This article has been accepted and published on J-STAGE in advance of copyediting. Content is final as presented. IEICE Communications Express, Vol., 1 6 Experimental evaluation of massive MIMO at GHz band in indoor environment Ryochi Kataoka 1a), Kentaro Nishimori 1b), Ngochao Tran 2, Koshiro Kitao 2 and Tetsuro Imai 2 1 Graduate School of Science and Technology, Niigata University, 8 2-nocho Ikarashi, Nishi-ku, Niigata 9-2181, Japan 2 Research Laboratories, NTT DOCOMO INC. 3-, Hikari-no-oka, Yokosukashi, Kanagawa, 239-836, Japan. b) nishimori@ie.niigata-u.ac.jp Abstract: Massive multiple input multiple output (MIMO) improves transmission rate without increasing the complexity on signal processing by employing a large number of antennas at a base station. The effect of applying massive MIMO to small cells at the 2-GHz band has been reported. However, the primary targets of massive MIMO are small cells at a high-frequency band, because antenna size is considerably large when considering massive MIMO in a macro frequency band. In this study, measurements are performed for actual propagation channels using a wideband channel sounder with a horn antenna in the - GHz band in an actual indoor propagation environment. Moreover, the performance of interference rejection is evaluated using a virtual circular array antenna with 24 elements. Keywords: massive MIMO, indoor propagation characteristics, - GHz band, circular array, maximum ratio combining, zero forcing Classification: Antennas and Propagation References IEICE 17 DOI: 1.187/comex.16SPL26 Received January, 17 Accepted January 17, 17 Publicized February 21, 17 [1] F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Marzetta, O. Edfors, and F. Tufvesson, Scaling Up MIMO Opportunities and challenges with very large MIMO, IEEE Signal Processing Magazine, Vol. 3, pp. 4 6, Jan. 13. [2] J. Hoydis, S. ten Brink, and M. Debbah, Massive MIMO in the UL/DL of cellular networks: How many antennas do we need?, IEEE Journal on Selected Areas in Communications, Vol. 31, No. 2, pp. 16 171, Feb. 13. [3] H. Yang and T. L. Marzetta, Performance of conjugate and zero-forcing beamforming in large-scale antenna systems, IEEE Journal on Selected Areas in Communications, Vol. 31, No. 2, pp. 172 179, Feb. 13. [4] S. Suyama, J. Shen, T. Obara, M. Sumi., M. Nakajima, and Y. Okumura, Basic performances of super high bit rate massive MIMO transmission using higher frequency bands, IEICE Technical Report, RCS13-348, March 14. 1
[] R. Kataoka, K. Nishimori, J. Miyazawa, N. Tran, and T. Imai, Performance evaluation of massive MIMO with analog-digital hybrid processing in a real microcell environment, 1 International Workshop on Antenna Technology, FIS-, Seoul, Republic of Korea, Mar. 1. [6] R. Kataoka, K. Nishimori, N. Tran and T. Imai, Interference reduction characteristics by circular array based massive MIMO in a real microcell environment, IEICE Trans. Commun., Vol. E98-B, No. 8, pp. 1447 14, Aug. 1. [7] T. Nakamura, S. Nagata, A. Benjebbour, and Y. Kishiyama, Trends in small cell enhancements in LTE Advanced, IEEE Communications Magazine, Vol. 1, No. 2, pp. 98 1, Feb. 13. [8] R. Kataoka, K. Nishimori, N. Tran and T. Imai, Basic Performance of Massive MIMO in Indoor Scenario At -GHz Band, Proc. of ISAP1, pp.87-9, Nov. 1. 1 Introduction Multiuser multiple input multiple output (MU-MIMO) improves system channel capacity by improving transmission rate between a base station (BS) and multiple user equipment (UE) using a small number of antennas at the UE. To further improve frequency utilization in future wireless systems with MU-MIMO transmission, the concept of massive MIMO has been recently proposed [1]. In massive MIMO systems, the number of antennas at a BS (N BS ) is considerably larger than the number of antennas at UE and the number of UE. Massive MIMO allows for low-complexity signal processing, because inter-user interference is easily mitigated by high beamforming resolution [2]. Representative studies on massive MIMO systems have focused on theoretical investigation of channel capacity [2], computer simulation using maximum ratio combining (MRC) and zero forcing (ZF) as linear control methods [3], and measurement of actual propagation characteristics using a 128- elements channel sounder [1]. Even though a linear or planar array in horizontal and vertical planes is used in these evaluations, to realize high beamforming resolution [4], such array configurations cannot create a service area for all directions. Using a circular or cylindrical array is one of the methods to resolve this issue. Therefore, we previously reported the performance of interference rejection by a cylindrical array using an actual outdoor propagation channel at the 2-GHz band [][6]. However, the primary targets of massive MIMO are small cells at a high-frequency band [7], because antenna size is considerably large when considering massive MIMO in a macro frequency band. In this study, we conducted measurements to obtain the basic performance of massive MIMO for a high-frequency band in a small cell environment [8]. Actual channel state information (CSI) was measured using a wideband channel sounder with a horn antenna in the -GHz band in an indoor environment. In addition to the results in [8], this paper presents the 2
Sleeve antenna B6 B A6 A S C 1. m Up. conv. Transmitter B4 D4 A4 C4 (b) Configuration of the UE Horn antenna B3 B2 A3 A2 Down. conv. Vertical rotation Horizontal rotation 2.3 m 4.m B1 3.m D1 2m A1 3m C1 Receiver m BS (a) Measurement environment (c) Configuration of the BS Fig. 1. Measurement environment and configuration of the UE and BS. characteristics of angular spread in a measured indoor environment. These characteristics significantly affect the transmission performance of massive MIMO [1][6]. The characteristics of interference rejection of MRC and ZF in an actual environment are evaluated using a virtual circular array antenna with 24 elements. We verify that ZF is essential for reducing interference when considering a circular array, even for the -GHz band. Moreover, the characteristics are verified for a smaller number of antennas in a horizontal plane to reduce the complexity of signal processing for the ZF algorithm. The remainder of this paper is organized as follows: Section2 describes the measurement environment and the configuration of the virtual circular array. Section3 describes the characteristics of interference rejection in an actual propagation channel. 2 Measurement and evaluation environment Fig. 1(a) shows the measurement environment. We conducted measurements for an office in an indoor environment (the office of the DOCOMO R&D Center, Kanagawa, Japan). Actual CSI was measured using a wideband channel sounder at 19.8 GHz. The bandwidth and transmit power were MHz and 1 W, respectively. A wideband QPSK-OFDM signal was used, and its FFT points and number of sub-carriers were 124 and 449, respectively. Fig. 1(b) and Fig. 1(c) show the configuration of the UE and BS, respectively. We measured uplink CSI of single input single output from the UE to the BS. As shown in Fig. 1(b), a sleeve antenna with vertical polarization 3
Direction of UEs #24 #1 Ce1 #2 #3 Ce19 #19 φ = :1 :36 #7 Ce7 Ce13 #13 (a) Configuration of the BS (b) Angular spread versus the measured location Fig. 2. Virtual circular array with 24 elements and angular spread. is used at the UE. The UE transmits a signal at 17 points from A1 D4, as shown in Fig. 1(a). As shown in Fig. 1(c), a horn antenna is used at the BS. This antenna (3 db beam width is ) receives a signal while rotating in steps of 1 in a horizontal plane. The distance between the BS and UE is.4 27.6 m. Fig. 2(a) shows the evaluation configuration. The characteristics of interference rejection are evaluated using a virtual circular array antenna with 24 elements. 24 points are selected in a horizontal plane at an interval of 1. As shown in Fig. 2(a), UE is located at ϕ = 18. Moreover, the element in the virtual circular array are numbered from 1 to 24 in the clockwise direction, starting from ϕ = 18. Fig. 2(b) shows the azimuth angle spread and elevation angle spread. As seen in the figure, the azimuth angle spread is considerably larger than the elevation angle spread. The average and standard deviation of the azimuth angle spread are 64.7 and 1.7, respectively. The average and standard deviation of the elevation angle spread are 3.6 and 3.1, respectively. As a higher transmission rate is typically obtained for larger angular spread in MIMO transmission [1][6], arranging array antennas in a horizontal plane is suitable for massive MIMO transmission in an indoor environment. 3 Characteristics of interference rejection in actual propagation channel In this study, we assume that there is one desired user (S) and one interference user (I). The signal-to-interference-plus-noise power ratio (SINR) characteristics are evaluated using MRC and ZF. MRC and ZF weights are calculated for each subcarrier. Moreover, thermal noise is artificially added. The average signal-to-noise power ratio (SNR) among subcarriers is 3 db at the reference antenna. This antenna is selected because it has the maximum the average received power among the subcarriers of the desired user. The evaluation is performed using the following eight user configurations : the desired user is located at a fixed point (A4) and the interference user is located at A1, 4
A2, A3, A, A6, B4, C4, and D4. Furthermore, measurement is conducted at five instances of time for each point. Fig. 3(a) shows the SINR calculated using MRC and ZF, for a cumulative density function (CDF) of 1 %. All elements are used to obtain the SINR. As seen in Fig. 3(a), the SINR obtained using ZF is more than 13 db regardless of the angle of arrival from the interference user. On the other hand, the SINR obtained using MRC is less than 2.2 db. Moreover, the SINR obtained using ZF is approximately 1 db higher than that obtained using MRC. Therefore, even in the -GHz band, ZF is essential for reducing interference when considering a circular array. Note that it is not possible to form a sufficiently narrow beam when applying MRC in a circular array in the 2- GHz band [6]. Next, we evaluate SINR characteristics with a smaller number of antennas at the BS to reduce the complexity of signal processing for the ZF algorithm. Fig. 3(b) shows the SINR obtained using ZF for CDF = 1 % versus the number of antennas in the horizontal plane. Here, the desired user is A4 and interference user is A1 (near the BS) or A6 (far from the BS). For simplification, evaluation is carried out for Ce1 Ce19 shown in Fig. 2(a). The antennas #1, #7, #13, and #19 for Ce1, Ce7, Ce13, and Ce19 in Fig. 2(a) are defined as center elements, respectively. Then, an antenna is added to each side of the center elements. For example, in the case of Ce7, the number of added antennas is increased as follows: (#6, #7, #8), (#, #6, #7, #8, #9). Fig. 3(c) shows the SNR and interference-to-noise power ratio (INR) versus the angular positions. The angular positions for #1, #7, #13, and #19 are shown in Fig. 3(c). It is observed that the SINRs do not change with the position of the interference user (A1 and A6). As seen in Fig. 3(b), the SINRs for Ce1 and Ce13 are considerably higher than those of Ce7 and Ce19 when the number of antennas is less than 9. The SINR for Ce1 is higher than that for Ce13 regardless of the number of antennas. The SINRs for Ce7 and Ce19 are almost equal. They are higher than that for Ce13 when the number of antennas is more than or equal to 11. These characteristics can be explained using Fig. 3(c). The SNRs of the antennas around #1 are considerably higher than those around #7 and #19. The average SNRs for Ce1 are higher than those for Ce13 regardless of the number of antennas. Thus, when the number of antennas is less than or equal to 11 and 13, the average SNRs for Ce13 and Ce1 are higher than those for Ce7 and Ce9, respectively. It can be observed from Figs. 3(b) and (c) that selecting an antenna with high SNR is significant for using a small number of antennas in a massive MIMO system. 4 Conclusion In this study, the performance of massive MIMO was investigated in an actual indoor small cell environment using a virtual circular array. An actual propagation channel was measured using a wideband channel sounder with a horn antenna in the -GHz band in the indoor environment. Sufficient in-
(a) SINR with CDF=1 % by the MRC and ZF SINR [db] 1 1 S : A4, I : A6 24 elements Ce1 Ce7 Ce13 Ce19 3 6 9 12 1 18 21 24 Number of antennas SINR [db] (b) SINR versus the number of antennas (ZF, CDF = 1%) 1 1 S : A4, I : A1 24 elements Ce1 Ce7 Ce13 Ce19 3 6 9 12 1 18 21 24 Number of antennas SNR[dB], INR[dB] #13 3 3 2 1 1 #19 #1 #7 #13 A6 (INR) A4 (SNR) 6 1 18 24 3 36 Angle [deg.] SNR[dB], INR[dB] 2 1 1 (c) SNR and INR versus angle Fig. 3. SINR characteristics. #13 #19 #1 #7 #13 3 A1 (INR) 3 A4 (SNR) 6 1 18 24 3 36 Angle [deg.] terference reduction was obtained using ZF, whereas MRC could not reduce interference sufficiently when all antennas (24 elements) were used, even in the -GHz band. Moreover, it was shown that it is important to select efficient antennas with high SNR for cases in which all antennas cannot be used because of hardware limitations. This was because SNR changed significantly with the direction of UE. 6