Introduction of LTE Rel8 and Further Evolution--Huawei（下） (2).ppt

HUAWEI CONFIDENTIAL,Page 103,内部资料 注意保密,MIMO,HUAWEI CONFIDENTIAL,Page 104,内部资料 注意保密,Outline for MIMOIntroductionTechnical considerations for MIMOUL MIMOBFUL TX diversityDL MU-MIMODL SU-MIMO,HUAWEI CONFIDENTIAL,Page 105,内部资料 注意保密,What to do in MIMO to meet those requirements?,HUAWEI CONFIDENTIAL,Page 106,内部资料 注意保密,UL MIMO(1)Research focus:UL MIMO MA scheme:SC-FDMA vs OFDMA SC-FDMA decided as uplink MU scheme also for MIMO at RAN1#55bisLow CM UL MIMO transmission scheme in SC-FDMA Transmit diversity scheme with low CM SU-MIMO precoding scheme with low CM,HUAWEI CONFIDENTIAL,Page 107,内部资料 注意保密,UL MIMO(2)UL MA decison related to MIMOIf OFDMA is used,then advanced receivesr such as QRM-MLD could be used to improve ULMIMO performance.However,in SC-FDMA,Turbo-SIC receiver(see structure below)is thedesired receiver to maximize performance.When comparing both performance and complexitythe SC-FDMA was selected as the MA scheme in favour of OFDMA.,HUAWEI CONFIDENTIAL,Page 108,内部资料 注意保密,UL MIMO(3)Low CM TxD in SC-FDMAEspecially for cell edge users,it is important to maintain the low CM for maximal PA utilization.A transmit diversity scheme denoted the modified SFBC scheme for four tranmit antennas wasfirst proposed in R1-083708,and it has the following desirable properties:,Keep the single carrier propertyHas better performance than both CDD and FSTDAvoids the pairing problem of STBC based schemesHas similar performance to STBC in most channel scenarios,and equal performance asSTBC at cell edge(single RB scheduled),Block of M,modulation,symbols,DFT,Sub-carriermapping,Sub-carriermapping,IFFT,Cyclic prefixinsertion,IFFT,Cyclic prefixinsertion,Zero insertion,ReversingCyclic shiftingConjugating,Adding minus sign,Modified-SFBC transmitter blockM DFT samples,HUAWEI CONFIDENTIAL,Page 109,内部资料 注意保密,UL MIMO(4)Defining Uplink antenna ports for MIMO Antenna mapping/precoding is pre-defined,implementation issue or controlledby eNB(in case of CL MIMO)DRS antenna ports The number of DRS antenna ports is equal to the transmission rank SRS antenna ports Should be configurable to support backward compatability operation of aRel.10 UE with a Rel.8 eNB.Configuration to a smaller number of SRS ports reduce the SRS overheadand is useful for uplink MU-MIMO.,HUAWEI CONFIDENTIAL,Page 110,内部资料 注意保密,UL MIMO(5)Layer shifting and bundlingEvaluate performance gain of Layer-shifting combined with HARQ bundling for MMSE andMMSE-SIC receiverIs there any evaluation on UL layer shifting with Turbo-SIC receiver?For UL-SM we need to clarify the following 3 points regarding link adaptation in thesimulations:1.effective SINR per subcarrier based on the type of receiver-MMSE for first stream andMRC for second2.method of mapping from SINR per sub-carrier to effective SINR EESM/3.Target PER per stream for each stream(aggressiveness on each stream)Evaluate different link adaptation schemes for Turbo-SIC,HUAWEI CONFIDENTIAL,Page 111,内部资料 注意保密,Uplink MIMO(6)Conclusion from R1#56 Same layer mapping as downlink LTE Rel-8 Maximum of 2 codewords(TBs)Spatial bundling of HARQ parameters desirable:Single shared downlink ACK/NAK(PHICH);single shared NDI,RV Impact on performance(including overhead)to be verified Final decision in the next meeting Number of MCS fields:FFS:one or two Layer shifting in time domain FFS:exact shifting pattern Possibility to configure with or without layer shiftingFurther refinement in#56bis Number of MCS fields:two Two modes of operation for further study No HARQ-ACK Spatial Bundling and no Layer Shifting HARQ-ACK Spatial Bundling with Layer Shifting,HUAWEI CONFIDENTIAL,Page 112,内部资料 注意保密,Uplink MIMO(7)Conclusion from 56bis:Codebook-based precoding is supported for FDD Single TPMI per UL component carrier(frequency non-selective in a componentcarrier)is supported Frequency-selective precoding(multiple TPMIs)in a component carrier FFS Size-1 codebook with identity precoding for full-rank transmission Dynamic rank(TRI)adaptation 2Tx codebook:Total size(1-layer+2-layer)8(3-bit codebook)FFS if(1/sqrt(2)is necessary for codebook index 4 and 5 of rank-1 case.4Tx codebook:Total size(1-layer+2-layer+3-layer+4-layer)64(6-bit codebook)At most 8PSK alphabet exp(j*p*n/4),n=0,1,7 for non-zero elements Non-8PSK but constant modulus alphabet FFS Precoding for TDD,continue discussion about codebook-based vs non-codebook based approaches.,HUAWEI CONFIDENTIAL,Page 113,内部资料 注意保密,Uplink MIMO(8)2Tx Codebook,HUAWEI CONFIDENTIAL,Page 114,内部资料 注意保密,Uplink codebook for 4Tx transmissionDefine uplink codebook for 4Tx at UE side Rank 1 codebook Rank 2 codebook Rank 3 codebook Rank 4 codebook,HUAWEI CONFIDENTIAL,Page 115,内部资料 注意保密,Rank-1 CodebookSize-24:16 constant modulus+8 antenna-pair turn-off vectors,HUAWEI CONFIDENTIAL,Page 116,内部资料 注意保密,Rank-2 CodebookSize-16:CM-preserving matrices,HUAWEI CONFIDENTIAL,Page 117,内部资料 注意保密,Rank-3 Codebook,A v erage T hroughput(k bps),HUAWEI CONFIDENTIAL,Page 118,内部资料 注意保密,2,4,6,8,10,12,14,16,18,20,22,1000,SNR(dB),Performance EvaluationRank=3,SCM channel,BS element distance:4,MS element distance:0.590008000700060005000400030002000,HUAWEI CONFIDENTIAL,Page 119,内部资料 注意保密,Rank-4 CodebookIdentity matrix,HUAWEI CONFIDENTIAL,Page 120,内部资料 注意保密,Multilayer BFOnly single layer BF is supported in LTE,but the throughput should be furtherimproved in LTE+.Nearly 20%improvement is obtained if 2 layers transmission is used.If multilayer BF is proposed,hardware calibration at UE side should done.MU-MIMO operation shall be studied.,F(x),HUAWEI CONFIDENTIAL,Page 121,内部资料 注意保密,Multilayer BF(2),0,0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,0,10.90.80.70.60.50.40.30.20.1,Empirical CDF,xCDF of user throughput(bps/Hz),HUAWEI CONFIDENTIAL,Page 122,内部资料 注意保密,DM RS DesignOverhead&CompatibilityMultiplexing methodsSpecified Pattern,HUAWEI CONFIDENTIAL,Page 123,内部资料 注意保密,Overhead&Compatibility12 RE is enough for dual layer beamforming(56b:R1-091259)Rank 1 and Rank 2 should use the same pattern(56b:Email Discussion)Conclusion(RAN1 57)Agree on the DMRS overhead as 12 REs Same set of RE used for Rank 1 and Rank 2 FFS whether or not Rel9 Rank1 pattern is different from the Rel8 Rank1 pattern Consider the forward compatible design that makes Rel-9 patterns a subset of Rel-10patterns,on the condition that the new pattern presents better performance than or atleast equivalent performance to the existing Rel-8 pattern,HUAWEI CONFIDENTIAL,Page 124,内部资料 注意保密,Multiplexing MethodThe reason to support CDMed DMRS for rank1 and rank 2 According to the conclusion in RAN1 57,12RE overhead is assumed forboth rank1 and rank 2 If the FDM is used,considering the fast rank adaptation in SU-MIMO if 6RE DMRS is used when rank degraded to 1,which will lead toworse performance than 12RE DMRS If 12 RE is used when rank degrade to 1,the channel estimationmodule couldnt be reused for both rank1 and rank 2 If the FDM is used,considering the MU-MIMO,and if user-paring lose if 6RE could be used for the served user,which will have bad impacton the performance If 12RE could be used for the served user,the channel estimationmodule couldnt be reused for both multi-user and signel usertransmission The only argument from the FDM is the 1dB performance gain overCDM when the velocity is high like(30km/h-120km/h),HUAWEI CONFIDENTIAL,Page 125,内部资料 注意保密,Specified PatternCDM 0 or CDM 1 is preferred for Rank1 and Rank2(57:R1-091792,R1-091796),HUAWEI CONFIDENTIAL,Page 126,内部资料 注意保密,Specified PatternDMRS patterns(Normal CP,Normal Subframe)for Rel-9 dual streambeamforming(RAN1 58)CDM multiplexing of two layers as shown is baseline same DM-RS structure(including REs,spreading and scrambling)will be used forrank 1-2 in Rel-10,HUAWEI CONFIDENTIAL,Page 127,内部资料 注意保密,Specified PatternDM-RS structure for DwPTS and/or extended CP may be different from regular sub-frames w/normal CP(RAN1 58)based on performance considerations,HUAWEI CONFIDENTIAL,Page 128,内部资料 注意保密,Rank AdaptationFast rank adaptation is physical layer control signaling based Already have the rank 1 transmission,why should we need fast adaptation?Rank 1 transmission mode is designed for some specific antennaconfiguration with high probability of supporting rank 1 transmission inorder to save the control signaling.Fast rank adaptation is also needed to support some scenario with rank 1and rank 2 adaptation as tm4 and tm6Slow rank adaptation is supported by the mode switching through the higherlayer control signalingThe back off mode is transmit diversityConclusion Fast rank adaptation is agreed in RAN1 57,HUAWEI CONFIDENTIAL,Page 129,内部资料 注意保密,PMI/CQI in Rel-9Considering the channel reciprocity and limited codebook size,PMI is not needed atall for TDD At most 2 CRS could be used to get the PMI The correlation between the cross-polarized antenna group is relatively low,so thegain from close loop mimo is very small compared with open loop mimo Only 2 antenna codebook could be used,in which the number of the precoding matrixis very limited and will further narrow the gain it could achieve The AS of SRS is optional,which means some UE will not support AS,how to supportchannel reciprocity in this scenario?（Qualcomm）CQI of Rel-8 is enough for Rel-9 CQI for TxD could be reused for SU-BF CQI for TXD could be resued for MU-BF?Conclusion(RAN1 57):PMI feedback is supported in FDD FFS whether or not to have PMI feedback for TDDConclusion(RAN1 58):TxD based and also PMI based feedback modes supported for both TDD and FDD Baseline:No PMI feedback for TDD and FDD PMI based feedback is also supported for TDD and FDD Re-use Rel-8 feedback modes as much as possible,HUAWEI CONFIDENTIAL,Page 130,内部资料 注意保密,CalibrationPhase Calibration If the max-SNR criterion is used,no need to do phase calibration If the max-SINR criterion is used,the phase calibration is also needed The interference matrix used in SINR criterion is for reducing the interference to the other UEsAmplitude Calibration Have large impact on the performance,because its negative effect on the CQI and RI prediction Considering the imperfect power control,the amplitude calibration have very minor impact on theperformance（Ericsson）Conclusion(RAN1 57)UE antenna calibration Need for UE antenna calibration FFS UE antenna calibration would not be specified in RAN1,it would be an implementationissue Evaluate the performance(loss)due to the UE TX-and RX-mismatch Carry out more discussion on the cost and method of different UE calibration method,HUAWEI CONFIDENTIAL,Page 131,内部资料 注意保密,Layer shiftingShould we introduce the layer shifting into the R9?Layer shifting in downlink will lead to performance loss In R8 layer shifting is for open loop MIMO,dual-layer beamforming is a kind of closeloop MIMO,so layer shifting is not needed for dual-layer beamforming The control signaling overhead is not a problem for dual-layer beamforming becausethe mode 4 already feedback 2 A/NConclusion(RAN1 57)Use Rel8 layer mapping as the baseline in this WI Further evaluate the gain of layer shift especially when spatial bundling is applied,HUAWEI CONFIDENTIAL,Page 132,内部资料 注意保密,Control SignalingNew transmission mode should be defined for single user dual layerbeamformingSingle-user beamforming has separate control signaling from that ofMulti-user beamforming for saving the control signaling overheadReuse format 2A to support the single-user beamformingConclusion(RAN1 58)Single transmission mode for Rel-9 Dual-layer beamforming could be based on DCI format 2A Dynamic indication of DM-RS port is supported in case of rank 1transmission To enable scheduling of two UEs with rank-1 transmission using differentorthogonal DMRS ports on the same PDSCH resources,HUAWEI CONFIDENTIAL,Page 133,内部资料 注意保密,R9 Multi User BeamformingWIDTransmission Scheme Transparent MU-MIMO Non-Transparent MU-MIMOControl Signaling,HUAWEI CONFIDENTIAL,Page 134,内部资料 注意保密,Transparent MU-MIMOvs.Non-Transparent MU-MIMOR1-092247(CASE 1),HUAWEI CONFIDENTIAL,Page 135,内部资料 注意保密,Control SignalingObtain the inter-user interference The DMRS of AP5 is specified by UE related n_RNTI,which means its hard to getthe inter-user interference without knowing the paired UE n_RNTISU/MU assumption(RAN1 58)no explicit signaling of the presence of co-scheduled UE in case of rank 1transmissions in case of rank-1 transmission,the UE cannot assume that the other DM RS antennaport is not associated with PDSCH assigned to another UE LS to RAN4 with action to consider defining test for such use case,HUAWEI CONFIDENTIAL,Page 136,内部资料 注意保密,UL TX diversityDesign considerations for LTE-A UL TX diversity(Huaweis views)Preserve the single carrier property The same requirement as LTE Extend cell coverage and improve cell edge spectrum efficiency.Avoid the paring problem The number of symbols used for data in uplink is not a constantbecause of the existence of LTE sounding reference signal.The UL TX diversity scheme should be independent of the number ofsymbols for data,otherwise two different TX schemes should be used inone sub-frame,which cause the increase of UE complexity.Scalable reuse of the receiver In uplink,Node B serves for multiple UEs with different configurations.The reuse of receiver structures between 2 and 4 TX antenna diversityshould be considered to reduce the Node B complexity.,HUAWEI CONFIDENTIAL,Page 137,内部资料 注意保密,Performance of the M-SFBC based scheme(1/3)For the candidates for UL TX diversity,M-SFBC based scheme is preferred for 2 TXantennas at UE due to the robust performance and low complexityThe required SNR dB to reach BLER=0.1.The lowest SNR is marked with green.,HUAWEI CONFIDENTIAL,Page 138,内部资料 注意保密,Performance of the M-SFBC based scheme(2/3)For 4 TX antennas ath the UE,the gain of 4 TX Modified SFBC+FSTDover other 4 TX schemes at BLER=0.01 is between 0.3-1.6 dB(see table),HUAWEI CONFIDENTIAL,Page 139,内部资料 注意保密,Performance of the M-SFBC based scheme(3/3)Performance with different number of DRS antenna ports In TU channel,it is better to operate with 2 DRS antenna portsThe required SNR in dB at BLER=0.01 for code rate 1/2 and TU channel,6 RBThe required SNR in dB at BLER=0.01 for code rate 1/2 and flat channel,6 RB,HUAWEI CONFIDENTIAL,Page 140,内部资料 注意保密,DL MU-MIMOCurrent LTE.R8 MU-MIMO suffers from Inaccurate CQI calculation Coarse codebook designs leading to high inter stream interference Not knowing whether to do interference cancellation or not Interference between co-scheduled UEs Limited throughput due to single stream per UETwo options for inter layer interference suppression UE suppression(preferable)Robust in uncorrelated,frequency selective(bad)channels Good if UE has multiple antennas eNB suppression(unfortunately necessary)Only good for highly correlated,narrow-band channels 8deg spread instead of 16deg High Error floor due to coarse CB and frequency fluctuations Good if UE has 1 Ant.CQI Estimation problem,sum rate(bits/sec/Hz),HUAWEI CONFIDENTIAL,Page 141,内部资料 注意保密,0,2,4,6,8,10,12,14,16,18,20,0,We can do better than LTE MU-MIMOA large performance gap exists between LTE MU-MIMO and the other