Spatia Characteristics of 3D MIMO Wideband Channe in Indoor Hotspot Scenario at 3.5 GHz ei Tian Jianhua Zhang Yuxiang Zhang State Key ab. of Networking and Switching Tech. Beijing Univ. of osts and Teecom. China. tianbupt@bupt.edu.cn jhzhang@bupt.edu.cn zhangyx@bupt.edu.cn Yi Zheng China Mobie Communications Corporation China. zhengyi@chinamobie.com Abstract Three dimension (3D) Mutipe Input and Mutipe Output (MIMO) with massive antenna eements due to the advantage of using eevation domain than D MIMO is regarded as a promising technique for the fifth generation (5G) mobie communication systems. This makes the channe characteristics of eevation domain as a basic requirement for the 5G channe mode. In this paper we conducted a 3D MIMO channe measurement at 3.5 GHz with 0 MHz bandwidth in indoor hotspot scenario which is one of the important scenario for 5G. Based on the channe measurement data spatia characteristics of eevation domain are anayzed and discussed incuding power anguar spectrum anguar spread and crosscorreation etc. The extracted channe parameters of eevation domain can be used in the 3D MIMO technoogy evauation and simuation for indoor hotspot scenario. Index Terms 3D MIMO channe measurement indoor hotspot eevation ange I. INTRODUCTION With the rapid and tremendous growing of the voume of mobie traffic the fifth generation (5G) mobie communication system aso named as IMT- 00 by Internationa Teecommunication Union Radiocommunication Sector (ITU-R) aims at higher user data rate short transmission atency and ow energy consumption etc. has aroused great attention a over the word [1]. In order to reach these requirements three dimension (3D) Mutipe Input and Mutipe Output (MIMO) is considered to be one of the most promising and practica technique. MIMO technique has been introduced into the fourth generation mobie communication system and improved the system spectrum efficiency to a arge extent by utiizing the channe spatia degree of freedom. Compared with the traditiona MIMO 3D MIMO especiay with massive antenna eements can take fu use of the spatia channe resources by introducing the eevation domain []. One of the fundamenta work to evauate and utiize 3D MIMO technique is to extend the D channe to 3D with adding the eevation domain characteristics. A ot of research has been focusing on the 3D MIMO channe to estabish a reiabe and precise 3D channe mode [3-8]. The report of Word Wireess Initiative New Radio + (WINNER+) summarise some measurement resuts of the eevation ange of arriva (EoA) at mobie station (MS) side and corresponding eevation ange of departure (EoD) at base station (BS) side for various scenarios [3]. most of the studies pay attention to the 3D channe in outdoor scenarios ike Urban Macro (UMa) and Urban Micro (UMi) and outdoor-to-indoor (OI) scenario. [7] anayzes the anguar spread and offset in eevation domain based on the measurement in Urban environment whereas [6] and [8] focus on the eevation characteristics in OI scenario. In eary 014 the 3 rd Generation artner roject (3G) issued a technica report TR 36.873 which is a 3D MIMO channe mode contributed by ots of companies [9]. However the report ony contains the 3D channe mode for UMa UMi and OI scenario. Thus there is few study on the 3D channe of indoor hotspot (InH) scenario. InH scenario referred to the environment with arge quantities of users and big mobie traffic oad such as shopping ma train station and airport is one of the most important scenario for IMT-00. Therefore considering that the knowedge of 3D channe is indispensabe to deveop 3D MIMO in InH scenario we conducted the 3D channe measurement campaign at 3.5 GHz with 0 MHz bandwidth in order to obtained the spatia characteristics of eevation domain in InH scenario. The extracted eevation ange parameters coud support the evauation and simuation of 3D MIMO technique. The remainder of the paper is organized as foows. The channe measurement system and scenario pan are described in Section II. Then the data processing procedure is expained in Section III. In Section IV the channe characteristics of eevation domain of InH scenario are anayzed and discussed in detaied. Finay the concusion is drawn in Section V.
II. CHANNE MEASUREMENT DESCRITION A. Measurement System The measurement campaign was conducted at a center frequency of 3.5 GHz with 0MHz bandwidth by using the Eektrobit ropsound channe sounder []. An auto-swicthing unit is equipped both at transimtter (Tx) and receiver (Rx) to contro the periotic pseudo random (N) binary signa transmitting between the Tx and Rx antenna paris in a time division mutipexing mode. One time period in which a the sub-channes are sounded once is regarded as a measurement cyce (snapshot). The Rx antenna with the height of 1.7 m is set up on a troey. The Rx represented by mobie station (MS) are marked by yeow and gray trianges for ine-of-sight (os) and non-ine-of-sight (NoS) respectivey. A tota of 60 Rx ocations are seected and 500 snapshots are coected at each Rx ocation. Fig.. Measurement pan of InH scenario. Fig. 1. Antenna arrays used at Tx (UA) and Rx (ODA). 3D antenna arrays are chosen to capture the 3D channe responses. A uniform panr array (UA) is appied at Tx whereas a omni-directiona antenna array is utiized at Rx. As iustrated in Fig.1 the UA consists of totay 16 antenna patches which are paced 4 4 uniformy in a square. Each antenna patch is a dua-poarized antenna connected with two antenna ports. The ODA has 8 antenna patches 4 of which are arranged in 8 adjacent sides with 3 in each side 4 or which are on the top. The detaied antenna specification and system configuration are isted in Tabe I. TABE I. ANTENNA SECIFICATION AND SYSTEM CONFIGURATION Configuration Item Vaue Antenna type of Tx / Rx UA ODA Number of antenna eements 3 56 Anguar range of antenna oarization Horizonta 45 45 70 ~ 70 180 ~ 180 Vertica 70 ~ 70 70 ~ 90 Center frequency 3.5 GHz Bandwidth 0 MHz ength of N code 55 Tx power 33 dbm B. Measurement scenario The InH scenario is seected in the ha of the third teaching buiding of Beijing University of osts and Teecommunications. Fig. shows the ayout of the InH measurement area. The Tx regarded as base station (BS) marked by the red triange is ocated in the center of the obby of the first foor with the antenna height of.61 m. III. DATA ROCESSING After the fied channe measurement channe impus responses (CIRs) are firsty be cacuated from the raw channe data. It is assumed that the number of Tx and Rx antenna eements are M and N respectivey. Then the number of sub-channe is M N which is 3 56 in our measurement. A. SAGE Agorithm After we obtain CIRs the space aternative generaized expectation maximization (SAGE) agorithm is utiized to extract the muti-path channe parameters due to its high precision and ow computaion compexity [11][1]. The SAGE agorithm provides a joint estimation parameter set of the th path which can be written as The A rx rx tx tx 1. (1) A and denote the compex poarization matrix propagation deay and dopper shift of the th path. Especiay and represent the azimuth and rx rx eevation anges of arriva at Rx side whereas tx and tx are the corresponding anges of departure at Tx side. is the maximum number of resovabe path to be estimated. It is empiricay set to 10 which gauratees that the paths with the strongest power can be extracted from the measurement data. B. ower Anguar Spectrum As we have obtained the parameter set of each path the power of each path can be cacuated by p1 p p11 p1 ()
where p 1 p is the poarization coefficient in the power anguar spectrum (AS) is written as where coud be A.Then (3) f( ) rx 1 rx tx or tx. Then the AS of AoA EoA AoD or EoD can be obtained correspondingy. In this paper we mainy focuse on the AS of eevation ange. Generay apacian distribution is used to fit the eevation ange AS whose prorabiity distribution function is given as 1 f ( x) exp x (4) where and denotes the mean and standard deviation of apacian distribution respectivey. C. Anguar Spread Root-mean-square anguar spread (rms AS) is an important parameter to characterize the dispersion in spatia domain of muti-path components. It is the second centra moment of AS. In order to avoid the ange ambuguity the circuar anguar spread (CAS) is often used and can be cacuated by D. Cross-correation of parameters As the eevation ange is introduced to the channe mode the cross-correations of eevation ange with other channe parameters incuding shadow fading (SF) Ricean K factor deay spread (DS) and azimuth anguar spread of departure and arriva (ASD ASA) are required. The cross-correation coefficient of two channe parameters X and Y can be cacuated by XY where and respectivey. X X X Y Y E E E (8) Y A. AS of Eevation Ange X Y dontes the variance of IV. RESUTS AND ANAYSIS X and Y The ASs of EoD and EoA for os and NoS cases are iustrated in Fig. 3. It can be observed that The eevation ange is fitted by the apacian distribution we. It shows the same with the eevation ange distribution of UMa and UMi defined in [9]. CAS 1 min AS min 1 (5) represent the shifted ange with where. certain anguar shift of. are defined as 1 1. (6) (a) AS of EoD in os InH It shoud be noted that both in (6) and in () are normized according to if if. if (7) (b) AS of EoA in os InH
(c) AS of EoD in NoS InH (d) AS of EoA in NoS InH Fig. 3. AS of eevation ange. B. Eevation Anguar Spread Fig. 5. ognorma fit for ESD and ESA distribution in NoS InH. Fig. 4 and 5 shows the distribution of ESD and ESA in os and NoS cases respectivey. ognorma distribution is used and found to fit the distribution of ESD and ESA we. The same distribution is aso used for DS ASA and ASD distribution fitting in [9]. The mean vaue and standard deviation of the ESD and ESA distribution is given in Tabe II. The mean ESD in NoS case is arger than that in os case. So is the mean ESA. However the difference of ESD or ESA between os and NoS is siight. This refects the eevation ange can be restricted to some extent by the foor and ceiing in the indoor environment. Comparing the resuts of indoor office in [3] with the measured resuts we can further demonstrate the concusion that the eevation anguar sparead coud decrease with the indoor space becoming smaer. TABE II. ESD AND ESA STATISTICS IN OS/NOS INH SCENARIO Measured InH WINNER + Indoor Office [3] Anguar Spread ESD og ([degree]) ESA og ([degree]) ESD og ([degree]) ESA og ([degree]) Indoor Hotspot os NoS 1.0 1.08 0.41 0.36 1. 1.6 0.3 0.67 0.88 1.06 0.31 0.1 0.94 1. 0.6 0.17 Fig. 4. ognorma fit for ESD and ESA distribution in os InH. C. Cross-correation of channe parameters Except for distribution of eevation ange and eevation anguar spread the cross-correation is aso needed when generating the 3D channe mode. Tabe III gives out the cross-correation of eevation ange with other important channe parameters both in os and NoS case. It can be seen from Tabe I that the ESD is generay higher correated with ASD than with other parameters both in os and NoS cases. The simiar reation can aso be found between ESA and ASA. This
interprets that the azimuth ange and eevation ange at either termina (BS or MS side) are more ikey to change with each other correspondingy. TABE III. CROSS-CORREATION OF EEVATION ANGUAR SREAD WITH OTHER ARAMETERS IN OS AND NOS OF INH SCENARIO Cross-correation Indoor Hotspot os NoS ESD vs SF 0. 0 ESA vs SF 0.3 0 ESD vs K 0 N/A ESA vs K 0.1 N/A ESD vs DS 0.1-0.7 ESA vs DS 0. -0.06 ESD vs ASD 0.5 0.35 ESA vs ASD 0 0.3 ESD vs ASA 0-0.08 ESA vs ASA 0.5 0.43 ESD vs ESA 0 0.4 V. CONCUSION As 3D MIMO technique is important to enhance the throughput in InH scenario it is definitey necessary to extend the D channe to 3D with adding the eevation ange parameters. Thus we performed an 3D MIMO wideband channe measurement at 3.5 GHz with 0 MHz bandwidth in InH scenario incuding both os and NoS cases. Further the AS anguar spread and cross-correation of eevation ange are anayzed. The distributions of EoD and EoA obey apace distribution which are simiar to those of UMa and UMi scenarios. The ESD and ESA can be modeed as ognorma distribution. Aso the cross-correation of ESD and ESA with other parameters are provided. The eevation ange parameters in this paper can be used to extend the D InH channe mode to 3D channe so as to support the 3D MIMO technique evauation and simuation. ACKNOWEDGMENT The research is supported by Nationa Key Technoogy Research and Deveopment rogram of the Ministry of Science and Technoogy of China (NO. 01BAF14B01) Nationa Science and Technoogy Major roject of the Ministry of Science and Technoogy (NO. 015ZX0300008) Ministry of Education-CMCC Science and Research Fund (MCM01605). The research is aso supported by the Exporatory roject of State Key ab of Networking and Switching Technoogy (NST017005). REFERENCES [1] ITU-R M.038 IMT-Vision Framework and overa objectives of the future deveopment of IMT for 00 and beyond Sept. 015. [] M. Shafi A. F. Moisch. J. Smith T. Haustein. Zhu. De Siva F. Tufvesson A. Benjebbour and G. Wunder 5G: A tutoria overview of standards trias chaenges depoyment and practice IEEE Journa on Seected Areas in Communications 017. [3] Wireess word initiative new radio + (WINNER+) D5.3 fina channe modes https://www.ist-winner.org. 0 [4] J. Zhang; Y. Zhang; Y. Yu; R. Xu; Q. Zheng;. Zhang 3D MIMO: How Much Does It Meet Our Expectations Observed from Massive Channe Measurements? in IEEE Journa on Seected Areas in Communications 017. [5] J. Zhang C. an F. ei G. iu and X. Cheng Threedimensiona fading channe modes: A survey of eevation ange research IEEE Communications Magazine vo. 5 no. 6 pp. 18-6 014. [6] V. Kristem S. Sangodoyin C. Bas M. Kaske J. ee C. Schneider G. Sommerkorn J. Zhang R. Thoma and A. Moisch 3D MIMO outdoor to indoor macro/microceuar channe measurements and modeing in 015 IEEE Goba Communications Conference (GOBECOM) pp.1-6 015. [7] F. ei J. Zhang C. an Eevation ange characteristics of Urban wireess propagation environment at 3.5 GHz in 013 IEEE Vehicuar Technoogy Conference (VTCfa 13) Sept. 013. [8] Y. Yu J. Zhang M. Shafi. A. Dmochowski M. Zhang and J. Mirza "Measurements of 3D channe impuse response for outdoor-to-indoor scenario: Capacity predictions for different antenna arrays" 015 IEEE 6th Annua Internationa Symposium on ersona Indoor and Mobie Radio Communications (IMRC) Hong Kong 015 pp. 408-413. [9] 3G TR 36.873 Study on 3D channe mode for TE (Reease 1) 014. [] ropsound mutidimensiona channe sounder. [Onine]. Avaiabe: http://www.propsim.com. [11] B. H. Feury. Jourdan and A. Stucki High-resoution channe parameter estimation for MIMO appications using the SAGE agorithm 00 Internationa Zurich Seminar on Broadband Communications Access - Transmission - Networking (Cat. No.0TH8599) Zurich 00 pp. 30-1-30-9. [1] X. Yin B. H. Feury. Jourdan and A. Stucki oarization estimation of individua propagation paths using the SAGE agorithm in 14th IEEE roceedings on ersona Indoor and Mobie Radio Communications (IMRC 003) 003 pp. 1795-1799.