Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Erik G. Larsson March 2014 MM Div. of Communication Systems Dept. of Electrical Engineering (ISY) Link¨ oping University Link¨ oping, Sweden www.commsys.isy.liu.se YS Massive MIMO k=1 M= x100 antennas! K terminals k=K ◮ Massive MIMO is massive multiuser MIMO: ◮ ◮ M ≫ K ≫ 1 (think M = ×00, K = ×0) coherent, but simple, processing ◮ Spatial multiplexing gain O(K), in favorable propagation ◮ √ Array gain (radiated energy efficiency) gain O(M ), or O( M ) 1/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Massive MIMO Deployment ◮ ◮ Testbeds: Argos (USA), 96× ∼10 FP7-MAMMOET, 100×? Massive MIMO is not mmWave MIMO! 2/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University λ 4v c ∆ Tc = Bc = frequency Coherence Interval τc , Bc Tc symbols time At 2 GHz carrier frequency: pedestrian v = 1.5 m/s indoors ∆ ∼ 30 meters Bc = 10 MHz Tc = 25 ms τc = 250000 vehicular v = 30 m/s Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? outdoors ∆ ∼ 1000 meters Bc = 300 kHz Tc = 25 ms τc = 7500 Bc = 300 kHz Tc = 1.25 ms τc = 375 Communication Systems Link¨ oping University 3/12 τc for a Coherence Bandwidth of Bc = 200 kHz fc = 1GHz fc = 2GHz fc = 6GHz fc = 60GHz τc = 15000 τc = 7500 τc = 2500 τc = 250 10 km/h τc = 5400 τc = 2700 τc = 900 τc = 90 50 km/h τc = 1080 τc = 540 τc = 180 τc = 18 100 km/h τc = 540 τc = 270 τc = 90 τc = 9 350 km/h τc = 154 τc = 77 τc = 26 τc = 3 pedestrian 1 m/s (with the convention that Tc = λ ) 4v 4/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Massive MIMO Operation ◮ Main challenge in MU-MIMO: getting CSI at the base station. ◮ FDD training: BS transmits pilots, terminals report CSI. ◮ ◮ ◮ ◮ ◮ ◮ Downlink training: each antenna transmits sequence ψm , of length τp , in each coherence interval! Optimal training on the downlink requires ( const 6= 0, m = m′ H ψm ψm′ = 0, m 6= m′ Require M ≤ τp ≤ τc ! In addition, uplink control channel requirements scale as O(M ) FDD operation is not scalable, and probably infeasible in most cases TDD training: terminals transmit pilots, BS estimates channel ◮ ◮ ◮ ◮ Rely on uplink-downlink reciprocity The K terminals send pilots of length τp in each coherence interval Require K ≤ τp ≤ τc ! Pilot overhead independent of M — TDD entirely scalable w.r.t. M 5/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Massive MIMO Operation ◮ On the uplink, ◮ ◮ ◮ ◮ acquire CSI from uplink pilots and/or blindly from data detect symbols M ≫ K ⇒ linear processing (MRC, ZF, MMSE) nearly optimal On the downlink, ◮ ◮ ◮ ◮ use CSI obtained on the uplink make necessary adjustments based on reciprocity calibration apply multiuser MIMO precoding simple precoders desirable (and very good!): MRT, ZF, R-ZF, ... 6/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Uplink and Downlink Models ◮ M antennas ◮ K terminals (single-antenna)  1  1 g1 · · · gK  ..  .. G ,  ... . .  ◮ g1M ◮ M · · · gK √ Uplink: y = Pu Gx + w, ◮ √ Downlink: y = Pd GT x + w. ◮ gkm = terminal 1 1 terminal K 2 m gkm M base station array terminal k √ βk h m k 7/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Receivers and Precoders ◮ MRC/MRT: ◮ ◮ ˆ= MRC receiver (uplink): x GH 2 kGk  ·y  ∗ G  ◮ MRT precoder (downlink): x =  q ·q E kGk2 X 1 ◮ Channel hardening: βk , M ≫ 1 kGk2 ≈ M k ◮ intracell interference treated as noise ◮ typical operating point is 1 bps/Hz/terminal; K bps/Hz total ◮ distributed implementation ZF: ◮ ZF receiver (uplink): x ˆ = (GH G)−1 GH · y ◮ ◮ ◮  G∗ (GT G∗ )−1   ZF precoder (downlink): x =  q ·q E trace{(GT G∗ )−1 } can cancel out (some) interference computationally somewhat more demanding Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? 8/12 Communication Systems Link¨ oping University Bc Nsmooth subcarriers Nslot OFDM symbols coherence interval frequency OFDM with TDD Operation Nsmooth Nslot Ts uplink Tslot time downlink τc = Nslot Nsmooth symbols τu τp τd UL data UL pilots DL data τc symbols 9/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Typical OFDM Parameters coherence bandwidth subcarrier spacing number of subcarriers within coherence bandwidth slot duration OFDM symbol duration number of OFDM symbols within one slot Bc Bs Nsmooth Tslot Ts Nslot 210 kHz 15 kHz 14 1 ms 71 µs 14 10/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Literature for this Lecture ◮ T. L. Marzetta, “Noncooperative MU-MIMO with unlimited numbers of base station antennas,” IEEE Trans. WC 2010. ◮ E. G. Larsson, F. Tufvesson, O. Edfors and T. L. Marzetta, “Massive MIMO for Next Generation Wireless Systems,” IEEE Comm. Magazine, Feb. 2014. ◮ F. Rusek, D. Persson, B. K. Lau, E. G. Larsson, T. L. Marzetta, O. Edfors, and F. Tufvesson, “Scaling up MIMO: Opportunities and Challenges with Large Arrays”, IEEE SP Magazine, Jan. 2013 ◮ C. Shepard et al., “Argos: practical many-antenna base stations”, in Proc. MobiCom 2012. 11/12 Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? Communication Systems Link¨ oping University Massive MIMO Explained is Based on Joint Work and Discussion with My Colleagues: ◦ ◦ ◦ ◦ ◦ ◦ ◦ Hien Q. Ngo (LiU, Sweden) Christopher Moll´en (LiU) Marcus Karlsson (LiU) Antonios Pitarokoilis (LiU) Hei Victor Cheng (LiU) Emil Bj¨ ornson (LiU) Daniel Persson (LiU) ◦ Fredrik Tufvesson (LU, Sweden) ◦ Ove Edfors (LU) ◦ Fredrik Rusek (LU) ◦ Thomas L. Marzetta (Bell Labs/Alcatel-Lucent, USA) ◦ Saif K. Mohammed (IIT, Dehli) Erik G. Larsson Massive MIMO Explained: Part 1: What is Massive MIMO and How Does It Work? 12/12 Communication Systems Link¨ oping University