MU-MIMO in LTE/LTE-A Performance Analysis. Rizwan GHAFFAR, Biljana BADIC

Transcription

1 MU-MIMO in LTE/LTE-A Performance Analysis Rizwan GHAFFAR, Biljana BADIC

2 Outline 1 Introduction to Multi-user MIMO Multi-user MIMO in LTE and LTE-A 3 Transceiver Structures for Multi-user MIMO Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 / 40

3 Outline Introduction to Multi-user MIMO 1 Introduction to Multi-user MIMO Multi-user MIMO in LTE and LTE-A 3 Transceiver Structures for Multi-user MIMO Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 3 / 40

4 Introduction to Multi-user MIMO Multi-user vs Single-user MIMO Advantages of Multi-user MIMO Low rank channels no longer a problem but an advantage Provides decorrelation of spatial signatures Provides multi-user diversity (less reliance on antenna diversity)[knopp. 1995] Mitigates the need for multiple antennas at mobile Allows for user- (in addition to stream-) multiplexing The Price. Channel State Information at the Transmitter (CSIT) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 4 / 40

5 Introduction to Multi-user MIMO Multi-user vs Single-user MIMO Advantages of Multi-user MIMO Low rank channels no longer a problem but an advantage Provides decorrelation of spatial signatures Provides multi-user diversity (less reliance on antenna diversity)[knopp. 1995] Mitigates the need for multiple antennas at mobile Allows for user- (in addition to stream-) multiplexing The Price. Channel State Information at the Transmitter (CSIT) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 4 / 40

6 Introduction to Multi-user MIMO CSIT? What if No CSIT enodeb cannot form spatial beams. Multiplexing Gain 1 Exception for terminals having enough antennas to remove interference conditioned on full rank individual user channels If CSIT available Linear Precoding Strategies MMSE Precoding [Hochwald. 005] Zero Forcing Precoding [Haardt 004] Non-Linear Precoding Strategies Dirty Paper Coding [M. Costa 1983] [Caire and Shamai. 003] Vector Perturbation [Hochwald. 005] Tomlinson-Harashima Precoding [Shamai. 00] How to get CSIT Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

7 Introduction to Multi-user MIMO CSIT? What if No CSIT enodeb cannot form spatial beams. Multiplexing Gain 1 Exception for terminals having enough antennas to remove interference conditioned on full rank individual user channels If CSIT available Linear Precoding Strategies MMSE Precoding [Hochwald. 005] Zero Forcing Precoding [Haardt 004] Non-Linear Precoding Strategies Dirty Paper Coding [M. Costa 1983] [Caire and Shamai. 003] Vector Perturbation [Hochwald. 005] Tomlinson-Harashima Precoding [Shamai. 00] How to get CSIT Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

8 Introduction to Multi-user MIMO CSIT? What if No CSIT enodeb cannot form spatial beams. Multiplexing Gain 1 Exception for terminals having enough antennas to remove interference conditioned on full rank individual user channels If CSIT available Linear Precoding Strategies MMSE Precoding [Hochwald. 005] Zero Forcing Precoding [Haardt 004] Non-Linear Precoding Strategies Dirty Paper Coding [M. Costa 1983] [Caire and Shamai. 003] Vector Perturbation [Hochwald. 005] Tomlinson-Harashima Precoding [Shamai. 00] How to get CSIT Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

9 Introduction to Multi-user MIMO How to acquire CSIT and how much? TDD system. Reciprocity is key to CSIT acquisition [Viswanath and Tse 003]. A feedback control channel improves CSIT. FDD systems. Dedicated feedback channel [Marzetta and Hochwald 006] Limited feedback channel Quantization partial CSIT [Sharif and Hassibi. 003] Vector quantization Dimension reduction Adaptive feedback Statistical feedback Opportunistic spatial division multiple access (SDMA) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 6 / 40

10 Introduction to Multi-user MIMO How to acquire CSIT and how much? TDD system. Reciprocity is key to CSIT acquisition [Viswanath and Tse 003]. A feedback control channel improves CSIT. FDD systems. Dedicated feedback channel [Marzetta and Hochwald 006] Limited feedback channel Quantization partial CSIT [Sharif and Hassibi. 003] Vector quantization Dimension reduction Adaptive feedback Statistical feedback Opportunistic spatial division multiple access (SDMA) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 6 / 40

11 Outline Multi-user MIMO in LTE and LTE-A 1 Introduction to Multi-user MIMO Multi-user MIMO in LTE and LTE-A 3 Transceiver Structures for Multi-user MIMO Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 7 / 40

12 Multi-user MIMO in LTE and LTE-A From Rel-8 to Rel-10 From Rel-8 to Rel-10: An Overview LTE Release CRS DM-RS CSI-RS Rel-8 Rel-9 Rel-10 Cell specific reference signals (CRS) Targeting channel estimation and demodulation Present only in PDSCH resource blocks and layers scheduled by enodeb User-specific DM-RS (Demodulation RS) Targeting PDSCH demodulation Present only in PDSCH resource blocks and layers scheduled by enodeb Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 8 / 40

13 Multi-user MIMO in LTE and LTE-A From Rel-8 to Rel-10 From Rel-8 to Rel-10: An Overview LTE Release CRS DM-RS CSI-RS Rel-8 Rel-9 Rel-10 Cell-specific CSI RS (Channel State Information RS). Designed for CoMP (Coordinated Multi-Point Transmission) Defined for f = 15kHz only Low overhead, transmitted sparse in time and frequency No mixed use of Rel-8 CRS and Rel-10 CRS for Rel-10 UE Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 9 / 40

14 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 8 Downlink Transmission Modes in LTE Rel 8 Possible Configurations SISO. 1 1 SIMO. 1, 1 4 MISO. 1, 4 1 MIMO., 4, 4 4 Downlink Transmission Modes in Release 8 UE DL transmission mode Transmission scheme of PDSCH CQI mode DCI format Mode 1 Single-antenna port CQI Only DCI format 1 Mode Transmit diversity CQI Only DCI format 1 Mode 3 Open-loop spatial multiplexing CQI Only DCI format A Mode 4 Closed-loop spatial multiplexing CQI, RI, PMI DCI format Mode 5 Multi-user MIMO CQI, PMI DCI format 1D Mode 6 Closed-loop Rank=1 precoding CQI, PMI DCI format 1B Mode 7 Beamforming Single-antenna port; port 5 CQI only DCI format 1 Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

15 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 8 Two approaches to multi-antenna transmission CQI PMI RI MIMO 4 4 MU-MIMO CQI Single layer beamforming MU-MIMO DRS CRS CRS MCS PMI RI SRS MCS PDSCH Channel estimation based common reference signals (CRS) Closed loop, codebook precoding PDSCH Channel estimation based dedicated reference signals (DRS) Open loop, non-codebook precoding Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

16 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 8 Multi-user MIMO transmission CQI PMI RI CQI PMI RI Multi-user MIMO users - Single layer CRS PDSCH Channel estimation based common reference signals (CRS) Closed loop, codebook precoding Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 1 / 40

17 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 8 LTE Codebook 3GPP LTE. Codebook based approach LTE Codebook. Low resolution and equal gain transmission (EGT) Transmit Antennas p {0, 1} Single layer transmission µ = 1 1 [ 1 1 ], 1 [ 1 1 ], 1 [ 1 j ], 1 [ 1 j ] Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

18 LTE Codebook Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 8 4 Transmit Antennas p {0, 1,, 3} Index u n µ = 1 µ = µ = 3 µ = 4 0 u 0 = [1, 1, 1, 1] T W {1} 0 W {14} 0 / W {14} 0 / 3 W {134} 0 / 3 u 3 = [1, j, 1, j] T W {1} 3 W {1} 3 / W {13} 3 / 3 W {314} 3 / [ 4 u 4 = 1, ( 1 j)/, j, (1 j)/ ] T W {1} 4 W {14} 4 / W {14} 4 / 3 W {134} 4 / [ 7 u 7 = 1, ( 1+j)/, j, (1+j)/ ] T W {1} 7 W {13} 7 / W {134} 7 / 3 W {134} 7 / 11 u 11 = [1, j, 1, j] T W {1} 11 W {13} 11 / W {134} 11 / 3 W {134} 11 / 15 u 15 = [1, 1, 1, 1] T W {1} 15 W {1} 15 / W {13} 15 / 3 W {134} 15 / µ denotes the number of layers. The precoding matrix W s n denotes the matrix defined by the columns given by the set {s} from the expression W n = I u nu H n /u H n u n where I is the 4 4 identity matrix. Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

19 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 9 Downlink Transmission Modes in Release 9 Possible Configurations SISO. 1 1 SIMO. 1, 1 4 MISO. 1, 4 1 MIMO., 4, 4 4 Addition of Transmission Mode 8 UE DL transmission mode Transmission scheme of PDSCH CQI mode DCI format Mode 1 Single-antenna port CQI Only DCI format 1 Mode Transmit diversity CQI Only DCI format 1 Mode 3 Open-loop spatial multiplexing CQI Only DCI format A Mode 4 Closed-loop spatial multiplexing CQI, RI, PMI DCI format Mode 5 Multi-user MIMO CQI, PMI DCI format 1D Mode 6 Closed-loop Rank=1 precoding CQI, PMI DCI format 1B Mode 7 Beamforming Single-antenna port; port 5 CQI only DCI format 1 Mode 8 Dual layer beamforming CQI, RI, PMI DCI format B Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

20 Transmission Mode 8 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 9 Spatial Multiplexing supported Upto layers per user (single-user MIMO) CQI Dual layer beamforming MU-MIMO Two orthogonal streams of UE-specific RS are supported for multi-user MIMO transmission Two users in multi-user MIMO. Upto 4 layers (Quasi-orthogonal multi-user MIMO with the aid of scrambling IDs) DM-RS CRS SRS MCS RI Allows for ZF-MU precoding PDSCH Channel estimation based demodulation reference signals (DM-RS) Open loop, non-codebook precoding Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

21 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 10 Downlink Transmission Modes in Release 10 Possible Configurations SISO. 1 1 SIMO. 1, 1 4, 1 8 MISO. 1, MIMO., 4, 4, 4 4, 8, 4 8, 8 8, 8, 8 4, 8 8 Addition of Transmission Mode 9 UE DL transmission mode Transmission scheme of PDSCH CQI mode DCI format Mode 1 Single-antenna port CQI Only DCI format 1 Mode Transmit diversity CQI Only DCI format 1 Mode 3 Open-loop spatial multiplexing CQI Only DCI format A Mode 4 Closed-loop spatial multiplexing CQI, RI, PMI DCI format Mode 5 Multi-user MIMO CQI, PMI DCI format 1D Mode 6 Closed-loop Rank=1 precoding CQI, PMI DCI format 1B Mode 7 Beamforming Single-antenna port; port 5 CQI only DCI format 1 Mode 8 Dual layer beamforming CQI, RI, PMI DCI format B Mode 9 Seamless switching between CQI and RI DCI format C SU and MU-MIMO upto rank 8 PMI not necessary Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

22 Transmission Mode 9 Multi-user MIMO in LTE and LTE-A 3GPP LTE Rel 10 Seamless switching between single-user and multi-user MIMO up to rank 8 Upto 8 layer transmission in single-user and upto 4 layer transmission for multi-user MIMO Max of codewords per UE. Max of 4 codewords at enodeb Frequency selective PMI. Precoding per subband basis Layered precoding for the case of 8 transmit antennas W 1 targeting wideband/long-term channel properties. For the case of and 4 antennas, it is identity W targeting frequency-selective/short-term channel properties. For the case of and 4 antennas, it is same as LTE Rel 8 precoders CQI CSI RI DM-RS CRS CSI-RS MCS RI Rank 8 Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

23 Outline Transceiver Structures for Multi-user MIMO 1 Introduction to Multi-user MIMO Multi-user MIMO in LTE and LTE-A 3 Transceiver Structures for Multi-user MIMO Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

24 Motivation Fundamentals of Multi-user MIMO Pre-subtracting, cancelling or mitigating Gaussian interference. Is Gaussian assumption valid? Gaussian alphabets are entropy maximizers but not Gaussian interference? Result is simplified receiver structures but void of exploiting the interference structure. Inputs are from discrete constellations which may significantly deviate from Gaussian idealization Finite alphabet interferers have structures that can be exploited in the detection process Precoders can be designed to manage the interference in a way that this interference can be exploited in the detection process at the receivers Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 0 / 40

25 Motivation Fundamentals of Multi-user MIMO Pre-subtracting, cancelling or mitigating Gaussian interference. Is Gaussian assumption valid? Gaussian alphabets are entropy maximizers but not Gaussian interference? Result is simplified receiver structures but void of exploiting the interference structure. Inputs are from discrete constellations which may significantly deviate from Gaussian idealization Finite alphabet interferers have structures that can be exploited in the detection process Precoders can be designed to manage the interference in a way that this interference can be exploited in the detection process at the receivers Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 0 / 40

26 Motivation Fundamentals of Multi-user MIMO Pre-subtracting, cancelling or mitigating Gaussian interference. Is Gaussian assumption valid? Gaussian alphabets are entropy maximizers but not Gaussian interference? Result is simplified receiver structures but void of exploiting the interference structure. Inputs are from discrete constellations which may significantly deviate from Gaussian idealization Finite alphabet interferers have structures that can be exploited in the detection process Precoders can be designed to manage the interference in a way that this interference can be exploited in the detection process at the receivers Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 0 / 40

27 Motivation Key Idea Key Idea Interference comes from discrete constellations and possesses a structure that can be exploited in the detection process [Ghaffar and Knopp 011] Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 1 / 40

28 System Model LTE System Model - Multi-user MIMO Mode enodeb UE-1 UE- y 1 = h 1 p 1 x 1 + h 1 p x + z 1 x 1 χ 1, x χ of powers σ 1 and σ respectively. h k C 1. h k CN (0, I), y 1 C and z 1 CN (0, N 0 ) LTE precoder codebook p = 1 [ 1 1 ], Low resolution precoders 1 [ 1 1 ], Based on equal gain transmission (EGT) 1 [ 1 j ], 1 [ 1 j ] Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 / 40

29 System Model LTE System Model - Multi-user MIMO Mode enodeb UE-1 UE- y 1 = h 1 p 1 x 1 + h 1 p x + z 1 x 1 χ 1, x χ of powers σ 1 and σ respectively. h k C 1. h k CN (0, I), y 1 C and z 1 CN (0, N 0 ) LTE precoder codebook p = 1 [ 1 1 ], Low resolution precoders 1 [ 1 1 ], Based on equal gain transmission (EGT) 1 [ 1 j ], 1 [ 1 j ] Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 / 40

30 Geometric Scheduling Strategy Scheduling based on Geometrical Alignment Channel of UE-1, h 1 = [h 11 h1 [ ] ] Precoder of UE-1, p 1 = 1 1 j UE- is scheduled with p = 1 [ 1 j h 1 p 1 = 1 (h 11 jh 1 ) ] h 1 p = 1 (h 11 + jh 1 ) h 11 h 1 UE-1 enodeb h 1 h 11 Channel of UE-1 Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 3 / 40

31 Geometric Scheduling Strategy Scheduling based on Geometrical Alignment Channel of UE-1, h 1 = [h 11 h1 [ ] ] Precoder of UE-1, p 1 = 1 1 j UE- is scheduled with p = 1 [ 1 j h 1 p 1 = 1 (h 11 jh 1 ) ] h 1 p = 1 (h 11 + jh 1 ) h 11 h 1 UE-1 enodeb h 1 jh 1 h 11 h 11 Channel of UE-1 Channel of desired signal of UE-1 Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 3 / 40

32 Geometric Scheduling Strategy Scheduling based on Geometrical Alignment Channel of UE-1, h 1 = [h 11 h1 [ ] ] Precoder of UE-1, p 1 = 1 1 j UE- is scheduled with p = 1 [ 1 j h 1 p 1 = 1 (h 11 jh 1 ) ] h 1 p = 1 (h 11 + jh 1 ) h 11 h 1 UE-1 enodeb h 1 h 11 jh 1 h 11 jh 1 h 11 Channel of UE-1 Channel of desired signal of UE-1 Channel of interference of UE-1 Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 3 / 40

33 Receivers for Multi-user MIMO Low Complexity Interference Aware Receiver Optimal bit metric Λ i y1 1(y 1, c) = min h x 1 χ i 1,c,x 1 p 1 x 1 h 1 p x χ Interference aware detector (Equal energy alphabets) Λ i { 1(y 1, c) = min ψa x,r (y x 1 χ i 1 x 1) R ψ B x,i } 1,c where( y j = h 1 j) p y1 is the MF output ( ) p 1 = h 1 p 1 h is the cross-correlation between two effective channels 1 p ψ A = p 1,R x 1,R + p 1,I x 1,I y,r ψ B = p 1,R x 1,I p 1,I x 1,R y,i Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 4 / 40

34 Receivers for Multi-user MIMO Low Complexity Interference Aware Receiver Optimal bit metric Λ i y1 1(y 1, c) = min h x 1 χ i 1,c,x 1 p 1 x 1 h 1 p x χ Interference aware detector (Non equal energy alphabets) { h Λ i 1(y 1, c) = min x 1 χ i 1 p 1 x 1 } + h 1 p x ψ A x,r (y 1 x 1) R ψ B x,i 1,c where x,r ψ A h 1 p x,i ψ B h 1 p Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 4 / 40

35 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE - No scheduling - SU Rx 1 QAM64 LTE - Geometric scheduling - SU Rx 10 LTE - Geometric scheduling - IA Rx I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

36 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE - No scheduling - SU Rx 1 QAM64 LTE - Geometric scheduling - SU Rx 10 LTE - Geometric scheduling - IA Rx I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

37 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE - No scheduling - SU Rx 1 QAM64 LTE - Geometric scheduling - SU Rx 10 LTE - Geometric scheduling - IA Rx I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

38 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE - No scheduling - SU Rx 1 QAM64 LTE - Geometric scheduling - SU Rx 10 LTE - Geometric scheduling - IA Rx I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

39 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE - No scheduling - SU Rx 1 QAM64 LTE - Geometric scheduling - SU Rx 10 LTE - Geometric scheduling - IA Rx I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

40 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE - No scheduling - SU Rx 1 QAM64 LTE - Geometric scheduling - SU Rx 10 LTE - Geometric scheduling - IA Rx I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 5 / 40

41 Simulation Results Simulations Dual antenna enodeb and single-antenna UEs bps/Hz MU MIMO IA Rx SU MIMO Transmit diversity MU MIMO SU Rx FER SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 6 / 40

42 Simulation Results Simulations Dual antenna enodeb and single-antenna UEs bps/Hz bps/hz MU MIMO IA Rx SU MIMO Transmit diversity MU MIMO SU Rx FER SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 6 / 40

43 Simulation Results Simulations Dual antenna enodeb and single-antenna UEs bps/Hz bps/hz 4bps/Hz MU MIMO IA Rx SU MIMO Transmit diversity MU MIMO SU Rx FER SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 6 / 40

44 Simulation Results Simulations Dual antenna enodeb and single-antenna UEs bps/hz 1bps/Hz bps/hz 4bps/Hz Slow Fading MU MIMO IA Rx SU MIMO Transmit diversity MU MIMO SU Rx FER SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 6 / 40

45 LTE Precoders - Constraints Precoding Feedback and Strategies Low angular resolution j Equal gain transmission - Same magnitudes Infinite angular resolution 1 1 Infinite magnitudes j Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 7 / 40

46 LTE Precoders - Constraints Precoding Feedback and Strategies Low angular resolution j Equal gain transmission - Same magnitudes Infinite angular resolution 1 1 Infinite magnitudes j Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 7 / 40

47 LTE Precoders - Constraints Precoding Feedback and Strategies Low angular resolution j Equal gain transmission - Same magnitudes Infinite angular resolution 1 1 Infinite magnitudes j Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 7 / 40

48 LTE Precoders - Constraints Precoding Feedback and Strategies Low angular resolution j Equal gain transmission - Same magnitudes Infinite angular resolution 1 1 Infinite magnitudes j Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 7 / 40

49 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE precoders Infinite angular resolution 1 10 QAM64 Infinite magnitudes Infinite angular resolution and magnitudes I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 8 / 40

50 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE precoders Infinite angular resolution 1 10 QAM64 Infinite magnitudes Infinite angular resolution and magnitudes I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 8 / 40

51 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE precoders Infinite angular resolution 1 10 QAM64 Infinite magnitudes Infinite angular resolution and magnitudes I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 8 / 40

52 Information Theoretic Perspective Information Theoretic Perspective Sum rate of downlink of single antenna UEs and dual antenna enodeb LTE precoders Infinite angular resolution 1 10 QAM64 Infinite magnitudes Infinite angular resolution and magnitudes I(bps/Hz) SNR (db) Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 8 / 40

53 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Single-user System y 1,k = h 1,k p 1,k x 1,k + z 1,k where h 1,k C 1 and h 1,k CN (0, I). y 1,k C and z 1,k CN (0, N 0 ) while x 1,k χ 1 is of power σ 1. For EGT, p 1,k = 1 [1 h 1,k h 11,k h 1,k h 11,k ] T. Pairwise Error Probability Analysis P(c 1 ĉ 1 ) 1 d free 48 ( )) σ ( d 1 1,min N 0 where d 1,min is the normalized minimum distance of the constellation χ 1. Full diversity for EGT in single-user systems Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 9 / 40

54 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Single-user System y 1,k = h 1,k p 1,k x 1,k + z 1,k where h 1,k C 1 and h 1,k CN (0, I). y 1,k C and z 1,k CN (0, N 0 ) while x 1,k χ 1 is of power σ 1. For EGT, p 1,k = 1 [1 h 1,k h 11,k h 1,k h 11,k ] T. Pairwise Error Probability Analysis P(c 1 ĉ 1 ) 1 d free 48 ( )) σ ( d 1 1,min N 0 where d 1,min is the normalized minimum distance of the constellation χ 1. Full diversity for EGT in single-user systems Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 9 / 40

55 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Single-user System y 1,k = h 1,k p 1,k x 1,k + z 1,k where h 1,k C 1 and h 1,k CN (0, I). y 1,k C and z 1,k CN (0, N 0 ) while x 1,k χ 1 is of power σ 1. For EGT, p 1,k = 1 [1 h 1,k h 11,k h 1,k h 11,k ] T. Pairwise Error Probability Analysis P(c 1 ĉ 1 ) 1 d free 48 ( )) σ ( d 1 1,min N 0 where d 1,min is the normalized minimum distance of the constellation χ 1. Full diversity for EGT in single-user systems Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 9 / 40

56 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Dual antenna BS and single antenna UEs. Slow fading channel SU - Infinite magnitudes & angles 10 0 bps/hz SU - Infinite angles 10 1 SU - LTE precoders MU - Infinite magnitudes & angles FER 10 MU - Infinite angles 10 3 MU - LTE precoders SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

57 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Dual antenna BS and single antenna UEs. Slow fading channel SU - Infinite magnitudes & angles 10 0 bps/hz SU - Infinite angles 10 1 SU - LTE precoders MU - Infinite magnitudes & angles FER 10 MU - Infinite angles 10 3 MU - LTE precoders SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

58 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Dual antenna BS and single antenna UEs. Slow fading channel SU - Infinite magnitudes & angles 10 0 bps/hz SU - Infinite angles 10 1 SU - LTE precoders MU - Infinite magnitudes & angles FER 10 MU - Infinite angles 10 3 MU - LTE precoders SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

59 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Dual antenna BS and single antenna UEs. Slow fading channel SU - Infinite magnitudes & angles 10 0 bps/hz SU - Infinite angles SU - LTE precoders MU - Infinite magnitudes & angles FER MU - Infinite angles MU - LTE precoders SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

60 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Dual antenna BS and single antenna UEs. Slow fading channel SU - Infinite magnitudes & angles 10 0 bps/hz SU - Infinite angles 10 1 SU - LTE precoders MU - Infinite magnitudes & angles FER 10 MU - Infinite angles 10 3 MU - LTE precoders SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

61 Performance Analysis of EGT Effect of EGT on single-user and multi-user modes Dual antenna BS and single antenna UEs. Slow fading channel SU - Infinite magnitudes & angles 10 0 bps/hz SU - Infinite angles 10 1 SU - LTE precoders MU - Infinite magnitudes & angles FER 10 MU - Infinite angles 10 3 MU - LTE precoders SNR Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

62 Performance Analysis of EGT Increase of one bit of feedback - Options Increase the resolution or increase the levels of transmission j -1 1 j -j j j j j -j -j -j Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

63 Proposed precoding codebook Performance Analysis of EGT Dual antenna BS and single antenna UEs bps/Hz bps/Hz FER 10 1 FER SNR SNR Mode 6 - additional bits Enhanced Levels Mode 6 - additional bits Angular Resolution Mode 6 LTE Mode 5 - Lower bound Mode 5 - additional bits Mode 5 - additional bits Mode 5 Enhanced Levels Angular Resolution LTE Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/011 3 / 40

64 Concluding Remarks Conclusions Modern wireless systems based on quantized CSIT Single-user detection highly sub-optimal Multi-user MIMO can work only if Interference Aware receivers are used Modern wireless systems based on fixed feedback Channel magnitude information more important than channel direction Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

65 Concluding Remarks Conclusions Modern wireless systems based on quantized CSIT Single-user detection highly sub-optimal Multi-user MIMO can work only if Interference Aware receivers are used Modern wireless systems based on fixed feedback Channel magnitude information more important than channel direction Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

66 Conclusions Thanks Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

67 Conclusions References D. Gesbert, M. Slim Alouini, How much feedback is multi-user diversity really worth? in Proc. of IEEE ICC 004, June 004. G. Caire and S. Shamai (Shitz), On the achievable throughput of a multi-antenna Gaussian broadcast channel, IEEE Trans. Info. Th., vol. 49, no. 7, pp. 1691Ű1706, July 003. R. Knopp and P. Humblet, Information capacity and power control in single cell multi-user communications, in Proc. IEEE ICC 1995, Seattle, WA, USA, pp. 331Ű335. M. Sharif and B. Hassibi, On the capacity of MIMO broadcast channel with partial side information, IEEE Trans. Info. Th., vol. 51, no., pp. 506Ű5, Feb. 005 C. B. Peel, B. M. Hochwald, and A. L. Swindlehurst, A vector-perturbation technique for near capacity multiantenna multiuser communication - part I: channel inversion and regularization, IEEE Trans. Comm., vol. 53, no. 1, pp. 195Ű0, Jan R. Zamir, S. Shamai (Shitz), and U. Erez, Nested linear/lattice codes for structured multiterminal binning, IEEE Trans. Info. Th., vol. 48, no. 6, pp. 150Ű176, June 00. N. Jindal, MIMO broadcast channels with finite-rate feedback, IEEE Trans. Inf. Theory, vol. 5, no. 11, pp , 006. Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

68 Conclusions References-Contd Q. H. Spencer, A. L. Swindlehurst, and M. Haardt, Zero-forcing methods for downlink spatial multiplexing in multiuser MIMO channels, IEEE Trans. on Signal Processing, vol. 5, no., pp. 461Ű471, Feb P. Viswanath and D. Tse Writing on dirty paper, IEEE Trans. on Information Theory, vol. 9, no.3, pp. 439Ű441, May P. Viswanath and D. Tse Sum capacity of the multiple antenna Gaussian broadcast channel and uplink-downlink duality, IEEE Trans. on Information Theory, vol. 49, pp. 191Ű191, August 003. T. Marzetta and B. Hochwald, Fast transfer of channel state information in wireless systems, IEEE Trans. on Signal Processing, vol. 54, pp. 168Ű178, April 006. R. Ghaffar and R. Knopp, Interference-aware receiver structure for Multi-User MIMO and LTE, EURASIP Journal on Wireless Communications and Networking, 011. Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

69 Conclusions 3GPP LTE Rel 8 Supports SISO, 1,, 4 1, 4 and 4 4 systems Only wideband PMI Mode 4: Single-user MIMO. Rank- transmission ( layers). So max of two codewords transmission to a UE Mode 5: Multi-user MIMO. Rank-1 transmission. Max of UEs can be scheduled Mode 6: Single-user MIMO. Rank-1 transmission Mode-7. Single-user MIMO. 1 layer Beamforming. Single layer UE-specific RS (DM-RS) based beamforming. PDSCH Channel estimation based on dedicated reference signals (DRS). Open loop, non-codebook precoding DoA beamforming in TDD/FDD while channel reciprocity can be used in TDD. Does not support multi-user MIMO Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

70 Conclusions Transmission Mode 8 Enhancement of transmission mode 7 of Rel. 8 Spatial multiplexing supported Two layers of UE-specific RS Upto layers per user (Single-user MIMO) Two orthogonal streams of UE-specific RS are supported for multi-user MIMO transmission Upto 4 layers in total (Quasi-orthogonal multi-user MIMO with the aid of scrambling IDs) They have the same overhead as Rel-8 one stream UE-specific RS Rank-1 multi-user MIMO Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

71 Conclusions Transmission Mode 9 In addition, supports 8 1, 8, 8 4 and 8 8 systems Seamless switching between single-user and multi-user MIMO up to rank 8 Frequency selective PMI DM-RS used for demodulation and it can support up to eight layer transmission in single-user Four codewords transmitted in multi-user MIMO mode so two users can be multiplexed with two codewords to each user The precoder index is not necessary to inform and the DL precoder is not specified Flexible precoder usage is possible in the enb design due to the precoded DMRS based transmission DCI format C Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40

72 Conclusions Multi-user MIMO in LTE-Advanced Max 4 UEs can be scheduled with max layers each Not more than 4 layer transmission per PRB Precoding per subband basis Layered precoding for the case of 8 transmit antennas W 1 targeting wideband/long-term channel properties. For the case of and 4 antennas, it is identity W targeting frequency-selective/short-term channel properties. For the case of and 4 antennas, it is same as LTE Rel 8 precoders In practice two streams will remain most common case as typical UEs would not have more then receive antennas New Transmission Mode 9 introduced Rizwan GHAFFAR (EURECOM) MU-MIMO in LTE/LTE-A 17/06/ / 40