3D Beamforminmg Methods with User-specific Elevation Beamfoming

3D Beamforminmg Methods with User-specific Elevation Beamfoming Zheng Hu, Shaoli Kang, Xin Su, Rongke Liu School of Electronic and Information Engineering, Beihang University, Beijing, China Key Laboratory of Wireless Mobile Communications, China Academy of Telecommunications Technology, Beijing, China Email: huzheng2868@sina.com Abstract In recent years, attentions have been focused on 3D (three-dimensional) beamforming which can improve cell average and edge user throughput and eliminate the interference to adjacent cells. In the work towards calibration of 3D channel model, it seems that elevation beamforming with fixed electrical downtilt of 2 degrees has a good SINR (signal to interference and noise ratio) performance in 3D-UMi (Urban Micro) and 3D- Uma (Urban Macro) scenarios. In order to improve the user s SINR of elevation beamforming, in this paper, two user-specific elevation beamforming methods are proposed to improve the performance, one (method I) selecting the EoD (elevation angle of departure ) as the electrical downtilt for each user, the other (method II) selecting the optimal electrical downtilt by considering user signal strength and inter-cell interference. Simulation results with the 3D channel model show that method I achieves a little improvement in 3D-UMi scenario but poor performance compared with the fixed 2-degree elevation beamforming method from SINR. Method II in both scenarios outperforms the fixed downtilt method from SINR. Index Terms 3D beamforming; electrical downtilt; 3D-UMi; 3D-; 3D channel model I. INTRODUCTION Long Term Evolution-Advanced (LTE-A) networks are being designed to improve the spectral efficiency and achieve higher peak data rates while achieving cost reduction and higher power efficiency []. Motivated by these requirements, with the introduction of the active antenna systems (AAS), recently there has been a significant interest in 3D beamforming. Traditionally, the beamforming was designed to support the adaptation in azimuth only. 3D beamforming attempts to create a user-specific elevation tilt to each spatial stream formed by the existing azimuth-only closed-loop SU/MU-MIMO methods [2-3]. In Release 2 of Third Generation Partnership Project (3GPP), there has been a significant focus on enhancing system performance through the use of antenna systems structure that provides adaptive control over both the elevation dimension and the azimuth dimension [4]. 3D beamforming takes advantage of elevation domain as well as the azimuth domain and provides additional degrees of freedom compared to conventional typical 2Dantenna beamforming [5]. User-specific elevation beamforming promises to increase the SINR by pointing the vertical antenna pattern in the direction of the user while spraying less interference to adjacent sectors by virtue of being able to steer the transmitted energy in elevation [6]. In [7], the author analyzed the elevation angle on MIMO capacity and shed light on the importance of the elevation dimension. In [6], three typical elevation beamforming scenarios are investigated and analysed for future LTE- Advanced. With and without the requirement for inter cell communications, [8] studied beam coordination methods for interference avoidance. In 3GPP RAN#74, for the work of initial calibration of large-scale channel model, it seems that the fixed electrical downtilt 2 degrees achieves a good SINR performance in elevation beamforming in both 3D- and 3D-UMi scenarios[9-]. Following the work mentioned above, herein we exploit two user-specific elevation beamforming methods to enhance the performance. One is to select the user s EoD as the electrical downtilt, the other tries to find the best optimal electrical downtilt with respect to user SINR and inter-cell interference. Based on 3D UMi scenario, it is necessary to introduce the 3D channel model of 3GPP to assess the benefits of user-specific methods that considers full dimensions. The 3GPP channel model is extended to incorporate the elevation dimension and it can be referred to [2]. In the 3D-UMi and 3D- regime, the two methods in this paper permit an improvement from a certain SINR with the large-scale 3D channel model setting compared to the fixed 2-degree elevation beamforming method. The remainder of the paper is organized as follows: in section II we describe the system model for 3D beamforming. In section III the two 3D beamforming methods for userspecific elevation beamforming are proposed. In Section IV we evaluate the performance of the proposed methods in 3D-UMi and 3D scenarios. Finally, we address the conclusion and our views on future research. Throughout the paper, the bold letter denotes a vector or a matrix. The superscript T denotes the transpose of a vector or a matrix. X is an L 2 norm of the vector X. II. SYSTEM MODEL In this paper, for simplicity, we consider a single-user downlink environment where the base station (BS) is just communicating with single-antenna user in a time-frequency resource block. We assume that each user suffers effective

interferers from neighbor cells. Each BS is equipped with the same uniform linear array (ULA) of M elements in elevation dimension. The elevation beamsteering method is based on 3D channel model of 3GPP standard. Here the system just introduces the large-scale 3D channel model ignoring the fast fading factors. So the received signal of the user in cell i can be expressed as: r = H i W i x i + N n=,n i H nw n x n + V () Where H i W i x i represents the desired signal. H i is forward link channel matrix between the user and transmitter. N n=,n i H nw n x n is the interference from adjacent cells. is the i.i.d Gaussian noise vector. W i is a M elevation beamforming vector and can be represented as: W i = [w w 2...w m...w M ] T, here the element w m is ω m = M exp ( 2π j (m ) dv λ cos (θ etilt) ) (2) m =, 2,... M where θ etilt is the optimal electrical downtilt. d v is the distance between antenna element. λ is the wavelength. In order to evaluate the performance of elevation beamsteering method, we present the antenna element radiation model and uniform linear array response. Also, the cells topological structure and user distribution are given in this section. A. antenna element radiation model In this paper, the 3GPP antenna model is adopted. Two formulas below are applied for horizontal and vertical radiation patterns: A E,H (φ) = min[2( φ ) 2, A m ], φ 3dB φ 3dB = 65, A m = 3 A E,V (θ) = min[2( θ 9 θ 3dB ) 2, SLA v ], θ 3dB = 65, SLA v = 3 where φ is the angle between the direction of interest and the boresight of the horizontal antenna. θ is the angle between the direction of interest and the vertical direction. A m is the front-to-back attenuation, and SLA v is side lobe attenuation. φ 3dB and θ 3dB are the horizontal HPBW (half power beam width) and vertical HPBW, respectively. The 3D antenna gain from two perpendicular cross-sections azimuth and elevation patterns is defined as below [9]: (3) (4) A E (θ, φ) = min[ [A E,V (θ) + A E,H (φ)], A m ] (5) The sum of horizontal and vertical patterns is limited for a common front-to-back attenuation, because it considers inaccuracies of the real world implementation and gives more an environmental view of the effect of the antenna pattern and corresponds better to e.g. the limited isolations of co-sited sectors typically found in field measurements [3]. B. uniform linear array (ULA) response The antenna system at the BS is a uniform linear array of M elements with half-wavelength spacing in the vertical direction. The antenna array response can be represented by [4]: A a (θ, θ etilt ) = log [(W it V i )] 2 = log M 2 w m v m m= (6) where V i = [v, v 2...v m...v M ] T and and the element v m is v m = exp( 2πi(m ) d v cos(θ)), m =, 2,...M (7) λ here V i denotes the phase shift due to element placement in the array. θ is the user s Eod. III. CELLS TOPOLOGICAL STRUCTURE AND USER DISTRIBUTION To perform the system-level simulation, it is necessary to introduce the cells topological structure and user distribution. Here we conduct the wrap-round technique to eliminate the border effect. The cell structure is depicted in Figure.Each hexagonal site is wrapped around by 8 hexagonal sites and is divided into 3 cells (or sectors). The arrow direction in each cell represents the boresight of antennas in horizontal direction. In each cell, the user is dropped uniformly. Fig.. layout of multi-cell system IV. 3D BEAMFORMING METHODS 3D beamforming exploits the additional degree of freedom given by the elevation beam-steering. In this section, we perform two methods to improve the elevation beamformng. They are interpreted below. A. Method I This mehod selects each user s EoD as the electrical downtilt, namely θ etilt = θ EoD. The idea is simple and it just tries to direct the beam to the serviced user rightly. So it will maximize the desired signal for the user, but it ignores the interference caused by the beamforming to other users in adjacent cells who occupy the same time-frequency resource blocks.

B. Method II If the terminal is closer to the BS, the interference to adjacent cells caused by elevation beamforming is reduced, while larger interference is generated if the terminal is closer to cell edge. So the second method follows the principle below to find an appropriate downtilt for per user is based [5]: Maximize the signal strength at the location of the desired user. 2 Minimize the mutual interference between adjacent cells, since current cellular radio systems are mainly interference limited. Based on formula (), and refer to [6][7], it formulates signal-to-leakage-and-noise ratio (SLNR) criterion: H i W i 2 2 max SLNR i = θ N etilt n=,n i H nw i 2 2 P (θ i, φ i, θ etilt ) = N n=,n i P (θ n, φ n, θ etilt ) + Z subject to : θ etilt 8 in (8), Z means the power of noise. P (θ i, φ i, θ etilt ) is desired signal strength when cell i selects θ etilt degrees as the electrical downtilt for elevation beamforming. P (θ n, φ n, θ etilt ) is the interference caused by cell i to the user in adjacent cell n. N is the number of adjacent cells around cell i. φ i is the azimuth of the user in cell i. Similarly, φ n is the azimuth of the user in adjacent cell n with respect the antenna boresight of cell i. θ i is the EoD of the user in cell i, while θ n is the EoD of the user in cell n relative to cell i. Both P (θ i, φ i, θ etilt ) and P (θ n, φ n, θ etilt ) include the path loss, shadow fading and penetration loss considering 3D channel model. The calculation of them are same and can be expressed as : P (θ n, φ n, θ etilt ) = P power + A E (θ n, φ n ) + A a (θ n, θ etilt ) (loss path + loss shad + loss penetreate ) (9) where P power is the transmit power; A E (θ n, φ n ) comes from equation(5); A a (θ n, θ etilt ) is referred to equation(6); loss path is the path loss; loss shad is the shadow fading; loss penetreate is the penetration loss. Utilizing the same approach as descried above, we can find the optimal electrical downtilt for each user. V. SYSTEM EVALUATION In this section, the two 3D Beamforming methods are evaluated compared to the fixed electrical downtilt 2-degree beamforming method in 3D-UMi and 3D- scenarios. In the simulation, we make statistics of the SINR and couplingloss of all users. Couplingloss is the net gain of the received signal including antenna element gain, array response, pathloss, shadow fading and penetration loss. The parameter setting of 3D-UMi and 3D- scenarios is listed below in table. 3D channel model parameter can be referred to [2]. As depicted in figure, there are 57 cells in the system (8) TABLE I PARAMETER DESCRIPTION OF SCENARIOS parameter Urban Micro cell Urban Macro cell (3D-UMi) (3D-) BS antenna height m m Total BS 4 dbm 4 dbm Tx Power formhz formhz Carrier frequency 2 GHz 2 GHz Min. UE-BS 2D distance m m UE antenna height.5m.5m Indoor UE fraction 8% 8% Inter-site distance 2m 5m Outdoor uniform uniform UEs in cell in cell Indoor uniform uniform UEs in cell in cell and the number of users attributed to each cell is. So the multi-cell system holds 57 users.the number of elements of uniform linear array is. Figure 2 is the cumulative distribution function () of SINR in 3D-UMi scenario. From the viewpoint of SINR, Method I outperforms the fixed downtilt method from the SINR of 2.5 db. But method I just obtains a little improvement. The second user-specific elevation beamforming method outperforms the method with fixed electrical downtilt from the SINR of 2.dB. Figure 4 is the simulation result of SINR in 3D- scenario. From the SINR of 6.2dB, method II is better than the fixed downtilt beamforming. However the performance of method I does not achieve improvement. Considering the couplingloss in 3D-UMi scenario, we can see from figure 3 that the two methods proposed in this paper are both better than the fixed 2-degree beamforming. And method I is superior to method II. In figure 5, the couplingloss performance of method II is a little worse than the mehtod with fixed electrical downtilt method in 3D- scenario. However method I still achieves a better performance in couplingloss. Figure 6 shows the average capacity of per cell with the methods under 3D-UMi and 3D- scenarios. The results correspond to the performance of SINR and help us learn the property of the three methods more conveniently. All of the results can be explained below. The second method is a suboptimal strategy which tries to maximize the signal of interest and suppresses the interference to the users in adjacect cells. But method I just selects the EoD as the electrical downtilt and neglects the interference to the users in adjacent cells. Of course it maximizes the desired signal strength. So both in 3D-UMi and 3D- scenarios, method I always achieves the best couplingloss performance. But in SINR performance,it performs worse than method II. In 3D scenario, the couplingloss performance of

.9.8.7.6.5.4 fixed 2 degree dwontilt method UMi.9.8.7.6.5.4 fixed 2 degree downtilt method.3.3.2.2.. 4 2 2 4 6 8 SINR/dB 6 4 2 2 4 6 8 SINR/dB Fig. 2. of SINR in 3D-UMi scenario Fig. 4. of SINR in 3D- scenario.9.8.7.6.5.4.3.2. fixed 2 degee downtilt mehtod UMi 6 4 2 8 6 4 Fig. 3. CouplingLoss/dB of couplingloss in 3D-UMi scenario.9.8.7.6.5.4.3.2. 8 6 4 2 8 6 4 Fig. 5. fixed 2 degree downtilt method couplingloss/db of couplingloss in 3D- scenario method II is a little worse than fixed downtilt method. It can be interpreted that method II sacrifices the couplingloss performance to boost the performance of SINR. Generally the performance of both methods in 3D-UMi scenario is better than that in 3D- scenario. The difference of geometric factors is the main reason. For example, in 3D- scenario, the EoD of all users is not less than 9 degrees. However, in 3D-UMi scenario, many users locations are higher than the BS antennas leading to the EoD less than 9 degrees. VI. CONCLUSION AND FUTURE WORK In this paper, two methods are proposed for user-specific elevation beamforming. Method I just simply selects the EoD of each user as the electrical downtilt. Method II is to find the optimal electrical downtilt for every user making a tradeoff between the desired signal strength and the interference to adjacent cells. The simulation is carried out in 3D- and 3D-UMi scenarios utilizing the latest 3D average capacity of per cell/bit/s/hz 5 45 4 35 3 25 2 5 5 Fig. 6. Umi fixd 2 degree downtilt method Three Methods Average capacity of per cell with 3 methods in both scenarios

channel model of 3GPP. The method is evaluated through SINR and couplingloss compared to the fixed 2-degree elevation beamforming method. Method I in 3D-UMi achieve a little improvement. But in 3D- scenario, it achieves poor performance. Method II outperforms the fixed 2- degree elevation beamforming method from SINR in both scenarios. But the performance of couplingloss of method II in 3D- scenario is a little worse than that with the fixed electrical downtilt method because the method sacrifices the couplingloss performance to boost the performance of SINR. Also both methods achieve better performance in 3D-UMi scenario than that in 3D- scenario. In the future work, we will consider the aspects of fast fading, the scheduling of radio resources and coordination scheme to develop the methods. [6] Mirette Sadek,Alireza Tarighat, and Ali H. Sayed, A Leakage-Based Precoding Scheme for Downlink Multi-User MIMO Channels, IEEE Trans.Wireless Communications,vol.6,no.5,pp.7-72,May.27 [7] Jai-Hoon Lee,Seokkwon Kim,Seung-Ri Jin and Dong-Jo Park, A Multi- User Beamforming Scheme in MIMO Downlink Channels for Multi-Cell NetworksA,IEEE International Conference on Consumer Electronics (ICCE),Jan 9-2,2,Las Vegas, NV ACKNOWLEDGMENT This research has been supported by 863 Project of China (No. SS24AA25) and the State Key Laboratory of Wireless Mobile Communication of China Academy of Telecommunication Technology (No.27DQ3556). REFERENCES [] Shanzhi Chen, Yingmin Wang, Weiguo Ma and Jun Chen, Technical Innovations Promoting Standard Evolution: From TD-SCDMA To TD-LTE and Beyond, IEEE Wireless Communications,Vol.9,PP.6-66,Feb.22 [2] Meilong Jiang, Mohsen Hosseinian, Moon-il Lee, Janet Stern-Berkowitz, D Channel Model Extensions and CharacteristicsStudy for Future Wireless ystems, PIMRC 23 IEEE 24th International Symposium,Sept 8-,23, London, United Kingdom [3] Frederick W. Vook, Timothy A. Thomas, Eugene Visotsky, Elevation Beamforming with Beamspace Methods for LTE,PIMRC 23 IEEE 24th International Symposium,Sept 8-,23, London, United Kingdom [4] Over view of 3GPP Release 2 V..9(23-6) [5] Yan Li, Xiaodong Ji, Dong Liang, and Yuan Li, Dynamic Beamforming for Three-Dimensional MIMO Technique in LTE-Advanced Networks, International Journal of Antennas and Propagation,vol.23,pp.C8,23 [6] Yang Song, Xiang Yun, Satoshi Nagata, Lan Chen,Investigation on Elevation Beamforming forfuture LTE-Advanced,Communications Workshops (ICC), 23 IEEE International Conference,June 9-3, 23, Budapest [7] Mansoor Shafi, Min Zhang, Peter J. Smith, Aris L. Moustakas and Andreas F.Molisch, The Impact of Elevation Angle on MIMO Capacity, ICC 26 IEEE International Conference,June 26,Istanbul [8] H.Halbauer, S. Saur, J. Koppenborg, C. Hoek, Interference Avoidance with Dynamic Vertical Beamsteering in Real Deployments, IEEE Wireless Communications and Networking Conference, April -4, 22, Paris, France [9] R-3328, Initial calibration results for 3D channel modelling, CATT, 3GPP TSG RAN WG Metting #74, Barcelona,Spain 9th Aug-23th AUG 23 [] R-33347, Initial calibration of Large Scale channel modelling,nec Group,3GPP TSG RAN WG Metting #74,Barcelona, Spain 9th Aug- 23th AUG 23 [] R-33383, Initial results and discussion for 3D-channel calibration,lg Electronics, 3GPP TSG RAN WG Metting #74,Barcelona, Spain 9th Aug-23th AUG 23 [2] 3GPP, TR 36.873 v..(23-9), 3D channel model for LTE (Release 2) [3] O. Yilmaz, S. Hamalainen, J. Hamalainen, Comparison of Remote Electrical and Mechanical Antenna Downtilt Performance for 3GPP LTE, In Proceedings of IEEE Vehicular Technology Conference,Sept 2-23,29,Anchorage-Alaska, USA [4] 3GPP, TR 37.84, Study of Radio Frequency(RF) and Electromagnetic Compatibility (EMC) requirements for Active Antenna Array System(AAS) base station (Rel.2),23 [5] Valeria DAmico and Hardy Halbauer, Innovative advanced signal processing algorithms for interference avoidance, ARTIST4G technical deliverable, 2