PowerConsumptionbyWireless Communication LinZhong ELEC518,Spring2011 Powerconsumption(SMT5600) Flight mode: Sleep, 3, 0% Lighting: Keyboard, 73, 3% Lighting: Display I, 148, 5% Cellular network, 17, 1% Lighting: Display II, 61, 2% LCD, 13, 0% Compute, 370, 13% Speaker, 45, 2% Bluetooth, 440, 16% GPRS, 1600, 58% 2 100 Power (mW) Powerconsumption(T-Mobile) Cellular Wi-Fi Bluetooth 10000 1000 10 1 Transmission Connected Transmission Connected Transmission Connected Paging Discoverable Speaker Keyboard lighting LCD lighting LCD Computing IDLE-Flight mode 3 Powerconsumption(Contd.) • Theoreticallimits – Receivingenergyperbit>N*10-0.159 • N:Noisespectralpowerlevel • Widebandcommunication PTX∝ PRX *da PRX Distance:d Propagation constant:a(1.81-5.22) 4 Powerconsumption(Contd.) • Whatincreasespowerconsumption – Governmentregulation(FCC) • Availablespectrumband(Higherband,higherpower) • Limitedbandwidth • Limitedtransmissionpower – Noiseandreliability – Highercapacity • Multipleaccess(CDMA,TDMAetc.) – Security – Addressability(TCP/IP) – More…… 5 Wirelesssystemarchitecture Networkprotocolstack Hardwareimplementation Application Transport Hostcomputer Network Datalink Baseband Physical Networkinterface RFfrontends 6 Powerconsumption(Contd.) Low-noiseamplifier Antenna interface LocalOscillator (LO) Intermediate Frequency (IF) signalprocessing PA Poweramplifier PhysicalLayer IF/Baseband Conversion LNA Baseband processor MACLayer& above >60%non-display powerconsumed inRF RFtechnologies improve muchslowerthanIC 7 Powerconsumption(Contd.) 1% 6% 8% PA FS 18% 67% Mixer Components Power (mW) Power amplifier (PA) 246 Frequency synthesizer (VCO/FS) 67.5 Mixer 30.3 LNA 20 Baseband Amplifier 5 Source:Lietal,2005 http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1579876 8 Low-noiseamplifier(LNA) • • • • • Bandwidth(sameasthesignal) Gain(~20dB) Linearity(IP3) Noisefigure(1dB) Powerconsumption Circuitpoweroptimization • Majorpowerconsumers Low-noiseamplifier Hugedynamic range105 High dutycycle Antenna interface LocalOscillator (LO) Almostalwayson Intermediate Frequency (IF) signalprocessing IF/Baseband Conversion LNA Baseband processor PA Poweramplifier High power consumption PhysicalLayer MACLayer& above 10 Circuitpoweroptimization(Contd.) • Reducesupplyvoltage – Negativelyimpactamplifierlinearity • Higherintegration – CMOSRF – SoCandSiPintegration • Power-savingmodes 11 Circuitpoweroptimization(Contd.) • Power-savingmodes – Completepoweroff • (Circuitwake-uplatency+networkassociationlatency) ontheorderofseconds – Differentpower-savingmodes • Lesspowersavingbutshortwake-uplatency 12 Power-savingmodes RadioDeepSleep Wake-uplatencyontheorderofmicro seconds Low-noiseamplifier Antenna interface LocalOscillator (LO) Intermediate Frequency (IF) signalprocessing PA Poweramplifier PhysicalLayer IF/Baseband Conversion LNA Baseband processor MACLayer& above 13 Power-savingmodes(Contd.) SleepMode Wake-uplatencyontheorderof milliseconds Low-rateclockwith savednetwork association information Low-noiseamplifier Antenna interface LocalOscillator (LO) Intermediate Frequency (IF) signalprocessing PA Poweramplifier PhysicalLayer IF/Baseband Conversion LNA Baseband processor MACLayer& above 14 Networkpoweroptimization • Usepower-savingmodes – Example:802.11wirelessLAN(WiFi) • Infrastructuremode:Accesspointsandmobilenodes – Example:Cellularnetworks 15 802.11infrastructuremode • Mobilenodesniffsbasedona“ListenInterval” – ListenIntervalismultipleofthe“beaconperiod” • Beaconperiod:typically100ms • DuringaListenInterval – Accesspoint • buffersdataformobilenode • sendsoutatrafficindicationmap(TIM),announcingbuffered data,everybeaconperiod – Mobilenodestaysinpower-savingmode • AfteraListenInterval – MobilenodechecksTIMtoseewhetheritgetsbuffered data – Ifso,send“PS-Poll”askingfordata 16 Buffering/sniffingin802.11 Gast,802.11WirelessNetwork:TheDefinitiveGuide 802.15.1/Bluetooth usessimilarpower-savingprotocols: HoldandSniff modes 17 Cellularnetworks • Discontinuoustransmission(DTX) • Discontinuousreception(DRX) Wirelessenergycost • Connection – Establishment – Maintenance • Transferdata – Transmitvs.receive 19 Energyperbittransfer Oppermann etal.,IEEEComm.Mag.2004 20 Wastefulwirelesscommunication Time Micropowermanagement Spectrum Efficiency-drivencognitiveradio Space Directionalcommunication 21 Spacewaste • Omnitransmissionèhugepowerbypoweramplifier(PA) 22 Timewaste • NetworkBandwidthUnder-Utilization – Modestdataraterequiredbyapplications • IE~1Mbps, MSNvideocall~3Mbps – Bandwidthlimitofwiredlink • 6Mbps DSLathome 1400 Data Size (Byte) 1200 1000 800 600 400 200 0 ) % ( e 80 m it y s 60 u b n i sl 40 av re t 20 n i el d I 0 100 0 0.2 0.4 0.6 Time (s) 0.8 1 User1 Time User2 User3 23 User4 Energy 23 Spectrumwaste 24 Observedfroman802.11guser Energyperbit Distribution ofobserved 802.11gthroughput 1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 1.E+07 Throughout(bps) 25 Temporalwaste Radio Activity 1 0 0 0.2 0.4 0.6 0.8 1 Time(s) 90%oftime&80%ofenergyspentinidlelistening 26 Four802.11glaptopusers,oneweek FundamentalproblemwithCSMA • CSMA:CarrierSenseMultipleAccess – Clientscompeteforairtime • Incomingpacketsareunpredictable 27 FundamentalproblemwithCSMA 28 Micropowermanagement(µPM) • Sleepduringidlelistening • Wakeupintimetocatchretransmission • Monitorthetrafficnottoabuseit • ~30%powerreduction • Noobservedqualitydegradation J.LiuandL.Zhong,"Micropowermanagementofactive802.11interfaces,"inProc.MobiSys’08. 29 Directionalwaste OngoingprojectwithAshutosh Sabharwal Directionalwaste Twowaystorealizedirectionality • Passivedirectionalantennas – Lowcost – fixedbeampatterns Desclos,Mahe,Reed,2001 • Digitalbeamforming – Flexiblebeampatterns – Highcost 32 Phased-arrayantennasystemfromFidelityComtech ChallengeI:Rotation!!! Solution: Don’t getridoftheomni directionalantennas Usemultiple directionalantennas Butcanweselecttheright antennaintime? 33 ChallengeII:Multipathfading 34 ChallengeIII • Canwedoitwithoutchanging theinfrastructure? 35 Characterizingsmartphonerotation • Howmuchdotheyrotate? • Howfastdotheyrotate? • 11HTCG1users,eachoneweek • Logaccelerometerandcompassreadings – 100Hzwhenwirelessinuse 36 Deviceorientationdescribedbythree Eulerangles • θ andφ basedontri-axisaccelerometer • ψ basedontri-axiscompassandθ andφ 37 Rotationisnotthatmuch • <120° persecond θ 0.3 0.2 0.1 ψ 0.4 100ms 1s 10s 0.3 PDF PDF 0.3 0.4 100ms 1s 10s PDF 0.4 φ 0.2 0.1 100ms 1s 10s 0.2 0.1 0 -4 -3 -2 -1 0 1 2 3 10 10 10 10 10 10 10 10 0 -4 -3 -2 -1 0 1 2 3 10 10 10 10 10 10 10 10 0 -4 -3 -2 -1 0 1 2 3 10 10 10 10 10 10 10 10 Rotational speed(°/s) Rotational speed(°/s) Rotational speed(°/s) 38 Directionalityindoor 5dBi 8dBi 39 8dBiantenna 5dBiantenna Measurementsetup • RSSImeasuredatbothends Datapackets ACKpackets 41 Directional channel still reciprocal NLOS ind. / 5dBi antenna RSS(dBm) -20 -30 -40 Dir-Client Dir-AP Omni-Client Omni-AP -50 -60 0 60 120 180 240 300 360 Direction(°) 42 Directional beats omni close to half of the time 5dBi 30 total time(%) 25 20 15 10 5 0 [0,0.1) [0.1,1) [1,10) [10,inf) superiority intervals(s) Fieldcollectedrotationtracesreplayed 43 RSSispredictable(toabout100ms) 5dBi Error(dB) 100 Zero order First order 1 0.01 10ms 100ms 1s 10s Prediction Intervals(s) 44 Multi-directionalantennadesign(MiDAS) • OneRFchain,oneomniantenna,multipledirectionalantennas • Directionalant.onlyusedfordatatransmitandACKReception – Standardcompliance – Tradeoffbetweenriskandbenefit 45 Packet-based antenna selection • Assessanantennabyreceivingapacketwithit – Leveragingchannelreciprocity • Continuouslyassesstheselectedantenna • Findoutthebestantennabyassessingthemone byone – Potentialriskofmissingpackets • Staywithomni antennawhenRSSchanges rapidly • Nochangein802.11networkinfrastructure 46 Symbol-basedantennaselection • AssessallantennasthroughaseriesofPHYsymbols – SimilartoMIMOantennaselection • NeedshelpfromPHYlayer Antenna training packet Regular packet SEL ACK 47 Tracebasedevaluation • Rotationtracesreplayedonthemotor • RSSItracescollectedforallantennas • Algorithmsevaluatedontracesoffline -45 RSS(dB) Dir 3 -50 Omni -55 -60 0 Dir 1 Dir 3 5 10 15 time(second) 20 48 Anearlyprototype 1 omni antenna 3 directional antennas WARP Laptop Controllable motor FinalistofMobiCom’08BestStudentDemo 49 Thebusierthetraffic,thebetter 6 Gain(dB) 5 Upper bound Symbol-based Packet-based 4 3 2 1 0 10ms 100ms 1s 10s Average Packet Interval 50 Two 5dBiantennasenough 6 Gain(dB) 5 Upper bound Symbol-based Packet-based 4 3 2 1 0 three two-opp two-adj one Antenna Configuration 51 Two5dBi antennasenough NLOS ind. / 5dBi antenna 6 Upper bound Symbol-based Packet-based 4 -30 -40 Dir-Client Dir-AP Omni-Client Omni-AP -50 -60 0 60 120 180 240 300 360 Direction(°) 3 NLOS ind. / 8dBi antenna -20 2 RSS(dBm) Gain(dB) 5 RSS(dBm) -20 1 0 5dBi 8dBi Antenna Gain -30 -40 Dir-Client Dir-AP Omni-Client Omni-AP -50 -60 0 60 120 180 240 300 360 Direction(°) 52 Real-timeexperiments:3dBgain Avg. RSS(dB) -45 Omni Multi antenna -60 -75 NLOS ind. LOS ind. Environment • • • • Packet-basedantennaselection Three5dBiantennas Continuoustraffic(1400bytepackets) Fieldcollectedrotationtrace 53 Throughputimprovement Throughput(Mbps) 4 Omni Multi antenna 3 2 1 0 NLOS ind. LOS ind. Environment 54 SNRvs.transmissionrate(802.11a) Goodput (Mbps) 35 30 25 20 15 10 6Mbps 9Mbps 12Mbps 18Mbps 24Mbps 36Mbps 48Mbps 54Mbps 5 0 0 10 20 30 SNR (dB) (D.Qiao,S.Choi,andK.Shin,2002) 55 MiDAS+rate adaptation+power control • RecallthatRSSisquitepredictableupto100ms 200 Goodput Gain-Upperbound Goodput Gain-MiDAS TXpowerreduction-Upper bound TXpowerreduction-MiDAS % 150 100 50 0 0 10 20 30 40 OmniSNR(dB) 56 Protocolwaste Cellularnetwork Transmission efficiency Connection Availability WLAN(Wi-Fi) Howtocombinethestrengthofboth Wi-FiandCellularnetwork? EstimateWi-Finetworkcondition WITHOUTpoweringonWi-Fiinterface 58 UsecontexttopredictWiFi availability • Visiblecellularnetworktowers • Motion • Timeoftheday,dayoftheweek Statisticallearning Context Wi-Fi Conditions P(WiFi|Context) AhmadRahmatiandLinZhong,"ContextforWireless:Context-sensitiveenergy-efficientwirelessdatatransfer," inProc.MobiSys’07. JournalversionwithnewresultstoappearinIEEETMC 59 Cellularnetworkoffersclues Cellularnetworkoffersclues Wedon’tmovethatmuch 50% 40% 30% 20% 10% 0% moving (1,5] (5,10] (10,30] (30,60] (60,120] (120,inf) Lengthofmotionlessperiod(minute) Datacollectedfrom 2smartphone users2006 Shoehorned smartphonewith accelerometer 62 Ourlifeisrepetitive ProbabilityofsameWi-Fiavailability (normalizedautocorreletaion) 1 0.9 0.8 0.7 0.6 0.5 0 1 2 3 4 Time(days) Datacollectedfrom 11smartphone users 63 PredictionaccuracyofWi-Fi availability WiFi availabilityisHIGHLYpredictable 1 0.9 0.8 0.7 0.6 0.5 0 120 240 360 Time(minutes) 480 600 • Application – MobileEKGmonitoring – 35%batterylifeimprovement(12to17hours) 64