POWER CONSTRAINTS <strong>OF</strong> WIRELESS SENSOR NODESBy Dave Freeman and Sriram Narayanan, Texas InstrumentsRecent developments in hardware and software designsfor wireless-sensor-network nodes motivate the need forsystem-level design methods. Figure A shows a network,along with the subsystems of each node. Considerationsfor ease of deployment and cost of installation requireINTERNETOREXTERNALNETWORKULTRA-WIDEBANDZIGBEEWIRELESSLANCONVERGENCEBASE STATIONWIRELESSLANMICRO-CONTROLLERSENSORSNODE 1WIRELESSLANMICRO-CONTROLLERSENSORSNODE Nthat the nodes be able to communicate wirelessly. Toreduce communication overhead and improve responsetime, the nodes should be able to locally process sensordata and control actuators. Performing routine maintenance,such as replacing batteries on a large numberof nodes, may be prohibitively expensive. Sensor nodesshould also last for several years by relying only onstored or harvested energy.The choice of sensors, radio, and microcontrollerdepends on the nature of the application.A sensor network in an office environmentaddresses applications such as energy management,security, or resource planning.TABLE A ACCRUAL RATESSetting Intensity (lux) Power density (μW/cm 2 )Conference 300 7.5roomOffice 500 12.5Hallway 400 10ENERGYSTORAGEENERGYSTORAGEACTUATORENERGYSTORAGEENERGYSTORAGEACTUATORFigure A Harvested energy powers sensor nodes that make autonomous decisionsabout their environment and can communicate using multiple protocols.Light energy is usually the most abundant form of ambientenergy in indoor environments. Modern solar cellscomprising amorphous silicon generate approximately 5μW/cm 2 when illuminated by a 200-lux fluorescent-lightsource. Table A lists estimates of energy-accrual ratesand suggests that a 10-cm 2 solar cell may generate70 to 120 μW.Microthermoelectric generators use a gradientin temperature to produce electricalenergy. To generate a power density of 15 μW/cm 3 , however, thermal harvesters may need athermal gradient of approximately 10°C. Manyapplications, particularly indoor environments,do not experience large swings in temperature.Thus, thermal harvesters have limited applicability.Today’s vibrational-energy harvestersneed acceleration of approximately 1.75 to2g—magnitudes usually absent in indoor environments—toproduce 60 μW of power.With limited onboard capacity to storeenergy and limited opportunities to harvestambient energy, the sensor node must befrugal in energy use. Consider, for example, a100-mAhr battery and a solar cell that accrues70 μW for 50% of a 10-year node lifetime. Thisnode must operate its various subsystems andconsume no more than 39 μW of average power.The microcontroller, radio, sensors, and actuatorsexhibit dramatically different power and performancecharacteristics. Meeting the system’s power budgetrequires the sensor node to optimally manage its subsystems.Modern low-power microcontrollers consumeapproximately 345 μW of peak power to operate at aTABLE B LOW-POWER ARCHITECTURESFrequency(GHz)Datarate(Mbps)Nominalrange(m)Peakpower(mW)Energyper bit(nJ/bit)Bluetooth 2.4 1 1 to 10 92.4 923.1 to 10.6 1 to 27 1 to 30 Less than10IEEE802.15.4aultrawidebandApproximately1Zigbee 2.4 0.25 10 to 100 29.7 119Wireless 2.4 54 100 851 15LANsuper in capacity, but they also leak likesieves,” says Skurla. It’s also difficult toget precise specifications for leakage andshelf-life characteristics. He says thatthe company bought and tested a bunchof supercapacitors that were availableon the market, running the tests overdays and weeks. The leakage current isSUPERCAPACITORSARE SUPER IN CAPAC-ITY, BUT <strong>THE</strong>Y ALSOLEAK LIKE SIEVES.proportional to the capacity: As the capacitanceincreases, the leakage currentalso increases. “What you really want isa 10F supercap with the leakage-currentcharacteristic of a 0.1F supercap,” saysSkurla. “They still have about 11-timesmore leakage than we like to see for anenergy-harvesting application.”32 EDN | FEBRUARY 3, 2011
clock speed of about 1 MHz. Given that the requirementsfor sensor processing are often modest, youcan use a microcontroller with an aggressive dutycycle—less than 1%, for example—to reduce averagepower consumption.Sensor nodes usually communicate informationabout physical phenomena and related control messagesat relatively low rates. Table B summarizesthe salient features of some key low-power wirelesscommunicationtechnologies. The power-consumptionquantities in the table serve only as general guidelinesfor system design. As transceiver designs evolve, theyconsume less power. When choosing a transceiverarchitecture, it is important to consider all aspectsof the design. A wireless-LAN transceiver consumeslower energy per bit than does a Zigbee transceiver,but the LAN transceiver has a higher data rate andconsumes more peak power.Sensors that may find use in indoor applicationsinclude thermometers, humidity sensors, microphones,and passive infrared sensors. Today’s temperatureand humidity sensors and microphonesconsume approximately 70 to 80 μW of peak power.Passive infrared sensors that can detect human activitytypically consume peak power of 100 to 500 μW.Temperature and humidity sensors monitor slowlyvarying phenomena and can operate on low dutycycles, but you cannot power off other sensors thatdetect motion without compromising detection performance.In many applications, the sensor demandsmore energy than the processing of the data or wirelesscommunication. Therefore, meeting the systempower budget requires innovation in the way you managethe sensors.The lack of an adequate power supply and energyremains a formidable challenge to implementing wirelesssensor networks, despite advancements in computation,communication, and sensing. Technologicaladvancements in energy harvesting and storagecontinue to ease power constraints, but the demandsof the end application continue to push their limits.Bridging this persistent power gap requires a systemlevel-designapproach that optimally trades off performancefor energy savings and that guarantees aminimum quality of service. Future wireless sensornodes will autonomously adapt to time-varying-applicationdemands and energy availability.See the Web version of this article at www.edn.com/110203cs for a list of references used in writingthis sidebar.FOR MORE INFORMATIONAdvanced LinearDeviceswww.aldinc.comAnalog Deviceswww.analog.comCap-XXwww.cap-xx.comHitachi Maxellwww.maxell.co.jpIoxuswww.ioxus.comJM Energywww.jmenergy.co.jpLinear Technologywww.linear.comOn Semiconductorwww.onsemi.comTexas Instrumentswww.ti.com