DESIGNANDIMPLEMENTATIONOFEFFICIENTENERGYHARVESTINGCIRCUITSFORULTRALOWPOWERANDIMPACTENERGYAPPLICATIONSByTaoFengADISSERTATIONSubmittedtoMichiganStateUniversityinpartialoftherequirementsforthedegreeofElectricalEngineeringŒDoctorofPhilosophy2016ABSTRACTDESIGNANDIMPLEMENTATIONOFEFFICIENTENERGYHARVESTINGCIRCUITSFORULTRALOWPOWERANDIMPACTENERGYAPPLICATIONSByTaoFengThebattery-poweredelectronicsystems,suchaswearableandmobiledevice,wirelesssensors,implantablebio-medicaldevices,etc.,havewidelyappearedinourdailylifetoimprovethelifequalityandworkefy.However,thedependenceonbatterieshasbroughtgreatchallengestocurrentelectronicdevicesinmanyaspects,suchassystemminiaturization,massivedeployment,devicelifetimeandenvironmentalpollution,etc.Toaddresstheseissues,theideaofharvestingtheenergyfromtheambientenvironmentinsteadofthebatteryhasbeenproposed.Variousenergyharvestingtechnologieshavebeenimplementedasthealternativesolutionsforthebattery-poweredelectronicdevices.However,lowPCE(powerconversionefy)preventstheseenergyhar-vestingtechnologiesfrombeingwidelyadoptedinpracticalapplications.Anothermajorconcernforenergyharvestingtechnologiesisthepowerthresholdwhichistheminimumenergythatcanbeharvestedbytheenergyharvestingsystems.Actuallythepowerthresholdsetsthelowerlimitofthecapabilityforenergyharvestingsystems,therefore,overcomingthethresholdeffectcouldimprovetheefyoftheenergyharvestingsystems.Theobjectiveofthisdisserta-tionistoinvestigatethepowerconversionefyandthepowerthresholdofenergyharvestingtechnologies,andthenproposeseveralsolutionstoovercometheseissues.Firstly,theRFenergyharvestingfront-endwhichconsistsofanantenna,matchingnetworkandisexaminedduetopervasivewirelesspowerintheenvironment.TheconventionalmatchingnetworkinthenearRFenergyharvestingfront-endneedstotrade-offbetweenthecouplingcoefandQ-factor.Inchapter2,ahigh-Qseriesresonantmatchingnetworkisproposedsothatbothofhighcouplingcoefandhigh-Qmatchingnetworkcanbeobtainedatthesametime,thereforethepowerconversionefyofthenearRFenergyharvestingcanbeimproved.Besidesthematchingnetwork,thealsoplaysakeyroleinRFenergyharvestingsystems.ThemajorbottleneckoftheRFisnearlyzeropowertransferefywhentheinputpowerisbelowthepowerthreshold.Suchoperatingregioniscalled"deadzone"fortheRF.Tofreethedeadzone,ahybridtechniquewhichutilizesbothofRFpowerandvibrationpowertoefharvesttheenergybelowthethresholdlevelisproposedinchapter3.Thistechniqueutilizeslow-frequencyandlow-powerpiezoelectricsignalasthebiasingcircuitoftheRFsothatthethresholdvoltageoftheRFrecanbeoptimizedandthuseffectivepowerthresholdisreduced.InadditiontotheRFenergyharvesting,vibrationenergyharvestingisalsocrucial,especiallyforthesensorsembeddedinsidethestructuraland/orunderneaththeground.Comparedtocontin-uousRFenergy,thevibrationenergyisimpulsivesothatitisdiffortheharvestertocapturethewholeenergyduringashortpulseperiod.However,iftheimpulsiveenergycanbestretchedinthetimedomainwhilekeepingthetotalenergyconstant,theharvestingdurationwillbecomelongersothatitispossiblefortheharvestertoharvestthewholeenergy.Basedonthisidea,anon-linearcompressivecircuit,named"time-dilationcircuit",isproposedinchapter4.Thistechniquecanquicklyrespondtotheimpulsiveenergysothatthedynamicharvestingrangeisextended.Comparingtoelectromagneticpropagation,acousticpropagationhaslessattenuationintheconductivemedia.Thus,theultrasonicenergyharvestingattractsmoreinterestforthesensorsem-beddedinsidecompositematerialsorhumanbody.Furthermore,theuseofultrasoundalsoallowsminiaturizationoftheembeddedtelemetrysystembyrelaxingthetransducersize.Inchapter5,aCMOSsystem-on-chip(SOC)ultrasonicreceivertagisimplementedtoinvestigatetheultrasonicenergyharvestingandtelemetry.Thissystemcannotonlyharvestitsoperationalenergydirectlyfromtheacousticinterrogationsignal,butalsoachievebi-directionalcommunicationusingacous-ticback-scattering,thusitissuccessfullyshownthattheultrasonicenergyharvestingandtelemetryisapropersolutionforsuchapplications.Finally,thecontributionsofthisdissertationandopenproblemsforfutureworkaresumma-rizedinchapter6.Thisdissertationisdedicatedtomyfamily:mydearmotherYingLiu,mypassedfatherYucaiFeng,mywifeLingSunandmylovelysonKevinJinyuanFengaswellasmyparents-in-lawYuanbiSunandYurongMa.ivACKNOWLEDGMENTSDuringmytimeatMSUIhadmanyuniqueandwonderfulexperiences,suchasdiscussingtheresearchwithmyadvisorandlabmates,learningnewsubjectsfromdistinguishedprofessorsandexcellentclassmates,participatingvariousimpressiveactivitieswithmyfriendsandfamily.ThusIwouldliketoexpressmyacknowledgmentstoallofthepeoplewhohavebeenwithmeduringmytimeatMSU,especiallytothosewhohavebeenpersistentlysupportingmyresearchwork.Firstandforemost,IwouldliketothankandexpressmydeepestappreciationandgratitudetomyresearchadvisorProfessorShantanuChakrabartty.Hisclearguidanceandgreatinsightwereinvaluabletomyresearch.Hisgreatpassiontotheresearchalwaysinspiredmetoexplorenewopportunitiesintheresearchjourney.Inaddition,hewasalwayshelpfulwheneverIfacedanyresearchproblems.IalotfromhismentorshipandguidanceandIreallyappreciatehisgreathelpineditingresearchpapers.Thankyouforgrantingmyresearchandenablingmetosuccessfullytheresearchwork.Inaddition,Iamalsoverythankfultoothercommitteemembers,Prof.TimHogan,Prof.PremChahal,Prof.NizarLajnef,Prof.MiZhang.Thankallofyoutoprovidevaluableadvicesandsuggestionstomyresearchwork.Itisknownthattheresearchworkisnotalwaysgoingsmoothly.However,Iwasveryluckytohavemylabmates'supportandencouragewhenIfeltdepressedintheresearch.Thetimewespenttogetherismemorableandourfriendshipisthemostcherishedthinginmylife.Finally,Ihavetomentionmyfamilywithcomplexmoodsincethemosthappiestthing-mysonwasbornin2014,andthemostsaddestthing-mydadwasgonein2015,happenedinmyPhDlife.Thecomingofmykidbroughtunbelievablehappinesstomyfamilybuttheleavingofmyfathermadeusdeeplyfellintosorrow.Thankmymotherandmywifeaswellasmyparents-in-lawtostandbehindmeinthepastyears.Withoutthem,Icouldn'timagehowmylifewouldbe.ThefamilyisalwaysthemostimportantthinginmylifeandIwillappreciateeverydaywiththem.vTABLEOFCONTENTSLISTOFTABLES.......................................viiiLISTOFFIGURES.......................................ixCHAPTER1INTRODUCTION...............................11.1IntroductionofEnergyHarvesting..........................11.2EnergyHarvestingSystemModel..........................31.3ResearchMotivation.................................61.4ThesisContributionsandOrganization........................8CHAPTER2HIGH-QSERIESRESONANTZ-MATCHINGNETWORK.........102.1Introduction......................................102.2ParallelResonantMethod...............................102.3ProposedSeriesResonantStrategy..........................132.4MeasurementResults.................................162.5Summary.......................................19CHAPTER3HYBRIDRECTIFICATIONTECHNIQUEFORRFENERGYHARVEST-ING.......................................203.1Introduction......................................203.2VCEandPCEofTheConventionalVoltage................203.3State-of-Art.................................243.3.1Passive...............................243.3.2Active...............................253.4ProposedPZTAssistedRF.........................273.4.1ThresholdCompensationInvestigation....................293.4.2CircuitImplementation............................313.4.3DeviceSizeAnalysisandOptimization...................333.4.4MeasurementResults.............................363.5Summary.......................................41CHAPTER4TIMEDILATIONTECHNIQUEFORIMPULSIVEENERGYHARVEST-ING.......................................434.1Introduction......................................434.2ModelingofthePiezoelectricTransducer......................454.3ProposedTime-dilationCircuit............................464.4DifferentialandRotationalImpact..................494.5PFGbasedSelf-poweredEnergyMeasurementandData-logging..........524.6MeasurementResults.................................544.7Summary.......................................60viCHAPTER5ACMOSSYSTEM-ON-CHIPFORPASSIVE,NEAR-FIELDULTRA-SONICENERGYHARVESTINGANDBACK-TELEMETRY.......615.1Introduction......................................615.2SystemOverview...................................625.2.1Ultrasonicpowertransferandtelemetrymodel...............625.2.2SystemArchitecture.............................665.3CircuitDesignoftheUltrasonicTag.........................675.3.1PowerManagementCircuits.........................675.3.2DataRecoveryCircuit............................725.3.3DigitalBasebandandManchesterEncoder..................735.3.4SensorDataAcquisitionCircuitry......................745.4MeasurementResults.................................765.5Summary.......................................78CHAPTER6CONCLUSION................................806.1SummaryofContributions..............................806.2OpenProblems....................................81BIBLIOGRAPHY........................................83viiLISTOFTABLESTable1.1Macro-scalev.s.micro-scaleenergyharvestingcomparison............2Table1.2Powerdensityoffourpopularenergysources[1]..................2Table1.3ElectricalCharacteristicsoffourpopularenergysources[2]............4Table2.1Coil.................................18Table4.1Equivalentcircuitparametersusedinsimulation..................46Table4.2CommandDescription...............................55Table4.3InterfaceoftheSelf-poweredIC..................56Table4.4Technicaldetailsofthehelmet...........................56Table5.1Materialparametersusedforlinksimulations....................65Table5.2MainforProposedUltrasonicReceiverIC..............77Table5.3PerformanceComparison..............................79viiiLISTOFFIGURESFigure1.1Generalenergyharvestingsystemmodel......................3Figure1.2ConventionalEnergyProcessingCircuits.....................4Figure1.3IllustrationofdeadzoneforAC-DCconversion..................5Figure1.4IllustrationofMPPTbyusingsolarcellapplication:(a)Typicalcircuitmodelforsolarcell;(b)Outputcurrentv.s.outputvoltage;(c)Outputpowerv.s.outputvoltage[3]..................................6Figure1.5(a)Astructuralhealthsensorembeddedundertheground;(b)Proposedhy-bridenergyharvestingmethod...........................7Figure1.6(a)Illustrationofstretchingimpulsiveenergyintimedomain;(b)Proposedtime-dilationmethod................................8Figure2.1Blockdiagramofatypicalinductivelycoupledwirelessenergyscavengingsystem........................................11Figure2.2Parallelmodelofinductivecoupling........................11Figure2.3Mathematicalmodelofthetranspondercircuitusingseriesmatchingstrat-egy:(a)Parallelformoftheloadand(b)theirequivalentseriesform;(c)showsanmoreconvenientmethodtotunethecircuitbyreplacingsingleC2withaseriesL1C2..................................14Figure2.4Improvingfactorsof(a)voltagegainfVand(b)powertransferefyfPv.s.themultiplicationofQL............................15Figure2.5Experimentalsetupusedforpoweringdistanceevaluation............16Figure2.6Comparisonofthevoltagegeneratedacrossaconstantloadusingparallelandproposedmatching...............................17Figure2.7ComparisonofthevoltagegeneratedbySchottky-multiplierusingtheparal-lelandtheproposedmatching:(a)18-stagemultiplierusedforanalogVDD;(b)12-stagemultiplierusedfordigitalVDD....................18Figure3.1Typicalhalf-wave(a)conventional;(b)withidealcompensatedvoltage.21Figure3.2VoltageandcurrentwaveformofthepasstransistorMP..............22ixFigure3.3Threepopularpassive(a)Differentialcross-coupled[4];(b)Pseudogate[5];(c)InternalVTHcancellation[6]..25Figure3.4(a)Comparatorbasedactivediode[7];(b)Offsetcontrolhighefyac-tive[8];(c)Illustrationofleakagecurrentissueinactive[8]...26Figure3.5Conceptualhybridrectifer[9]...........................27Figure3.6Theconceptofahybridvoltage-multiplieranditspotentialapplicationtoastructuralhealthmonitoringsensorthatscavengesRFandvibrationenergy...28Figure3.7Complementarycross-coupled.......................29Figure3.8DCboostingeffectsimulationsetup(VRF=0.3Vpkpk,fRF=13.56MHz,RL=20KW,CL=1nF)................................30Figure3.9OptimalbiasvoltagesversusinputRFsignal...................30Figure3.10ProposedHR:(a)mainrect(b)DCbiasgeneratorforNMOSgateter-minalsand(c)forPMOSgateterminals......................32Figure3.11Simulatedwaveformofcriticalnodesinbiasvoltagegenerators.Fpiezo=10KHz,Vpiezo=1V,FRF=13.56MHz,VRF=0.3Vpkpk...............33Figure3.12VOUTversusdevicesizeofthemain(VRF=0.3Vpkpk,L=0.1mm,CC=10pFandIOUT=10mA)...........................34Figure3.13EquivalentcircuitoftheupperbranchofNMOSbiasgenerator..........34Figure3.14SixstageHR....................................36Figure3.15Micrographofvoltagemultiplierchip.......................37Figure3.16Measurementsetupofthetestchip.........................38Figure3.17DCoutputvoltagevesusinputRFvoltage.....................39Figure3.18ComparisonofmeasuredPCEversusinputpowerPIN(fRF=13.56MHz,FPZT=10kHzandVPZT=2V)..........................40Figure3.19DCoutputvoltageoftheHRversePZTinputamplitudeatdifferentPZTfrequency(VRF;pkpk=300mV,FRF13.56MHzandRL=330KW)........41Figure4.1(a)PoweringandsensingmechanismofthePFGhead-impactmonitoringsensor;(b)time-dilationapproachwheretheimpactenergyisspreadintime;and(c)compressiveresponseofatime-dilatedimpactmonitoringsensor.....44xFigure4.2(a)Equivalentcircuitofthepiezoelectrictransducer;(b)Simulatedpulseresponseofthepiezoelectrictransducerwith1MWload..............45Figure4.3Equivalentcircuitofaself-poweredsensor:(a)without,and(b)withthetime-dilationcircuit.................................47Figure4.4Comparisonbetweentheoutputvoltagesofthefullwavewith(redline)andwithout(blueline)time-dilationcircuitswithrespecttotheinputpulseof(a)100Vand(b)300V..........................48Figure4.5Comparisonbetweenthestoredenergywith(redline)andwithout(blueline)time-dilationcircuitastheinputpulselevelincreases...............49Figure4.6(a)Illustrationofthelinearandrotationalaccelerations;(b)Measuredpulseresponsesofthesensorsinthreepositionswithrespecttothedirectandlat-eralimpacts,respectively..............................50Figure4.7(a)Apairofpiezoelectrictransducersareconnectedindifferentialrationtotesttherotationalacceleration;(b)Thesimulatedoutputvoltageoftheforthecommon-modeanddifferential-modeinputsignalwhenusingdifferentialconnectedpiezoelectrictransducerswithtime-dilationcircuit.51Figure4.8Schematicofthedata-loggingcircuitwithaoating-gatebasedlinearinjec-torandaspikinganalog-to-timeconverter....................53Figure4.9(a)SystemarchitectureforthePFGsensor.(b)MicrographofthesensorICintegratingdifferentmodules:1.FloatingGateArray;2.DigitalDecoder;3.TunnelingVoltageCharge-pump;4.LevelShifter;5.InjectionControl;6.DiodeProtection&7.VoltageReferences;8.TunnelingCharge-pump;9.InjectionCharge-pump;10.RingOscillator;11.Analog-to-timeConverter;12.Supportingcircuitry,power-onreset,buffers,etc.........55Figure4.10Experimentalsetupusedinthedrop-tests:aCOTSfootballhelmetwithem-beddedIntegratedPFGsensorprototypes(inset)..................56Figure4.11MeasuredresultsshowingVoutforthesensorwith:(a)notime-dilationcir-cuit;(b)time-dilationcircuitwithCS=50nF;and(c)time-dilationcircuitwithCS=1mF...................................57Figure4.12SignalrecordedattheoutputofaPZT-5Hpiezoelectrictransducer(10MWload)whenthehelmetisdroppedfrom1foot(heightA)and2feet(heightB),respectively...................................58Figure4.13Measuredoutputfromthreeofthesensorchannelswhenthehelmetisre-peatedlydroppedfrom(a)1foot;(b)differentheights...............58xiFigure4.14(a)Measuredimpactdistributionusingthreesensors:sensor1isinthefront,sensor3isinthesideandsensor2isinbetween;(b)RotationalimpacttestusingadifferentialPFGsensor...........................59Figure5.1Illustrationofanultrasonicbasedtelemetrysystemusingacousticcouplingthroughthemetalbarrier..............................62Figure5.2UltrasonicpoweringandcommunicationsystembasedonMasonmodel:(a)systemdiagram;(b)equivalentcircuitmodel...................63Figure5.3ThesimulationofpowertransferefybasedonMasonmodelasafunc-tionofmetalthickness...............................65Figure5.4Proposedultrasoniccommunicationsystem....................66Figure5.5Powermanagementmoduleswhichincludeavoltagemultiplier,avoltagelimiterandaregulator................................66Figure5.6Themeasurementresultsfordifferenttypesofvoltagemultipliers(12-stage,18-stageand24-stage)undertheloadresistorof10MWand1MW........67Figure5.7MeasurementsofLDO:(a)dropoutvoltagefordigitalblock;(b)dropoutvoltageforanalogblock;(c)lineregulationforanalogblockwhenVIN=0Vto10V;(d)linetransientresponseforanalogblockwithIL=0.4mA,VIN=0Vto6V;(e)linetransientresponseforanalogblockwithIL=0.4mA,VIN=6Vto0V;(f)ripplerejectionforanalogblockwithIL=0.4mA;(g)linetransientresponseforanalogblockwithIL=400mA,VIN=0Vto6V;(h)linetransientresponseforanalogblockwithIL=400mA,VIN=6Vto0V;(i)ripplerejectionforanalogblockwithIL=400mA...............68Figure5.8(a)Blockdiagramand(b)detailedcircuitdesignofdatarecoverycircuit.....69Figure5.9Communicationprotocol:(a)PIE;(b)Manchester.................71Figure5.10(a)currentreference;(b)ringoscillatorwithpulseshaping;(c)power-onresetcircuit;(d)8-bitsingle-slopeADC......................72Figure5.11Measuredresultofthe8-bitADC.........................75Figure5.12Measurementsetupoftheultrasoniccommunicationsystem:(a)theultra-sonicreaderconsistsofXilinxFPGAandTIAnalogFrontEnd(AFE);(b)TxPZTandTxPZTareseparatedby2mmthickAlmetalbarrier;(c)micro-graphofthefabricatedultrasonictagIC......................75Figure5.13(a)MeasuredtransmitpowerfromtheTxPZTand(b)receivepowerfromTxPZT.......................................76xiiFigure5.14MeasuredresultsshowingprotocolsynchronizationbetweenthereaderandthetagIC......................................76xiiiCHAPTER1INTRODUCTION1.1IntroductionofEnergyHarvestingEnergyharvesting,alsoknownaspowerharvestingorenergyscavenging,isgenerallyastheprocessofscavengingvariousenergiesfromambientenvironment.Theharvestedenergiesarethenaccumulatedandstoredforeitherimmediateorlateruse[10].TheconceptofenergyharvestingisnotbrandnewanditshistorycandatebacktoB.C.whenthewindmillsandwatermillswereinvented.Suchkindsofenergyharvestingmethodscanbeasmacro-scaleenergyharvestingsincetheenergysourcesarerenewableenergysothattheavailablepowerlevelisusuallyabovekilowatt.Inrecentyears,astherapiddevelopmentofsemiconductordevicesandintegratedcircuits,variousenergyharvestingtechnologieshavebeeninventedandimplemented,suchaselectromagnetic,thermoelectricandvibrationenergyharvesting,etc.Thesetechnologieshavedrivenenergyharvestingintoamicro-scaleleveltomeetthegrowingenergydemandoftheworld.Differentfromitsmacro-scalecounterpartofwhichthecommongoalistofeedtheutilitygrid,themicro-scaleenergyharvestingtechnologiesaimtoprovidetheperpetualdevicewithoutfeedingthegrid.Itisknownthatbattery-poweredsolutionisalsoabletofreetheelectronicsystemsfrompowercord.Butthelifetimeofbattery-poweredsolutionislimited(usuallyseveralyears),sothemaintenanceeffortcouldexponentiallyincreaseasthesensornetworkexpands.Energyharvestingisaverypracticalandusefulsupplementarymethodforlifetimeextensionofthebattery-poweredsystems.Itevencanreplacethebatteriessothatthelifetimeofthesystemonlydependsonthede-viceitself.Inaddition,environmentalpollutionisanincreasingconcerninoursociety.Bygettingridofthebattery,energyharvestingisaverypromisingsolutiontotheenvironmentprotection.Lastbutnotleast,theelectronicsystemcanbefurtherminiaturizedwithoutthebattery,somoreandmoreapplicationsandmarketscouldbeexplored,suchaswearableandimplantabledevices,etc.1Table1.1Macro-scalev.s.micro-scaleenergyharvestingcomparison.Macro-scaleMicro-scaleEnergySourcesRenewableenergy(e.g.,so-lar,wind,etc.)Environmentalenergy(e.g.,vibration,electromagneticra-diation,etc.)EnergyLevelabovekilowattbelowmilliwattPurposeReducethedependencetonaturalresources(suchasoil,gas,coal,etc.)Prolongsystem'slifetimeandultimatelyprovideperpetualdevicesUbiquitousenvironmentalenergysources,suchassolar,thermal,vibrationandelectromag-neticenergy,havebeenexploredanddevelopedinmanyresearchstudiesduringthelastdecades.Comparedtomacro-scaleenergytransducerswhicharecharacterizedbytheenergydensity,themicro-scaleenergytransducersusethepowerdensityasthemainTable.1.2showsthepowerdensityoffourpopularenergysources.Itisobviousthatthesolarenergyhasthehighestpowerdensityandisquitesuitableforoutdoorapplications,butthedisadvantageisalsoquitecon-spicuousduetotheneedoflight.Thermalenergyhasmediumpowerdensityandissuitablefortheindustryapplications,especiallyforthemachineswithhugetemperaturedifference,butitisnotapplicablefortheconstanttemperatureenvironmentsuchasofbuilding.Vibrationenergyandelectromagneticenergyareusuallyubiquitousintheenvironment,butitisagreatchallengefortheenergyharvesterstocapturethemduetothesmallpowerdensity.ThustheuseoftheenergyTable1.2Powerdensityoffourpopularenergysources[1].EnergySourcesPowerDensitySolarOutdoors10mW/cm2;Indoors10mW/cm2ThermalHuman25mW/cm2;Industrial1-10mW/cm2VibrationHuman4mW/cm2;Industrial100mW/cm2ElectromagneticAmbientSource(GSM):0.01mW/cm2(100m)-0.3mW/cm2(25m);DedicatedSource(EIRP4W):1.27mW/cm2(5m)2Figure1.1Generalenergyharvestingsystemmodel.harvestingtechnologiesdependonthepracticalapplicationsandavailableenergysources.1.2EnergyHarvestingSystemModelAnyenergyharvestingsystemcanbemodeledasatransducerconnectingwiththeloadimpedancewhichrepresentsthepowerconsumptionoftheloadsystem,asshowninFig.1.1.ThetransducerconvertstheambientenergytoelectricalenergywhichisrepresentedbyvoltagesourceVOC.ThetransduceritselfhasanequivalentelectricaloutputimpedanceZSwhichvarieswithsourcevoltageVOCanditsoperatingfrequencyfS.AsitcanbeseeninTable.1.3,eachsourcehasawiderangeofopencircuitvoltageanditsoperatingfrequencycouldalsovary,sothesourceimpedanceisnoted.ItisknownthatZLandZSshouldbeaconjugatedmatchformaximumpowertransfer.However,ZLisalsovariablebecausethepowerconsumptionoftheloadsystemisnotalwaysconstant.Inaddition,ZLisalsoaffectedbythesourcevoltage.Therefore,howtoefharvesttheenergyisthemajortaskforeveryenergyharvestingsystem.ConventionalenergyprocessingcircuitsconsistofAC-DCconversionandDC-DCconver-sionasshowninFig.1.2.TheACpowersourcessuchasvibrationalandelectromagneticenergyshouldfeedtheAC-DCconverterandthenperformDC-DCconversion,whiletheDCpowersourcessuchasthermalandsolarenergycanbedirectlyDC-DCconverted.ForAC-DCconver-3Table1.3ElectricalCharacteristicsoffourpopularenergysources[2].EnergySourcesTypicalOpenCir-cuitVoltageVOCOperatingFre-quencyfSTypicalSourceImpedanceZSSolar0.5V˘5VDCVariableimpedance:1kW˘100kWThermal10mV˘10VDCConstantresistiveimpedance:1W˘100WVibration10V˘50V0.1Hz˘1KHzConstantcapacitiveimpedance:10kW˘100kWElectromagnetic100mV˘5V100KHz˘5GHzConstantinductiveimpedance:1kW˘10kWFigure1.2ConventionalEnergyProcessingCircuits.sion,althoughthereexistmanydifferentcircuittypologies,thebasicpartactuallycanbethoughtasasimplehalf-wave,whichconsistsofadiodeandastoragecapacitor.Oncetheinputvoltageisabovethediode'sthresholdvoltage,thecurrentcanchargethestoragecapacitor.How-ever,belowthethreshold,thereversecurrentiscomparabletotheforwardcurrent,leadingtoaverylowpowertransferefy,thusthisareaisalsocalleddead-zonefortheAD-DCconver-sion,asshowninFig.1.3.ThesecondissueisrelatedtotheDC-DCconversionwhosefunctionistoefentlyconvertthevariableDCpower.SincetheinputDCpowerisvariable,toobtainthemaximumpowertransfer,aspecialtechniquecalledmaximumpowerpointtracking(MPPT)isusuallyused.4Figure1.3IllustrationofdeadzoneforAC-DCconversion.MPPTisimplementedbythesophisticatedcontrolalgorithmandcanbeillustratedbyusingthesolarcellcircuitmodelasshowninFig.1.4.Theequivalentcircuitmodelofasolarcellmainlyconsistsofacurrentsourceandaforwardbiaseddiode.ThecurrentsourceIPH;SCrepresentsthegeneratedphotocurrent,RSistheparasiticseriesresistance,RPistheequivalentshuntresistance.IPHandVPHareoutputcurrentandterminalvoltageofthesolarcell.ThereforetheI-VequationforthiscircuitmodelcanbeexpressedasIPH=IPH;SCISAT(eqVPH;SCAKT1)VPH;SC=RP(1.1)whereVPH;SC=VPH+IPHRS.Fig.1.4(b)plotstheoutputcurrentversetheoutputvoltageforthesolarcellusingthisequationfordifferentlightintensities.Fig.1.4(c)presentstheoutputpowerofthesolarcellasafunctionofitsterminalvoltage.Itcanbeseenthatforagivenlightintensity,thereexistsanoptimaloutputvoltageforthesolarcellatwhichthemaximumpowercanbeobtained.Thispointiscalledmaximumpowerpoint(MPP).ItisnotedthattheMPPchangesasthelightintensitychanges.ThereforevarioussophisticatedcontrolalgorithmshavebeenproposedtotracktheMPPtoensurethesolarcellcanalwaysoutputthemaximumpower.However,suchkindofsophisticatedcontrolalgorithmsrequirenotonlymorepowerconsumption,5Figure1.4IllustrationofMPPTbyusingsolarcellapplication:(a)Typicalcircuitmodelforsolarcell;(b)Outputcurrentv.s.outputvoltage;(c)Outputpowerv.s.outputvoltage[3].butalsoadditionalsetuptime,thusitisdiftocaptureimpulsiveenergywhosemagnitudecouldbeveryhighbutdurationisshort.1.3ResearchMotivationAsmentionedinprevioussection,theavailableenergythatcanbeharvestedintheenvironmentcouldbequitelowandintermittent,thustheefytransferredfromtheharvestedenergyintotheeffectiveelectricalenergyisakeyfactorforthesuccessfuldeploymentoftheenergyharvestingsystems.Sophisticatedcontrolalgorithmsandcircuitscouldhelptoimprovetheefy,buttheirpowerconsumptioncouldnotbeaffordableforself-powereddevices.Basedonthisanalysis,tworesearchquestionsareproposed:(1)Cantheenergybeharvestedmoreefforultralowpowerapplications,eg.,whentheenergysourceisindead-zone?(2)Cantheenergybeharvested6Figure1.5(a)Astructuralhealthsensorembeddedundertheground;(b)Proposedhybridenergyharvestingmethod.moreefforimpactenergyapplications,eg.,whentheenergysourceisimpulsive?Toanswerthequestion,ahybridenergyharvestingtechniqueisproposedtoreducethethresholdeffectasshowninFig.1.5.Assumingthereexistmultipleenergysourcesintheen-vironment,forexample,RFpowerisusuallythemainpowersourceforstructuralhealthsensor,however,ifthestructuralhealthsensorisembeddedunderground,thestrainpowercanbeusedasanauxiliarypowersourcetoimprovetheefy.,themajorsourceVOC1,whichisRFpowerinthiscase,deliversthemostoftheenergythatisconsumedbytheload,whiletheauxiliarysourceVOC2,isusedtodecreasethethresholdeffect,sothepowertransferefyofthemainpathisabletobeimproved.Oneproposedsolutionforthesecondquestionistouseatime-dilationtechniqueasshowninFig.1.6.Consideringtwofootballplayerscomeintoahead-oncollision,thetransientenergycouldbeverylargebutthedurationisveryshort,soconventionalenergyharvestingsystemisunabletoperformMPPTduringsuchshortperiod.However,ifthechargingdurationcanbestretchedwhilethetotalenergyiskeptconstant,thefollowingDC-DCconvertermayhavemoretimetoperformMPPTalgorithm.Actuallythisideaistomaketheenergystoragecapacityadaptivethroughasimplesensingcircuit,,smallstoragecapacitycanachievethefastresponsewhilethe7Figure1.6(a)Illustrationofstretchingimpulsiveenergyintimedomain;(b)Proposedtime-dilationmethod.largestoragecapacityisabletofullycapturetheenergy.1.4ThesisContributionsandOrganizationThisdissertationwillfocusonthedesignandimplementationoftheefenergyharvestingcircuitsforultralowpowerandimpactenergyapplications.Themajorcontributionsinthisdisser-tationincludes:1.Ahigh-Qseriesimpedancematchingtechniqueisproposedinchapter2toimprovethepowertransferefyfornearRFenergyharvestingapplications.Theconventionalpar-allelmatchingtechniqueiswidelyusedincurrentRFenergyharvestingapplications.However,onemajorlimitationisthetrade-offbetweentheQ-factoroftheimpedancematchingnetworkandthecouplingcoefoftheenergytransferlink.Basedontheseriesmatchingnetwork,8theproposedimpedancematchingtechniquecannotonlyachievehighQ-factorandlowloss,butalsoreleasetheconstrainbetweentheQ-factorandthecouplingcoefsothattheenergyharvestingefyisimproved.2.Ahybridtechniqueisproposedinchapter3toimprovethepowertransferefyofthevoltageforultralowpowerapplications.Inthehybridtech-nique,themainenergyharvestingsourceisRFenergywhiletheauxiliarysourceislow-frequencyandlow-powervibrationenergy.Inultralowpowerapplications,theconventionalvoltagefacesthethresholdissue,thatistosay,thepowertransferefyofthevoltageisneartozerowhentheinputpowerisbelowthepowerthreshold.Toovercomethisissue,theproposedhybridtechniqueutilizesthelowfrequencyvibrationsignalastheDCbiasofthehighfrequencyRFsignal,thustheeffectivethresholdisreducedandthepowertransferefyisimproved.3.Anon-linearcompressivecircuitisproposedinchapter4toimprovethepowertransferefyoftheharvestersforimpactenergyapplications.Thiscompressivecircuitconsistsofafullwaveandatime-dilationcircuit.Whentheimpulsivesignalexceedsthepresetthreshold,thetime-dilationcircuitistriggeredtoprovideanadditionalwaytostoretheimpulsiveenergy,thustheenergycanbeharvestedasmuchaspossible.Inaddition,thisdissertationalsopresentsonesystem-leveldesignforultrasonicenergyhar-vestingandback-telemetryinchapter5.ThissystemimplementationistoprovethatthenearpassiveultrasonictelemetrysystemcouldbeusedforembeddedmonitoringinsidetheconductivestructuresandionicmediawheretheRFattenuationisFinally,chapter6summarizesthecontributionsofthisresearchandalsopointsouttheopenproblemsthatcouldbefuturedirectionsforthiswork.9CHAPTER2HIGH-QSERIESRESONANTZ-MATCHINGNETWORK2.1IntroductionAstherapiddevelopmentofthewearable,implantableandembeddeddevices,wirelessenergyscavengingtechniqueshavebecomeincreasinglyimportantoverthepastfewyears[3],especiallyintheapplicationsofbodyareanetworks[11],[12],implantablebio-medicaldevices[13],[14],andstructuralhealthmonitoring[15],[16],etc.Fig.2.1showsatypicalRFenergyharvestingsys-temwhichiswidelyusedinthenearapplications.ThepoweriswirelesstransferredfromthereadertothetransponderthroughinductivelycouplingofthetwocoilsandthenanAC-DCrecti-(orvoltagemultiplier)convertsthecoupledACpowertoDCpowerforproperoperationoftheprocessingcircuit.Thushighefyenergyconversionisdesiredinsuchenergy-constrainedsystem.Inthischapter,ahigh-Qseriesresonantimpedancematchingnetworkisproposedtoimprovethepowertransferefyofthewirelessenergyharvestinglink.2.2ParallelResonantMethodInductivecouplingiswidelyusedinwirelesspowertransfersystem,particularlyinthenearregion.Thesystemconsistsofareaderandatransponderandutilizestwoweeklycoupledcoilstoformaninductivelink,asshowninFig.2.2.Theinductivelinkcanbemodeledasatransformerwithtworesonantcircuits.Thevoltagecanthusbeinducedinthetranspondercoilandthenrecti-toaDCvoltageforproperoperatingthefollowingprocessingcircuit.Betweentheprocessingcircuitandthecoil,amatchingnetworkisneededtoresonatethesystem.Amongvariousmatch-ingstrategiesonthetransponderside,parallelmatchingisthemostpopularmethodduetoitseaseintuning.However,onemajorlimitationisthatthetotalqualityfactorQTofthetransponderisdependentonitscoilinductanceLt.ThisimpliesthatifLtisconstrainedbythesizeofthecoil,QT10Figure2.1Blockdiagramofatypicalinductivelycoupledwirelessenergyscavengingsystem.Figure2.2Parallelmodelofinductivecoupling.andhencethepowertransferefyandvoltagegainareconstrained.Inthissection,wewillrevisitthekeyparameterswhichaffecttheperformanceinparallelmatchingandthenproposeanewmatchingtechniquetoincreaseitsQ-factortooptimizetheefy.WeexaminethemathematicalmodelinFig.2.2.ThetransformerisformedbythemutualinductancesoftwocoilsM=kpLrLt,wherekiscouplingcoefwhichstronglydependsongeometryofthetwocoilsandtheirrelativespacingandalignment.Therefore,oncethesizeandthedistancebetweenthetwocoilsareed,themaximumvalueofkisusuallyconstrained.TheinducedvoltageinthesetwocoilscanbemodeledastwodependentvoltagesourcessMitandsMir.Forthesakeofsimplicity,theRFreaderismodeledasaseriesLC-circuit,whereLrrepresentsthereadercoilinductance,RrrepresentsthereadercoilresistanceandVSisthereadersourcevoltage.Similarly,LtandRtrepresenttheinductanceandtheresistanceofthetranspondercoil.Amatchingnetworkisusuallyrequiredatthetranspondertomaximizethepowertransfertothetransponder11loadandineffectboostthesmallinducedvoltagesMir.Duetoeasytuning,parallelmatchingiswidelyusedininductivepowerlink.Toresonatethecapacitiveinputimpedanceofthesensor,themaximuminductanceofthetranspondercoiliscon-strained.However,inmostapplications,thecoilsizeisrelativelysmallandhenceLtisalsosmall,soanotherparallelcapacitorCPisstillneededtotunetheresonantcircuit,asshowninFig.2.2.Forthesakeofconvenientcalculations,weuseCLp=CL+CPandthentransformtheparallelRLCLptotheirseriesrepresentation,whereCLsˇCLpandRLs=RL=Q2L,underthereasonableassumptionofQL˛1.Attheresonantfrequency,seriesLtandCLscanceleachother,thustheinducedvoltagesMirislydividedbetweenRtandRLsandthenboostedapproximatelyQLtimesbecauseofthecapacitorCLs.AssumingRLs=aRt,thevoltagegainAVtandthepowertransferefyhPtatthetranspon-dersidearegivenbyAVt=VLsMir=1jwCLs+RLsRt+RLsˇa1+aQL(2.1)hPt=PLPjsMirj=RLsRt+RLs=a1+a(2.2)Torelatethetranspondersidewiththereaderside,thedependentvoltagesourcesMitismodeledasthetransformedtransponderimpedanceZtr=b=(1+a)Rr,whereb=k2QtQr.WiththehelpofderivationofZtr,itismoreintuitivetoestimatethevoltagegainAVrandpowertransferefyhPratthereaderside,AVr=sMirVS=wMRr+Ztr=wMRr1+a1+b+a(2.3)hPr=PZtrPS=ZtrZtr+Rr=b1+b+a(2.4)Assumingthelossthroughthetransmissionmediumcanbeignoredattheresonantfrequency(PZtrˇPjsMirj)andhencethetotalvoltagegainAVandpowertransferefyhPcanbeobtainedbycombining(2.1)and(2.3),(2.2)and(2.4)asAV=AVtAVr=wMRra1+b+aQL(2.5)12hP=hPthPr=a1+ab1+b+a(2.6)Substitutinga=Qt=QLandb=k2QtQrinto(2.5)and(2.6),AVandhPcanberearrangedintermsofQ-factorsas,AV=kQrQT1+k2QrQTrLtLr(2.7)hP=k2QrQTk2QrQT+1QtQt+QL(2.8)whereQT=QtQL=(Qt+QL)istotalQ-factorofthetransponderside.ItisshownthatbothAVandhPimproveastheincreaseofQT.SincegenerallyQL˝Qt(supposeQt>100andQLis5˘10),QTismainlydeterminedbyQLandthusincreasingQLismorepracticaltoimproveAVandhP.However,forparallelmatchinginwhichQL=RL=(wLt),onceRL(orpowerconsumptionoftheload)ised,increasingQLcanonlybeachievedbydecreasingLt,whichisequivalenttoshrinkthecoilsizeorreducethenumberofturns,buteitherwaywoulddegeneratethecouplingcoefk,whichactuallyisthemostimportantfactoraffectingtheefyoftheinductivelink.Therefore,itisindicatedthatthetrade-offbetweenQLandkisonemajorlimitationforparallelmatchingstrategytoachievehighefy.2.3ProposedSeriesResonantStrategyToovercomethelimitationimposedbythetranspondercoilLtintheparallelmatchingstrategy,aseriesmatchingstrategyisproposed[17],asshowninFig.2.3(a).Inthisseriesmatchingnetwork,C1issettobemtimesofCLpwhilekeepingRLconstant,thusQ0LincreasestomQLandtheseriesformoftheloadresistorR0LsdecreasestoRLs=m2,resultinginashrinkingto1=m2times.C2ischosentobem=(m1)CLpforresonance.Totunethecircuitmoreeasily,ahigh-QinductorL1isseriesinsertedbetweenC2andC1inthepracticalcircuit,asshowninFig.2.3(c).FirstlyweletC2=CLptoresonatewithLt,thisresonanceiseasilyobtainedwhentheimpedanceoftheleftseriesLtC2canonlyshowpureresistanceRtinSmithChart;thenthesameprocedurecanbeappliedtotherightL1C1matching.Finally,bothoftheleftandrightpartsarepermittedto13Figure2.3Mathematicalmodelofthetranspondercircuitusingseriesmatchingstrategy:(a)Parallelformoftheloadand(b)theirequivalentseriesform;(c)showsanmoreconvenientmethodtotunethecircuitbyreplacingsingleC2withaseriesL1C2.beconnectedtogetheronlywhenitissurethateachpartresonatesverywell.AftertransformingtheparallelC1RLtoaseriesC1RLs,thesameanalysisprocedurefortheparallelmatchingstrategyisstillapplicabletotheseriesmatchingstrategy,inwhichaandQLaresubstitutedwitha=m2andmQL,respectively.AnotheristhattheparasiticlossimposedbyL1shouldbeconsidered.AssumingL1'sparasiticlossR1=gRt,thenthevoltagegainA0Vandpowertransferefyh0ParegivenbyA0V=wMRra(1+g+b)m+a=mQL(2.9)h0P=ab(1+g)(1+g+b)m2+a2=m2+(2+2g+b)a(2.10)Tocomparetheperformanceoftwomatchingstrategies,wetwo(FOMs)14Figure2.4Improvingfactorsof(a)voltagegainfVand(b)powertransferefyfPv.s.themultiplicationofQL.fVandfPasfV=A0VAV=1+b+a(1+g+b)m+a=m(2.11)fP=h0PhP=(1+a)(1+b+a)(1+g)(1+g+b)m2+a2=m2+(2+2g+b)a(2.12)ItiseasilyshownthatwhenmV=qa1+g+b,fV;max=1+b+a2pa(1+g+b)ˇ12ra1+g+b=12mV(2.13)andwhenmP=4ra2(1+g)(1+g+b),fP;max=1+aa1+b+a(p1+g+p1+g+b)2ˇa(p1+g+p1+g+b)2(2.14)ToestimatefVandfPasafunctionofm,thecouplingcoefkissettobe0.01andQt=Qr=150,QL=5,whicharethesameparametersusedinthemeasurements.However,thevalueofgisdiftoobtain,sowehavetoestimateitaccordingto:g=Qt=(mQL1),if10m1andQt=150,then150/gQL115/g.ThuswecanusedifferentgtomodelQL1:ultrahigh-Q(g=0.1,15Figure2.5Experimentalsetupusedforpoweringdistanceevaluation.1500QL1150),moderatehigh-Q(g=1,150QL115)andlow-Q(g=10,15QL11.5).TheQ-factorofsmallvalueSMDinductoraround13.56MHzisabout25˘50,thustheassumptionofg=1ismorereasonable.FromtheresultsshowninFig.2.4,itisnotedthatbothoffVandfPincreasewithmandreachamaximum.Then,bothfVandfPdecreasebecauselargermresultsinsmallerequivalentRLsandhencemostoftheinducedvoltageandpowerareconsumedbytheparasiticlossRtandR1.Itcanalsobeseenthatitisdiftoimprovetheefyiflow-Qinductorisused.Fig.2.3(c)showstheoftheproposedmatchingstrategy,whichusesalargecoilLttoimprovethecouplingcoefkandsmallL1insteadofLttoincreasetheloadqualifyfactorQL=RL=(wL1).Consequently,theconstraintimposedbyLtinparallelmatchingisabletobereleasedinthisproposedmatchingstrategyandthustheefyoftheinductivelinkisimproved.2.4MeasurementResultsTheperformanceoftheproposedmatchingstrategyhasbeenvusingtheexperimentalsetupshowninFig.2.5,inwhichthe13.56MHzreaderandthesensorareseparatedbyanadjustabledistance.ThereaderandthesensorcoilshavebeenfabricatedonaplanarPCBandthe16Figure2.6Comparisonofthevoltagegeneratedacrossaconstantloadusingparallelandproposedmatching.tionsoftherespectivecoilsaresummarizedinTable.5.2.ThereadercoilisdrivenbyTITRF7960chipset[18],whosemaximumoutputpowerissetto200mW.ForthisworkweuseaconventionalSchottky-basedDicksonvoltagemultiplierasthefront-endcircuit.Tocomparetheperformanceoftheproposedmatchingnetwork,aconventionalparallelmatchingnetworkisusedasabench-mark.At13.56MHz,theinputimpedanceofthesensorismeasuredtobeCL=56pFinparallelwithRL=350W.TheinductanceLtofthetranspondercoilis0.84mHandthetheoreticalCLp=164pF.Forparallelmatching,CPissetto110pF;forproposedmatching,fromFig.2.4,itisseenthatbothoffVandfPhaveadistinctimprovementwithintherangeofmˇ3˘5,thusL1ischosentobe0.18mH(m=Lt=L1ˇ4.7)whileC2=166pFandC1=783pF,respectively.Inthesetofexperiments,aresistorRL=350WandcapacitorCL=56pF(equaltotheinputimpedanceofthesensor)werechosenasthestaticloadfortheparallelandtheproposedmatchingnetwork.ThemeasuredresultsareshowninFig.2.6,whichplotsthevoltageacrosstheloadwithrespecttothedistanceDrsbetweenthereaderandthesensor.TheresultsshowthatgivenaedDrstheboostedvoltageintheproposedapproachisdoublethatoftheparallelmatchingapproach.AlsonotethatthethresholdoftheSchottkymultiplier(VL=VONˇ300mV)isreachedatrespective17Figure2.7ComparisonofthevoltagegeneratedbySchottky-multiplierusingtheparallelandtheproposedmatching:(a)18-stagemultiplierusedforanalogVDD;(b)12-stagemultiplierusedfordigitalVDD.Table2.1CoilParametersReaderTransponderTurnsN43Outerlengtha(cm)12.610.7Outerwidthb(cm)7.81.4Tracewidthw(mm)1.271.27Tracespaces(mm)0.50.25Tracethicknesst(mm)3535Self-inductanceL(mH)5.60.84UnloadedQ150150distancesof13cmfortheparallelmatchingand17cmfortheproposedapproach.Forthesecondsetofexperiments,theedimpedancewasreplacedbytheSchottkyvoltagemultipliers.Theresultsofan18-stagemultiplierinthisexperimentareshowninFig.2.7(a)wheretheinsetshowsthesameresponseinlogarithmicscale.Theseresultsagainshowthattheproposedapproachcanboostthevoltagecomparedtotheconventionalparallelmatching.NotethatthemaximumoutputvoltageoftheSchottkymultiplierislimitedbytheover-voltageprotectiondiodesintegratedonthesensorIC.Fig.2.7(b)comparesthemeasured18responseobtainedfroma12-stagemultiplier,andshowsthattheimprovementinvoltage-boostingandhencethepoweringdistancecanstillbeachievedfortheproposedmatching.Therefore,itisconcludedthattheproposedmatchingstrategycanimprovethelinkperformance.2.5SummaryThepowermatchingnetworkbetweenthetranspondercoilandthevoltageisnecessaryforwirelesspowertransfer(WPT)system.Itisshown[19]thattheWPTsystem'sPCEisheavilydeterminedby:(1)thecouplingcoefkbetweenthereadercoilandthetranspondercoil;(2)thequality-factorsoftheircorrespondingmatchingnetworks.Inmostactualapplications,thedimensionofthetranspondercoilisusuallyconstrained;furthermore,misalignmentbetweenthesetwocoilsalwayshappens.Thusthecouplingcoefcouldbesoweakthatthepowerpropagationrangeisverylimited.TocompensatesuchweaklycoupledlinkandthedegenerationoftheQ-factor,ahigh-Qmatch-ingnetworkisdesirableandthecommunicationbandwidthcanbetraded-offifthethroughputisnothigh.Itiswellknownthatthecouplingcoefisproportionaltothecoilinductance.How-ever,intraditionalparallelmatchingnetwork,thecoilinductanceisinverselyproportionaltotheQ-factor,thussmallcoilinductanceresultsinhighQ-factorbutweekcouplingcoefwhichisunwantedinWPTsystem.ToreleasetheconstrainbetweentheQ-factorandk,thischapterpro-posesanovelseriesmatchingnetworkwhichuseslargecoiltoimprovekandsmallseriesinductortoboostQ-factor.TheboostedACvoltagehasanotherthatcanovercomethedead-zoneofthetoimprovethepowerconversionefy(PCE).19CHAPTER3HYBRIDRECTIFICATIONTECHNIQUEFORRFENERGYHARVESTING3.1IntroductionThepreviouschapterillustratestheimportanceoftheimpedancematchingnetworkinthewirelessenergyharvestingsystem.Besidestheimpedancematchingnetwork,theRFrectorvoltagemultiplieralsoplaysanimportantroleinefyimprovement.Thekeyparameterthataffectstheperformanceofanyisthethreshold-voltageoftherectifyingdevicethatsetsthelower-limitontheamplitudeoftheinputRFsignalatwhichthecanstartharvestingenergy.Thatistosay,iftheinputRFsignalisbelowthethresholdvoltage,thepowerconversionefyoftheisnearlyzero.Thustheminimumenergythatisabletobeharvesteddependsonthesthresholdvoltage.Topushthelimitofthestartupenergy,thischapterproposesahybridtechniquewhichutilizesbothofRFpowerandvibrationpowertoefharvesttheenergybelowthethresholdlevel.Firstly,thevoltageconversionefy(VCE)andpowerconversionefy(PCE)oftheconventionalareanalyzedanddiscussed.Thenseveralstate-of-artarepresentedtoprovethatthelimitationisnotcompletedsolved.FinallythehybridPZTassistedRFisproposedtoaddressthislimitation.3.2VCEandPCEofTheConventionalVoltageVoltageisalsonamedAC-DCconverter.Atypicalconventionalhalf-wavevoltageisshowninFig.3.1(a)withtheloadcapacitorCLandloadcurrentIL.ThepasstransistorMPisimplementedinPMOStoavoidthebody-biasissueinthestandardsinglewellCMOSprocesswhencascadingmultiplestages.AsshowninFig.3.2,whentheinputRFvoltageVRFisabovetheoutputDCvoltageVOUT,thetransistorisforwardbiased.DependingontherelationshipbetweenVRFandVOUT,thetransistor20Figure3.1Typicalhalf-wave(a)conventional;(b)withidealcompensatedvoltage.canbeeitherinthesaturationregionorsubthresholdregion.IfVsg<jVTHj,thetransistorworksinsubthresholdregionwherethecurrentisISubˇISexp(VSGjVTHjnVT)(1exp(VSDVT))(3.1)whereISandnareempiricalparameterswithn1;VSG=VSD=VRFcosqVOUT,VTHandVTarethethresholdvoltageandthermalvoltage,respectively.Tosimplifytheanalysis,thesubthresh-oldcurrent(region2)canbelinearizedasISub=ISf2qf2f1(3.2)wheref1=cos1(VOUT+jVTHjVRF)andf2=cos1(VOUTVRF),areconductinganglesforthesaturationregionandforwardbiasedregion,respectively.Similar,thesaturationcurrent(region1)canbelinearizedasISat=IF(IFIS)qf1(3.3)whereIF=12mpCox(WL)(VRFVOUTjVTHj)2isthemaximumsaturationcurrent.WhentheinputRFvoltageVRFisbelowtheoutputDCvoltageVOUT,thetransistorisreversebiasedwithVsg=0,whichindicatesthereverseleakagecurrentisinthesubthresholdregionbutwithzerosourcetogatevoltage,IRevˇISexp(jVTHjnVT)(1exp(VSDVT))(3.4)21Figure3.2VoltageandcurrentwaveformofthepasstransistorMP.Thelinearizationmethodisstillapplicabletothereverseleakagecurrent(region3)asIRev=IRqf2pf2(3.5)whereIR=ISexp(jVTHjnVT)isthemaximumleakagecurrentatwhichVSD=VOUT+VRF3VTsothattheeffectofVSDcanbeignored.Afterlinearizingthecurrents,thechargesindifferentregionscanbeeasilycalculated,thusinoneperiodQSat=(IF+IS)f1(3.6)QSub=IS(f2f1)(3.7)QRev=IR(pf2)(3.8)Insteadystate,accordingtothechargeconservationQSat+QSub=QRev+QL,whichcanbe22extendedby(IF+IS)f1+IS(f2f1)=IR(pf2)+2pIL(3.9)UsingTaylorseriesexpansioncos1qˇp=2qforf1andf2,thevoltageconversionefy(VCE)canbeestimatedbyVOUTVRF=p2IFIF+IS+IRjVTHjVRFpIR+2ILIF+IS+IR(3.10)ThepowerconversionefyisasPCE=POUTPIN=POUTPOUT+PLOSS(3.11)wherePOUT=VOUTIL,isoutputpowerofthe.PLOSSispowerconsumptionofthepasstransistorandcanbeestimatedbyPLOSSˇPforward+Preverseˇ1p(Zf10ISatVSDdq+Zf2f1ISubVSDdq+Zpf2IRevVSDdq)(3.12)withZf10ISatVSDdq=Zf10(IF(IFIS)qf1)VRF(cosqcosf2)dq=VRFIFf1sin2f22+VRFIS(sinf1f1cos2f22)(3.13)Zf2f1ISubVSDdq=Zf2f1ISf2qf2f1VRF(cosqcosf2)dq=VRFIS(f1+f22sinf1f2f12cosf2)(3.14)Zpf2IRevVSDdq=Zpf2IRqf2pf2VRF(cosf2cosq)dq=VRFIR(1+2+pf22cosf2)(3.15)233.3State-of-ArtInordertoovercomethelimitationsduetotheVthandtoimprovethePCE,severalpassiveandactiverectitopologieshavebeenreportedinliterature[14,15,20Œ23].Inthispartsomeofthesestate-of-arttechniquesareexamined.3.3.1PassiveManytechniqueshavebeenproposedtogeneratethecompensationvoltageforpassiveThemostpopularoneisthedifferential-drivencross-coupled,whichactivelybiasesthegateofthetransistorwithpositiveandnegativeRFsignal,asshowninFig.3.3(a).Duetosimpleandrobuststructure,thistechniqueiswidelyusedinUHFenergyharvestingapplications,howeveritalsoneedsthestart-upvoltageabovethethresholdvoltagetoturnonNMOSorPMOSattheinitialstage.AnothermethodshowninFig.3.3(b)istousepseudogatedevicewhichactuallyconnectsthegateofthediode-tiedtransistorandthegateoftheMOScapacitortoformahigh-impedancenodetotrapchargeonthegate.ThechargeonthegateisthereforeedwhichresultsinaedvoltagebiasacrosstheMOScapacitor.Thechargesthataretrappedinsidethegatedeviceactasagate-sourcebiastopassivelyreducetheeffectivethresholdvoltageofthetransistor.Althoughthecircuitstructureusingpseudogateisnotcomplex,itrequiresinitialprogrammingoneachinggatebeforeworking,andtheamountofchargeishardtocontrol,andadditionalcircuitryisneededtoinjectorremovechargefromthegate.ThemethodshowninFig.3.3(c)iscalledinternalthresholdcancellation,inwhichthecapac-itorCBPholdsthethresholdsvoltageoftheMOSdiodeMP1byreplicatingitsthresholdvoltagewithMPB.AnotherequivalentcircuitisappliedtotheMOSdiodeMN2.ThisequivalentcircuitcanaccuratelytracktheprocessandtemperaturevariationinthesediodesbymatchingMPBtoMP1andMNBtoMN2.Inaddition,theleakagecurrentsofalldiodesareminimizedbythelargebiasresistorRb.Actuallytheinternalthresholdcancellationtechniqueneedssuflargeinputto24Figure3.3Threepopularpassive(a)Differentialcross-coupled[4];(b)Pseudogate[5];(c)InternalVTHcancellation[6].startupthecircuitandthebiasresistorRBthatisaroundtheMWrangeisalargevalue.Duetothesimplecircuitarchitecture,thepassivecanworkinultrahighfrequencywithlowstart-uppower,soitissuitableforlongdistanceapplication.However,thepowertransferefyislowandsometimesspecialprocessisneededtoimprovetheefy.3.3.2ActiveTheotherkindofisactivewhichreplacesdiodewithactivelycontrolledswitchtoimprovetheefy.TheswitchusuallyimplementedinMOSFETwhichhasverylowresistancewhenconducting.AsshowninFig.3.4(a),thecontrolcircuitforactiveusuallyusescomparatortosensetheACinputandopenthetransistoratthecorrecttimetoallowthe25Figure3.4(a)Comparatorbasedactivediode[7];(b)Offsetcontrolhighefyactive[8];(c)Illustrationofleakagecurrentissueinactive[8].currenttowinthecorrectdirection.Soitcanobtainhighpowerconversionefywithhighconductingcurrent.However,thecomparatorhasatradeoffbetweenthespeedandpower,soitisverydiftoachievefastresponsewithlittlepower.Inaddition,theswitchisusuallyimplementedinlargesizeforlowon-resistancesothatitsgatecapacitancecouldbeverylargeandthecomparatorneedsstrongdrivingcapabilitywhichisalsoverydiftoachieveunderlowpowercondition.Sotheactiveisdiftooperateinatultrahighfrequencieswheretheavailablepowerisverylimited.Anotherissuethatcannotbeneglectedforanactiveisthetiming.Duetotheoffsetandpropagationdelayofthecomparator,theswitchcannotbecompletedturnedoffwhentheinputACsignalislowerthantheoutputDCvoltage(showninFig.3.4(b)and(c)),causingthereverseleakagecurrenttoconductsothattheefyisdecreased.26Figure3.5Conceptualhybridrectifer[9].3.4ProposedPZTAssistedRFBasedtheanalysisabove,itcanbeseenthatalloftheexistingmethodswithoutspecialprocessesstillrequiretheinputamplitudetobeatleastequaltotheVTHoftheMOSFET,althoughtheycanachieveexcellentpowerconversionefy.Toovercomethethreshold-voltageinputamplitudelimitation,weproposeasolutionbyusinghybridrecttechnique.Thetechniquereliesonthefactthatmostenergy-scavengingsensorsoperateinanenvironmentwheremultiplesourcesofpowercouldbeavailable.Typically,theenergyfromeachsignalsourcecouldbeharvestedusingaseparateoperatingindependently,asshowninFig.3.5(a).Theiroutputscouldthenbecombinedinseriestoboosttheoutputvoltage.Thisapproach,however,requiresthateachenergysourcehassimilarelectricaldrivingcapabilitywhichisdeterminedbythefrequencyf,27Figure3.6Theconceptofahybridvoltage-multiplieranditspotentialapplicationtoastructuralhealthmonitoringsensorthatscavengesRFandvibrationenergy.theamplitudeVandtheoutputimpedanceZofthesource,whichmightnotbepossiblewhenthephysicsandefyofenergytransductionfordifferentenergysourcesaredissimilar.AnexampleofasensorwithmultiplepowersourcesisanRF(RFID)sensorthatisattachedtoamechanicalstructure[15]whereoneofthesourcesofenergycouldbeambientvibrationinadditiontotheRFsignalusedforpoweringandinterrogationofthesensorasshowninFig.3.6.TheenergythatcouldbeharvestedfromtheambientvibrationmightbemuchlowerthantheenergythatcouldbeharvestedfromtheRFsource.Also,apiezoelectrictransducerusedforvibrationenergyscavengingispurelycapacitivewithverylimitedcurrent-drivingcapability,whereasanearelectromagnetictransducer(forscavengingRFsignals)ispurelyinductivewithalimitedvoltage-drivingcapability.Insuchcases,ahybriddesign(asshowninFig.3.5(b))offersasolutionthatcanexploittheelectricalproperties(f,V,Z)ofthesourcesinconjunctionwitheachotherandnotindependentofeachother.Forexample,duetotheircapacitivenaturepiezoelectrictransducerscangeneratelargeopen-loadvoltagesandthereforecouldbeusedforbiasingthegatesoftheMOSFETsinthe.ThisinturncouldreducetheminimumamplituderequirementsontheincidentRFsignal.Thiswouldthenimplythatasmallersizeinductorcoilcouldbeusedforbio-telemetryapplicationsorinthecaseofafarRFIDsensor,thepowering-distanceofthesensorcouldbeincreased.28Figure3.7Complementarycross-coupled.ThereforeIwillpresentanovelstructurewhichusesapurelycapacitive,highampli-tudeandlowfrequencypiezosignaltoboosttheDCcomponentoftheACcontrolledgatevoltageofrectifyingtransistors.3.4.1ThresholdCompensationInvestigationAtalowinputRFsignallevel,complimentarycross-coupledcellasshowninFig.3.7ispreferredbecauseitcanachievehighestpowerconversionefythankstodifferentiallydynamicbiasscheme[21].Thegate-to-sourcevoltageofPMOS(NMOS)deviceVGP(VGN)isdeterminedasVGP=VRFCCCC+Cpar(3.16)whereCparisthetotalparasiticscapacitanceofPMOSandproportionaltothesizeofthetransistor(WidthLength).VGPneedstobelargerthanVthtoturnonthedevicesinthe.SeveralstagescanbecascadedtogenerateasufhighDCoutputvoltage(VDC).Thenumberofstages(N)isdetermineasN=VDCVout(3.17)29Figure3.8DCboostingeffectsimulationsetup(VRF=0.3Vpkpk,fRF=13.56MHz,RL=20KW,CL=1nF).Figure3.9OptimalbiasvoltagesversusinputRFsignal.ToovercomethethresholdlimitoftheCMOSdevicesatlowRFinputamplituderange,theDCcomponentsoftheirgatevoltagesneedtobecompensatedi.ebiasingapositiveDClevelfortheNMOSsandanegativeDClevelforthePMOSs.AninvestigationofDCbiasingeffectisperformedwithaCCCRisshowninFig.3.8(a).Fig.3.8(b)showsanexampleoftheDCoutputvoltageversusDCbiaslevelswhenVRF=300mVpkpkandfRF=13.56MHz.ItrevealsanoptimalDCbiasingvoltagepairforNMOSsand30PMOSs(VBN=500mV,VBP=-500mV)forapeakDCoutputvoltageof180mV.WhentheDCbiasinglevelsgobeyondtheoptimalvalues,theDCoutputvoltagedecreasesgradually.Thisisbecausewhenthebiasinglevelincreases,thereversedraincurrentincreasesfasterthantheforwarddraincurrentsincethetransistorsenterstrioderegionintheconductionperiodatlargeDCbiaslevels.Thiseffectworsenstherectifyingbehaviorofthe.Thisbehavioriswell-knownandanalyzedinpreviousreports[4,22].AsetofsimulationrunsareperformedforawiderangeofVRFtotheoptimalbiasinglevelsandtheresultisshowninFig.3.9.ItiscleartoseethatDCbiasvoltagesdecreaseswiththeincreaseofRFamplitude.3.4.2CircuitImplementationTogeneratetheDCbias,thepiezoelectricsignalcanbetakenintoconsiderationsinceitsavailablefrequencyresponseusuallyfallsbelow10kHz,whichthuscanbeconsideredasaDCsignalwithrespecttoRFsignal.AnotherconsiderationtousePZTsignalislowpowerconsumptionsinceitonlyrequiresverylittlepowertodrivethegateofdevice.Therefore,basedonthedifferentphysicalpropertiesofthePZTsignalandRFsignal,weproposeaHRinwhichthegatesofthedevicearenotbiasedbytheRFsignalbutbytheclampedpiezoelectricsignals.TheCCCRstructurehasbeenextendedherebyaddinganappropriateDCbiastoenableittoworkatextremelylowinputRFsignallevels.Apairofbiasingvoltagegenerators(Fig.3.10(b)and(c))generatethegatebiasingvoltagesfortheNMOSandPMOSdevicesrespectively.TheNMOSbiasingvoltagegenerator(Fig.3.10(b))consistsoftwosymmetricalbranchesthatgeneratethetwo180ophaseshiftedbiasingvoltagesVGN+andVGNforthecrosscoupledNMOStransistorsinFig.3.10(b).Forsimplicity,onlyupperbranchisconsidered.TheRFinputVRF+providestheACcomponentanalogoustothemaincircuit(Fig.3.10(a)).ThePZTsignalVPZ+isboundedinitsamplitudetothegatetosourcevoltageVGSoftheverticaldiodeconnectedPMOSandthePNjunctionbetweenbulkandthelowerS/DterminalthatformsabipolardiodeinparallelpointingtheotherwayP1N(aPMOSknownasTobi-element[24]).ThentheresultantvoltagesVPZN+is31Figure3.10ProposedHR:(a)main(b)DCbiasgeneratorforNMOSgateterminalsand(c)forPMOSgateterminals.bythehorizontaldiode-connectedPMOSP2NtoprovidetheDCbiasleveltoVGN+.Inotherwords,thelowfrequencyclippedPZTsignalsVPZN+issuperimposedwiththeRFsignal.ThisDCboostedRFsignalisthenusedtosynchronouslydrivethegatesoftheCCCR,insteadofdirectlyusingtheincidentRFsignals,i.e.theRFamplitudenolongerneedtobelargerthanVthasintheCCCR.ThelowersymmetricalbranchandthebranchesinFig.3.10(d)generatingthebiasingvoltagesVGP+andVGPfortherectifyingPMOSsoperateanalogously.Fig.3.11showssomeoftherelevantsimulatedwaveformsfromthebiasinggenerators.Asintended,theextremesofVPZN+andVPZP+areboundedduetoTobi-elements(P1N;1P)inthestageofbiasgenerators.TheVGSoftheP1N;1PandtheVthofofbulk-S/DP-Ndiodedetermines32Figure3.11Simulatedwaveformofcriticalnodesinbiasvoltagegenerators.Fpiezo=10KHz,Vpiezo=1V,FRF=13.56MHz,VRF=0.3Vpkpk.thepositiveandnegativeboundinglimits,respectively.ThisVGSisdependedonthesizeoftheP1N;1Pandcanbeadjustedtotheoptimalvalue,makingtheabletoharvestRFenergyefatultralowRFamplitudes.3.4.3DeviceSizeAnalysisandOptimizationThewidth(W)andlength(L)ofMOSdevicesarethedesignconstrainsofloadingconditionandPCEwhilethechipareaismainlydeterminedbythecouplingcapacitorCCvalue.ThenumberofstageNintheneedstobeminimaltoachievehigherpeakPCEsincethebodyeffectinmultistagetopologyincreasesVTHofNMOSsinthelaststages[25].Therefore,theoutputvoltageismaximizedthroughoutfollowingoptimizedsizinganalysis.FordraincurrentmatchingbetweenNMOSandPMOS,theaspectratio(W=L)PMOS=2(W=L)NMOSarechosen.Assumethat1VDCoutputvoltageand10mAoutputcurrentat300mVpkpkRFvoltageanda10KHzPZTarethedesigntarget.ForCC=10pF,theDCoutputvoltageofonestageasafunctionofdevicesizesisshowninFig.3.12.Whenthetransistorsizeincreases,theconductivestrengthofthetransistorincreasesinaccordancewiththeincreaseofCparwhichreducestheeffectiveVGPandVGNas33Figure3.12VOUTversusdevicesizeofthemain(VRF=0.3Vpkpk,L=0.1mm,CC=10pFandIOUT=10mA).Figure3.13EquivalentcircuitoftheupperbranchofNMOSbiasgenerator.statedinEq.3.16.Therefore,thereexistsanmaximalVout=200mVatanoptimalW/L=100.Inaddition,thesizeofthedevicesinthebiasgeneratorshouldalsobeoptimized.ConsidertheupperbranchofaNMOSbiasgeneratorasshowninFig.3.13.CParisthetotalparasiticscapacitortogroundatnoteVGN+whichisthetotalparasiticscapacitanceofP2NandNMOSinthemain.Intheinitialstate,apartofcurrentfromPZTsignal(IP2N)isthroughP2NtochargetotheparasiticscapacitorCPar2toincreasetheDCcomponentofVGN+.Atthesteadystate,wheretheDCcomponentofVGN+reachestoapproximatelypositiveclippedPZTvoltage34VPZN+,mostcurrentfromPZTsignalIPZTwsintodiode-connectedPP1Ni.eIPZT'IP1N.ThecurrentIP1Nismodelas[26]IP1N=8><>:Iso:WL:eVgsnVT(1eVdsVT)Vgs<Vthb:WL(VgsVth)2VgsVth(3.18)whereIsoisthesubthresholdunitcurrent,nisthesubthresholdregionswingparameter,VTisthethermalvoltage(typicallyis26mVat27oC)andbisprocessconstant.BecauseP1Nisconnectedasadiode,Vds=Vgs=VPZN+ofwhichamplitudeisinarangeof300-800mV,muchlargerthanVT.Therefore,Eq.3.18isrewrittenasIP1N=8>><>>:Iso:WL:eVPZN+nVTVPZN+<Vthb:WL(VPZN+Vth)2VPZN+Vth(3.19)SincePZTcurrentIPZTiscalculatedasIPZT=CP:VPZT:2:p:fPZT(3.20)whereVPZTandfPZTarevoltageamplitudeandfrequencyofPZTsignal.From(3.19)and(3.20),IPZT'IP1N,theamplitudeofVPZN+canbesolvedVPZN+=8>>><>>>:n:VT:ln(CP:VPZT:2:p:fPZTIso:WL)VPZN+<VthsCP:VPZT:2:p:fPZTb:WL+VthVPZN+Vth(3.21)Itcanbeseenfrom(3.21)thattheamplitudeofVPZN+isdependedonbothPZTelectricalprop-erties(voltageamplitudeandfrequency),PZTcouplingcapacitorCPandthesizeoftransistorP1N(W/L).ThisvoltageamplitudedeterminestheDCbiaslevelofNMOSgatevoltageVGN+,therebyaffectingtotheoutputvoltagegeneratedbythemainasanalyzedpartII.SincethecurrentIP2Nisverysmall,thevoltagedroppingondiodeP2Ncanbenegligible.WithVBN;optistheoptimalDCbiasvoltageforNMOSs,theoptimalamplitudeofVPZN+'VBN;opt.ThereforefromEq.(3.21),wecantheoptimalvaluepair(W/LandCP)foreachbiasgeneratorbranchatagivenPZTsource.Forexample,at0.3VRFvoltage,VBN+;opt=0.5Vas35Figure3.14SixstageHR.investigatedinpartII,andwithaPZTsource(VPZT=2VandfPZT=10kHz),theoptimalvaluepairscanbefoundas(1.25mm/1mm;0.1pF),(5mm/1mm;0.4pF)or(10mm/1mm;1ppF).ThediodeconnectedP2N;2PservesasatotheclippedPZTsignalsVPZN;PZPandcanbechosenwithasmallsize.TheRFcouplingcapacitorCRFneedstobesuflargecomparedwiththeparasiticscapacitanceofP2N;2PtomaximizetheamplitudeofVGN;GPduetothecapacitivevoltagedivisionsasshowninEq.3.16.ThedevicesizeW/L=5mm/0.2mmforP2N;2PandCRF=0.4pFareselected.3.4.4MeasurementResultsWithadesigntargetofVDC=1VandIOUT=10mA,asixstageHRasshowninFig.3.14alongwithasixstageCCCRforcomparisonareimplementedonthesamedieina90nmCMOSprocess.Fig.3.15showsthechipmicrographofbothTheactivesiliconareaofthetwohavethesamesizeofapproximately0.137mm2(550mm250mm).ThisisbecausethecouplingcapacitorsCCareimplementedbymetal-in-metal(MIM)capsandspanmostlytheareaofthethebiasgeneratorsinHRcontributeignorablyadditionalsiliconarea.ThetestchipwasimplementedwithanESDprotectioncircuit.Thiswoulddegradetheperformanceofthehoweverthetwoareputunderthesameconditionandtheperformancecomparisonisstillrational.Aimpedancematchingnetworkcanbedesignedtoimprovethetotalperformanceofbothtwobutdesigningamatchingnetworkisoutofthescopeofthispaper.36Figure3.15Micrographofvoltagemultiplierchip.ThemeasurementsetupforthechiptestisshowninFig.3.16.Asingle-endedRFsignalfromasignalgeneratorisconvertedintodifferentialsignalbyusingatransformerbalunCX2147withinsertionlossof0.3dB.Toemulatethedifferentialpiezoelectricsignal,signalsfromtwochannelsofTGA1421aresynchronizedwithaphaseshiftof180o.Theseoutputsignalsareconnectedinserieswithbigresistersof1MWtolimittheoutputcurrentwingintothecircuit.TomeasuretheinputcurrentoftheRFsignals,twosmallresistorsof100WareconnectedinserieswiththebalunandtheinputcurrentiscalculatedbymeasuringdifferentialvoltageattwopointsXandY[27].SincethemeasuringprobehasaloadingcapacitorCprobeof15pF,theintrinsiccurrentwingintothechipiscalculatedasIINintr=IINIprobe=VXYRsensVrms;XX0(2:p:fRF:Cprobe)(3.22)whereIintristheintrinsicinputcurrentwingintothechipandfRFistheRFfrequency.Theinputpoweriscalculatedthroughtheintrinsicinputcurrentandinputvoltage.TheinputRFpower37Figure3.16Measurementsetupofthetestchip.andPZTpowerarecalculatedasPRF;IN=ZT0IIN;intrVrms;XX0dx(3.23)PPZT=N:2:V2PZT:2:p:fPZT:CP(3.24)withNisthenumberofstagesofthe.TheoutputpoweriscalculatedthroughDCoutputvoltageandloadingresistorRL.ThePCEiscalculatedasPCE=POUTPIN=V2DCRLPRF;IN+PPZT(3.25)Fig.3.17showsthefunctionofDCoutputvoltageversusinputRFvoltageatdifferentloadingconditions.ItisworthnotingthattheparasiticinductorandcapacitorofthepackageandbondingwiredocauseadistortionfortheinputRFvoltageandtheDCoutputvoltage.Therefore,abetterintrinsicperformancecanbepredicted.Inthismeasurementsetup,thepseudoPZTsignalsarekeptatamplitudeof2Vandafrequencyof10kHz.TheintrinsicinputpowerofthepseudoPZTsignaliscalculatedat1.2mW(-29.2dBm).ItiscleartoseethattheHRhasdeadzonefreeandisabletoharvestpoweratultralowinputvoltagefrom100mV.Ontheotherhand,thedeadzoneexistsintheCCCRandlimitsitsability38Figure3.17DCoutputvoltagevesusinputRFvoltage.toharvestpowerataRFinputvoltageamplitudehigherthanVTH(around0.5V)oftheCMOSdevices.Forexample,at300mVinputvoltagewiththeloadingresisterof330KW,theHBRcangenerateanDCvoltageof1VwhiletheCCCRisalmostsilentandcanonlygeneratethesameDCvoltagelevelattheinputvoltageof650mV.Thisoutputpowerlevel(3mW)canbesufforsomesubthresholdmixedsignalblockswhicharewidelyusedinsensorapplications.WhenCCCRturnson,itsDCoutputvoltageincreasesatafasterrateandovercomestheHRatlargeinputvoltagelevels.ThereasonisthatDCbiaslevelsinbiasgeneratorareconstantwiththeRFamplitudewhilethehighertheRFinputvoltagelevel,thelowertheoptimalDCbiaslevelsformaximaloutputvoltage.HoweverthisperformancedegenerationathighinputRFlevelsislessbecauseasensorsystemneedsonlyasupplyvoltageof1.2Vfornormaloperations(in90nmCMOStechnology).Inaddition,providingtoolargeDCsupplyvoltageatlargeRFinputsignalwilldecreasethepowerefyofthetelemetrysensorsystem.Fig.3.18showsthePCEofthetwoatdifferentloadingresistorswhentheRFfre-quencyandpseudoPZTsignalarekeptconstant.WecanseethattheHRoutperformstheCCCRinlowinputpowerrange.At100kWloadingresistorand-12.9dBminputpower,thePCEoftheCCRisonly9.7%whileitisforHRfourtimeslargerwith40.3%.ThePCEoftheHRachieves39Figure3.18ComparisonofmeasuredPCEversusinputpowerPIN(fRF=13.56MHz,FPZT=10kHzandVPZT=2V).apeakvaluebeforeitdecreasesathigherinputpowerlevels.ThepeakvalueofPCEsincreaseswiththedecreaseofloadingresistorvalue.ItmeansthattheHRworksmoreeffectivelyatlargerloadingcurrentcondition.Itisbecausetherectifyingtransistorsaresizedathighaspectratio(W/L),hence,theyaremoreeffectiveingeneratinglargercurrents.However,comparedwiththeCCCR,thepeakPCEoftheHRbecomeshigherthanthepeakPCEofCCRwhenRLgetslarger.WithRLis680KW,forinstance,thepeakPCEoftheHRis14.36%at-14.6dBminputpowerwhilethepeakPCEoftheCCRis13.67%at-6.5dBminputpower.Itisalsonotedthatwhenloadingresistorincreases,thePCEcurveoftheHBRmovestowardtothesmallinputpowerregionandtheHRcanstartharvestingpoweratsmallerinputpowerlevel.AnothersetofexperimentswithaedRFsignalandvaryingamplitudeandfrequencyofthe40Figure3.19DCoutputvoltageoftheHRversePZTinputamplitudeatdifferentPZTfrequency(VRF;pkpk=300mV,FRF13.56MHzandRL=330KW).PZTsignalareconducted.Fig.3.19showsthedependencyoftheDCoutputvoltageonVPZTandfPZT.TheDCbiasofthehybridVMmaynotbeproperlygeneratedifthePZTamplitudeistoosmall.ItiscleartoseethattheDCoutputvoltageincreaseswithincreasingamplitudeofVPZTuptoanoptimalvalue.ThisoptimalVPZTvaluedecreaseswiththeincreaseoffPZT.Thereafter,itdecreasesgraduallywiththeincreaseofthePZT'samplitude.ThisphenomenonbecausetheDCboostinglevelofthegatecontrolsignalsincreaseswiththePZTamplitude.This,inturn,increasesthereversecurrentandlowerstheDCoutputvoltage.AnotherbehavioroftheHRisthatthemaximalDCoutputvoltageoftheHRdecreaseswithfrequencyofPZT.ThisisbecausethelowerPZTfrequency,thelargerthedriftoftheDClevelofthegatecontrollingvoltagesduetochargeleakageintothesubstrate,thusthelowerthemaximalDCoutputvoltage.3.5SummaryThischapterpresentsanovelCMOShybridthatcanefscavengeenergyfromacombinationofaradio-frequency(RF)signalandalow-frequency,low-energysignalfroma41piezoelectric(PZT)transducer.Thepiezoelectricsignalisusedforbiasingacomplimentarycross-coupledr(CCCR)chaintoanoperatingpointwheretheRFsignalenergycanbeefharvestedevenifitsamplitudeiswellbelowthethresholdvoltage(Vth)ofthedevice.TheoptimaldevicesizeshavebeenthoroughlyinvestigatedtoachievethemaximalDCoutputvoltage.Themeasurementresultshowsthatthe6-stagehybridcangeneratea1VDCoutputvoltageanda3mAoutputcurrentfromacombinationofa300mVpeak-to-peak,13.56MHzRFsignalanda10KHz-2VPZTsignal.TheproposedHRcaneffectivelyeliminatethedeadzoneandyieldsaimprovementinpowerconversionefy(PCE)forlowinputpowerascomparedtoaconventional(CCCR)onthesamedieina90nmCMOStechnology.42CHAPTER4TIMEDILATIONTECHNIQUEFORIMPULSIVEENERGYHARVESTING4.1IntroductionInadditiontothewirelessenergyharvesting,thevibrationenergyisalsoubiquitousintheenvi-ronment.Thusitisanaturalsourcefortheenergyharvestingdevices,especiallyforthesensorsthatareattachedtoorembeddedinsidethestructure.However,differentwiththewirelessenergythatisalmostconstantandcontinuous,thevibrationenergyischaracterizedtobevariableandintermittent.Insomeparticularconditions,thevibrationcausedbytheimpactcouldbeimpulsivesothatitisdiffortheharvestertocapturethewholeenergyduringshortpulseperiod.Inthischapter,anon-linearcompressivetechniqueisproposedandthecorrespondingtime-dilationcircuitispresentedforextendingtheenergyharvestingduration.Self-poweredsensorsoperatebyharvestingenergydirectlyfromthesignalbeingsensed[15].Forexample,apiezoelectrictransducercouldbeusedforsensingvibrationsorvariationsinme-chanicalstrainandtheenergyinthestrainvariationscouldalsobeusedforthecomputation,datastorageanddatatransmission.Oneofthechallengesindesigningself-poweredsensorsisthethresholdeffectwhichlimitsthesensing,poweringandhencetheoperatingrange[28].Thisop-eratingrange(insensingandpowering)isdeterminedbytwoenergythresholds:(a)minimumenergyrequiredtoactivatethesensor;and(b)maximumenergyabovewhichthesensorcouldincurelectricaldamage.Forexample,inhead-impactmonitoring,presenceoftheseenergythresh-oldslimitsthemeasurementrangeofimpactenergyandhencetheaccelerationlevelsthatcouldeasilyexceed100g[29].Ifapiezoelectrictransducer(forexample,PZTceramic)isusedasatransducer,suchhigh-levelsofaccelerationcouldeasilyproduceopen-loadvoltagelevelsinhun-dredsofvolts.Topreventdamagetothesensor,atypicalarchitecturewilldissipatemostoftheimpulseenergythroughtheover-voltageprotectioncircuits;andhencethesensorwouldbeunabletoaccuratelymeasurethemagnitudeofenergy[15].Anotherapproachistouseacompressive43Figure4.1(a)PoweringandsensingmechanismofthePFGhead-impactmonitoringsensor;(b)time-dilationapproachwheretheimpactenergyisspreadintime;and(c)compressiveresponseofatime-dilatedimpactmonitoringsensor.fishuntflregulatortypearchitecture[28]whichdissipatesamajorfractionoftheenergythrougharesistorandisthusabletocompresstheinputrange(asshowninFig.4.1(c))withintheminimumandmaximumthresholdlevels.Thismethodensuresthatthesensorcanbequicklyactivatedbysmallmagnitudeimpacts,whereascompressiveresponsebeyondanenergythresholdEminensuressensoroperationwithinelectronicsafetycompliancelevels.However,thedissipativenatureofthisapproachmakesitunsuitableforhead-impactmonitoringbecausethedurationoftheimpactistooshort(intheorderofafewmilliseconds)foraccurateenergymeasurement.Therefore,thetime-dilationapproachisproposedtoachieveacompressiveresponsesimilartoFig.4.1(c)andatthesametimeenableaccurateenergymeasurement.Theprincipleexploitstheimpulsivenatureofhead-impactswhichoccurinfrequently,asillustratedinFig.4.1(b).Thesparsitywithrespecttotimeisexploitedbythetime-dilationapproachwheretheimpulsivesignalisstretchedoutintime(asshowninFig.4.1(b))whileretainingasmuchoftheenergy(areaunderthecurve)astheoriginalimpulse.Beforethetime-dilationcircuitisdiscussed,it'sworthtodiscusstheequivalentcircuitmodelofapiezoelectrictransducerwhichhasbeenusedforoursimulationstudies.44Figure4.2(a)Equivalentcircuitofthepiezoelectrictransducer;(b)Simulatedpulseresponseofthepiezoelectrictransducerwith1MWload.4.2ModelingofthePiezoelectricTransducerForthiswork,apiezoelectrictransducerhasbeenusedforharvestingimpactenergy;therefore,anysimulationoranalysisofthetime-dilationbasedenergymeasurementhastoincorporateanequivalentcircuitofthepiezoelectrictransducer.ThemostcommonlyusedequivalentcircuitmodelthatcapturesthetransientresponseofapiezoelectrictransducerisshowninFig.4.2(a).ThevoltagesourceVS(µFµa)modelstheforceFappliedtothetransducerduetothehead45Table4.1Equivalentcircuitparametersusedinsimulation.CMRMLMNCPRL227.4nF57.87W1H0.0541780.08nF1MWimpactandisproportionaltotheaccelerationa.ThecapacitorCMmodelsthemechanicalstiffnessofthetransducer,theresistorRMmodelsthemechanicaldamping,andtheinductorLMmodelsthemechanicalmass.Themechanicaldomainiscoupledtotheelectricaldomain(accordingthefunctionofthetransducer)throughatransformerwithaturnratioofN.Intheelectricaldomain,CPistheequivalentintrinsiccapacitanceofthetransducerandRLrepresentsaresistiveloadconnectedatthetransducer'soutputterminals.Table.4.1[30]listthevaluesofthesecircuitparameterschosenforthesimulationstudy.Themechanicalimpacthasbeenmodeledasashort-durationpulsewhoseamplitudeisproportionaltotheappliedforce.Themagnitudeofthepulseisalsoproportionaltotheaccelerationlevelsandtheimpactenergy.TheresponseoftheequivalentcircuitofthepiezoelectrictransducerissimulatedasshowninFig.4.2(b)fora75Vinputpulsewithadurationof1ms.Theresonantresponseoftheoutputisdeterminedbytheresonantfrequency(productoftheequivalentinductanceandcapacitance)andthedecayisdeterminedbythetotalequivalentresistance.ThetransducerresponseandhencetheparametersoftheequivalentcircuitparametershavebeenvusingmeasuredresultsasshownintheinsetofFig.4.2(b)foraloadof1MW.4.3ProposedTime-dilationCircuitTheequivalentcircuitofthepiezoelectrictransducercanbeusedtostudytheresponseofatime-dilationcircuitshowninFig.4.3.Forcomparison,Fig.4.3alsoshowsanequivalentcircuitwithoutthetime-dilationcircuit.Inbothofthecircuits,ILmodelstheloadcurrentofthesensor,CListheloadcapacitance,andDLrepresentsazenerdiodewhichmodelsthefunctionalityofanover-voltageprotectioncircuit.Inpractice,thecurrentIListriggeredonlywhenthevoltageVOUT46Figure4.3Equivalentcircuitofaself-poweredsensor:(a)without,and(b)withthetime-dilationcircuit.exceedsaminimumthresholdlevelVMIN.Thefull-waveisachievedusingthediodesDR1DR4formedusingabulk-drivencross-coupledpMOScircuitwhichwasreportedin[31].Forthecircuitwithoutthetime-dilation(showninFig.4.3(a)),thecurrentofIPin-creasesthevoltageVOUTpasttheminimumthresholdVMINwhichthenactivatesthesensorandhencethecurrentIL.IfthecurrentIPexceedsthesensorcurrentIL,thevoltageVOUTwillincreaseuntilthezener,DL,startsdrawingtheextracurrent.Thus,forlargeimpulsecurrentsmostoftheimpactenergyisdissipatedthroughthezener.Thetime-dilationcircuitshowninFig.4.3(b)mit-igatesthisprobleminthefollowingmanner.Initially,thesensorvoltageVOUT=0impliesthatthenMOStransistorNTisOFFandhencethestoragecapacitorCTisdisconnected.Therefore,duringstart-upthebehaviorofcircuitinFig.4.3(b)isidenticaltothestart-upforthecircuitinFig.4.3(a).ButwhenthevoltageVOUTexceedsKVTH,withKbeingthenumberofdiodesinthechainD1ŒDKandVTHbeingthethresholdvoltageofthetransistorNT,thestoragecapacitorCTisconnectedinparallelwiththesensor.Thus,theadditionalpiezoelectriccurrentnowchargesupthestoragecapacitorwithoutdissipatingenergythroughthezener.Whentheimpulsesubsides(aftertheimpact),thechargestoredonthecapacitorCTdrivesthesensorcurrentuntilVOUTfallsbelowtheactivationvoltageVMIN.Thus,byusingtime-dilationcircuit,thesensorcanmoreaccuratelymeasurethelevelofimpactenergy.Fig.4.4comparesthesimulationresultsoftheoutputvoltageofthefullwavewithandwithouttime-dilationcircuitforinputpulseofmagnitudes100Vand300V,respectively.Itis47Figure4.4Comparisonbetweentheoutputvoltagesofthefullwavewith(redline)andwithout(blueline)time-dilationcircuitswithrespecttotheinputpulseof(a)100Vand(b)300V.clearlyshownthatwithouttime-dilationcircuit,thefullwavejustfollowstheinputbutthemaximumvoltageislimitedbytheprotectioncircuit,thusitisunavoidabletowastetheenergythroughtheprotectioncircuit.Incontrast,byusingthetime-dilationcircuit,theoutputvoltagecanbebuiltupprogressivelyonceNTisturnedONandcanbeheldbythestoragecapacitorCTevenaftertheinputsubsides.Theover-voltageprotectionisdeterminedbythebreak-downvoltageofthereversep-njunctionwhichfora0.5mmCMOSprocessisapproximately10V.TheminimumvoltageVMINisdeterminedbythevoltagerequiredtoactivatehot-electroninjectionontheateinjectors.Thisvoltagehasbeenexperimentallydeterminedtobearound6V,basedonourpreviouswork[31].Thethresholdvoltageofthediodes(proposedandmeasuredin[31])isapproximately1V,whichleadstothenumberofdiodesinFig.4.3(b)asK=6.Thus,thereissuftimeforthesensortomeasuretheleveloftheimpactenergy.Inaddition,thetimethattheenergyisstoredonthecapacitoralsoincreasesastheinputpulselevelincreases.Itisalsoshowninthesimulationthatthestart-upresponsesareidenticalforbothcircuits.Fig.4.5comparestheenergythatcanbemeasuredforthesensorwithandwithouttime-dilationcircuit.Astheinputpulselevelincreases,theenergythatcanbemeasuredmonotonicallyincreasesforbothHowever,itisclearlyshownthat:byusingthetime-dilationcircuitthesensorcanharvestmoreenergyfromtheinputimpulseandhenceshouldbeableto48Figure4.5Comparisonbetweenthestoredenergywith(redline)andwithout(blueline)time-dilationcircuitastheinputpulselevelincreases.activatemoresensorfunctions.4.4DifferentialandRotationalImpactAsdescribedintheprevioussection,asinglepiezoelectrictransducerbasedtime-dilationcircuitcouldbeusedtomeasuretheenergyofanimpactthatisspatiallylocalizedaroundthepositionofthetransducer.ThisisshowninFig.4.6(a)whichcorrespondstothecasewhenaregionofthehelmetexperienceslinearacceleration.However,forconcussionstudiesthestatisticsofrotationalaccelerationisalsoimportant.Thehelmetexperiencesarotationalacceleration(aboutsomereferenceaxis-typicallythenecklineoftheplayer)whenitexperiencesalateralimpactasshowninFig.4.6(a).Inthiscase,regionsofthehelmetthataresymmetricaboutthereferenceaxiswillexperiencedifferentlevelsofimpactenergiesandwillresultintherotationalmotion(orsnap)ofthehead.Thecaseforlinearaccelerationandrotationalaccelerationcanbediscriminatedusingpiezoelectrictransducersplacedatdifferentspatiallocations.InFig.4.6(b)weshowmeasuredresponsesfrompiezoelectrictransducersplacedatthreelo-cationsonthehelmet.Foradirecthead-onimpact,thefrontpiezoelectrictransducergeneratesalargeoutputpulsesignal,andthepiezoelectrictransducerslocatedatthesidesofthehelmetshow49Figure4.6(a)Illustrationofthelinearandrotationalaccelerations;(b)Measuredpulseresponsesofthesensorsinthreepositionswithrespecttothedirectandlateralimpacts,respectively.similarbutweakresponses.Forthelateralimpact,thepiezoelectrictransducersonthetwosidesgeneratestrongandcomplementaryresponsesasexpected,andthefrontpiezoelectrictransducershowaveryweakresponsebecausetheimpactisparalleltothefronttransducer.Basedontheseobservations,energyduetoalinearaccelerationimpactcanbemeasuredusingasingletransducerandhasalreadybeenvinthesimulationshowninFig.4.4.Theimpactenergycausingtherotationalaccelerationcanbemeasuredusingapairoftransducersinadiffer-entialasshowninFig.4.7(a).Thepositiveterminalsofthetransducersareusedastwoinputsofthewhilethenegativeterminalsofthetransducersareconnectedtogether.Thus,thedirectandlateralimpactsshowninFig.4.6(b)canbeseenasthecommon-modeanddifferential-modeinputsforthefullwave,respectively.Similartotheprevioussimulationstudy,theequivalentcircuitinFig.4.7(a)isusedtounder-standthetime-dilatedresponseformeasuringrotationalimpactenergy.Fig.4.7(b)presentsthe50Figure4.7(a)Apairofpiezoelectrictransducersareconnectedindifferentialtotesttherotationalacceleration;(b)Thesimulatedoutputvoltageoftheforthecommon-modeanddifferential-modeinputsignalwhenusingdifferentialconnectedpiezoelectrictransducerswithtime-dilationcircuit.simulatedoutputvoltageoftheforthecommon-modeanddifferential-modeinputsignalwhenusingdifferentialconnectedpiezoelectrictransducerswiththetime-dilationcircuit.ItcanbeseenthattheofpiezoelectrictransducersinFig.4.7(a)onlyrespondstothediffer-51entialsignal(duetorotationalimpact)butnottothecommon-modesignal(thatcausedbylinearacceleration).4.5PFGbasedSelf-poweredEnergyMeasurementandData-loggingTheate(PFG)basedself-poweredsensingwasreportedin[15]andisde-scribedhere.PFGbasedsensorsexploitsaninterestingpropertyoftheceramicandpolymerbasedpiezoelectricmaterials(e.g.PZTandPVDF),whichisabletogeneratelargeopen-sourcevoltage(>10V/me),withpico-to-nanoamperesofcurrentdrivingcapability(duetolargecapacitivecou-pling).Thisproperty,however,isidealfortriggeringimpact-ionizedhot-electroninjection(IHEI)insilicontransistorsrequiringhigh-voltageswhileoperatingwithultra-lowcurrents.Ifthegateofthesilicontransistorisisolatedbythehighqualityinsulatingoxide,suchasgatetransistortheinjectedelectronsremainthereforlongintervalsoftime(>8yearsfor8bitsprecision).Asthepiezoelectricelementisperiodicallyexcited,moreelectronsareinjectedontothegateandthetotalamountofchargeonthegateindicatesthedurationandextentofthemechanicaldisturbance.WehaveshownthatthePFGself-poweringdeviceiscapableofoperatingatpico-watt(1012Œ109W)powerdissipationlevels,withitsresponseremaininginvariantformechanicalstrain-levelsdowntoafewmicro-strains(me).ThisenablesthePFGdevicetopotentiallysense,computeandstoreatthefundamentallimitsoftheharvestableimpact-energy.Inthiswork,thePFGprincipleisusedtomeasurethemagnitudeofimpactenergywhichdeterminestheamountofchargethatisdepositedonthetime-dilationcapacitorinFig.4.3(b).Theamountofchargecanbeestimatedbythetimeittakestodischargethetime-dilationcapacitor.However,forself-poweredoperationthemeasureddataalsoneedstobestoredonanon-volatilememoryforsubsequentretrieval.Self-poweredandnon-volatiletime-measurementcanbereadilyimplementedusingourpreviouslyreportedlinearateinjectortopology[32].ThecircuitlevelschematicofthelinearateinjectorisshowninFig.4.8.ThecircuitconsistsofaatepMOStransistorMfgwhosesourceisdrivenbyaconstantcurrentsourceIrefwhichis52Figure4.8Schematicofthedata-loggingcircuitwithaatebasedlinearinjectorandaspikinganalog-to-timeconverterpoweredbyeitherapiezoelectrictransducer(duringbattery-lessoperation)orbyanenergysourceVext(whenthedataisbeingretrievedbyareader).NotethattheenergysourcesareisolatedbyadiodewhichallowsVexttosupersedethesignalgeneratedbythepiezoelectrictransducer.TheopampA(connectedtoVref),theconstantcurrentsourceIref,andtheatetransistorMfgformanegativefeedbackwhentheswitchSPisopen,thusensuringthatthesourceVsandgateVgvoltagesremainconstant.SincethedrainvoltageofMfgistiedtoground,ifthereferencevoltageVrefexceeds4:2V,thenegativefeedbackwillcausethesource-to-drainvoltageofMfgtoalsoexceed4:2V,whichcauseselectronstoinjectontotheate.Becauseallterminalparametersoftheatetransistorareheldconstantduringtheinjectionprocess,theinjectioncurrentIinjremainsconstant.Thus,theamountofchargeinjectedontothegate,orthedecreaseinatevoltageVfg,isproportionaltothedurationforwhichthesourcecurrentIrefisactivatedandSPisopen.Thiscanbeexpressedas:DVfg=1CTZT0Iinjdt=IinjCTt(T);(4.1)wheretisthedurationofinjectionandCTisthetotalatecapacitancewhichincludesCfg,tunnelingcapacitance,andotherparasiticcapacitancesassociatedwiththenode.ThechangeinatevoltageDVfgismeasuredbyclosingtheswitchSPwhichbreaksthefeedbackloopbyshortingtheotherterminalofCfgtoground.BecausethesourcecurrentIrefisconstant,DVout=DVfgwhichisread-outusingaanalog-to-timeconverterasshowninFig.4.8.The53analog-to-timeconverterusesamultiplexertoswitchthevoltagereferenceofacomparatortooneofthreevoltagesVout,Vcal1andVcal2.ThecalibrationvoltagesVcal1andVcal2aretheoutputsoftwolinearinjectorswhosevaluesremainedandareprogrammedonlyduringthesensorinitiation.Thepurposeofthesecalibrationvoltagesistocompensateforanygainandoffseterrorsintheanalog-to-timeconverter.TheoutputDoutofthecomparatorperiodicallytogglesthepMOSandthenMOSswitches,oneatatime.ThusthecurrentIPchargesthecapacitorCADCtothevoltageVsafterwhichthecurrentIListurnedONwhichdischargesCADCandgeneratesaspike,asshownintheinsetofFig.4.8.Thetime-differenceToutbetweentwospikesisproportionaltothevoltageinputwhichinturnisproportionaltotheatevoltageVfg.Forcalibration,theMUXthenselectsthecalibrationvoltageVcal1andmeasuresthecorrespondingtime-differenceTcal1afterwhichitselectsthenextcalibrationvoltageVcal2andmeasuresthecorrespondingtime-differenceTcal2.TheoutputvoltageVoutcanbedeterminedintermsofToutandthecalibrationparametersas:Vout=Vcal1Vcal2Tcal1Tcal2(ToutTcal1)+Vcal1:(4.2)4.6MeasurementResultsThesystemarchitectureofthePFGsensorICisshowninFig.4.9(a).Itconsistsof21linearateinjectorchannelswhere14ofthechannelsareprogrammedtotriggeratdifferentlevelsofimpacts.The7remainingchannelsareusedforstoring:(a)anioncodeforthesensor;(b)placementlocationonthehelmet;and(c)calibrationvoltagesfortheanalog-to-timeconverter.Thesystemarchitectureisdividedintotwomainmodules:(a)theself-poweringmodule;and(b)theprogrammingmodule.Duringtheprogrammingmode,anexternalsupplyVddD(˘2V)activatesthehigh-voltagecharge-pumpswhichgeneratethesupplyvoltageVddatodrivethesensorandtheread-outbuffers.Thehigh-voltagecharge-pumpsarealsousedforgener-atingthetunnelingvoltageVTun(˘16V)whichisusedforerasingthechargeontheatesofthelinearinjectors.APoweronReset(PoR)generatesthereset(RST)signalwhichisusedto54Figure4.9(a)SystemarchitectureforthePFGsensor.(b)MicrographofthesensorICintegratingdifferentmodules:1.FloatingGateArray;2.DigitalDecoder;3.TunnelingVoltageCharge-pump;4.LevelShifter;5.InjectionControl;6.DiodeProtection&7.VoltageReferences;8.TunnelingCharge-pump;9.InjectionCharge-pump;10.RingOscillator;11.Analog-to-timeConverter;12.Supportingcircuitry,power-onreset,buffers,etc.Table4.2CommandDescriptionRESETClearallregistersandselectthechannelNEXTSelectthenextchannelINJECTReducethevalueofcurrentchannelTUNNELIncrementthevalueofallchannelsinitializethecontentsandstateofallthedigitalregistersandstate-machines.Adigitalprocessorcontrolsthehot-electroninjection,tunnelingandread-outthroughdigitalcontrollines(RESET,NEXT,INJECTandTUNNEL).Thefunctionalitiesofeachofthesecontrolsignalsaresumma-rizedinTable4.2.Thecircuitlevelimplementationformanyoftheread-outandcharge-pumpmoduleshavebeenpreviouslyreported[15]andisomittedhereforthesakeofbrevity.Apro-totypePFGsensorICwasfabricatedina0.5mmCMOSprocessanditsmicrographisshowninFig.4.9(b)alongwithitsmeasuredsummarizedinTable4.3.ThesensorICwasthenintegratedonanexternalprintedcircuitboardasshowninFig.4.10(b)whichhoststhesensorIC,thetime-dilationcapacitor,thepiezoelectricinterfaceandtheprogram-minginterface.Thesizeofthesensoris2cm1.5cmandeasilyinbetweenthehelmet'scushionpads.Forourvalidationstudies,theprototypesofthesensorboardswereattachedtoa55Figure4.10Experimentalsetupusedinthedrop-tests:aCOTSfootballhelmetwithembeddedIntegratedPFGsensorprototypes(inset).Table4.3InterfaceoftheSelf-poweredICProcess0.5mmCMOSprocessSize2.5mm2.5mmEnergyRequirementforSelf-poweredMode100nJPowerDissipationforRead-outMode75mWPowerDissipationforProgrammingMode150mWSupplyVoltage1.5VŒ2.5VRead-outResolution<8bits@10KspsTable4.4TechnicaldetailsofthehelmetBrandNameRiddellItemWeight4.8poundsItemSize1300L9:500W1000HMaterialTypePolycarbonLexanShell-SteelandPolyvinylCoatedMaskfootballhelmet(technicalsummarizedinTable4.4)asshowninFig.4.10(a).Fig.4.11showsmeasuredresultsfromafabricatedprototypecomparingthevoltageVOUTunderthreedifferentconditions:(a)withoutanytime-dilation,basedontheschematicshowninFig.4.3(a);(b)time-dilationusingCT=50nF;and(c)time-dilationusingCT=1mF.AsshowninFig.4.11(a),thezenerclipstheoutputvoltageatVMAX=10Vandthesensorshutsdownaftertheimpulsedecays.Forthetime-dilationcircuitinFig.4.11(b),thestoragecapacitorCTholdsthe56Figure4.11MeasuredresultsshowingVoutforthesensorwith:(a)notime-dilationcircuit;(b)time-dilationcircuitwithCS=50nF;and(c)time-dilationcircuitwithCS=1mF.extracharge.NotethatinFig.4.11(b),thestart-upresponseremainsunaffectedwhichisimportantforloggingdifferentlevelsofimpact.TheresultinFig.4.11(c)showsthatchoosingtherightrangeofvaluesforCTisimportantasalargestoragecapacitor(eventhoughitcanstoremorecharge)willtakealongertime(hencedurationofimpulseevent)topushthevoltageVOUTbeyondVMINandactivatethesensor.Themeasurementsetupemulatedasmallerscaledrop-testprocedurereportedin[33].Thehelmetwiththeintegratedsensorwasdroppedfromtwoheightlevels:HeightA(1foot)andHeightB(2feet).Forlinearimpacttest,thehelmetwasdroppedheadalongaverticalaxispassingthroughthetopofthehelmet.Fig.4.12showstheoutputofthepiezoelectrictransducer(withoutthesensorattached)whenthehelmetisdroppedfromheightsAandB.Theresponseclearlyshowstheimpulsivenatureofthepiezoelectricsignalgenerationandthevoltagelevelclearlyshowstheneedforthetime-dilationapproach.Fig.4.13(a)showsthedatarecordedfromthesensorwhenthehelmetisdroppedrepeatedlyfrom1foot.Notethatevenatthisheightthethreechannelsrecordthelevelofimpactindi-57Figure4.12SignalrecordedattheoutputofaPZT-5Hpiezoelectrictransducer(10MWload)whenthehelmetisdroppedfrom1foot(heightA)and2feet(heightB),respectively.Figure4.13Measuredoutputfromthreeofthesensorchannelswhenthehelmetisrepeatedlydroppedfrom(a)1foot;(b)differentheights.catedbythechangeintheiroutputvoltage.Thelinearityoftheresponseshowsthatthevoltagemeasurementscouldbecalibratedtodifferentimpactenergylevelswithcorrespondingaccelera-tionlevels.Fig.4.13(b)showsdatarecordedfromthesensorwhenthehelmetwasdroppedfrom3differentheights.TheControlthatthehelmetwasjusttappedwhileatrest.Theresultshowsthatthesensor'sgainincreasesastheheightincreasesandhencetheconcussionenergy,58Figure4.14(a)Measuredimpactdistributionusingthreesensors:sensor1isinthefront,sensor3isinthesideandsensor2isinbetween;(b)RotationalimpacttestusingadifferentialPFGsensor.validatingthesensor'sabilitytologdiverselevelsofimpact.Fig.4.14(a)presentsthedistributionofthedatameasuredusingthreesensorswhere:sensor1waslocatedinthefrontofthehelmet,sensor3waslocatedonthesideandsensor2waslocatedinbetween.Duringthedroptest,thefrontpositionofthehelmetdirectlyimpactstheground,thusthechangeofthesensor1outputisthelargestofallthree.Thesensor3wasmountedinthesideofthehelmetanditstransducerisalignedparalleltothedirectionoftheimpactforce,thus,itsoutputremainsrelativelyunchanged.Mountedbetweensensor1and3,thesensor2alsorecordstheoccurrenceoftheimpactbutshowsalowergainthansensor1.TherotationalimpacttestwasalsoperformedinwhichapairofPZTtransducersweremountedontwosidesofthehelmetandtheiroutputsweredifferentiallyconnectedasshowninFig.4.7(a).Forthistest,thehelmetwasdroppedalonganaxispassingthroughthesideofthehelmet.Asdescribedbefore,thedirectandlateralimpactscanbeseenasthecommon-modeanddifferential-modeinputsforthesensor,respectively.TheresultinFig.4.14(b)clearlyshowsthatthesensorcanrecordtheinstancesofrotationalimpact(acceleration)andremainsunaffectedbytheinstancesofdirectimpact(orlinearacceleration).594.7SummaryThischapterpresentsaminiaturesensorsystemthatcanbeusedtomonitorthelevelandfrequencyofheadimpactsinhelmetedsports.Atime-dilationcircuitenablesthesensortomeasurehighenergyimpulsesandalinearateinjectorenablesthesensortorecorddataonnon-volatilememory.Thesensorisself-poweredandoperatesbyharvestingtheenergyfromthehead-impactwithnoneedforbatteries.Itisdemonstratedthatdifferentandplacementsofthesensorinsideahelmetcanbeusedtosense,measureandrecordspatialandtemporalstatisticsoflinearandrotationalheadimpacts.Itisenvisionedthatduetoitssmallform-factoranditslow-cost,alargearrayoftheseself-poweredsensorscouldbeembeddedatdifferentlocationsinsidethehelmet.Thehistoricalinformationoftheimpactsandthetypeandnatureoftheimpactsrecordedbythesensorscouldbeimportantmappingdatafortraumaticbraininjury(TBI)researcherstoprognosticateconcussionsduringthecourseofnormalplay.Notethattheactivationthresholdofthesensor(minimumimpactenergy)canbeadjustedbychangingthesizeofthetime-dilationcapacitor.Inthismanner,thesensorcouldbecustomizedfordifferentspatiallocationsandfordifferentplayers.60CHAPTER5ACMOSSYSTEM-ON-CHIPFORPASSIVE,NEAR-FIELDULTRASONICENERGYHARVESTINGANDBACK-TELEMETRY5.1IntroductionWhilepassiveradio-frequency(RF)basedtaggingandback-telemetrysystemsareattractiveforembeddedandimplantedmonitoringofenvironmentalandbiologicalprocesses[15,34Œ38],thetechnologycannotbeusedinsideconductiveorionicmediawhereRFattenuationcouldbesig-AnalternativesolutiontoRFbasedtelemetryisacousticorultrasonicbasedtelemetrywhichhasbeendemonstratedtoexhibitlowattenuationinsideconductivemedia[39Œ46].Theuseofultrasoundalsoallowsminiaturizationoftheembeddedtelemetrysystembyrelaxingthesizerequirementsonthepiezoelectrictransducer.Also,unlikeRFtransmissionwhichisregulatedaccordingtoFCCrequirements,ultrasonicpowerrequirementisonlylimitedbystructure'sme-chanicalcompliance.Intheliterature,severalultrasonictelemetrysystemshavebeenreportedforuseinmetallicstructures[47Œ49]andforin-vivoapplications[50].In[41Œ46]passiveultrasonictelemetrysystemshavealsobeenreportedwhicheliminatetheneedforbatteriesonthesensors.However,thesesystemsareeitherbulkyorpowerhungryandthusarenotsuitableformassivedeploymentinpracticalapplications.Thischapterpresentsthesystem-on-chipimplementationofanearpassiveultrasonictelemetrysystemthatcouldbeusedforembeddedmonitoringinsidetheconductivestructuresandionicmedia.Nearoperationallowsoptimizationoftheback-telemetrylinkviaresonancetuningatthereadersothatthecomplexityofthetagcanberelaxed.Inaddition,oncetheresonantpowerandtelemetrylinkhasbeenestablished,nearoperationmakestheinterrogationandcommunicationimmunetothemulti-pathfromthematerialboundariesanddisconti-nuities.AnillustrationofanultrasonictaggingsystemisshowninFig.5.1whereanultrasonicreaderscansthesurfaceofaconductivestructuresurroundinganarrayofembeddedsensors/tags.61Figure5.1Illustrationofanultrasonicbasedtelemetrysystemusingacousticcouplingthroughthemetalbarrier.Theroleofthesesensors/tagsistoconveymeasurementsofsomephysicalpropertyofinterestforexample,temperature,pH,corrosionorhumiditybacktothereader.Undernearoperatingconditions,theacousticimpedanceofthepiezoelectrictransduceratthesensor/tagisasanequivalentelectricalimpedanceonthereader'stelemetrycircuits.Toestablishtwo-waycommunications,thesensor/tagharvestsenergyfromthereadersignalandthenelectricallymodulatestheacousticimpedanceofitspiezoelectrictransducer.Achangeinthesensor'sacousticimpedanceisasachangeinthedrivingcurrentthroughtheultrasonicreaderandthenthecurrentchangeisdemodulatedtorecoverthesensordata.5.2SystemOverview5.2.1UltrasonicpowertransferandtelemetrymodelThemostcommonultrasonictransducersarethepiezoelectrictypematerialslikeleadzirconimumtitanate(PZT)andcanconvertacousticenergyintoelectricalenergyorviceversa.Fig.5.2(a)showsaninterfacecomprisingofatransmitPZTtransducerandareceivePZTtransducerwhichareseparatedbyametallicbarrier.ThetransmitPZTisexcitedbyanACvoltagesourceVINdriventhroughasourceresistanceRS.ThegeneratedultrasonicwavepropagatesthroughthemetalbarrierandimpingesonthereceivePZTwheretheacousticstresswaveisconvertedinto62Figure5.2UltrasonicpoweringandcommunicationsystembasedonMasonmodel:(a)systemdiagram;(b)equivalentcircuitmodel.anelectricalsignalwhichthendrivesaloadresistorRL.Forthesakeofsimplicity,inFig.5.2(a)anairbackinghasbeenusedasareplacementforthefrontandtailmass[51].Foradisk-shapedtransducerwithdiameterDtheultrasonicneardepthLisgivenby[52]L=D2f4v(5.1)wherefistheultrasonicfrequencyandvisthevelocityoftheultrasonicwaveinthebarrierma-terial.Thus,foraeddiameterofthetransducerandthevelocityoftheultrasound,alarger63neardepthcanbeachievedathigherfrequency.Inthispaperwehavechosentheoperatingfrequencyf=13.56MHzwhichforatransducerwithD=22mmandv=6400m/s(forAluminum),leadstoaneardepthofL=256mm,suffortelemetryacrossmillimetersizedbarriers.Anotheradvantageofusingahigheroperatingfrequencyisthatitallowstheuseofon-chipcapac-itorsintheenergyharvestingcircuitsaswillbedescribedinsectionIII.,a13.56MHzoperatingfrequencyisalsocompatiblewiththeISMbandRFIDstandardwhichenablesreuseofcommercialoff-the-shelfRFIDreadersfortheproposedultrasonicback-telemetry.Theefyofthisnearultrasoniclinkcanbeanalyzedusingacousticwaveequationsasdescribedin[41].However,theanalyticalsolutionstothewaveequationsinpiezoelectricmaterialsareverycomplex,thusforone-dimensionalanalysistheacousticlinkcanbeusingtheMason'sequivalentlumpedcircuitmodel[53]asshowninFig.5.2(b).ThetransformersN1andN2connecttheacousticalandtheelectricaldomainswhereallthemechanicalparametersofthepiezoelectriclayerandthebackinglayercanbetransformedtoanequivalentT-networkasshowninFig.5.2(b).TheelectricalimpedancesZTkandZSkcorrespondtothekthelementandcanbeestimatedforafrequencyfasZTk=jZ0ktan(2pftk2vk)ZSk=jZ0ksin1(2pftkvk)(5.2)whereZ0kandtkarethecharacteristicimpedanceandthethicknessofthekthlayerandvkisthevelocityoftheultrasoundinthekthlayer.ThecharacteristicimpedanceZ0kcanbeestimatedasZ0k=rkvk(5.3)withrkbeingthedensityofthekthlayer.Thepowertransferefy(PTE)fortheultrasonicmodelcanbeestimatedbyinsertingthematerialpropertiesofthePZTtransducerandtheacousticpropertiesofthemetallicbarrier(summarizedinTable5.1)intothemodel.NotethatthePTEisgivenbyPTE=POUTPIN=Re(VLIL)Re(VINIIN)(5.4)64Table5.1Materialparametersusedforlinksimulations.ParametersPZTAluminumDensityr(kg/m3)78002700Velocityv(m/s)28206400Thicknesst(mm)0.250.1˘4AreaA(mm2)380380PermittivityeS33(F/m)4:78108N1N233.93Figure5.3ThesimulationofpowertransferefybasedonMasonmodelasafunctionofmetalthickness.wheretheloadcurrentILandloadvoltageVLaredeterminedusingcircuitsimulations.Fig.5.3showsthesimulationresultwherethePTEofthemodelinFig.5.2(b)hasbeencom-putedformetalthicknessrangingfrom0.1mmto4mm.ThetransmitPZTandthereceivePZTwerechosentohavesamedimensionsandexhibitedsimilaracousticalpropertiesasaPZT-5Hmaterial.Thefrequencyoftheinputsourcewassetto13.56MHzandtheloadresistorRLwascho-sentobe100W.Itcanbeseenfromtheresultthatthepowertransferefydropsrapidlyasthedistanceincreasessothattheenergyharvestingdistanceisverylimited,orthepowertransferefyistoosmall.65Figure5.4Proposedultrasoniccommunicationsystem.Figure5.5Powermanagementmoduleswhichincludeavoltagemultiplier,avoltagelimiterandaregulator.5.2.2SystemArchitectureThesystemarchitectureofaultrasonicback-telemetrysystemisshowninFig.5.4.Thereadercomprisesofthedigitalcontroller,theanalogfront-endandtheimpedancematchingcircuit.Thedigitalcontrollerimplementsastatemachinethatsendsandreceivescommandstoandfromthetag.Theanalogfront-endmodulatesthedigitalsignalwithultrasoniccarrieranddemodulatesthebackscattersignal.Theimpedancematchingcircuitisusedtocreatearesonantnetworkcompris-ingofthethetransmitPZT,thetransmissionmediumandthereceivetag.ThereceivertagconsistsofareceivePZTandamatchingcircuitwhichformsapartoftheultrasonicresonantnetwork.66Figure5.6Themeasurementresultsfordifferenttypesofvoltagemultipliers(12-stage,18-stageand24-stage)undertheloadresistorof10MWand1MW.Throughthisnetwork,thetagreceiveselectricalpowerandapowermanagementmodulecompris-ingofvoltagemultipliersandvoltageregulatorsgeneratestablesupplyvoltagesforotheron-chipmodules.Ademodulatorextractstherawdatareceivedovertheresonantnetworkandadigi-talstate-machineperformserror-correctionanddecodesthecommandsreceivedfromthereader.Thestate-machinealsocontrolsananalog-to-digital(ADC)converterwhichsamplesanddigitizesthesensorsignalfortransmissionbacktothereader.ReversetransmissionisachievedthroughbackscatteringbyturningONandOFFaswitchconnectedtothereceivePZT.Othermodulesinthesystem-on-chipdesignincludesapower-on-reset(POR)circuitwhichisusedforinitializingallthestatesofthedigitalmodules.Anintegratedring-oscillatorprovidesstableclockfordigitalbasebandandtheADC.5.3CircuitDesignoftheUltrasonicTag5.3.1PowerManagementCircuitsDuetotheattenuationoftheultrasonicsignalinsidethesolidmediumandthetransducer'slimitedenergyconversionefyaswellastheimpedancemismatchatvariousinterface,themagni-67Figure5.7MeasurementsofLDO:(a)dropoutvoltagefordigitalblock;(b)dropoutvoltageforanalogblock;(c)lineregulationforanalogblockwhenVIN=0Vto10V;(d)linetransientresponseforanalogblockwithIL=0.4mA,VIN=0Vto6V;(e)linetransientresponseforanalogblockwithIL=0.4mA,VIN=6Vto0V;(f)ripplerejectionforanalogblockwithIL=0.4mA;(g)linetransientresponseforanalogblockwithIL=400mA,VIN=0Vto6V;(h)linetransientresponseforanalogblockwithIL=400mA,VIN=6Vto0V;(i)ripplerejectionforanalogblockwithIL=400mA.tudeofthesignalinducedatthereceivePZTistypicallybelow1V.Thereforeavoltagemultiplierisrequiredtotransferchargeandboostvoltageonastoragecapacitor[54,55].Forthisimplemen-tationwehaveusedastandardDicksontypemultiplier[56]asshowninFig.5.5.ThemultiplieriscomprisedofSchottkyjunctiondiodes(withapproximatethresholdvoltageofVthˇ300mV)andproducesanoutputvoltageVOUTˇN(VINVth),withNbeingthenumberofmultiplierstages.68Figure5.8(a)Blockdiagramand(b)detailedcircuitdesignofdatarecoverycircuit.ThemeasurementresultsinFig.5.6(a)and(b)showstheoutputofthemultiplierwithdifferentnumberofstagesandmeasuredunderdifferentloadingconditions.Thepowertransferefyofthevoltagemultiplierhighlydependsontheinputpowerlevelandtheloadingcondition.Forinputpowerlevelaround-10dBm,thevoltagemultiplier'efyisaround10%with100kWofloadresistor.Inaddition,tolimitthemaximumvoltage(inthiscase11V)generatedbythemultiplier,theoutputvoltageisclampedusingadiodechainasshowninFig.5.5.IfthemultiplieroutputVDCisbelow11V,thevoltagelimiteronlyconsumesnAlevelcurrentsothatthelimiter'sefycanbethoughtnearly100%.Consideringthatthemultiplieroutputwillwiththemagnitudeofthereceivedpowerandissusceptibletohigh-frequencyripples,aregulatorhasbeenusedtoobtainastableDCoutput69voltage.Inthiswork,wehaveimplementedahybridtopologycombiningtheefyofaswitchingregulatorwiththesimplicityofripple-rejectionabilityofalinearregulator.ThecircuitisshowninFig.5.5andconsistsofacurrentreferencecircuitformedbytransistorsM1˘M10andaregulatorcircuitformedbytransistorsM11˘M13andMD1˘MDN.TransistorsM1,M2andC1formthestart-upcircuitinFig.5.5andtransistorsM3˘M10,C2andRgenerateasupply-insensitivecurrentwhichisgivenbyIrefˇUT=Rlog(M)withMbeingtheratiooftheaspectratiosfortransistorsM9andM10.Oncethecurrentreferenceissetup,M2isturnedonandM1isturnedofftominimizethepowerconsumption.Notethataminimumvoltageisrequiredforthecurrentreferencetobeenabled.BelowtheminimumvoltagelevelthecurrentreferenceisOFF.Thisleadstoaswitchingresponsesimilartoaswitching-typeregulatorwhereiftheloadvoltagegoesbelowacertainthreshold,theloadisdisconnectedandthestoragecapacitorisrechargedtillthevoltagelevelexceedsthethreshold.Oncetheinputvoltagelevelisgreaterthantheminimumstartupvoltage,thecircuitformedbythetransistorsM11˘M13andMD1˘MDNactasalinearregulator.BecausethecurrentwingthroughthediodechainMD1˘MDNisheldconstantbythecurrentreference,thevoltageVDDisheldconstantbythefeedbacktransistorM12andtheoutputtransistorM13.NotetheloadcurrentwsthroughtheoutputtransistorM13,thereforethedrivingcapabilityofM13hastobeadjustedbyincreasingitsaspectratio.TheefciencyofthislinearregulatorisdeterminedbythepowerdissipationduetoM13whichisafunctionoftheloadcurrentandthevoltagedropVDCVDD.C3isusedtostabilizetheoutputvoltagewhentheloadcurrentIL6=0,anddisablethecurrentthroughM13whenIL=0.Forthisdesign,theregulatorwasdesignedforamaximumloadcurrentof1mA,sothesizeofM13waschosentobe300mm/0.6mmandthesizeofC3waschosentobe1pF.SomeexperimentalresultscorrespondingtotheproposedregulatorareshowninFig.5.7wheretworegulatorsweredesignedonefordrivingtheanalogmodulesandtheotherfordrivingthedigitalmodules.Thepresetoutputvoltagesfordigitalandanalogregulatorswerechosentobe2Vand4V,respectively.TheinputtotheregulatorwasatrianglewaveasshowninFig.5.7(a)and(b)anddropoutvoltageswerearound0.8Vand1.2Vfordigitalblockandanalogblock,respectively.70Figure5.9Communicationprotocol:(a)PIE;(b)Manchester.ItisnotedthattheresultsofregulatorforanalogblockareshowninFig.5.7becauseregulatorfordigitalblockexhibitthesimilarbehavior.Thelineregulationmeasuredfortheanalogregulatorwas120mVwhenVIN=5Vto10V,whichcorrespondsto2.4%ofdeviationasshowninFig.5.7(c).Forthedigitalregulatorthedeviationwasmeasuredtobe2.2%undertheconditionVOUT=1.88Vto2.04VwhenVIN=2.7Vto10V.Fig.5.7(d),(e)and(g),(h)showthelinetransientresponsesforloadingcurrentof0.4mAand400mA,respectively.Itisnotedthatthestart-upresponsetimearesimilarforbothcases(148msforIL=0.4mAand146msforIL=400mA,whichcorrespondsto6.7KHzbandwidth),buttheshut-downtimeisdeterminedbytheloadingcurrent(1000msforIL=0.4mAand1.46msforIL=400mA)becausethedischargingtimetµC3VDD=IL.Fig.5.7(f)and(i)showtheripplerejectionperformanceoftheanalogregulatorwhendrivingasmallloadcurrent(0.4mA)andwhendrivingalargeloadcurrent(400mA),respectively.Theinputis6VDCvoltagewith1Vrippleat13.56MHz,andtheanalogregulatorcansuppressthevoltagerippleto192mVforIL=0.4mAand416mVforIL=400mA.Theefyoftheregulatorcanbeestimatedash=IOUTVOUT(IOUT+Iq)VIN(5.5)whereIqisthequiescentcurrent(alsocalledgroundcurrent).Inthisdesign,IqisnAlevelsothatitcanbeneglectedwhencomparingtoIOUTwhichisaroundmAlevel.Thus,theregulator's71Figure5.10(a)currentreference;(b)ringoscillatorwithpulseshaping;(c)power-onresetcircuit;(d)8-bitsingle-slopeADC.efyisjusttheratioofoutputvoltagetoinputvoltage.Fortheregulatorthatdrivesthedigitalmodules,theinputvoltageis2.7Vandtheoutputvoltageis1.9V,thusitsefycanbeestimatedtobe70.4%.Fortheregulatorthatdrivestheanalogmodules,theinputvoltageis5Vandtheoutputvoltageis3.8V,thusitsefycanbeestimatedtobe76%.5.3.2DataRecoveryCircuitThedatarecoverycircuitisshowninFig.5.8(a)anddemodulatesapulse-intervalencoded(PIE)modulationsignal.InaPIEcodealongerdurationbetweentwodigitalpulsesrepresentslogic`1'andashortdurationbetweentwopulsesrepresentslogic`0'.Theenvelopeofthemodulatedpiezo-72electricsignalisextractedbyavoltagedoublerfollowedbyalow-pass.AcomparatorthencomparesthesignalwithV2whichisthemid-pointofthesupplyvoltage.SincetheDCcomponentoftheV1equalsV2,thecomparatorextractsthePIEsignal.Notethatahystereticcomparatorhasbeenusedinthisdesigntoreducetheeffectofnoiseandinterference.Fig.5.8(b)showstheschematicofthecomparatorwheretransistorsM3˘M6formapositivefeedbackwhichdeterminestheupperandlowerhysteresistransitionpoints.TheoutputofthecomparatorcontrolsthegateofM11whichinturncontrolsthechargingordischargingthecapacitorC.Forlongdura-tionbetweentwobinarypulses(logic`1'inPIEcode),M11isturnedoffsuflongsothatthethecapacitorischargedtoavoltageexceedingtheinverterthresholdvoltage-thusproducingalogicfihigh"output.Forshortdurationbetweenthebinarypulses(logic`0'),theoutputoftheinverterremainslow.Thus,thecircuitinFig.5.8(a)effectivelyrecoversthetransmitteddigitaldata.5.3.3DigitalBasebandandManchesterEncoderThedigitalbasebandmoduleincludesthepreamblecircuit,theADCcontrollerandtheManchesterencoder.ThePIEdatatransmittedfromthereadertothetagisencapsulatedinaframeconsistingof4-bitpreamblebits,3-bitcommandbitsand1-bitCRCcode,asshowninFig.5.9(a).The4-bitpreambleischosentobe`1100',3-bitcommandencodeseightpossibleinstructionsthatcanoperatethetagandthesensor,and1-bitCRCisoddparityerrorchecking.Thepreamblemodulevthe4-bitpreambleandthenoutputsthecorrespondingcommandifthe1-bitCRCcodeiscorrect.Basedonthereceivedcommand,thebasebandcircuitcanhavedifferentresponses,suchasreadtheADCdataorjustsendaseriesofvcodes.ThereversetransmissionfromthetagtothereaderutilizestheManchesterencodingschemewherethedatasymbol`1'isencodedas`10'andthedatasymbol`0'isencodedas`01',asshowninFig.5.9(b).Similartotheforwardlink,theprotocolforthereverselinkalsoconsistsof4-bitpreamble,8-bitdataand3-bitCRCcode.The4-bitpreambleisagainsettobe`1100',andthe8-bitdatarepresentstheADCoutputoraseriesof`0'ifjustvcommandisreceived.ThemodulatorisasimpleNMOSswitch73whichisturnedON/OFFbytheManchesterdatastream,thustransmittingultrasonicwavecanbeback-scatteredtothereader,asshowninFig.5.4.5.3.4SensorDataAcquisitionCircuitryApower-onreset(POR)showninFig.5.10(c)detectsthechangeinthesupplyvoltageandgen-eratesadelayeddigitalpulsewhichisusedtoinitializeallthedigitallogicandinternalregisters.ThedelaytimetdisdeterminedbythechargingcurrentIchandcapacitorsizeCc,td=CcVthIch(5.6)whereVthisthethresholdoftheinverter.Inaddition,thecurrent-starvedinvertershavebeenusedtominimizethepowerdissipation.Aninternalring-oscillatorisusedforgeneratingthesystemclockforthedigitalbasebandandtheclockisalsousedinthesingleslopeADC.A3-stagecurrentstarvedringoscillatorisproposedwithpulseshapingcircuit,asshowninFig.5.10(b).ThestarvedcurrentiscontrolledbythecurrentreferenceinwhichtheresistorRcanbeusedtotunetheclockfrequency.TheschematicofasingleslopeADCusedinthisdesignisshowninFig.5.10(d).Inthesamplingphase,S1isclosedandS2isopensothattheinputvoltageVINissampledonthecapacitor.Inthecountingphase,S1isopenandS2isclosed,thecapacitorbeginstodischargethroughconstantcurrentsourceIREF.Meanwhile,an8-bitcounteristriggerforcountingthedischargingtimeuntilthesampledvoltagereachesVREF.OncethesampledvoltageisbelowVREF,thecomparatorgeneratesapulsetostopthecounter.ThereforethedigitaloutputcanbeassumedasN=C(VINVREF)IREFf(5.7)wherefisthefrequencyofthecountingclock.ItisnotedthatfandIREFshouldbetunedtomeetthe8-bitresolutionrequirementforthefullrange.AsshowninFig.5.11,theADC'sinputrangeis0.9V˘3.2Vandtheresolutionisabout12.8mVperbit.TheSNRwasmeasuredtobeapproximately47dBwhichresultsinaneffectivenumberofbitstobe7.5bits.Themaximumsam-plingrateoftheADCis250KS/s.Also,notethatthereferencecurrentIREFisgeneratedusinga74Figure5.11Measuredresultofthe8-bitADC.Figure5.12Measurementsetupoftheultrasoniccommunicationsystem:(a)theultrasonicreaderconsistsofXilinxFPGAandTIAnalogFrontEnd(AFE);(b)TxPZTandTxPZTareseparatedby2mmthickAlmetalbarrier;(c)micro-graphofthefabricatedultrasonictagIC.PTATandhencevarieslinearlywithtemperature.Thecounterclockfrequencyfisgeneratedus-ingaring-oscillatorshowninFig.5.10(b)whosefrequencyalsovarieslinearlywithtemperature.Thus,accordingtoequation5.7,theADCoutputshouldtheoreticallybetemperaturecompensated.However,transistormismatchinthePTATcircuitandtheoffsetmismatchbetweentheinFig.5.10(d)willleadaweakdependencyontemperature.Notethattheresponseofthepiezo-electrictransducerwillalsovarywithtemperature,soanytemperatureormismatchcompensationhastobeachievedatasystemlevelusingcalibrationtables.75Figure5.13(a)MeasuredtransmitpowerfromtheTxPZTand(b)receivepowerfromTxPZT.Figure5.14MeasuredresultsshowingprotocolsynchronizationbetweenthereaderandthetagIC.5.4MeasurementResultsTheultrasonictagIChasbeenimplementedandfabricatedina0.5-mmstandardCMOSprocesswithadieareaof3mm3mm,asshowninFig.5.12(c).ThemainofthisSOCislistedinTable5.2.Thetotalpowerconsumptionofthesystem-on-chipwasmeasuredtobe76Table5.2MainforProposedUltrasonicReceiverIC.Fabricationprocess0.5-mmstandardCMOSDiesize3mm3mmTotalpowerconsumption22.3mWADC14.9mWDigitalbaseband3.4mWPOROscillatorandDemodulator3.3mWVoltageregulator0.7mWUltrasonicmodulationscheme100%ASKCarrierfrequency13.56MHzInterrogationdistance2mmSensorinputvoltage0.9V-3.2V22.3mWoutofwhichtheADCconsumes14.9mW,digitalbasebandconsumes3.4mW,theregu-latorconsumes0.7mW,andtheremaining3.3mWisconsumedbythedemodulator,thePORandtheoscillator.Thecustom-madeultrasonicreadercomprisesofanFPGAboard(XilinxSpartan-3)andaCOTSTexasInstrumentschipset(TRF7960).Sincethecommunicationprotocolinthisdesignhasbeencustomized,thebuilt-incommunicationprotocoloftheTRF7960wasbypassedandonlytheanalogfront(AFE)wasused.TheFPGAboardisusedforcommandandcontroloftheTRF7960chipset.ThesystemsetupisshowninFig.5.12wheretheoutputpowerofTRF7960issetto200mWandthecarrierfrequencyis13.56MHzwith100%ASK.TheofusingASKasopposedtoothermodulationschemeslikeFSKorPSKisthatthedemodulationcircuitsarerelativelysimpletoimplementwhichisimportantforapowerconstrainedenergyharvestingapplication.AnimpedancematchingnetworkisinsertedbetweenthePZTandthecommunicationcircuitstomaximizepowertransmission.TheinputimpedancesfortheTxPZTandRxPZTmeasuredat13.56MHzarefoundtobe3.2+8.7jWand2.7+13.6jW.ThusT-typematchingnetworkisusedsothattheQ-factorofthematchingnetworkcanbeadjustedandtheinputimpedanceisappropriatelytransformedto50W.ThecommunicationandpowertransferchannelisformedbythemetalplatebetweentwoPZTs.Thepathlossofthechannelshouldischaracterizedtodeterminethe77optimalplacementofreceivertagrelativetothereader.WhenthePZTsarerespectivelyimpedancematched,thepowerreceivedbytheRxPZTatadistanceof2mmisfoundtobe-10.2dBmwhenthetransmitpowerfromtheTxPZTismeasuredtobe17.3dBm,asshowninFig.5.13.Thusthechannelattenuationisestimatedtobe27.5dBwhichcorrespondsto0.18%ofpowertransferefy.ItshouldbementionedthatthemeasuredpowertransferefyisveryclosetothetheoreticalcalculationbasedonMasonmodelwhichshows0.21%for2mmthicknessAlinFig.5.3.AsnapshotofthesuccessfulcommunicationsequencebetweenthereaderandthetagisshowninFig.5.14.First,thetransmitterissuesaPIEencoded`Query'command(topwaveform)`11000010'(thepulseisignored).Thefourbits`1100'correspondtothepreambleandthetrailingthreebits`001'indicatesthecommand,andthelastbit`0'isCRC.Then,thereceivercorrectlydemodulatesandrecoversthecommand(secondandthirdwaveform),andsendsbackan`Ac-knowledgement'(bottomwaveform),whichisthenreceivedandrecoveredbythetransmitter.The`Acknowledgement'signalisManchesterencodedas`110000000000000'.SimilartoPIEcode,thefourbits`1100'isalsopreamble,andtheremainingeightbits`00000000'is`Acknowl-edgement'signal,thelastthreebits`000'istheCRC.Oncetheprotocolsynchronizationhasbeenestablishedbetweenthereaderandtag,thereadercanissueadditionalcommandstocontrolandmonitorthetag.5.5SummaryThischapterpresentsthedesignofacompletednearultrasonicreceivertagSoC.Thetagharvestsitsoperationalenergydirectlyfromtheacousticinterrogationsignalgeneratedbyareaderwhichisimplementedusinga13.56MHzCOTSdevice.Duetothenearoperation,thetagcanachievebi-directionalcommunicationusinganacousticback-scattering,wherethepropertiesofthecommunicationlinkcanbeoptimizedbythereader.Theoperationofthesystemhasbeenvusingprototypesfabricatedina0.5mmCMOSprocess.Whiletheuseofnearop-78Table5.3PerformanceComparison.Thiswork[42][43][57]Carrierfrequency(MHz)13.561powerchannel:11communication:4PZTsize(mm2)380506powerPZT:349228communication:506Systemsize(mm2)1031.75Œ96.8ModulationSchemeASKASKOFDMASK,PSK,FSKBarrierAlSteelSteelSkinInterrogationdistance(mm)2304.863.5Powerconsumption(mW)0.022100Œ1.8BidirectionalcommunicationYesYesYesYesEnergyharvestingYesYesYesYesSensingexternalsignalYesYesYesYesSOCYesNoNoNoerationlimitstheinterrogationdistanceoftheproposedsystem,thecomplexityofthetagcanbereduced.Theneardepth,however,canbeadjustedbychoosingahigherop-
eratingfrequencywhichalsoallowstheuseofon-chipcapacitorsforharvestingacousticenergy.
Table5.3comparestheproposeddesignwithotherultrasonicenergyharvestingandtaggingsys-
tems.Thecomparisonshowsthatthepowerdissipationoftheproposedsystemisanorderof
magnitudelowerthanothersystems.Ifa(FOM)isasFOM=DComSPZTSSysPSys(5.8)whereDComisthecommunicationdistance,SPZTandSSysaresizesofPZTandsystem,respec-tively,andPSysispowerconsumptionofthesystem,theproposeddesignistwoordersofmagni-tudesuperiorthanthatof[42].NotethattheFOMcouldnotbecomputedforotherreferencesinTable5.3becausesomeoftheexperimentalparameterswerenotreported.79CHAPTER6CONCLUSION6.1SummaryofContributionsEnergyharvestingtechnologyisplayingamoreandmoreimportantroleintheadvancedelectronicsystemsthatrequiresmallerform-factor,lowercostandlongerlife-time.However,efconvertingtheharvestedenergyfromtheenvironmentintoeffectiveelectricalenergyisakeyfactorforthesuccessfuldeploymentoftheenergyharvestingsystems.Thisdissertationthusfocusesontheimprovementofthepowerconversionefyoftheenergyharvestingcircuits.,thisdissertationpresentsseveralsolutionstoanswertheresearchproblemsthatareproposedinsection1.3:howtoefharvestthepowerfor(1)ultralowpowerapplicationsand(2)impactenergyapplications.Themajorcontributionsofthisdissertationaresummarizedasbelow.1.Toharvestasmuchpoweraspossiblefortheultralowpowerapplications,ahigh-Qse-riesmatchingnetworkisproposedandvComparedtotheconventionalparallelmatchingnetworkwhichneedstotrade-offbetweentheQ-factorandthecouplingcoeftheseriesmatchingnetworkusesanextrasmallinductortoboosttheQ-factorsothatthecoilinductancecanbereleasedfromthematchingnetwork,andthenthesizeofthecoilcanbeenlargedtoachievelargercouplingcoefTherefore,theconstrainbetweentheQ-factorandcouplingcoefcanberelaxed.Withhigh-Qmatchingnetwork,thecoupledACvoltagecanbeboostedtoover-comethedeadzoneofthevoltage,leadingtotheefyimprovementoftheoverallRFpowerharvestingsystems.Themeasurementresultsshowthatgivenaeddistancetheboostedvoltageintheseriesmatchingnetworkisnearlydoubleofthatintheparallelmatchingnetwork.2.Tominimizethethresholdeffectoftherectifyingdevice,ahybridtechniqueisproposedandvThehybridrectechniquecanefharvesttheRFsignallowerthanthethresholdwiththehelpofthePZTenergywhichisusedasDCboostvoltagefortheRFsignal.Comparedtotheconventionalcross-coupled,thepeakpowerconversion80efyoftheproposedhybridisabletobeincreasedby30%.Inaddition,thehybridworksmoreefatlargerloadingconditionbecauseoflargesizeoftherectifyingtransistor.3.Toextendtheharvestingdurationfortheimpactenergyapplications,atime-dilationcircuitisproposedandvIntime-dilation,impactenergyisspreadandconservedintimesuchthattheself-poweredsensorcanaccuratelymeasuretheenergylevelwhileoperatingwithintheelectronicsafetycompliancelevels.Anonlinearcompressivecircuitenablesthesensortoquicklyrespondtotheimpactwhileenhancingthedynamicrange(rangeofimpactlevelsthatcanbemeasured)ofthePFGsensor.4.Toinvestigatetheenergyharvestingapplicationsintheconductivestructuresorionicmedia,anultrasonicreceiversystemisdesignedandimplementedinstandardCMOSprocess.Thesystemisself-poweredbyharvestingtheultrasonicenergyfromaremotevibrationsource.Inaddition,thissystemoperatesintheacousticnearsothatthebi-directionalcommunicationcanbeachievedbyusingacousticback-scattering.Duetothenearoperation,theinterrogationdistanceisrelativeshort,butthepowerconsumptionofthesystemislowwhencomparingtootherultrasonicenergyharvestingsystems.6.2OpenProblemsAlthoughthisresearchhasproposedseveraleffectivesolutionstoimprovetheefyforenergyharvestingsystems,therearestillmanyopenproblemsthatcanleadtofurtherimprovementinthisarea.Someoftheseopenproblemsarediscussedbelow.Oneoftheseproblemsthatcanbeexploredisthepossibilityofauto-tuningimpedancematch-ingnetworkfortheRFenergyharvesting.Auto-tuningimpedancematchingnetworkcouldhelptomaintaintheefyconsideringtheavailableRFpowerisvariableinthefrequencyandam-plitude.Oneofthemainproblemswithauto-tuningimpedancematchingnetworkistheneedofthecontrolcircuitsthatwillincreasethepowerconsumptionofthesystem.Analternativemethod81istousewidebandimpedancematchingnetwork,howeverthehighQ-factorisdiftoobtainforwidebandnetwork,thusthisisstillaninterestingproblemforfutureresearch.AnotheropenproblemassociatedwiththeRFenergyharvestingistoexploremorehybridtechniquestoovercomethedead-zoneissue.Thisresearchproposestousevibrationenergyastheauxiliarybiasforthevoltagetoreducethethresholdeffect,however,otherambientenergysourcescanalsobeconsidered,suchasthermalorsolarenergy,dependingonthepracticalapplications.Asmentionedinsection1.1,theofthethermalandsolarenergyisthehigherpowerdensitycomparedtovibrationsource.Integratingsolarenergyharvestingfunctionintothewirelesssensorscouldbequiteusefulforoutdoorwirelesssensornetwork,suchastheenvironmentalmonitoringsensorsthatcanusetheDCpowerharvestedfromthesolarcellasthebiasfortheRF.Thisdissertationpresentsanearultrasonicenergyharvestingsystemwithbacktelemetry.Operationintheacousticnearlimitsthecommunicationdistancebetweenthetagandthereader.Toextendthetelemetrycapability,itisthusworthtoinvestigatethefarultrasonicenergyharvestingsystem.Inaddition,theultrasonicsystemminiaturizationandpowerreductionarealsoattractiveforfuturework,especiallyfortheimplantablemedicalapplicationswheretheuseoftheultrasonicisconsideredtobesafe.82BIBLIOGRAPHY83BIBLIOGRAPHY[1]S.SudevalayamandP.Kulkarni,fiEnergyharvestingsensornodes:Surveyandimplications,flCommunicationsSurveys&Tutorials,IEEE,vol.13,no.3,pp.443Œ461,2011.[2]fiCymbetcorporation,flwebsite.[Online].Available:http://www.cymbet.com/products/energy-harvesting.php#[3]C.Lu,V.Raghunathan,andK.Roy,fiEfdesignofmicro-scaleenergyharvestingsys-tems,flEmergingandSelectedTopicsinCircuitsandSystems,IEEEJournalon,vol.1,no.3,pp.254Œ266,2011.[4]S.MandalandR.Sarpeshkar,fiLow-powercmosdesignforapplications,flCir-cuitsandSystemsI:RegularPapers,IEEETransactionson,vol.54,no.6,pp.1177Œ1188,2007.[5]T.Le,K.Mayaram,andT.Fiez,fiEffarradiofrequencyenergyharvestingforpassivelypoweredsensornetworks,flSolid-StateCircuits,IEEEJournalof,vol.43,no.5,pp.1287Œ1302,2008.[6]H.Nakamoto,D.Yamazaki,T.Yamamoto,H.Kurata,S.Yamada,K.Mukaida,T.Ninomiya,T.Ohkawa,S.Masui,andK.Gotoh,fiApassiveuhfrfcmostagicusingfer-roelectricramin0.35-mmtechnology,flSolid-StateCircuits,IEEEJournalof,vol.42,no.1,pp.101Œ110,2007.[7]C.Peters,J.Handwerker,D.Maurath,andY.Manoli,fiAsub-500mvhighlyefactiveforenergyharvestingapplications,flCircuitsandSystemsI:RegularPapers,IEEETransactionson,vol.58,no.7,pp.1542Œ1550,2011.[8]S.GuoandH.Lee,fiAnefy-enhancedcmoswithunbalanced-biasedcompara-torsfortranscutaneous-poweredhigh-currentimplants,flSolid-StateCircuits,IEEEJournalof,vol.44,no.6,pp.1796Œ1804,2009.[9]T.T.Nguyen,T.Feng,P.,andS.Chakrabartty,fiHybridcmosbasedonsynergisticrf-piezoelectricenergyscavenging,fl2014.[10]J.Ugwuogo,fiOn-demandenergyharvestingtechniques-asystemlevelperspective,fl2012.[11]J.Yoo,L.Yan,S.Lee,Y.Kim,andH.-J.Yoo,fiA5.2mwwearablebodysensornetworkcontrolleranda12mwwirelesslypoweredsensorforacontinuoushealthmonitoringsystem,flSolid-StateCircuits,IEEEJournalof,vol.45,no.1,pp.178Œ188,2010.[12]L.Yan,J.Yoo,B.Kim,andH.-J.Yoo,fiA0.5mv12mwwirelesslypoweredpatch-typehealth-caresensorforwearablebodysensornetwork,flSolid-StateCircuits,IEEEJournalof,vol.45,no.11,pp.2356Œ2365,2010.84[13]P.Cong,N.Chaimanonart,W.H.Ko,andD.J.Young,fiAwirelessandbatteryless10-bitimplantablebloodpressuresensingmicrosystemwithadaptiverfpoweringforreal-timelaboratorymicemonitoring,flSolid-StateCircuits,IEEEJournalof,vol.44,no.12,pp.3631Œ3644,2009.[14]D.Yeager,F.Zhang,A.Zarrasvand,N.T.George,T.Daniel,andB.P.Otis,fiA9ma,address-ablegen2sensortagforbiosignalacquisition,flSolid-StateCircuits,IEEEJournalof,vol.45,no.10,pp.2198Œ2209,2010.[15]C.HuangandS.Chakrabartty,fiAnasynchronousanalogself-poweredcmossensor-data-loggerwitha13.56mhzrfprogramminginterface,flSolid-StateCircuits,IEEEJournalof,vol.47,no.2,pp.476Œ489,2012.[16]C.Huang,N.Lajnef,andS.Chakrabartty,fiCalibrationandcharacterizationofself-poweredateusagemonitorwithsingleelectronpersecondoperationallimit,flCircuitsandSystemsI:RegularPapers,IEEETransactionson,vol.57,no.3,pp.556Œ567,2010.[17]T.FengandS.Chakrabartty,fiAnalysisanddesignofhighefyinductivepower-linksusinganovelmatchingstrategy,flinCircuitsandSystems(MWSCAS),2012IEEE55thInter-nationalMidwestSymposiumon.IEEE,2012,pp.1172Œ1175.[18]TRF7960,fihttp://www.ti.com/product/trf7960.fl[19]M.W.BakerandR.Sarpeshkar,fiFeedbackanalysisanddesignofrfpowerlinksforlow-powerbionicsystems,flBiomedicalCircuitsandSystems,IEEETransactionson,vol.1,no.1,pp.28Œ38,2007.[20]K.KocerandM.Flynn,fiAnewtransponderarchitecturewithon-chipadcforlong-rangetelemetryapplications,flIEEEJ.Solid-StateCircuits,vol.41,no.5,pp.1142Œ1148,May2006.[21]K.Kotani,A.Sasaki,andT.Ito,fiHigh-efydifferential-drivecmosforuhfflIEEEJ.Solid-StateCircuits,vol.44,no.11,pp.3011Œ3018,2009.[22]G.Papottoetal.,fiA90-nmcmosthreshold-compensatedrfenergyharvester,flIEEEJ.Solid-StateCircuits,vol.46,no.9,pp.1985Œ1997,2011.[23]H.XuandM.Ortmanns,fiAtemperatureandprocesscompensatedultralow-voltageinstandardthresholdcmosforenergy-harvestingapplications,flIEEETrans.onCircuitsandSystemsII,vol.58,no.12,pp.812Œ816,2011.[24]C.T.Delbrück,fiAnalogvlsiadaptivelogarithmicwide-dynamic-rangephotoreceptor,flinProc.IEEECircuitsandSystemsconf.,1994,pp.844Œ847.[25]P.weietal.,fiHigh-efydifferentialrffront-endforagen2tag,flIEEETrans.onCircuitsandSystemsII,vol.58,no.4,pp.189Œ194,2011.[26]B.R.Group,BSIM3v3.3MOSFETModelUser'sManual.Dept.Elect.Eng.Comput.Sci.,Univ.ofCalifornia.85[27]H.-M.LeeandM.Ghovanloo,fiAnintegratedpower-efactivewithoffset-controlledhighspeedcomparatorsforinductivelypoweredapplications,flIEEETrans.onCircuitsandSystemsI,vol.58,no.8,pp.1749Œ1760,2011.[28]P.SarkarandS.Chakrabartty,fiCompressiveself-poweringofatemechanicalimpactdetectors,flCircuitsandSystemsI:RegularPapers,IEEETransactionson,vol.60,no.9,pp.2311Œ2320,2013.[29]S.Rowson,G.Brolinson,M.Goforth,D.Dietter,andS.Duma,fiLinearandangularheadaccelerationmeasurementsincollegiatefootball,flJournalofbiomechanicalengineering,vol.131,no.6,p.061016,2009.[30]Y.YangandL.Tang,fiEquivalentcircuitmodelingofpiezoelectricenergyharvesters,flJour-nalofIntelligentMaterialSystemsandStructures,vol.20,no.18,pp.2223Œ2235,2009.[31]P.Sarkar,C.Huang,andS.Chakrabartty,fiAnultra-linearatestrain-gaugeforself-poweredmeasurementofquasi-static-strain,flBiomedicalCircuitsandSystems,IEEETransactionson,vol.7,no.4,pp.437Œ450,2013.[32]C.HuangandS.Chakrabartty,fiCompactself-poweredcmosstrain-ratemonitoringcircuitforpiezoelectricenergyscavengers,flElectronicsletters,vol.47,no.4,pp.277Œ278,2011.[33]S.RowsonandS.M.Duma,fiDevelopmentofthestarevaluationsystemforfootballhel-mets:integratingplayerheadimpactexposureandriskofconcussion,flAnnalsofbiomedicalengineering,vol.39,no.8,pp.2130Œ2140,2011.[34]C.Occhiuzzi,A.Rida,G.Marrocco,andM.Tentzeris,fiRFIDpassivegassensorintegratingcarbonnanotubes,flIEEETrans.Microw.TheoryTech.,vol.59,no.10,pp.2674Œ2684,Oct.2011.[35]H.Reinisch,M.Wer,S.Gruber,H.Unterassinger,G.Hofer,M.Klamminger,W.Pribyl,andG.Holweg,fiAmultifrequencypassivesensingtagwithon-chiptempera-turesensorandoff-chipsensorinterfaceusingEPCHFandUHFRFIDtechnology,flIEEEJ.Solid-StateCircuits,vol.46,no.12,pp.3075Œ3088,Dec.2011.[36]J.Yin,J.Yi,M.K.Law,Y.Ling,M.C.Lee,K.P.Ng,B.Gao,H.C.Luong,A.Bermak,andM.Chan,fiAsystem-on-chipEPCGen-2passiveUHFRFIDtagwithembeddedtemperaturesensor,flIEEEJ.Solid-StateCircuits,vol.45,no.11,pp.2404Œ2420,Nov.2010.[37]S.-Y.Lee,C.-H.Hsieh,andC.-M.Yang,fiWirelessfront-endwithpowermanagementforanimplantablecardiacmicrostimulator,flIEEETrans.Biomed.CircuitsSyst.,vol.6,no.1,pp.28Œ38,Feb.2012.[38]S.-Y.Lee,C.-J.Cheng,andM.-C.Liang,fiAlow-powerbidirectionaltelemetrydevicewithanearchargingfeatureforacardiacmicrostimulator,flIEEETrans.Biomed.CircuitsSyst.,vol.5,no.4,pp.357Œ367,Aug.2011.86[39]M.-K.Byeon,B.-W.Kim,J.-H.Jeon,andS.-J.Park,fiDesignandimplementationofhigh-speedcommunicationmodemusingultrasonicsensorsforunderwatersensornetworks,flinProc.ofIEEE/MTSOCEANS,Sep.2008,pp.1Œ4.[40]M.-K.Byeon,J.-H.Jeon,andS.-J.Park,fiPoint-to-multipointultrasoniccommunicationmo-demforubiquitousunderwaterwirelesssensornetworks,flinProc.ofIEEE/IFIPEUC2008,Dec.2008,pp.685Œ688.[41]Y.Hu,X.Zhang,J.Yang,andQ.Jiang,fiTransmittingelectricenergythroughametalwallbyacousticwavesusingpiezoelectrictransducers,flIEEETrans.Ultrason.Ferroelectr.Freq.Control,vol.50,no.7,pp.773Œ781,Jul.2003.[42]J.Ashdown,K.Wilt,T.Lawry,G.Saulnier,D.Shoudy,H.Scarton,andA.Gavens,fiAfull-duplexultrasonicthrough-wallcommunicationandpowerdeliverysystem.flIEEETrans.Ultrason.Ferroelectr.Freq.Control,vol.60,no.3,pp.587Œ595,Mar.2013.[43]T.J.Lawry,K.R.Wilt,J.D.Ashdown,H.A.Scarton,andG.J.Saulnier,fiAhigh-performanceultrasonicsystemforthesimultaneoustransmissionofdataandpowerthroughsolidmetalbarriers,flIEEETrans.Ultrason.Ferroelectr.Freq.Control,vol.60,no.1,pp.194Œ203,Jan.2013.[44]D.Shoudy,G.Saulnier,H.Scarton,P.Das,S.Roa-Prada,J.Ashdown,andA.Gavens,fiAnultrasonicthrough-wallcommunicationsystemwithpowerharvesting,flinProc.ofIEEEInt.Ultrason.Symp.,Oct.2007,pp.1848Œ1853.[45]G.Saulnier,H.Scarton,A.Gavens,D.Shoudy,T.Murphy,M.Wetzel,S.Bard,S.Roa-Prada,andP.Das,fiThrough-wallcommunicationoflow-ratedigitaldatausingultrasound,flinProc.ofIEEEInt.Ultrason.Symp.,Oct.2006,pp.1385Œ1389.[46]K.Wanuga,M.Bielinski,R.Primerano,M.Kam,andK.R.Dandekar,fiHigh-data-rateul-trasonicthrough-metalcommunication,flIEEETrans.Ultrason.Ferroelectr.Freq.Control,vol.59,no.9,pp.2051Œ2053,Sep.2012.[47]D.Dondi,A.DiPompeo,C.Tenti,andT.Rosing,fiShimmer:Awirelessharvestingembed-dedsystemforactiveultrasonicstructuralhealthmonitoring,flinProc.ofIEEESensors,Nov.2010,pp.2325Œ2328.[48]J.-L.Shih,M.Kobayashi,andC.-K.Jen,fiFlexiblemetallicultrasonictransducersforstruc-turalhealthmonitoringofpipesathightemperatures,flIEEETrans.Ultrason.Ferroelectr.Freq.Control,vol.57,no.9,pp.2103Œ2110,Sep.2010.[49]V.Giurgiutiu,fiTunedlambwaveexcitationanddetectionwithpiezoelectricwaferactivesensorsforstructuralhealthmonitoring,flJ.Intell.Mater.Syst.Struct.,vol.16,no.4,pp.291Œ305,Apr.2005.[50]A.Sanni,A.Vilches,andC.Toumazou,fiInductiveandultrasonicmulti-tierinterfaceforlow-power,deeplyimplantablemedicaldevices,flIEEETrans.Biomed.CircuitsSyst.,vol.6,no.4,pp.297Œ308,Aug.2012.87[51]S.Sherrit,M.Badescu,X.Bao,Y.Bar-Cohen,andZ.Chang,fiEfelectromechani-calnetworkmodelforwirelessacoustic-electricfeed-throughs,flinProc.ofSPIEonSmartStructuresandMaterialSystems,May.2005,pp.362Œ372.[52]J.A.JensenandN.B.Svendsen,fiCalculationofpressurefromarbitrarilyshaped,apodized,andexcitedultrasoundtransducers,flUltrasonics,FerroelectricsandFrequencyControl,IEEETransactionson,vol.39,no.2,pp.262Œ267,1992.[53]W.P.Mason,Electromechanicaltransducersandwaves.VanNostrandReinhold,1946.[54]C.Lu,C.-Y.Tsui,andW.-H.Ki,fiVibrationenergyscavengingsystemwithmaximumpowertrackingformicropowerapplications,flVeryLargeScaleIntegration(VLSI)Systems,IEEETransactionson,vol.19,no.11,pp.2109Œ2119,2011.[55]X.-D.Do,H.-H.Nguyen,S.-K.Han,D.S.Ha,andS.-G.Lee,fiAself-poweredhigh-efywithautomaticresettingoftransducercapacitanceinpiezoelectricenergyharvestingsystems,flVeryLargeScaleIntegration(VLSI)Systems,IEEETransactionson(accepted).[56]J.F.Dickson,fiOn-chiphigh-voltagegenerationinmnosintegratedcircuitsusinganim-provedvoltagemultipliertechnique,flIEEEJ.Solid-StateCircuits,vol.11,no.3,pp.374Œ378,Jun.1976.[57]J.-Y.Tsai,K.-H.Huang,J.-R.Wang,S.-I.Liu,andP.-C.Li,fiUltrasonicwirelesspoweranddatacommunicationforneuralstimulation,flinProc.ofIEEEInt.Ultrason.Symp.,Oct.2011,pp.1052Œ1055.88