Performance Evaluation of Limited Feedback Scemes for 3D Beamforming in LTE-Advanced System Sang-Lim Ju, Young-Jae Kim, and Won-Ho Jeong Department of Radio and Communication Engineering Cungbuk National University E10-601, 1, Cungdae-ro, Seowon-gu, Ceongju-si, Cungbuk SOUTH KOREA imaward@naver.com KYUNGSEOK KIM Department of Information and Communication Engineering Cungbuk National University E10-601, 1, Cungdae-ro, Seowon-gu, Ceongju-si, Cungbuk SOUTH KOREA Corresponding Autor, kseokkim@cbnu.ac.kr Abstract: - Te tree-dimension (3D) multiple input and multiple output (MIMO) system one of te key tecnologies studied for some advanced wireless communication systems suc as Long Term Evolution- Advanced (LTE-A) In te 3D MIMO systems, uniform planar array (UPA) equipped at BS. It enable elevation beamforming as well as azimut beamforming. T paper investigates two limited feedback scemes for 3D beamforming. Te spectral efficiency evaluated for two limited feedback scemes in various antenna configurations. Key-Words: - 3D beamforming, 3D MIMO, Limited feedback sceme, LTE-A, Spectral efficiency, UPA 1 Introduction Massive multiple input and multiple output (MIMO) employs a large number of antennas at base stations (BS) and can simultaneously serve multiple user equipment (UE) [1]. It being considered to improve te system trougput and te spectrum efficiency in some advanced wireless communication systems suc as Long Term Evolution-Advanced (LTE-A) [2-3]. In legacy MIMO systems, a uniform linear array (ULA) usually considered at BS. However, since it difficult to apply ULA due to te limited room in massive MIMO, Two-dimensional (2D) antenna arrays suc as a uniform planar array (UPA) are considered. Antenna elements in UPA are placed vertically and orizontally at te BS. Tey enable 3D beamforming, elevation beamforming as well as azimut beamforming. Recently, te 3rd Generation Partnersip Project (3GPP) as released a 3D MIMO system and a 3D spatial cannel model (SCM) [4-5], and many studies for 3D MIMO systems are investigations [6-8]. To perform 3D beamforming, a BS requires cannel state information (CSI) for orizontal and vertical dimensions. However, te feedback overead increases as te number of antennas and more computational complexity required for te best code word selection. For massive MIMO, effective limited feedback sceme needs to minimize te feedback overead and to acieve te ig spectral efficiency. In t paper, we investigate two limited feedback scemes for 3D beamforming in in downlink 3D MIMO system. And, te performance by teir two scemes analyzed in 3GPP 3D SCM scenario. T paper organized as follows. In section 2, we describe te system model. Section 3 describes two limited feedback scemes. Simulation results are presented in Section 4. Section 5 concludes te paper. 2 System Model We consider a downlink 3D-MIMO system were a BS equipped wit a UPA aving N t = N v N antennas based on LTE-A. K UEs are equipped wit a single antenna, as sown in Figure 1. Antenna elements at BS are placed in te vertical and ISSN: 2367-8887 51 Volume 2, 2017
orizontal direction. N t te number of total transmit antenna elements. N and N v are te number of orizontal and vertical antenna elements, respectively. Since frequency-divion duplexing (FDD) used in advanced wireless communication systems suc as LTE-A, to minimize te feedback overead need. To reduce feedback overead, LTE-A use dcrete Fourier transform (DFT) based LTE codebooks [11]. In FDD massive MIMO systems wit a ULA, te DFT based codebook usually used for te beamforming since it can provide a good performance in igly correlated cannels [12]. In 3D MIMO, elevation beamforming sould be supported as well as orizontal beamforming. Tus, 3D codebook, Kronecker product codebook (KPC), proposed for 3D beamforming, in wic te DFTbased 2D codebook as been expanded [13]. It generated by Kronecker product of two DFT code words for vertical and orizontal domains as follows. v C v = [w 0 w v v 1,, w Mv 1] (2) Figure 1. 3D MIMO beamforming wit UPA. To evaluate te performance of 3D MIMO system, a 3D cannel model considered. Te 3GPP as developed a 3D SCM based on 2D SCM [9-10]. It considered two different scenarios of cannel environments, Urban Macro (3D-UMa) and Urban Micro (3D-UMi), for 3D MIMO systems. T paper considers 3D-UMa scenario. It assumed tat te BS (eigt: 25m) well above te eigt of UEs of te surrounding buildings, and outdoor line-of-sigt (LOS), outdoor non-line-ofsigt (NLOS), indoor LOS, and indoor NLOS are considered. 2.1 Received signal model In 3D MIMO systems, te signal y k received by k- t UE can be expressed as follows: K y k = ρh k W k s k + ρh i W i s i + N k (1) i=1,i k Were s k te transmit signal. ρ te average transmit power. H k C (N v N ) 1 te cannel matrix from BS to k-t UE. For 3D MIMO, te cannel coefficient matrix tree-dimensional by antennas of vertical dimension. W k C 1 (N v N ) te precoding matrix for k-t UE, corresponding to K 3D cannel. i=1,i k H i W i s i denotes te undesired signal. N k denotes Gaussian noe. 2.2 Codebook for 3D beamforming C v te a DFT based codebook for te vertical domain. Were k-t codeword w k v w v k = 1 [1, e j2π N v T M v ], k M v,, e j2π(n v 1)k k = 0,1,, M v 1 (3) N v te number of vertical antennas, and M v a codebook lengt, 2 B v = M v, were B v te codebook bit for te vertical domain. Were a DFT based codebook for te orizontal domain, C, and k-t codeword w k w k = 1 [1, e j2π N C = [w 0 w 1,, w M 1] (4) T M ], k M,, e j2π(n 1)k k = 0,1,, M 1 (5) N te number of orizontal antennas, and M a codebook lengt, 2 B = M, were B te codebook bit for te orizontal domain. C 3D = C v C = [w v 0 w 0 w v v 0 w 1 w Mv 1 w M 1] 3D 3D 3D 3D = [w 0 w 1 w 2 w Mv M 1] (6) C 3D te Kronecker product-based 3D codebook. Te lengt of a 3D codebook M v M. ISSN: 2367-8887 52 Volume 2, 2017
For example, if bot B v and B are 5 bits, 32 code words are generated for vertical and orizontal domains, respectively. For a 3D codebook, 1024 code words are generated. 3 Limited feedback scemes Eac UE selects its own best code word, and an index of selected code word, called precoding matrix index (PMI), reported to BS. An PMI for orizontal domain reported in legacy LTE-A. However, it need to fed back te precoding information for te orizontal and vertical domains for 3D beamforming. In t paper, we introduce two different limited feedback scemes, one PMI feedback sceme and two PMI feedback sceme. First, in one PMI feedback sceme, 3D codebook C 3D = [w 3D 0, w 3D 3D 1,, w Mv M 1] in (6) required in bot BS and UEs. And, assume te cannel matrix H k C (N v N ) 1 from k-t UE to BS. In k-t UE, te best code word can be selected using w 3D p = arg max H kw 3D l 2 (7) l=0, M v M 1 were p denotes PMI, and k UE feeds back it to BS. In t sceme, M v M computation required to select te best code word in eac UE. Te transmit signal beamformed by te code word corresponding to feedback PMI. Next, in two PMI feedback sceme, two DFT based codebooks for vertical and orizontal domains are required in (2) and (4). Te best code word selected for two domains, respectively, and UE feeds back two PMI. Assume te cannel matrix H k (8) Te cannel H v,k for first column antenna elements called te vertical cannel and te cannel H,k for first row antenna elements called te orizontal cannel in t paper. Te best code word for te vertical domain selected using w v i = arg max H v,kw v l 2 (9) l=0, M v 1 Te best code word for te orizontal domain selected using w j = arg max H,kw l 2 (10) l=0, M 1 i and j are selected PMIs for two domains, respectively. In t sceme, M v + M computation required to select te best code words in eac UE. Kronecker product computed for te two code words corresponding to feedback PMIs in BS as follows w 3D = w i v w j 4 Simulation results Table 1. Simulation parameters. Parameters Carrier frequency System bandwidt TTI lengt Noe power density Scenarios Inter-site dtance Network layout Value 2.1 GHz 10 MHz 100 ms 174 dbm/hz 3D-UMa 500 m 1 sector in a site (11) Number of UEs per cell 2, 4, 6, 8, 10 UE dtribution Uniformly dtributed BS eigt 25 m BS transmit power BS Antenna configuration (N t, N v, N ) Vertical antenna element spacing Horizontal antenna element spacing Codebook Sceduler UE eigt UE max floor 8 UE Antenna elements 1 Receiver type UE speed Cannel estimation Feedback CSI 46 dbm UPA (8,2,4), (16,2,8), (16,4,4), (32,4,8) λ 2 λ 2 5bits DFT based codebook Max C/I 3*(UE_floor-1)+1.5 m MMSE 3 km/ Ideal PMI, CQI, RI PMI and CQI reporting per 5ms ISSN: 2367-8887 53 Volume 2, 2017
In t section, te average spectral efficiency of two limited feedback scemes evaluated and compared. To simulations, 3GPP 3D SCM scenario considered, and 3D-UMa adopted [4]. A 120 sector in a single cell considered were UEs are uniformly dtributed. We consider various transmit antenna configurations, 2 4, 2 8, 4 4, and 4 8. Two 5 bits DFT codebooks are used for vertical and orizontal domain. We assume te perfect cannel knowledge. Detailed simulation parameters are lted in Table 1. Figure 2 sows te spectral efficiency as te antenna configuration equipped at BS (K = 10). Te spectral efficiency of one PMI feedback sceme were iger tan two PMI feedback sceme in most cases. In te case (16,2,8), te spectral efficiency difference approximately 0.1 bps/hz, and it approximately 0.2 bps/hz in te cases (16,4,4) and (32,4,8). In te case (32,4,8), te spectral efficiency igest, and it 7.73501 bps/hz for one PMI feedback sceme. Compared wit te case (16,2,8), it sows te performance improvement of 1.3 bps/hz. Figure 3 sows te results of te spectral efficiency analys for an increasing number of UEs wen te antenna configuration equipped at BS te case (32,4,8). Te spectral efficiency of one PMI feedback sceme were iger tan two PMI feedback sceme in Figure 3 in common wit Figure 2. Te spectral efficiencies of one and two PMI feedback scemes were same for 2 UEs and 6 UEs, and 5.29309 bps/hz and 5.00845 bps/hz for 4 UEs, and 6.59408 bps/hz and 6.42691 bps/hz for 8 UEs, respectively. Figure 2. Spectral efficiency as transmit antenna configurations; K=10. 5 Conclusion In t paper, we investigate two limited feedback scemes for 3D beamforming in LTE-A. One PMI feedback sceme required ig computation complexity to select te best code word tan two PMI feedback sceme. Te performance of two scemes were compared troug average spectral efficiency analys. Te simulation results sow tat one PMI feedback sceme as te igest spectral efficiency. In our future work, we will perform furter optimization of 3D MIMO system, and te limited feedback sceme to attain te iger spectral efficiency will be developed. Acknowledgments T researc was supported by Basic Science Researc Program troug te National Researc Foundation of Korea(NRF) funded by te Mintry of Education (No.2017R1D1A1B03032420) Figure 3. Spectral efficiency for an increasing number of UEs; N t = 32, N v = 4, N = 8. References: [1] T. L. Marzetta, Noncooperative cellular wireless wit unlimited numbers of base station antennas, IEEE Trans. Wireless Commun., vol. 9, no. 11, pp. 3590 3600, 2010. [2] H. Q. Ngo, E. Larsson, and T. Marzetta, Energy and spectral efficiency of very large multiuser MIMO systems, IEEE Trans. Commun., vol. 61, no. 4, pp. 1436 1449, Apr. 2013. ISSN: 2367-8887 54 Volume 2, 2017
[3] H. Ji, et al., "Effect of 3-Dimensional Beamforming on Full Dimension MIMO in LTE-Advanced," Globecom Worksops (GC Wksps), 2014. [4] 3GPP Tecnical Reports TR36.873, Study on 3D cannel model for LTE. [5] 3GPP Tecnical Reports TR36.897, Study on Elevation Beamforming/Full-Dimension (FD) MIMO for LTE. [6] Y. Kim, et al., "Full Dimension MIMO (FD- MIMO): Te Next Evolution of MIMO in LTE Systems, IEEE Wireless Commun., vol. 21, sue 3, 2014. [7] Y. Coi, et al., "System-Level Performance of Limited Feedback Scemes for Massive MIMO," ETRI Journal, vol. 38, no. 2, pp. 280-290, Apr. 2016. [8] H. Ji, et al., "Overview of Full-Dimension MIMO in LTE-Advanced Pro," IEEE Commun. Magazine, 2017. [9] WINNER+ Final Cannel Models, Deliverable D5.3 V1.0, 30 Jun. 2010. [10] B. Mondal, et al., "3D cannel model in 3GPP," IEEE Commun. Magazine, vol. 53, no. 3, pp. 16-23, 2015 [11] 3GPP Tecnical Specification TS 36.211, Pysical cannels and modulation. [12] B. Clerckx, G. Kim, and S. Kim, Correlated fading in broadcast MIMO cannels: curse or blessing?, in IEEE Global Telecommun. Conf. (GLOBECOM 08), pp. 1 5, IEEE, 2008. [13] Y. Xie et al., A Limited Feedback Sceme for 3D Multiuser MIMO Based on Kronecker Product Codebook, in IEEE 24t Intern. Symp. on Personal Indoor and Mobile Radio Commun. (PIMRC), pp. 1130-1135, 2013. ISSN: 2367-8887 55 Volume 2, 2017