Field Test of Uplink CoMP Joint Processing with C-RAN Testbed

212 7th International ICST Conference on Communications and Networking in China (CHINACOM) Field Test of Uplink CoMP Joint Processing with C-RAN Testbed Lei Li, Jinhua Liu, Kaihang Xiong, Peter Butovitsch Ericsson (China) Communications Co. Ltd, Beijing, China lei.m.li; jinhua.liu; jack.xiong; peter.butovitsch@ericsson.com Abstract Uplink Coordinated Multi-Point (CoMP) joint processing has the potential to improve uplink capacity by joint processing the received signal from more antennas and nodes. With C-RAN architecture, centralized and coordinated baseband processing among base stations become feasible. This paper introduces LTE uplink CoMP joint processing -- the principles, simulation evaluations, field testbed and test results. The field test is done with C-RAN field testbed which is built around Ericsson office in Beijing. Both the simulation evaluations and field test results show that uplink CoMP has great gain at the cell edge especially when there is interference, so it can be a good solution to improve the cell edge throughput of the LTE system. Keywords - LTE; LTE-A; multiple-input-multiple-output (MIMO); coordinated multi-point (CoMP); inter-cell interference coodination (ICIC); interference rejection combination (IRC); field test; uplink; heterogenous network I. INTRODUCTION Inter-cell interference severely damages cell-edge user throughput. In the future heterogeneous network deployment, by introducing Micro or Pico cells into LTE Macro cell grid, interference problem between Macro and Micro or Pico becomes critical [1]. Uplink Coordinated Multi-Point (CoMP) joint processing has the potential to improve the interference situation both in the macro grid cell edge and in the heterogeneous network scenarios. By combining received signal from more antennas and nodes interference rejection may be further improved [2]. Uplink CoMP joint processing, however, needs real time signals and great amount of data exchange between enodebs and centralized data processing. C-RAN architecture shows the advantage for centralized and coordinated baseband processing among enodebs, thus uplink joint processing becomes feasible [3] [4]. In this paper the following content will be introduced, firstly the C-RAN field testbed which was deployed in Beijing around Wangjing Ericsson Tower; secondly the uplink CoMP joint processing principle and simulation evaluation; thirdly the field test results. In the field tests, two uplink CoMP receptions are compared: MRC (Maximum Ratio Combining) and full IRC (Interference Rejection Combining). The target with the tests was to evaluate if the uplink joint processing could show gain in both urban and suburban network environment. Both the simulation results and test results show that uplink joint processing has significant gain either with or without interferer. Hence it can be a good solution to improve the uplink performance at the cell edge. II. C-RAN TESTBED AND NETWORK CONFIGURATIONS A C-RAN field testbed was set up in Beijing, which is used for proof of concept to new features and solutions with C-RAN architecture. Ericsson provided an end-to-end C-RAN solution for the field test, consisting of multi-standard base stations, EPC (Evolved Packet Core) network, OSS (Operation Support System) software. The tests are conducted using precommercial LTE terminals. The network is located around Ericsson office in Wangjing Science and Technology Park, which is a combination of urban and suburban radio environments. The network and base station configurations are shown in Figure 1 and Table 1 separately. There are four Macro cells, including one three-sector site (, 2, 3) and one one-sector site (). Distance between and, 2, 3 is 46 meters. Each cell has two Tx/Rx antennas. This testbed can support TD-LTE, FDD LTE and GSM as well. CDK S3 Ericsson Figure 1. C-RAN testbed: network 753 978-1-4673-2699-5/12/$31. 212 IEEE

T R T R T R T R T R R TABLE I. BASE STATION CONFIGURATIONS Outdoor sites S3 Max. Tx Power Antenna Configuration Height WDM-PON CPRI backhaul PtoP fiber CPRI backhaul WDM WDM WDM WDM WDM WDM PtoP fiber CPRI backhaul E-band Microwave CPRI backhaul indoor sites Figure 3. C-RAN testbed: Transport solutions The uplink CoMP joint processing in this paper is tested under TD-LTE, 2.3GHz frequency band, 1MHz bandwidth system, using point to point (PtoP) fiber for CPRI (Common Public Radio Interface) transport. Despite we use TD-LTE in this test, the uplink CoMP joint processing can also work on FDD LTE systems. The C-RAN testbed hardware is shown in Figure 2. In order to emulate the baseband pool environment, one BBU (Base Band Unit) slice is used for one cell. For the CPRI transport between RRU (Remote Radio Unit) and BBU, we now have three alternative solutions: PtoP fiber, WDM-PON (Wavelength Division Multiplexing-PON) and E-band Microwave, which are shown in Figure 3. We also provide a SIU (Site Integration Unit) based LTE and GSM integrated backhaul solution from BBU to Core network in this testbed. RRU RRU RRU CPRI RRUs 2TX2RX III. UPLINK COMP JOINT PROCESSING A. Uplink CoMP introduction Figure 4 shows a basic physical layer processing procedure for uplink CoMP joint processing. It consists of four parts: 1) Channel estimation is performed individually per each receive antenna; 2) Antenna selection is performed to potentially remove some unsuitable links; 3) Covariance estimation is to capture interference estimation in a good way; 4) Minimum Mean Square Error (MMSE) equalizer for IRC receiver is finally performed to gain a system by uplink joint processing with involved receive antennas. Optical Transmission Equipment Baseband 9 x RBS 622 1 x BBU per sector Ethernet Switch Existing EPC CORE + O&M Figure 2. C-RAN testbed: Hardware Figure 4. Basic physical layer processing procedure B. Simulation results Table II shows the basic link level simulation assumptions. 754

TABLE II. SIMULATION ASSUMPTIONS Parameters Description System bandwidth 5MHz @ 2GHz UE category Rel-8, support of SIMO Base station antenna configuration 2Rx/cell: one pair of X-pol Channel model Extended TU, 3km/h Scheduled bandwidth 6 consecutive resource blocks Channel estimation Realistic based on DM-RS Link adaptation On HARQ On, max. 4 Power control for targeted SNR Desired UE: db or 8dB Interfering UE: fixed 8dB Figure 5 shows the simulation results of uplink joint processing involving two cells with each cell two receive antennas. It can be concluded that uplink joint processing can efficiently improve performance, especially when the link quality difference between serving link and supporting link is in the range of 4dB. When the supporting link is further worse, to combine the link for uplink joint processing is not necessary due to the negative impact by estimation errors. The larger gain is investigated in the case of low SNR situation, where energy combination is a more efficient way to boost performance. Thoughput 3.5 x 16 3 2.5 2 1.5 1.5 No CoMP CoMP,Sup.Link=dB CoMP,Sup.Link=-4dB CoMP,Sup.Link=-8dB CoMP,Sup.Link=-1dB SNR=dB 2BS case,etu-3km/h SNR=8dB Figure 5. Uplink CoMP performance: two cells of each two receive antennas C. Test scenario and results In the field tests, two sectors are selected to do uplink joint processing. There are two kinds of uplink CoMP receivers: 1) MRC, that is MRC reception is utilized for all four receive antennas from two cells; 2) full IRC, that is IRC reception is utilized for all four receive antennas from two cells. For the baseline, uplink CoMP is turned off and IRC reception is used at each individual cell. We have two test scenarios: intra-site CoMP and inter-site CoMP. We will introduce them in the following subsections. The location of the test spots together with user RSRP and NRSRP (Neighbor cell RSRP or support link RSRP) are denoted in Figure 6 RSRP: -9 NRSRP: -9 RSRP: -96 NRSRP: -96 Figure 6. Intra-site uplink CoMP test scenario RSRP: -11 NRSRP:-11 We test intra-site uplink CoMP both in non-interference scenario and in interfering scenario. During the interfering test, we add one interfering user. The interfering user is attached to the neighbor cell that is. As shown in Figure 6 with the purple dot. The interfering user is located in the place where the RSRP to and are both -95dBm. The interfering user sends uplink traffic with full PUSCH power 23dBm. During the test, both the test user and the interfering user upload full buffer traffic. It should be noted that, there are high buildings along the street, some of the buildings are higher than the enodeb antennas, therefore some of the test spots and interfering spots are in NLOS locations. Because of the uncertainty of radio link quality there are some fluctuations in RSRP during the test. We try to note a long time average on RSRP in Figure 6. Intra-site test results without interference are shown in Figure 7 and Figure 8. In Figures 7, baseline throughput decreases with RSRP decreases. At the test spot with -9dBm RSRP and -9dBm NRSRP, the baseline throughput is good enough and almost reaches to peak throughput, and there is very little CoMP gain due to there is very small margin to increase the user data rate. In Figure 8, the CoMP gain increases when the radio condition becomes worse. The throughput gain reaches to 5% when the RSRP and NRSRP are around -11dBm. a) Intra-site CoMP For the intra-site CoMP test, sector and are selected to do uplink joint processing, shown in Figure 6. In the test, three test spots are tested to compare uplink throughput with MRC and full IRC CoMP to the uplink throughput when uplink CoMP. At all the test spots, the test user is attached to sector. 755

1 9 8 7 6 5 4 3 2 1 Throughput (Kbps) N Intra-site UL CoMP without Interference -- Throughput Non-CoMP IRC -9-95.5-11 -9-96.8-11.7 18.% 16.% 14.% 12.% 1.% 8.% 6.% 4.% 2.%.% NRSRP(dBm) Intra-site UL CoMP with Interference -- Throughput Gain -94.2-97.8-1.5-11 -13-11.8 Figure 7. Intra-site uplink CoMP throughput without interference 6.% 5.% 4.% 3.% 2.% 1.%.% N Intra-site UL CoMP without Interference -- Throughput Gain -9-95.5-11 -9-96.8-11.7 Figure 8. Intra-site uplink CoMP throughput gain without interference Intra-site test results with interference are shown in Figure 9 and Figure 1. Like the above results, in Figure 9, the baseline throughput decreases with radio conditions getting worse. Compare to the baseline throughput in Figure 7, the throughput is lower because of interference. Because of interference and relatively worse supporting link compared the serving link, MRC CoMP almost shows no gain for the first two test spots. Full IRC CoMP, however, shows expected capability in interference mitigation. The CoMP gain is significant when RSRP and NRSRP are around -1dBm and reaches to 15%. This can be due to that with 4 receiver antennas, the IRC receiver can efficiently suppress one dominant interference. 7 6 5 4 3 2 1 Throughput (Kbps) NRSRP(dBm) Intra-site UL CoMP with Interference -- Throughput Non-CoMP -94.2-97.8-1.5-11 -13-11.8 Figure 9. Intra-site uplink CoMP throughput with interference Figure 1. Intra-site uplink CoMP throughput gain with interference b) Inter-site CoMP For the Inter-site CoMP test, sector and are selected to do uplink joint processing, shown in Figure 11. In this test, three test spots are selected to compare uplink throughput with MRC and full IRC CoMP respectively to the uplink throughput when without CoMP. At all the test spots, the test user is attached to sector. The location of the test spots together with their RSRP and NRSRP (The cell RSRP of the support link RSRP) are denoted in Figure 11. The inter-site uplink CoMP is tested both in noninterference scenario and in interfering scenario. During the interfering test, we add one interfering user. The interfering user is attached to the neighbor cell that is, as shown in Figure 11 with the purple dot. The interfering user is located where the RSRP from and are the same: -95dBm. The interfering user sends uplink full buffer traffic with full PUSCH power 23dBm. The test user also send full buffer traffic in the uplink. CDK Interference UE RSRP:-15 NRSRP:-13 RSRP:-95 NRSRP:-95 Figure 11. Inter-site uplink CoMP test scenario RSRP:-12 NRSRP: -12 Again, it should be noted that, there are high buildings along the street and some of the buildings are higher than the enodeb antennas, therefore some of the test spot and interfering spots are in NLOS locations. Because of the uncertainty of radio link quality there are some RSRP fluctuations during the tests. We try to give a long time average on RSRP in Figure 11. 756

Inter-site test results without interference are shown in Figure 12 and Figure 13. From these two figures we can see similar trends for baseline throughput as intra-site scenario and CoMP throughput gain increases as the RSRPs decreases. At the second test spot there is little CoMP gain especially for full IRC is because the supporting link is 9dB worse than the serving link, hence unbalanced receive signals contribute minor combination gain, which is fit for the theory expectations. In Figure 13, CoMP throughput gain reaches to 7% when the RSRP and NRSRP are around -13dBm. 8 7 6 5 4 3 2 1 Throughput (kbps) N Inter-Site UL CoMP with Interference -- Throughput Non-CoMP IRC -95-12 -15-95 -1-13 Inter-site UL CoMP without interference -- Throughput Figure 14. Inter-site uplink CoMP throughput with interference 9 8 7 6 5 4 3 2 1 Throughput (kbps) N -95-98 -14-96 -17-13 Non CoMP IRC Figure 12. Inter-site uplink CoMP throughput without interference 16 14 12 1 8 6 4 2 N Inter-Site UL CoMP with Interference -- Throughput Gain -95-12 -15-95 -1-13 Gain (% ) 8.% 7.% 6.% 5.% 4.% 3.% 2.% 1.%.% N Inter-site UL CoMP without interference -- Throughput Gain -95-98 -14-96 -17-13 Figure 13. Inter-site uplink CoMP throughput gain without interference Inter-site test results with interference are shown in Figure 14 and Figure 15. The CoMP gain reaches 15%, which is similar to the results for intra-site CoMP. Generally, inter-site CoMP gain is similar to intra-site CoMP, either without interference or with interference. Figure 15. Inter-site uplink CoMP throughput gain with interference IV. CONCLUSIONS In this paper, C-RAN field testbed, uplink CoMP joint processing principle and field test results are introduced. From both the simulation and the field test results, it can be concluded that there is significant gain with uplink CoMP at cell edge both for intra-site CoMP and inter-site CoMP. CoMP gain increase at the poor radio conditions. When there is no interference the CoMP throughput gain is 3~5%. When there is one strong interference the gain can be even higher, and sometimes can reach 15%. Full IRC CoMP shows clear advantages when there is interference. With heterogeneous network deployment in the future, uplink CoMP can be further promising for interference suppression between Macro cell and Pico or Micro cells and eliminating uplink/downlink imbalance. We have the further plan to test uplink CoMP in heterogeneous network scenarios. REFERENCES [1] Madan R., Borran J., Sampath A., Bhushan N., Khandekar A., and Tingfang Ji, Cell association and interference coordination in heterogeneous LTE-A cellular networks, IEEE Journal on Selected Areas in Communications, vol. 28, no. 9, pp. 1479 1489, Dec. 21. [2] Alberto Zanella, Marco Chiani, and Moe Z. Win, MMSE reception and successive interference cancellation for MIMO systems with high spectral efficiency, IEEE Transactions on Wireless Communications, vol. 4, no. 3, pp. 1244 1253, May 25. [3] Wang Xiaoyun, Huang Yuhong, Cui Chunfeng, Chen Kuilin, Chen Mo. C-RAN: Evolution toward Green Radio Access Network, China Communications, vol.3, pp. 17-112, Jan. 21. [4] China Mobile Research Institue. C-RAN White Paper, v2.5, Oct. 212. http://labs.chinamobile.com/report/view_59826 757