Beamforming on mobile devices: A first study Hang Yu, Lin Zhong, Ashutosh Sabharwal, David Kao http://www.recg.org
Two invariants for wireless Spectrum is scarce Hardware is cheap and getting cheaper 2
3.2 cm Passive directional antennas 3.2 cm Ardalan Amiri Sani, Lin Zhong, and Ashutosh Sabharwal, "Directional antenna diversity for mobile devices: characterizations and solutions," in Proc. ACM MobiCom, September 2010. 4
Findings: ~3 db gain Multifold throughput increase at network edge ~50% TX power reduction at network center 5
Can we go beyond 3 db? 6
Beamforming? Studied in the past for use on cellular base station, 802.11 access points, vehicles, and even wireless sensor nodes, e.g., MobiSteer (MobiSys 07), R2D2 (MobiSys 09), DIRC (SIGCOMM 09) 7
Beamforming primer 8
Beamforming primer Fixed transmission power
Beamforming primer Fixed transmission power
Beamforming primer Fixed transmission power
Beamforming primer Fixed transmission power
Is beamforming practical? Beamforming Mobile devices Antenna array Small form factor Narrow beam Rotate and move Power hungry Battery powered 13
Peak beamforming gain (db) Form factor? 7 6 5 4 3 2 1 4 antennas 3 antennas 2 antennas 0 0 0.1 0.2 0.3 0.4 0.5 Antenna spacing (wavelength) 0.3-0.4 λ : 4.5-6 cm at 2 GHz
Form factor! 18 cm 6 cm 24 cm 12 cm 0.3-0.4 λ (4.5-6 cm at 2 GHz) 15
Rotation? Client Node Infrastructure Node 16
Beamforming gain (db) Rotation? Indoor 6 Max Static 90d/s 180d/s 3 0 N=2 N=4 CSI estimation every 100 ms
Beamforming gain (db) Rotation! Indoor 6 Max Static 90d/s 180d/s 3 0 N=2 N=4 CSI estimation every 10 ms
Power? (uplink only) P Circuit P PA =P TX / η Baseband Signal DAC Filter Mixer Filter PA 1 Frequency Synthesizer N P Shared Baseband Signal DAC Filter Mixer Filter PA N P = P shared + N P Circuit + P TX / η 19
Tradeoff No. 1 P=P shared + 1 P Circuit + P TX / η Fixed receiver SNR
Tradeoff No. 1 P=P shared + 2 P Circuit + P TX / η Fixed receiver SNR
Tradeoff No. 1 P=P shared + 3 P Circuit + P TX / η Fixed receiver SNR
Tradeoff No. 1 P=P shared + 4 P Circuit + P TX / η Fixed receiver SNR
Tradeoff No. 1 Optimal number of antennas for efficiency N opt = a P O /P Circuit b P O
Transmitter Power Consumption (mw) Hardware is cheap & getting cheaper P = P shared + N P Circuit + P TX / η 1200 1000 800 600 400 200 SISO 2x2 MIMO Sources: 0 2002 2004 2006 2008 2010 Year IEEE Int. Solid-State Circuits Conferences (ISSCC) and IEEE Journal of Solid-State Circuits (JSSC)
Power! Beamforming with state-of-the-art multi-rf chain realization is already more efficient! Tradeoff No. 1 is increasingly profitable!
Beyond a single link 27
What the carrier wants: Use all your antennas! 28
What you want: N opt = a P O /P Circuit b P O 29
Tradeoff No. 2 Network capacity vs. client efficiency 30
How can clients figure out its N without talking to each other? 31
BeamAdapt Distributed algorithm to minimize TX power under uplink capacity constraints No explicit inter-client cooperation Iterative Guaranteed to converge Converge in a few iterations in practice Converge to a good solution in practice Can be built on top of uplink power control in cellular networks 32
WARPLab-based prototype Infrastructure Node 1 Infrastructure Node 2 Ethernet Router Uplink (Wireless) Uplink (Wireless) Client Node 1 Client Node 2 Laptop with MATLAB 33
Beamforming size SINR (db) Received SNR stable 20 10 0 Client Node 2-10 0 5 10 Time (s) 4 3 2 1 0 5 10 Time (s) Link SNR constraint: 5 db 34
Power close to optimal 1500 1000 500 0 Power consumption (mw)2000 5dB BeamAdapt Genie-aided I/S I/M O/S O/M I: Indoor O: Outdoor S: Stationary M: Mobile / Rotational Link SNR constraint: 5 db
4km 4km UMTS; Client movement: 0-70 mph; Client rotation: 0-120 /s
Client Power Consumption (mw) Power reduced 1000 800 Beamforming/Omni BeamAdapt 600 400 200 0 N=1 N=2 N=4 N=8 CBR traffic
Network Throughput (b/s) Network throughput maintained 1.5 2 x 10 5 Beamforming/Omni BeamAdapt 1 0.5 0 N=1 N=2 N=4 N=8 CBR traffic
Conclusions Beamforming is feasible for mobile devices Lower-power uplink for mobile devices Distributed optimization feasible
Looking forward Benefits of beamforming orthogonal to other spectrum efficiency technologies such as network MIMO Network capacity implications
Treating interference as noise Strong interference regime: Far from optimal from information theoretic perspective
Treating interference as noise Weak interference regime: Existing architecture yields close to optimal capacity
http://www.recg.org