Evolving 4G to the Next Level

Transcription

1 Evolving 4G to the Next Level A. Paulraj Stanford University Beceem Communications Inc. GCOE Workshop on Adv. Wireless Signal Processing and Networking Technology

2 How dense are Wireless networks 2 Internet Users Billions Mobile Phone Subscribers 4 (Billions) Source: IDC, The Digital Economy Fact Book

3 Exciting Promise Mobile internet represents a new computing cycle Mainframe to Mini to PC to PC Internet to Mobile internet Mobile Internet is the way Billions will connect to the internet. Paralleling trends in voice telephony

4

5 4G Requirements (NGMN) Throughput > 100 Mbps Peak BS Radio > 10 Mbps Peak / Terminal Spectral Eff. > 2.5 bps/hz/cell Cell range > 1 Mile dense urban Low delay < 10 ms RT Delay All IP 4G Technologies: WIMAX 16e R2 (2008), 3G LTE R8 (2011)

6 WIMAX Equipment Infrastructure Motorola, Samsung, Nortel-Alvirion, Alcatel-Lucent, NEC, Fijitsu Chips Beceem, Intel, Sequans, GCT, Samsung*

7 Sprint Network Herndon, VA FTP Throughputs 12.8 Mbps (DL Peak) 6.9 Mbps (DL Avg) 3.2 Mbps (UL Peak) 2.3 Mbps (UL Avg) Infra: Mot, Samsung, NSN Terminals ZTE / Mot (both use Beceem Chipsets) 7

8 Clearwire Network - Portland Coverage Test (145 km2) UDP Throughputs 17Mbps (DL Peak) 9 Mbps (DL Avg) 2Mbps (UL Avg) Infra: Motorola Terminals Mot (Beceem Chipsets) 8 C:\Documents and Settings\jchen.BECEEM\M

9 Performance Goals 3G 4G IMT Adv.?? Peak Mbps / Term Spec. Eff bps/hz/cell Rnd Trip Delay ms Higher Speed, Improved Spectrum Efficiency, Lower Delay, Better QoS, Better Coverage,

10 BS Peak Data Rates 10 Tbps Evolution of wireless and wired networks 1 Tbps Wired LAN Peak single-user uplink PHY data rate 100G 100 Gbps 10G BaseT PCIe3 10G Fiber 10 Gbps PCIe2 PCI2.2 60GHz 1G BaseT PCIe1a USB3 VHT 1G BaseX F 60GHz 1 Gbps ATA5 ATA6 SATA150 SATA300 WiMedia FW n 100 BaseT USB2 WiMedia ATA4 FW Mbps DOCSIS3 DOCSIS2 USB1 VDSL2 DOCSIS g WiMax2/3GLTE 10 Mbps VDSL b WCDMA HSDPA Bluetooth1 WiMax1 1 Mbps Bluetooth ADSL ADSL2 EDGE 100 Kbps GPRS V90 10 Kbps IS95A Wireless LAN 60GHz Wireless WAN Wireless WAN WWAN fit Wireless LAN WLAN fit Wireless PAN Wired WAN Wired LAN LAN fit Wired PAN 1 Kbps Intel document, Sumeet Sandhu and Ed Casas, Intel CorporationYear of Standardization Source: S Sandhu / Intel

11 PHY Level Tools Bandwidth Modulation, Multiple access, MIMO, Opportunistic scheduling, Relay, Cooperation, Interference mitigation, H -ARQ,

12 Bandwidth Wider band At low SNR (cell edge), more bandwidth does not increase data rates Getting more bandwidth Multi-band OFDMA Rx compression, RF leakage Cognitive access to spectrum

13 Modulation OFDM most favorable Improving PA efficiency Adaptive Per tone vs per FEC Block Hierarchical Data over data Used in Broadcast / Multicast Hierarchical Pilot over data Useful in Unicast

14 Multiple Access OFDMA, TDMA for DL, OFDMA, DFT Coded OFDM for UL Scalable / Adaptive FFT size, CP, Symbol period

15 MIMO Conventional p2p MIMO is a great success! MIMO OFDM is a good marriage! Codes that are optimal (diversity multiplexing gain) as well as easily decodable remains open for innovation Fast Rx decoding is also open for innovation sphere decoding, iterative decoding,..

16 MIMO Number of Antennas Today: BS 4, MS 2 Future: Increasing MS antennas have tradeoffs RF chain power Per antenna power constraint reduces array gain Answer depends on Low vs High SNR Good vs Poor CSI-Tx Full vs Low rank channels Antenna EM issues

17 MIMO Relay Wired and wireless relay create composite MIMO channels Space time coding for location and directionally inhomogeneous antenna arrays Wired Wireless Composite MIMO Channel

18 MIMO Broadcast Multi-cell broadcast from directional sector antenna arrays. Delivering directionally homogenous service using space-time coding. SFN Networks One omni antenna per cell Cellular Networks 2-4 antennas per sector

19 Multi-Hop Relaying Provides 1.5-3X gains in throughput for cell-edge users Also Multi-hop diversity

20 Opportunistic Scheduling Choosing best user for a resource such as time slot, x frequency sub-channel x antenna, based on some metric SNR or SIR, capacity Prop Fair / Max-percentile Joint vs independent scheduling

21 Opportunistic Scheduling SISO channel We can get log(k) scaling in capacity for interference limited scheduling vs log log (K) scaling for noise limited scheduling

22 Interference Management Reuse controlling interference that the user sees Tx interference avoidance, Rx interference cancellation BS Cooperation & MS Cooperation Interference averaging

23 Interference in Broadband Networks C/I histogram Growing density of subscribers per unit area smaller cell increasingly interference dominant

24 Reuse and Power Control Single Reuse Class Dual Reuse Class Multiple Reuse Classes: Interference can be varied by controlling loading and power control, 1x3x1 1x3x1.5 1x3x x x x x x x x x x x x x x x x x x

25 Reuse and Power control When there is a dominant interference, the strategy usually suggests no reuse (time sharing) Full reuse may be optimal only for high SIR

26 Spatial Filtering - Avoid / Cancel Interference avoidance (linear precoding) at Tx MMSE or ML interference cancellation at Rx. Cardinality and SIR

27 BS Cooperation - Independent Encoding Base stations use independent encoding (interference channel) Weak interference treat it as noise Medium interference rate splitting Strong interference can be decoded and stripped out

28 Infra Cooperation - Joint Encoding Base stations cooperate in encoding (multi-user channel) Dirty Paper coding (with individual power constraint) Both support only one user (~ soft handoff) User is served only one of BTS at a given time (FBSS)

29 Interference Diversity and Repetition Coding Make a codeword see many diverse interferers by use of different PN seeds for codeword permutation (PUSC in WIMAX). Also symbol repetition Reduces interference variability

30 WIMAX Technology Ops / Sec : WIMAX is ~10 GOPS ( x10,000 GSM) Team: 250 man years per chip generation Area / Power: 2 Tx, 2 Rx, multi-band DCR radio, PMU, PHY, MAC 9x9 x1.5 mm, 400 mw + PA Power and Throughput Management: Clock, voltage and power islanding, power constrained processing,

31 Beyond 4G Challenges Goals are indeed challenging, but there many approaches to getting there (and there is always hope for fundamentally new ideas!) Any solution must meet constraints on battery life, infrastructure cost and coverage reliability.

32 Thank You