Spectral Efficiency of Massive MIMO Systems with D2D Underlay

Size: px

Start display at page:

Download "Spectral Efficiency of Massive MIMO Systems with D2D Underlay"

Beatrix Miles
6 years ago
Views:

1 Robert W. Heath Jr. (15) Spectral Efficiency of Massive MIMO Systems with DD Underlay Xingqin Lin*, Robert W. Heath Jr. #, Jeffrey G. Andrews # * Radio Access Technologies, Ericsson Research, San Jose, CA, USA # Wireless Networing Communications Group, Department of Electrical Computer Engineering, The University of Texas at Austin IEEE ICC 15, London, UK

2 Robert W. Heath Jr. (15) Massive MIMO DD: Complimentary technologies Massive MIMO: ~1 antennas at base stations to serve ~1 users per cell DD DD u Massive MIMO: multi-user MIMO w/ lots of BS antennas [Marzetta1] ª Benefits: increased rate, reliability, reduced TX power, area spectral efficiency, etc u Device-to-device: direct communication between nearby mobile users ª Benefits: increased rate, low power, low delay, offloading, new services, etc. Underst the interactions between massive MIMO & DD

3 Robert W. Heath Jr. (15) Related wor u Stochastic geometry analysis for cellular [Andrews11, Dhilon1, Heath13] u Massive MIMO ª Networs of finite size: [Marzetta1, Hul1, Ngo13, Hoydis13, Truong13] u DD ª Networs w/ spatially distributed nodes: [Madhul13, Bai14, Liang15] ª Qualcomm FlashLinQ [Wu13] ª Much wor on single-antenna [Lin14], some on multi-antenna [Janis9, Min11] ª Several analyses related to spatially distributed nodes u Massive MIMO + DD [Yin14] ª Use DD to enable local CSI exchange in a FDD massive MIMO system No related wor on massive MIMO + DD + stochastic geometry 3

4 Robert W. Heath Jr. (15) Questions answered in our wor How does DD impact the spectral energy efficiency of massive MIMO? How does massive MIMO impact DD spectral efficiency? Assume perfect CSI (imperfect CSI is here* ) Asymptotic non-asymptotic results in large # of antennas * X. Lin, R. W. Heath Jr., J. G. Andrews, The interplay between massive MIMO underlaid DD networing, IEEE Transactions on Wireless Communications, to appear. 4

5 Robert W. Heath Jr. (15) System model g i h g ii h i g ir M Rx antennas at the BS u DD underlaid multi-cell cellular networ (shares uplin) ª Each cell has K romly distributed uplin cellular UEs ª PPP distributed DD TXs with density λ ª Each DD TX has a DD RX located at distance D away from the TX u SIMO scenario ª TXs (either cellular or DD) use a single antenna ª Each BS has M receive antennas ª Each DD RX has N receive antennas Single cell 5

6 y X p Pc x c K X p h u Robert W. Heath Jr. (15) Baseb channel model receive processing X p TX power Position Cellular signals DD signals Noise Pathloss exponent Fast fading K + X i p Pd x i c Data symbol h i u i + v, u The total received signal at the central BS (similar model for DD users) u Partial zero-forcing (PZF) receivers ª BS cancels nearest m c nearest cellular m d nearest DD interferers ª DD receiver cancels nearest n c nearest cellular n d nearest DD interferers 6

7 w is cellular interferers we denote by Kr PZF the setfilter of uncanceled Cellular Spectral Efficiency r the set of uncanceled DD interferers at the DD we have used different pathloss exponents c S Robert W. Heath Jr. (15) z with s {c, d} z denoting r r or the -th cellular UE, the post-processing SINR with the receiver r. SINR, Sreceiver UE-UE (cf.the(1)pathloss ()) due to their r, d > lins denotes filter w is I + I + w N III. C ELLULAR S PECTRAL E FFICIENCY lins, gr,characteristics. gri C N 1 are Specifically, the vector pagation the antenna llular transmitter to the receiver A. the Asymptotic Cellular SpectralEfficiency c macro BS is tens of DD meters, while typical S where S P x w c D transmitter i to the DD receiver r h denotes the d SINR, (3) For the -th cellular UE, the post-processing SINR with the ht at UE is undergaussian m. Asnoise a result, both terminals C Na 1 is complex signal power of cellular UE, I I respec I spectral + I with + w N u Average efficiency is the main performance metric PZF scattering filter w is in are low see similar near street "!# DD interference p denote the cochannel cellular cell c used is different from radio ewhich Sdifferent Ppathloss wthe environment denotes the desired c x exponents c h experiencedsby cellular UE SINR, (3) are given by E-UE lins (cf. (1) ()) due to their R E log 1 + cro BS []. alcharacteristics. power of cellular UE the, Iantenna I respectively II + I+ + w+x I Specifically, NN (s) user c I P x w h` er, we assume Gaussian signaling, i.e., {u }, s c te the cochannel cellular DD interference powers ` c BS is tens of meters, while the typical where S P x w h denotes the desired c i.d. zero-mean complex Gaussian UE terminals are given by Erienced is under bym.cellular As a unit-variance result, both `Krespectively signal power of cellular UE, I I (s) (s) X X street all see similar near scattering dowthat the vector channels,, s h c denote g the cochannel cellular DD interference I Pdpowers xi c w hi. I fromthe radiopenvironment w h`r isi.d. different c x` by cellular UE are given by CN (, 1) elements, independentexperienced across trans ]. X `K i (s) lows that the favorable propagation condition [3] c I Pc x` w h` sume Gaussian signaling, i.e., {u }, s X c assive in our spectral efficiency of the -th cellular UE is define systems Pholds wmodel: hi. (4) -mean IMIMO unit-variance complex Gaussian The d x `K i (s) (s) X h h i l the vector channels, gr, s i I Pd xi c whi. (4) 1 if s s r `; a.s. (s ) s) (, h 1) elements, across transr EDD log(1users + SINR ),!independentcellular users Uncancelled ` Uncancelled i otherwise, tspectral the favorable propagation condition [3] efficiency of the -th cellular UE is defined as (come from (distributed a PPP) IMO systems holds in our model:set) hfinite The ispectral efficiency of the -th cellular UE isas defined as h i E rlog(1 (5) 1 if R s s `; + SINR ), a.s. IN. symptotic Performance evaluation for cellular users! otherwise, R E log(1 + SINR ), (5) Compute rate assuming perfect CSI 596 7

8 Robert W. Heath Jr. (15) Cellular user large antenna regime u Spectral efficiency goes to infinity (asymptotic orthogonality & perfect CSI) u If P c Θ(1/Μ), a limiting finite cellular spectral efficiency is achieved R P without DD! log(1 + SNR! (m, ) ( ) P d (m +1 ), ( )N (m) R 1 ) with DD a fixed m d lim M!1 R log 1+ SNR scaling up m d with Θ(log(Μ))! (m d, c )+1 Mean DD canceled interf. R! log(1 + SNR ) Lie having no DD interference No loss of spectral efficiency power saving due to the DD underlay if m d scales appropriately 8

9 Cellular user large but finite antenna regime u With M m c + m d +1 m c > c R (c,lb) (M m c m d 1)SNR P SNR ` + (m d, c )+1 `K 1 m c m d 1 A Robert W. Heath Jr. (15) uncancelled cellular interference m c mean uncancelled DD ference, interference thus lowe density (m d, c ) c the distances of th TX power m d 9

10 Robert W. Heath Jr. (15) DD user large but finite antenna regime u With R (d,lb) r N n c + n d +1 of DD Tx-Rx pair P n d > c 1 (N n c n d 1)SNR A r ) d + (nd, d )+1 K r P c N (d 1 n c n d (19 uncancelled cellular interference n c non-homogenous (depends on locations of cellular users) mean uncancelled DD interference density TX power m d 1

11 Cellular user performance comparison BS coverage radius R c 5 m DD lin length d m # cellular UEs K 4 4 Density of DD UEs R m c # BS antennas M 1 # UE Rx antennas N 6 UE-BS PL exponent c 3.76 UE-UE PL exponent d 4.37 UE-BS PL reference C c, 15.3 db UE-UE PL reference C d, 38.5 db Cellular Tx power P c 3 dbm DD Tx power P d 13 dbm Channel bwidth 1 MHz Noise PSD 174 dbm/hz BS noise figure 6 db UE noise figure 9 db TABLE I Spectral Efficiency (bit/s/hz) Robert W. Heath Jr. (15) no DD canceling increasing # DD users canceling fixed # DD users m c, No DD (m c,m d ) (,) (m c,m d ) (, M 1/ ) m c 3, No DD (m c,m d ) (3,) (m c,m d ) (3, M 1/ ) M: # of BS Antennas 11

12 Robert W. Heath Jr. (15) Cellular spectral efficiency W/ constant cellular TX power W/ scaled cellular TX power Perfect CSI Imperfect CSI Imperfect CSI w/ inac7ve DD in the training Unbounded Scaling law: 1/M DD- to- cellular interference can be eliminated by scaling up m d Bounded reduced due to DD underlay contamina;on Should not be scaled down Scaled cellular TX power results in vanishing spectral efficiency Bounded no effect of DD underlay Scaling law: 1/M.5 DD- to- cellular interference in the data transmission persists X. Lin, R. W. Heath Jr., J. G. Andrews, The interplay between massive MIMO underlaid DD networing, IEEE Transac*ons on Wireless Communica*ons, to appear. 1

13 Robert W. Heath Jr. (15) References u Stochastic geometry for cellular [Andrews11] J. G. Andrews, F. Baccelli, R. K. Ganti, "A tractable approach to coverage rate in cellular networs", IEEE TCom, vol. 59, no. 11, pp , nov. 11. [Dhilon1] H. Dhillon, R. K. Ganti, F. Baccelli, J. G. Andrews, "Modeling analysis of K-tier downlin heterogeneous cellular networs", IEEE JSAC, vol. 3, no. 3, pp , Apr. 1. [Heath13] R. W. Heath, Jr., M. Kountouris, T. Bai`` Modeling heterogeneous networ interference using Poisson point processes,'' IEEE Trans. on Signal Processing, vol. 61, no. 16, pp , Aug. 13. u Massive MIMO [Marzetta1] T. L. Marzetta, Noncooperative cellular wireless with unlimited numbers of base station antennas, IEEE Twireless, vol. 9, no. 11, pp , Nov. 1. [Huh1] H. Huh, G. Caire, H. C. Papadopoulos, S. A. Ramprashad, Achieving massive MIMO spectral efficiency with a notso-large number of antennas, IEEE TWireless, vol. 11, no. 9, pp , Sep. 1. [Ngo13] H. Q. Ngo, E. Larsson, T. Marzetta, Energy spectral efficiency of very large multiuser MIMO systems, IEEE TCom, vol. 61, no. 4, pp , Apr. 13. [Hoydis3] J. Hoydis, S. ten Brin, M. Debbah, Massive MIMO in the UL/DL of cellular networs: How many antennas do we need? IEEE JSAC, vol. 31, no., pp , February 13. [Truong13] K. T. Truong R. W. Heath, Jr., Effects of channel aging in massive MIMO systems, Journal of Communications Networs, Special Issue on Massive MIMO, vol. 15, no. 4, pp , August

14 References Robert W. Heath Jr. (15) u Massive MIMO (cont d) [Madhu13] P. Madhusudhanan, X. Li, Y. Liu, T. Brown, Stochastic geometric modeling interference analysis for massive [Bai14] [Liang15] MIMO systems, in Proceedings of WiOpt, May 13, pp. 15. T.Bai R. W. Heath Jr, Asymptotic coverage probability rate in massive MIMO networs, in Proceedings of IEEE GlobalSIP, December 14, pp N. Liang, W. Zhang, C. Shen, An uplin interference analysis for massive MIMO systems with MRC ZF receivers, Proc. of WCNC, 15. u DD [Wu13] X. Wu, S. Tavildar, S. Shaottai, T. Richardson, J. Li, R. Laroia, A. Jovicic, FlashLinQ: A synchronous distributed [Lin14] scheduler for peer-to-peer ad hoc networs, IEEE/ACM Trans. Networing, vol. 1, no. 4, pp , Aug. 13. X. Lin, R. Ratasu, A. Ghosh, J. G. Andrews, Modeling, analysis optimization of multicast device-to-device transmissions, IEEE TWireless, vol. 13, no. 8, pp , Aug. 14. [Janis9] P. Janis, V. Koivunen, C. B. Ribeiro, K. Doppler, K. Hugl, Interference-avoiding MIMO schemes for deviceto-device radio underlaying cellular networs, in Proceedings of IEEE PIMRC, 9, pp [Min11] H. Min, J. Lee, S. Par, D. Hong, Capacity enhancement using an interference limited area for device-todevice uplin underlaying cellular networs, IEEE TWireless, vol. 1, no. 1, pp , Dec. 11. u Massive MIMO + DD [Yin14] H. Yin, L. Cottatellucci, D. Gesbert, "Enabling Massive MIMO Systems in the FDD Mode thans to DD Communications", in Proc. of the Asilomar Conference on Signals, Systems, Computers, Nov

Analysis of massive MIMO networks using stochastic geometry

Analysis of massive MIMO networks using stochastic geometry Tianyang Bai and Robert W. Heath Jr. Wireless Networking and Communications Group Department of Electrical and Computer Engineering The University