Test strategy towards Massive MIMO Using LTE-Advanced Pro efd-mimo Shatrughan Singh, Technical Leader Subramaniam H, Senior Technical Leader Jaison John Puliyathu Mathew, Senior Engg. Project Manager
Abstract User throughput is key performance indicator for any wireless network. With the advancement in mobile technology, demand for high user throughput is increasing. To fulfil this requirement, 3rd generation partnership project (3GPP) introduced Multiple Input Multiple Output (MIMO) technology from its initial releases of LTE. Till release-9, MIMO was upto 4-layers. In release-10 (LTE-Advanced), this got extended to 8- layers with the introduction of new reference signal: CSI-RS (upto 8 ports). So, we see an enhancement made in MIMO between LTE and LTE-Advanced. This trend continued in LTE- Advanced Pro as well, where a path towards Massive MIMO is introduced using FD- MIMO/eFD-MIMO with the support of upto 16 CSI-RS ports in 3GPP release-13 and upto 32 CSI-RS ports in release-14. The intent of this paper is to set a test strategy platform and emulate test scenarios of efd- MIMO thereby validating key performance and functional indicators which could be further utilised and extended for Massive MIMO. Motivation behind efd-mimo/massive MIMO 3GPP release-10 introduced 8x8 MIMO with new UE category Cat-8 supporting upto 8 spatial layers and base station supporting upto 8 CSI-RS antenna ports. Now UE can perform channel estimation based on CRS or CSI-RS antenna ports. CSI-RS ports introduced because increasing CRS ports were not feasible as it takes fix amount of downlink resources which cannot be used for data. Also, increasing CRS will introduce more interference in network. CSI-RS has no such restriction and it can be configured based on the need. Number of layers in MIMO scheduling depends on rank reported by UE based on 8-layer CSI-RS channel estimation. And, with different channel conditions of real network, most of the UEs are not able to report higher rank 6,7 or 8 to base station and so unable to get better throughput. This is the driving factors for efd-mimo/massive MIMO, where with increase in CSI-RS antenna ports, base station facilitates UEs to identify better channels in current traffic scenario. This enables more and more UEs to report better rank and reach upto 8-layers data scheduling in downlink. The main motivation is to improve: 1. Cell coverage 2. Cell edge user experience 3. Average spectrum efficiency 4. Optimise load-balancing between cells Massive MIMO Massive MIMO is a feature where base stations have very large numbers of antennas (in hundreds). With 3GPP release-10 eight layer beamforming, channel estimation is based on CSI-RS rather than CRS. To fulfil the need of better spectrum efficiency and overall increased downlink throughput, this approach got enhanced even further as a whole defined as Massive MIMO. In Massive MIMO
design, it s possible that base station has more antenna than active users. More number of antenna increases beamforming coverage by sharply focusing transmission on particular point thereby increasing overall network coverage and facilitating better channel estimation by UE. This will enhance possibility for UE to report better CSI with bigger rank/cqi resulting into increased user throughput and spectrum efficiency in downlink. Massive MIMO scheduling will be based on CSI-RS antenna ports. The evolution path for Massive MIMO is mentioned in below table: The evolution path for Massive MIMO is: 3GPP release R10-R12 LTE-A R13 LTE-A Pro R14 LTE-A Pro R15-5G MIMO Technology 2D MIMO FD-MIMO efd-mimo Massive MIMO No. of CSI-RS ports (Base station) 8 16 32 >=64 (assumed) Transmission Mode TM9, TM10 TM9, TM10 TM9, TM10 TM9, TM10 CSI-RS transmission NonPrecoded NonPrecoded/ Beamformed No. of layers supported by UE NonPrecoded/ Beamformed (assumed) NonPrecoded/ Beamformed (assumed) 8 8 8 8 (assumed) Since we planned this paper to be based on standards of 3GPP and 3GPP specification for release-15 5G massive MIMO is not yet ready, we have chosen release-14 efd-mimo configuration as a candidate for this paper. FD-MIMO/eFD-MIMO FD-MIMO introduced in 3GPP release-13 with upto 16 CSI-RS antenna ports supported. Later, in release-14 efd- MIMO introduced which is superset of FD-MIMO. efd-mimo supports upto 32 CSI-RS antenna ports. Increased number of CSI-RS antenna helps to cover vertical as well as horizontal propagation using 2D-AAA (Active antenna array) system applied with Kronecker Product (KP). 2D-AAA is nothing but configuration of number of antennas in horizontal (H) and vertical (V) dimension of the array making total antenna H x V. This results into better cell coverage and increased MIMO layers for more UEs because of better/optimal channel estimation by UE. On receiving the CSI-RS configuration from base station, UE performs channel estimation measurement on these resources and sends feedback to base station in
uplink. The better the reported rank, the more number of layers gets scheduled by base station to UE resulting into increased downlink throughput to UE. efd-mimo comes with two types of CSI-RS transmission type: 1) CLASS-A (Non-Precoded) 2) CLASS-B (Beamformed) Class-A or Class-B is configurable by base station using higher layer parameter CSI-RS- ConfigEMIMO as mentioned below: 3GPP 36.331 Section 6.3.2 CSI-RS-ConfigEMIMO information elements CSI-RS-ConfigEMIMO-v1430 ::= CHOICE { setup CHOICE { nonprecoded-v1430 CSI-RS-ConfigNonPrecoded-v1430, beamformed-v1430 CSI-RS-ConfigBeamformed-v1430 } } Test strategy for non-precoded CSI-RS transmission (Upto 32 antenna base station) Non-Precoded (Class-A) transmission: This is similar to the previous releases, where CSI-RS is cell specific parameters and is not applied with any precoding. Previously only 8 antenna ports (15-22) were supported with this. However, with efd-mimo there will be upto 32 antenna ports (15-46) supported and configurable by enodeb. Upto 32 CSI-RS ports configuration gets achieved by doing aggregation of multiple CSI-RS configuration, which is explained in table mentioned below:
From 3GPP 36.211 Table 6.10.5-1: Aggregation of CSI-RS configurations. Total number of antenna ports N CSI CSI res N ports Number of antenna ports per CSI-RS configuration CSI N ports Number of CSI-RS configurations CSI N res 12 4 3 16 8 2 20 4 5 24 8 3 28 4 7 32 8 4 Class-A supports only one CSI-RS resource configuration. Same resource configuration gets repeated in multiple numbers to achieve upto 32 ports as one CSI-RS resource configuration can have maximum 8 ports. Number of CSI-RS ports can be determined by higher layer parameters (codebookconfign1, codebookconfign2) configured by base station as below: 3GPP 36.331 Section 6.3.2 CSI-RS-ConfigEMIMO information elements CSI-RS-ConfigNonPrecoded-v1430::= SEQUENCE { codebookconfign1-v1430 ENUMERATED {n5, n6, n7, n10, n12, n14, n16}, codebookconfign2-r1430 ENUMERATED {n5, n6, n7 }, From 3GPP 36.213 Table 7.2.4-9: Supported configurations of ( O 1,O 2 ) and ( N ) 1, N 2 Number of CSI-RS antenna ports, P ( 1, N 2) ( ) 1,O 2 8 (2,2) (4,4),(8,8) 12 (2,3) (8,4),(8,8) (3,2) (8,4),(4,4) (2,4) (8,4),(8,8) 16 (4,2) (8,4),(4,4) (8,1) (4,-),(8,-) (2,5) (8,4) 20 (5,2) (4,4) (10,1) (4,-) (2,6) (8,4) (3,4) (8,4) 24 (4,3) (4,4) (6,2) (4,4) (12,1) (4,-) (2,7) (8,4) 28 (7,2) (4,4) (14,1) (4,-) (2,8) (8,4) 32 (4,4) (8,4) (8,2) (4,4) (16,1) (4,-) Above table describes all the possible combinations of (N1, N2) and (O1, O2) w.r.t different CSI-RS antenna ports {8, 12, 16, 20, 24, 28 and 32}. Where, (N1, N2) corresponds to number of antenna ports per polarization in dimension x and (O1, O2) corresponds to spatial oversampling rate in dimension x as used for transmission of CSI reference signals.
UE calculates total number of CSI-RS ports configured for measurement which is 2xN1xN2. Based on number of CSI-RS ports configured and UE capability w.r.t number of spatial layers, UE performs channel estimation and reports back optimal CSI and codebook to base station as defined in 3GPP 36.213 Table 7.2.4-10 to Table 7.2.4-17. Base station schedules data to UE based on the received CSI feedback from UE. Note: Release-8 and release-9 UEs will not benefit from 8x8 MIMO/FD-MIMO/eFD-MIMO and Massive MIMO as all these MIMOs are based on CSI-RS ports (release-10) channel estimation. Since the evolution path for advance MIMO is based on 8x8 rel-10 model, we have not added l feature interaction details. Implicitly, all the combinations available with 8x8 will be valid in FD-MIMO/Massive-MIMO as well. For release-10 and later UEs, maximum number of UE-specific DMRS ports are 8 (ports 7 to 14) and maximum supported layers by UE is still upto 8 even in release-14. So, UE can report only maximum 8 layers (Rank) to enodeb. This clearly indicates that advance MIMO principles are focused on utilising upto 8-layers MIMO and making it a common scenario for most of the UEs within cell. Test strategy for beamformed CSI-RS transmission (Upto 8 antenna base station) Beamformed (Class-B) transmission: This is new method used to transmit CSI- RS where CSI-RS will be precoded like DMRS and specific to UE. This configuration allows only upto 8 number of CSI-RS antenna ports (15-22) with one or more CSI-RS resource configuration. Since this is beamformed precoded signalling, it s not a common signalling but dedicated to selected UE. Based on the number of configured CSI-RS ports, base station selects codebook to precode CSI-RS as defined in 3GPP 36.213 Table 7.2.4-18 to 7.2.4-20. If UE is configured in this class then it measures the entire beam (CRI-upto 8 beam) configured and reports back optimal CRI (CSI-RS resource index also called beam index) and CSI to enodeb. Base station selects this beam to schedule data to UE. Total number of beam configured to UE is informed by base station higher layer parameter as mentioned below: 3GPP 36.331 Section 6.3.2 CSI-RS-ConfigEMIMO information elements CSI-RS-ConfigBeamformed-v1430::= SEQUENCE { csi-rs-confignzp-aplist-r14 SEQUENCE (SIZE(1..8)) OF CSI-RS-ConfigNZP-r11
The mapping of CSI-RS resource element (k,l) is given in 36.211 section 6.10.5.2 Scenario of Class-A and Class-B transmission Class-A This scenario is applicable when base station does not have current awareness of UE channel condition (during scheduling request for DL traffic). Base station may configure maximum number of upto 32 CSI-RS ports to facilitate UE report highest possible rank. This is applicable for cell-center UEs. Class-B This scenario is applicable when base station does have current awareness of UE channel condition (during on-going traffic with moving users) or UE reporting bad CSI in class-a because of cell edge, user movement, interference etc. Base station may configure different beamformed CSI-RS ports upto 8 layers to facilitate UE report best beam. Based on the feedback, base station schedules data on this beam. This also increases cell coverage area. Conclusion: efd-mimo and future massive MIMO is solid step towards IMT-2020 requirement which demands increase in peak data rate, average data rate, spectrum efficiency, network energy efficiency, area traffic capacity and connection density. FD-MIMO simulation results explained in 3GPP TSG RAN WG1 Meeting #78bis R1-143730 have shown good improvements over legacy MIMO as summarised below: a) Performance gain with beamforming in elevation dimension (20% gain in cell edge and 5% in cell average) b) Performance gain of load balance with EBF/FD-MIMO in Het-Net (27% average gain and 39% cell edge gain) c) Performance gain of MU-MIMO with EBF/FD-MIMO - 3D-UMa : 98.2% and 91.2% cell average and cell edge performance gain - 3D-UMi : 117.3% cell average and about 116% cell edge performance gain With the introduction of Massive MIMO we will be able to see bigger improvements in all the above areas and possibly meet IMT-2020 requirement. Abbreviations: FD-MIMO : Full Dimensional Multiple Input Multiple Output efd-mimo : Enhanced FD MIMO MU-MIMO : Multi user-mimo CRI : Channel State Information Resource Index CRS : Cell Reference Signal CSI-RS : Channel State Information Reference Signal DMRS : Demodulation Reference Signal DL : Downlink EBF : Elevated Beamforming IMT : International Mobile Telecommunications LTE : Long Term Evolution
3D 3GPP UE UMa UMi : 3 Dimensional : 3rd Generation Partnership Project : User Equipment : Urban Macro : Urban Micro References: 1. 3GPP TS 36.331, Radio Resource Control (RRC); Protocol specification (Release 14) 2. 3GPP TS 36.306, User Equipment (UE) radio access capabilities (Release 14) 3. 3GPP TS 36.211, Physical channels and modulation (Release 14) 4. 3GPP TS 36.212, Multiplexing and channel coding (Release 14) 5. 3GPP TS 36.213, Physical layer procedures (Release 14) 6. 3GPP TR 36.897, Study on elevation beamforming / Full-Dimension (FD) Multiple Input Multiple Output (MIMO) for LTE (Release 13) 7. 3GPP TSG RAN WG1 Meeting #78bis R1-143730 8. 3GPP TSG RAN WG1 Meeting #82bis R1-156217 9. 3GPP TSG RAN WG1 #78bis R1-143883 10. Recommendation ITU-R M.2083-0 (IMT-2020) Authors Biography: Shatrughan Singh Working with Aricent for past 5+ years and total Telecom domain experience of 10+ years. Currently located in UK, working on LTE advanced releases of protocol and feature testing needs of client Subramaniam H Working with Aricent for past 8+ years and total Telecom domain experience of 12+ years. Currently located in UK, working on LTE protocol and feature testing needs of client Jaison John Puliyathu Mathew Working with Aricent for past 7+ years and has total Telecom domain experience of 14+ years. Currently located in UK and managing onsite testing team of engineers THANK YOU!