcellular Environment at 5.25 GHz Mikko Alatossava, Student member, IEEE, Attaphongse Taparugssanagorn, Student member, IEEE, Veli-Matti Holappa, Student member, IEEE, Juha Ylitalo Centre for Wireless Communications, P.O. Box 4500, University of Oulu, Finland Elektrobit Ltd., Tutkijantie 7, FIN Oulu, Finland Abstract In this paper, the results of distributed multipleinput multiple-output antenna system (MIMO DAS) capacity measurements are presented. Until this point, literature about MIMO DAS considers only theoretical channels. The novelty of this paper comes from applying actual measured radio channel to MIMO DAS. In MIMO DAS, the coverage is obtained by using several largely separated MIMO antenna ports (AP) in the area. This way the advantages of shadowing diversity can be exploited and the link performance can be improved as theoretically shown in [1]. The measurements for the analysis were conducted in the downtown of Oulu, Finland, by generating a 4x16 MIMO configuration with EB Propsound CS TM as a channel sounding device. The results show that the capacity of MIMO DAS is improved from the conventional system and the diversity aspect in shadowing between the separated APs is large. I. INTRODUCTION Multiple-input multiple-output (MIMO) antenna configuration will most likely be adopted for future communication systems to satisfy the need for increased spectrum efficiency. Several information theoretic studies have shown that capacity of a MIMO system is directly proportional to the amount of antennas in the system [2]. This is valid assumption as long as the multipaths of the channel are uncorrelated, i.e., the environment enables rich scattering with uncorrelated small scale fading between the antenna elements [3]. Therefore, with the above mentioned conditions, MIMO offers spatial microscopic diversity. In this paper MIMO has been applied in a distributed antenna systems (DAS). DAS is first introduced by Valenzuela et al. in [4] and shown to be advantageous in terms of delay spread and power attenuation in comparison to conventional systems. Figure 1 shows the idea behind DAS. On the left side a cell of one base station (BS) with radius r leading to area of πr 2 is depicted and referred to as conventional system. On the right hand side of the figure, the same area is covered with seven smaller sized cells, each having own antenna port (AP) operated commonly by the central BS. Recently in [5], a generalized DAS applying multielement antennas (MIMO DAS) in the APs of the small cells has been presented. Later in [1], the effect of small scale fading and large scale fading in terms of capacity was considered in distributed MIMO system. China, among other countries, has shown great interest in MIMO DAS where it is considered as one of the corner stones of beyond 3G systems [6]. The greatest advantage in MIMO DAS is that, in addition to microscopic fading diversity, macroscopic fading (shadowing) diversity due to separated APs improves the system performance as well. Several publications have discussed this topic and they provide large variety correlation values ranging from 0.1 to 0.6 between the shadowing of different, largely separated, BSs [7], [8]. The lack of the above mentioned MIMO DAS studies is that no channel measurements have been performed. The results for MIMO DAS are obtained through Monte Carlo simulations without realistic channel measurement information. BS r Fig. 1. On the left side conventional method for the coverage of a cell with radius r is depicted. Right hand side of the figure shows how the same area could be covered with seven small cells, each controlled with the same central BS. This paper has the following contributions: Instead of obtaining results through theoretic Monte Carlo simulations, actual measured radio channel is applied. Capacity of the measured MIMO DAS will be investigated. Furthermore, the impact of small scale fading on capacity will be discussed. Correlation between the shadowing of different antenna ports will be studied based on measured radio channel. The rest of this paper is organized as follows: Section II discusses the general channel model applied in literature for BS /08/$ IEEE 430

2 MIMO DAS. Section III presents the measuring device and the environment. The results are given in Section IV and finally conclusions are drawn. II. SYSTEM MODEL OF DISTRIBUTED MIMO Theoretical channel matrix H with dimensions NL M for a(m,n,l) MIMO DAS depicted in Figure 2 is given in [1] as H = H SF (d)h SSF, (1) where H SSF and H SF are NL M channel matrix for small scale fading and NL NL matrix for shadow fading, respectively. Central BS consists of N APs, each having L antenna elements. M is the number of antenna elements in the mobile (MS) and d is the distance between MS and BS. Port 2 1.M Port N A. Measurement Equipment and Settings EB Propsound CS TM, [9], was used to measure the MIMO radio channel. In the sounder, multiple single-input singleoutput channels are switched in such a speed that the sounder can be said to operate in a quasi-simultaneous way and to measure a MIMO channel. MIMO antennas, presented in Figure 3, at the frequency of 5.25 GHz were applied in the measurements. The properties of the antennas are presented in Table I, where ODA denotes omni-directional antenna and DP stands for dual polarized. Mutual coupling between antenna elements is avoided by selecting the distance between adjacent elements to be λ/2. TABLE I PROPERTIES OF THE ANTENNAS Property Rx antenna (BS/AP) Tx Antenna (MS) Frequency [GHz] Bandwidth [MHz] Azimuth coverage [deg] ±70 ±180 Elevation coverage [deg] ±70 ±90 Antenna type DP ±45 DP ±45 Number of elements Arrangement of elements 4x4 square 2x9+7 ODA Port 1 Mobile Port 3 Central Base Station Unit Fig. 2. (M, N, L) MIMO DAS system. H SSF is calculated as H SSF = R Rx H w R Tx, where R Rx and R Tx are the spatial correlation matrices for receiver (Rx) and transmitter (Tx) terminals in a form of R = diag(r 1 R 2...R N ), respectively. H w is a NL M complex matrix with zero mean and unit variance. H SF (d) takes into account the path loss at each AP (expressed with path loss exponent α and the shadowing standard deviation σ SF ). The element of matrix H SF (d) forapnis calculated as h SF,n (d) = d α/2 n χ n 20, (2) d ref where χ is normal distributed random variable with zero mean and σ 2 SF variance. Finally H SF(d) = diag(h SF,1 (d 1 )I,h SF,2 (d 2 )I,..., h SF,N (d N )I), where I is identity matrix of size L L. III. MEASUREMENT EQUIPMENT AND ENVIRONMENT In this section, the measurement equipment, applied settings for measurement and the measurement procedure and environment are presented. Fig. 3. Applied antennas in the measurement. The figure shows the receiver antenna on the left and the transmitter antenna on the right. The applied settings in measurements are shown in Table II. A term cycle, is denoted as a procedure where the sounder switches through all the antenna elements in Tx and Rx. Used bandwidth was 0 MHz and the applied code length was 255 chips. This corresponds to a delay resolution T d =1/B = ns and a code duration of 2.55 us. TABLE II SETTINGS OF THE CHANNEL SOUNDER IN THE MIMO DAS MEASUREMENTS Frequency [GHz] 5.25 Bandwidth (B) [MHz] 0 Length of the code [chips] 255 Used antenna configuration 4 16 Transmitting power [dbm] 26 Cycle duration [ms]

3 B. Measurement Environment In Figure 4, the measurements conducted in the urban micro cellular environment is shown. The measurement location is in downtown Oulu, Finland. Red circles show the Tx (acting as MS) spots and the Rx (BS) spots are shown with blue lines. The arrows in the BS and MS marks denote the direction of the antenna zero angle. The measurement was conducted so that in each of the BS spots, the MS was located in each of the MS spots in turn and 20 channel snapshots were stored. Basically the MS was moving along predefined route making a stop of 20 snapshots in each marked spot. The same route was driven four times, ones for each BS spot. The MS spots were marked in the ground to guarantee that for each measurement, exactly the same MS spots were used. Height of the BS antenna and MS antenna were 5 and 1.7 meters, respectively. A. Channel Capacity We consider uplink transmission system with M =4, N = 4, L =4and L C =16. Narrowband ergodic channel capacity is calculated as C = E(log 2 [det (I + ν M H meash H meas)]), (3) where the dimensions of I are NL NL and ν is the received signal-to-noise ratio (SNR). H meas denotes the measured channel matrix. First we calculate the capacity when the SNR is fixed to db for both systems, MIMO DAS and C-MIMO. C-MIMO is analyzed separately for four different cases, each having different BS (BS1...BS4). Numbering for the four BS (Rx) and 12 MS (Tx) sites starts at the bottom of the Figure 4 moving counter clockwise until all the spots are measured. As the SNR is fixed, the differences in the ergodic capacity come only from the channel matrix H and the diversity aspects of it. From Figure 5 it can be seen that the overall level of the capacity with MIMO DAS is superior to the capacity obtained with only one BS. This implies that the shadowing diversity improvement obtained from MIMO DAS is significant istances in meters Wooden fence Car Hedgerow, low Lawn Building Walkpath High lamp Traffic sign Big tree Normal size tree BS and zero angle direction MS and zero angle direction DAS BS1 BS2 BS3 BS Fig. 4. Distributed antenna system measurement conducted in downtown Oulu, Finland. IV. RESULTS To enable fair comparison between conventional MIMO (C- MIMO) system and (M,N,L) MIMO DAS, the conventional system (N =1)usesL C = NL receiving antenna elements, whereas in MIMO DAS each AP uses only L antenna elements. The terminology difference in C-MIMO and MIMO DAS is that, in C-MIMO each cell has own BS whereas in MIMO DAS each cell, with decreased size, is covered by APs controlled by the central BS. In this paper, the Rx terminals (BS spots in Figure 4) are referred as BS in C-MIMO and AP in MIMO DAS. Fig. 5. Capacity comparison with fixed SNR between MIMO DAS and C-MIMO with four different BSs. Second capacity comparison takes into account the received ν. With MIMO DAS, the mobile is assumed to be often in line-of-sight (LoS) situation with one of the APs and, hence, the improved SNR is expected to give capacity gain compared to conventional MIMO system. Figure 6 shows that this assumption is true. The received SNR was extremely high due to short distances between MS and the APs and this is seen as high capacity values. Figure 7 shows the measured impulse responses from each of the four individual base stations obtained at the first Tx location. In DAS, the system takes advantage of the combined impulse responses which allows more diversity than an individual base station. 432

4 74 72 TABLE III PATH LOSS EXPONENTS α AND SHADOWING STANDARD DEVIATIONS σ SF AT THE ANTENNA PORTS DAS 60 BS1 BS2 58 BS3 BS Fig. 6. Capacity comparison with measured SNR between MIMO DAS and C-MIMO with four different BSs Parameter AP1 AP2 AP3 AP4 α σ SF [db] Measured ρ=0 ρ=0.2 ρ=0.4 ρ=0.6 ρ=0.8 ρ= BS 1 BS 2 BS 3 BS Fig. 8. Measured and theoretical MIMO DAS capacities at the 12 Tx spots. Power [db] Delay [chips] Fig. 7. The measured channel impulse responses from each of the four base stations obtained at the first Tx location. B. Fading Correlation When analyzing the impact of the correlation of small scale fading between the antenna elements of an individual AP on capacity, we generate a theoretical channel matrix H as depicted in Section II and in [1]. The impact of totally uncorrelated (ρ = 0) and fully correlated (ρ = 1) small scale fading on channel capacity is calculated and showed in Figure 8. For reference, the measured MIMO DAS capacity and theoretical capacity with correlation coefficients ρ =[ ] are shown. It can be seen that in the measured MIMO DAS, the correlation coefficient of small scale fading varies quite a lot between different Tx locations. The shape of the curves comes from different distances between Tx and APs and from H SF (d) which takes the measured distance related path loss and shadowing at four APs, presented in Table III, into account. Small scale fading between APs is assumed independent due to large distances between the ports. Shadowing correlation between the AP i and j in MIMO DAS is calculated as ρ i,j = E( (β i µ i )(β j µ j ) σ SF,i σ SF,j ), (4) where µ is the mean received power calculated as µ = αlog (d) and β = µ+χ, where χ is presented in Section II. In this environment, the resulting correlation matrix R SF for shadow fading between four APs was found to include relatively small values and, therefore, to imply that shadowing diversity could be used to improve the performance of the system. Typical correlation coefficients ρ was found to be less than 0.2. V. CONCLUSION In this paper, measured MIMO DAS capacity was analyzed and compared to conventional MIMO system. Until this point, the results found in the literature for MIMO DAS are based on theoretical analysis, not actual measured data as is the case in this paper. This paper has the following contributions: Instead of obtaining results through theoretic Monte Carlo simulations as is the case in literature of MIMO DAS until now, actual measured radio channel was applied in this paper. Capacity of the measured MIMO DAS was investigated with fixed SNR and with measured SNR. Furthermore, the impact of small scale fading on the channel capacity was discussed and compared with measured MIMO DAS capacity. Correlation between the shadowing of different APs in MIMO DAS was studied based on measured radio channel. 433

5 The results showed that capacity is increased also in measured MIMO DAS and, therefore, MIMO DAS improves system performance compared to conventional MIMO system. The reason for this is the increased diversity due to largely separated APs and the increased received SNR due to more frequent LoS condition between Tx and one of the APs. Furthermore, the shadowing correlation study between the APs of MIMO DAS showed that the impact of shadowing diversity is relatively high in MIMO DAS. ACKNOWLEDGMENT This work has been performed in the framework of the project FRACTA. The work of Mikko Alatossava was supported by Tekniikan edistämissäätiö, Kaupallisten ja teknillisten tieteiden tukisäätiö and Oulun yliopiston tukisäätiö. REFERENCES [1] Z. Ni and L. Daoben, Effect of Fading Correlation on Capacity of Distributed MIMO, 15th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, September, 2004, vol. 3, pp , [2] G. J. Foschini and M. J. Gans, On Limits of Wireless Communications in Fading environments when Using Multiple Antennas, Wireless Personal Communications, 1998, vol. 6, pp , [3] J. Ylitalo, J.-P. Nuutinen, J. Hämäläinen, T. Jämsä, and M. Hämäläinen, Multi-dimensional Wideband Radio Channel Charactarisation for 2-6 GHz Band, Wireless World Research Forum 11 th Meeting, Services and Applications Roadmaps - Invigorating the Visions, Oslo, June - 11, 2004, p.11. [4] A. Valenzuela, A. Rustako, and R. Roman, Distribute Antennas for Indoor Radio Communications, IEEE Transactions on Communications, December, 1987, vol. 35, [5] W. Roh and A. Paulraj, Outage Performance of the Distributed Antenna Systems in a Composite Fading Channel, IEEE 56th Vehicular Technology Conference, September, 2002, vol. 3, pp , [6] X.-H. Yu, G. Chen, M. Chen, and X. Gao, Towards Beyond 3G - A FuTURE Project In China, IEEE Communications Magazine, January, 2005, vol. 1, pp. 1 5, [7] E. Perahia, D. Cox, and S. Ho, Shadow Fading Cross Correlation Between Basestations, IEEE VTS 53rd Vehicular Technology Conference, 6-9 May, 2001, vol. 1, pp , [8] K. Zayana and B. Guisnet, Measurements and Modelisation of Shadowing Cross-Correlations Between Two Base-Stations, IEEE 1998 International Conference on Universal Personal Communications, 5-9 October, 1998, vol. 1, pp. 1 5, [9] L. Hentilä, P. Kyösti, J. Ylitalo, X. Zhao, J. Meinilä, and J.-P. Nuutinen, Experimental Characterization of Multi-Dimensional Parameters at 2.45 GHz and 5.25 GHz Indoor Channels, Proceedings of Wireless Personal Multimedia Communications, September