Article A Long-Distance RF-Powered Node with Adaptive Power Management for IoT Applications

Matteo Pizzotti 1,2,* ID , Luca Perilli 1 ID , Massimo del Prete 2, Davide Fabbri 2, Roberto Canegallo 3, Michele Dini 1, Diego Masotti 2 ID , Alessandra Costanzo 1,2, Eleonora Franchi Scarselli 1,2 ID and Aldo Romani 1,2 ID 1 Advanced Research Center on Electronic Systems, University of Bologna, Via Toffano 2/2, Bologna 40126, Italy; [email protected] (L.P.); [email protected] (M.D.); [email protected] (A..); [email protected] (E.F.S.); [email protected] (A.R.) 2 Department of Electrical, Electronic, and Information Engineering, University of Bologna, Via Risorgimento 2, Bologna 40126, Italy; [email protected] (M.d.P.); [email protected] (D.F.); [email protected] (D.M.) 3 STMicroelectronics, Via Camillo Olivetti, Agrate Brianza 20864, Italy; [email protected] * Correspondence: [email protected]; Tel.: +39-054-733-9537

Received: 30 June 2017; Accepted: 26 July 2017; Published: 28 July 2017

Abstract: We present a self-sustained battery-less multi-sensor platform with RF harvesting capability down to −17 dBm and implementing a standard DASH7 communication interface. The node operates at distances up to 17 m from a 2 W UHF carrier. RF power transfer allows operation when common energy scavenging sources (e.g., sun, heat, etc.) are not available, while the DASH7 communication protocol makes it fully compatible with a standard IoT infrastructure. An optimized energy-harvesting module has been designed, including a rectifying antenna (rectenna) and an integrated nano-power DC/DC converter performing maximum-power-point-tracking (MPPT). A nonlinear/electromagnetic co-design procedure is adopted to design the rectenna, which is optimized to operate at ultra-low power levels. An ultra-low power controls on-board sensors and wireless protocol, to adapt the power consumption to the available detected power by changing wake-up policies. As a result, adaptive behavior can be observed in the designed platform, to the extent that the transmission data rate is dynamically determined by RF power. Among the novel features of the system, we highlight the use of nano-power , the implementation of specific hardware/ wake-up policies, optimized algorithms for best sampling rate implementation, and adaptive behavior by the node based on the power received.

Keywords: wireless sensor networks; RF power transfer; energy harvesting; nano-power DC/DC converter; rectifying antenna; ultra-low power ; adaptive power management

1. Introduction Increasing interest in distributed sensor networks [1] and IoT applications has driven research into the scope of energy harvesting mechanisms towards a more precise field of application. The combination of smart nodes, able to interact with standard wireless communication infrastructures, and energy scavenging modules, which allow nodes to work in a standalone scenario, has proved crucial for the development of both technologies [2]. Since autonomous nodes are expected to operate in very dissimilar surroundings with different energy sources and power densities, the importance of finding efficient ways to exploit available energy is evident, as in many cases the power available from the environment is in the order of microwatts [3] or less. In this context, -frequency power harvesting [4,5] represents both a fascinating solution, due to the opportunity of selectively providing energy through dedicated RF

Sensors 2017, 17, 1732; doi:10.3390/s17081732 www.mdpi.com/journal/sensors Sensors 2017, 17, 1732 2 of 21

Sensors 2017, 17, 1732 2 of 21 energy showers augmented by smart power beaming techniques [6,7], and a tough challenge because of the limitations imposed by regulations causing very low voltage and power levels as the energydistance showers from the augmented source increases. by smart As power a matter beaming of fact, techniques the most [ 6prohibitive,7], and a tough obstacle challenge to the becausediffusion ofof the RF limitations harvesting imposed nodes byis the regulations lack of causingdedicated very power low voltageconverters and powerable to levels operate as theunder distance these fromconditions. the source However, increases. simply As a developing matter of fact, specific the most power prohibitive converters obstacle or designing to the diffusion more efficient of RF harvestingrectennas nodesis not issufficient: the lack ofpower dedicated management power converters must go ablehand to in operate hand undernot only these with conditions. obvious However,requirements simply in developingterms of ultra-low specific power, power convertersbut also with or designing the development more efficient of specific rectennas policies is not for sufficient:adjusting powerthe behavior management of the node must according go hand into the hand availability not only withof energy. obvious Active requirements interaction in between terms ofthe ultra-low power module power, butand alsosmart with node the is development integral to achieving of specific this policies target. forAnother adjusting important the behavior aspect that of thereinforces node according the need to for the such availability interaction of energy.is the cons Activeiderable interaction difference between between the the power power module profiles and of smartthe harvesting node is integral source to and achieving the active this node; target. while Another the importantformer can aspect be considered that reinforces constant the needor slowly for suchchangeable, interaction the is latter the considerable is characterized difference by short between peak the consumptions, power profiles followed of the harvesting by long sourceinactive andperiods. the active This node; mismatch while thecalls former for canthe beintroducti consideredon constantof storage or slowlyelements, changeable, e.g., capacitors the latter isor characterizedsuper-capacitors, by short which peak must consumptions, be dimensioned followed carefully by long inactiveand considered periods. Thisas an mismatch integral calls part for of theoptimization introduction policies. of storage elements, e.g., capacitors or super-capacitors, which must be dimensioned carefullyThus, and a consideredholistic approach as an integral can best part tackle of optimizationthis type of issue. policies. On the one hand, circuit design aims at minimizingThus, a holistic the approachpower consumption can best tackle of this electron type ofic issue.devices On and, the one on hand,the other circuit hand, design optimized aims at minimizingbehavioral thepolicies power for consumption active node of electronicmodules devicesneed investigating and, on the other with hand, a view optimized to exploiting behavioral the policiesavailable for energy active nodeat its modulesbest. The need solution investigating presented with is designed a view to from exploiting an ultra-low the available power energyapproach, at itswith best. the The introduction solution presented of a custom is designed rectenna, from a dedicated an ultra-low nano-power power approach, DC/DC withconverter the introduction designed to ofoperate a custom down rectenna, to 250 a mV dedicated and 1 μ nano-powerW, and a multi-sensor DC/DC converter node with designed a low-power to operate profile down and to adopting 250 mV anda standard 1 µW, and wireless a multi-sensor communication node with protocol a low-power for IoT. profile In order and adoptingto implement a standard the necessary wireless communicationinteractions between protocol the for IoT.power In ordermanagement to implement subsection the necessary and the interactions active sensor between node, the powerspecific managementcircuitry has subsection been designed, and the dynamically active sensor changing node, specific the behavior circuitry of the has system been designed, according dynamically to different changingscenarios the of behavioravailable of environmental the system according energy. toThe different peculiar scenarios achievement of available obtained environmental through such energy. active Theinteraction peculiar is achievement the possibility obtained to modulate through the transm such activeission interaction data-rate as is a the consequence possibility of to variations modulate in thethe transmission power harvested, data-rate so that as a consequencehigher amount of of variations power causes in the an power automatic harvested, increase so in that the adata-rate higher amountand vice-versa, of power in causes such a anway automatic that the highest increase feasib in thele data-rate is and always vice-versa, obtained. in suchSuch aan way outcome that themay highest have feasiblea considerable data-rate rebound is always on obtained. those applic Suchations an outcome where may high have transmission a considerable rates rebound are not onstrictly those indispensable applications wherebut can high help transmission in building a ratesmore areextensive not strictly information indispensable database but and can therefore help in a buildingmore precise a more behavioral extensive model information for the database monitored and system. therefore a more precise behavioral model for the monitoredIn Figure system. 1 the overall system presented consists of two sub-systems: a harvesting module and a sensingIn Figure node.1 theThe overall interaction system presented presented is consists obtained of twothrough sub-systems: specific acontrol harvesting signals module exchanged and abetween sensing node.the power The interactionmodule and presented the microcontrolle is obtainedr, throughdesigned specific to provide control information signals exchanged about the betweencurrent thestate power of the module harvesting and the module, microcontroller, e.g., the voltage designed level to provideand the informationamount of power about theand current energy stateextracted. of the harvestingBased on module,this, the e.g., microcontroller the voltage level can and adapt the amountits behavior of power in order and energy to optimize extracted. the Basedtransmission on this, therate. microcontroller can adapt its behavior in order to optimize the transmission rate.

FigureFigure 1. 1.General General scheme scheme of of the the overall overall system. system.

TwoTwo different different operating operating frequencies frequencies are are used used for for harvesting harvesting and and communication communication channels channels (868(868 MHz MHz and and 433 433 MHz MHz respectively). respectively). On On one one hand, hand,harvesting harvesting module moduleis is optimized optimized for for a a constant constant RF RF source,source, so so communication communication signal signal cannot cannot be be effectively effectively used used and and two two separate separate antennas antennas are are required. required. OnOn the the other other hand, hand, using using two different two different antennas antennas for harvesting for harvesting and communication and communication allows optimization allows

Sensors 2017, 17, 1732 3 of 21 of the harvesting antenna to best fit the DC/DC specifications, at the expense of a reasonable increase in area. Using a single antenna would be more complicated and would result in worse harvesting performance, mainly because of the losses introduced by the multiplexing circuitry. An overview on recent rectenna solutions for generic RF wireless energy transfer applications is available in [8], whereas a comprehensive perspective on RF harvesting techniques specifically devoted to wireless sensor nodes can be found in [3]. In this field of application most solutions concentrate the greatest effort in optimization of the harvesting module, while other studies are proposed to define advanced algorithms to find most efficient transmission rates in complex scenarios. Differently, this work aims at optimizing the full energy chain from the rectenna to power management. Enhanced rectenna designs are proposed in [9] with 40% (simulated) efficiency at −20 dBm input power at 868 MHz, in [10] with 20% (measured) efficiency at −20 dBm input power at 2450 MHz, and in [11], where both solar and RF sources are deployed by a compact structure. Concerning integrated RF-to-DC converters, an interesting solution is presented in [12] with a 65% efficiency at −20 dBm in UHF band and output voltage up to 1.6 V. Another notable solution is presented in [13], where 1.0 V output is obtained at 27 m distance. The nano-power design of an integrated 1 µW – to 5 mW power management circuit is shown in [14]. The DC/DC converter proposed achieves a peak conversion efficiency of 77% and a minimum start-up voltage of 223 mV. The trend for quiescent power consumption of commercial and academic PMIC implementation is shown in [15]. A complete solution with harvesting module and sensing node is proposed in [16], showing an operative range of 5 m with −8 dBm input power. Furthermore, in [6] a simple adaptive solution is adopted, with solely two possible transmission data-rates and a solar cell as harvesting source. Advanced adaptive solutions are studied in [17], for a general-purpose sensor node in a multi-antenna power transfer scenario. These solutions lie on the development of complex recursive algorithms based on the knowledge of the evolution of stored energy in time, which implies the use of additional power-consuming electronics incompatible with micropower scenarios. In fact, experimental results demonstrate just a 5-m operative range. A similar approach is used in [18], with an accurate analysis of harvest-and-use and harvest-store-use schemes. Again, the resulting optimization algorithm requires the development of dedicated energy meters interacting with the node MCU to estimate the correct transmission rate. Both solutions are based on an adaptive estimation of best data-rate accomplished by the MCU on the basis of complex monitoring of available energy. Although DASH7 protocol performs a communication range up to 5 km, it is also widely used for indoor applications and, in combination with RF harvesting, it can provide an optimum solution for enclosed sensors like smart meters placed inside walls or generic structures. Moreover, having two different interfaces for harvesting and communication allows placing the power source regardless the location of the receiving gateway, which can be even far farther away if enough energy is provided. The proposed solution is a fully autonomous RF energy-harvesting node, where all components (rectenna, DC/DC converter and sensing platform) are jointly optimized in order to obtain the best performances within distances up to 17 m. This work also proposes a micropower-compatible adaptive power management, based on both dedicated circuitry and microcontroller firmware. The introduction of smart voltage supervisors allows the harvesting module to adaptively configure the data transmission period of the node, and to accordingly change the sleep policy. The combination of a nano-power DC/DC converter, smart supervisors and a complex wireless sensor nodes allows operation with negligible input power and state-of-the-art operative distances up to 17 m.

2. Materials and Methods

2.1. The Harvesting Module Incident RF power in the standard 868 MHz band (scavenged or intentionally transmitted) is captured from the environment by a rectifying antenna, or rectenna, which converts it into DC power. Due to the unregulated nature of the output voltage and current, and to the strong dependence of the Sensors 2017, 17, 1732 4 of 21 Sensors 2017, 17, 1732 4 of 21 dependenceoutput voltage of the on theoutput load voltage current, on a the power load management current, a power module management is also required module to is regulate, also required store, toand regulate, distribute store, the and harvested distribute power the toharvested the sensor power node. to the sensor node.

2.1.1. Rectenna Rectenna This section describes the rectenna unit we deviseddevised and reports on itsits performance.performance. Figure 22 shows a photo of the discrete-component prototype:prototype: it it consists consists of of a a PIFA PIFA (Planar Inverted F Antenna)-like printedprinted antenna antenna which which is matchedis matched to the to RF the detector RF detector input port input by meansport by of ameans multistage of a multistagedistributed-element distributed-el matchingement network. matching network.

FigureFigure 2. RectennaRectenna prototype prototype with hi highlightedghlighted different sections.

The first first step towards rectenna optimization is antenna design. To To satisfy the space constraints of typical sensors for WSN applications [19,20], [19,20], we selected a compact PIFA-like antenna designed to operate in the 868 MHz band (Figure 22).). TheThe topologytopology isis similarsimilar toto thatthat proposedproposed forfor aa completelycompletely different type type of of application application [21] [21 ]in in which which the the target target was was sensitivity sensitivity rather rather than than joint joint maximization maximization of outputof output power power and and voltage. voltage. It cons It consistsists of two of two branches branches in which in which the the lengths lengths have have been been optimized optimized to tuneto tune the the antenna antenna to tothe the required required resonant resonant frequenc frequencyy and and to to meet meet the the specifications specifications on radiation efficiency,efficiency, radiation pattern and input return loss. The The overall substrate size size of of the final final design is 73 × 55 mm2.2 The commercial substrate chosen was a Roger 4350B (εr = 3.55, tan δ = 0.0031, thickness 73 × 55 mm . The commercial substrate chosen was a Roger 4350B (εr = 3.55, tan δ = 0.0031, thickness = =1.5 1.5 mm). mm). Figure Figure3a shows3a shows the simulatedthe simulated and measured and meas inputured reflectioninput reflection coefficient coefficient of the PIFA of the antenna. PIFA antenna.In addition, In addition, Figure3b Figure shows 3b the shows simulated the simulat E- anded H-plane E- and at H-plane the operating at the operating frequency. frequency. As expected, As expected,the antenna the has antenna an omnidirectional has an omnidirectional radiation pattern radiation in the pattern horizontal in the plane: horizontal the low-directivity plane: the low-directivitybehavior is a desirable behavior feature is afor desirable RF energy feature harvesting for RF scenarios, energy whereharvesting the direction scenarios, of the where incoming the directionpower is typicallyof the incoming unknown. power The is corresponding typically unkn peakown. gain The achievedcorresponding is 1.4 dBi.peak All gain simulations achieved is were 1.4 dBi.carried All outsimulations using a commercial were carried full-wave out using simulator a commercial (CST Microwavefull-wave simulator Studio 2016 (CST). Microwave Studio 2016).The second step consists in designing the rectifying section, which involves choosing a rectifier topologyThe second and a nonlinear step consists design in designing for the rectifier-antenna the rectifying section, matching which network. involves As choosing regards the a rectifier former, topologysingle-stage and full-wave a nonlinear voltage design doubler for the rectifier-an topology representstenna matching the best network. option As for regards ultra-low the former, power single-stageapplications, full-wave as is well voltage documented doubler in thetopology literature represents [22,23]. the For best this reason,option for a full-wave ultra-low Dickson power applications,rectifier based as on is low-threshold well documented Skyworks in the SMS7630-079 literature [22,23]. Schottky For diodesthis reason, was selected. a full-wave Dickson rectifier based on low-threshold Skyworks SMS7630-079 Schottky diodes was selected.

Sensors 2017, 17, 1732 5 of 21 Sensors 2017, 17, 1732 5 of 21

0 H-plane E-plane 0 868MHz 1 -10 a) b) 0,5 θ° -20 (dB) 270 0 90 Simulated

-30 E (V/m) Measured

-40 820 840 860 880 900 (MHz) 180

FigureFigure 3.3. PIFA-likePIFA-like antennaantenna performance:performance: ((aa)) reflectionreflection coefficientcoefficient andand ((bb)) EE andand H-fieldradiationH-fieldradiation patternpattern atat 868 868 MHz. MHz.

OnceOnce thethe rectifierrectifier topologytopology isis selected,selected, anan optimumoptimum matchingmatching networknetwork betweenbetween thethe rectifierrectifier andand thethe antennaantenna needsneeds toto bebe designeddesigned byby nonlinearnonlinear circuitcircuit techniques,techniques, inin orderorder toto ensure,ensure, throughoutthroughout thethe powerpower rangerange available,available, thatthat maximummaximum RFRF powerpower entersenters thethe rectifierrectifier input.input. TheThe topologytopology chosenchosen isis thethe distributeddistributed matchingmatching networknetwork shownshown inin FigureFigure4 a:4a: it it consists consists of of two two impedance impedance steps steps and and two two stubs,stubs, thethe firstfirst beingbeing short-circuitedshort-circuited toto groundground whereaswhereas thethe secondsecond isis open. open. ImpedanceImpedance stepssteps areare employedemployed toto guaranteeguarantee broadbandbroadband matchingmatching betweenbetween thethe RFRF detectordetector (i.e.,(i.e., thethe rectifyingrectifying circuit)circuit) andand thethe antenna.antenna. For For the the operating operating frequency frequency (868 (868 MHz) MHz) chosen chosen here andhere forand the for targeted the targeted RF input RF power, input rangingpower, fromranging−20 from to − −1020 dBm, to −10 a nonlineardBm, a nonlinear regime is regime optimized is optimized accounting accounting for the dispersive for the dispersive behavior ofbehavior the linear of sub-networkthe linear (antennasub-network and matching(antenna network),and matching represented network), by a frequency-variablerepresented by a complexfrequency-variable reflection coefficient complex computedreflection coefficient by full-wave computed simulation. by Thefull-wave nonlinear simulation. model of The the nonlinear diodes is completedmodel of withthe theirdiodes package is completed model, which with istheir essential package for accurate model, optimization which is essential of the entire for rectennaaccurate atoptimization these operating of the frequencies. entire recte Optimizationnna at these is basedoperating on the frequencie Harmonics. BalanceOptimization (HB) method is based and on aims the atHarmonic maximizing Balance the RF-to-dc (HB) method conversion and aims process, at maximi throughzing optimization the RF-to-dc of conversion rectenna efficiency, process, definedthrough asoptimization [24]: of rectenna efficiency, defined as [24]: VRECT · IRECT ηRF−dc = ∙ (1) = PAV (1) where VRECT and IRECT are the dc components of the rectified voltage and current at the output port where VRECT and IRECT are the dc components of the rectified voltage and current at the output port when a certain load is applied, while PAV is the power available at the rectenna location, and represents when a certain load is applied, while PAV is the power available at the rectenna location, and the maximum power the antenna is able to deliver to the rectifying circuit. Commercial simulator ADS represents the maximum power the antenna is able to deliver to the rectifying circuit. Commercial was used for the integrated nonlinear/EM design optimization. Figure4a shows the circuit layout, simulator ADS was used for the integrated nonlinear/EM design optimization. Figure 4a shows the with details on the microstrip line dimensions and on the optimum load resistor. Such a rectenna circuit layout, with details on the microstrip line dimensions and on the optimum load resistor. Such system has been fabricated and the corresponding performance is reported in Figure4b, where the a rectenna system has been fabricated and the corresponding performance is reported in Figure 4b, simulated and measured behavior of the RF-to-dc conversion efficiency, computed as in (1), and the where the simulated and measured behavior of the RF-to-dc conversion efficiency, computed as in output dc voltage are plotted as a function of the input power (PAV): the rectenna exhibits a conversion (1), and the output dc voltage are plotted as a function of the input power (PAV): the rectenna exhibits efficiency better than 25% throughout the entire range of low RF power, starting from −20 dBm. a conversion efficiency better than 25% throughout the entire range of low RF power, starting from In particular, at −18 dBm the efficiency is greater than 30%, corresponding to a dc power of about 6 µW, −20 dBm. In particular, at −18 dBm the efficiency is greater than 30%, corresponding to a dc power of which represents a practical limit for current state-of-the-art DC/DC converters [25]. At the same about 6 μW, which represents a practical limit for current state-of-the-art DC/DC converters [25]. At time, a dc output voltage greater than 200 mV has been experimentally verified for PAV = −20 dBm. the same time, a dc output voltage greater than 200 mV has been experimentally verified for All these values were obtained with respect to an optimum dc load of 22 kΩ. PAV = −20 dBm. All these values were obtained with respect to an optimum dc load of 22 kΩ.

SensorsSensors 20172017,, 1717,, 1732 1732 66 of of 21 21

RF Input Matching Network 60 Simulated Measured 1.2 a) 820μm b) 15mm 1060μm 50 1 9.8mm 650μm 820μm 23mm 15mm 40 0.8 (V) (%)

850μm RECT 15mm RF-dc 30 0.6 V D1 10pF η IRECT μ VRECT 600 m 20 0.4 17mm 10pF D2 ROPT (22kOhm) 890μm 9mm 10 0.2 -20 -18 -16 -14 -12 -10 Dickson Rectifier PAV (dBm)

FigureFigure 4. 4. Rectifying sectionsection performance:performance:( a(a)) matching matching network network and and rectifier rectifier schematic schematic topologies topologies and and(b) RF-to-dc(b) RF-to-dc conversion conversion efficiency efficiency and and output outp dcut voltage dc voltage versus versus input input poweravailable. poweravailable.

InIn order order to to estimate estimate the the maximum maximum achievable achievable di distancestance of of a a sensor sensor equipped equipped with with the the proposed proposed rectennarectenna from from an an RF RF source, source, the the system system was was tested tested in in a 15 × 1515 m m22 laboratorylaboratory environment. environment. ExperimentsExperiments took placeplace purposely purposely outside outside an an anechoic anechoic chamber chamber in order in order to account to account for a more for a realistic more realisticscenario scenario including including limited multipathlimited multipath effects; neverthelesseffects; nevertheless measured measured received received power was power tolerably was tolerablyclose to the close theoretical to the theoretical value derived value from derived Friis from equation. Friis equation. TheThe measurement measurement procedure procedure was was carried carried out out in in two two steps: steps: (i) (i) first, first, the the receiver-to-transmitter receiver-to-transmitter distancedistance was was set; set; (ii) (ii) then, then, measurement measurement of of the the ou outputtput dc dc voltage voltage on on the the rectenna rectenna load load was was carried carried out.out. The The distances distances involved involved in inthe the measurements measurements ranged ranged from from 4 to 4 13 to m. 13 The m. transmitted The transmitted power power was selectedwas selected following following the current the current regulations regulations for forRFID RFID applications applications [26]: [26 0.5]: 0.5 W W in in the the 863–870 863–870 MHz MHz frequencyfrequency range range and and 2 2 W W for for a a narrow narrow band band starting starting from from 869.4 869.4 MHz MHz to to 869.65 869.65 MHz. MHz. The The output output DC DC voltagevoltage on on the the optimum optimum load load R ROPTOPT (22(22 k kΩΩ) )was was measured measured by by a a digital digital multimeter. multimeter. This This setup setup lends lends itselfitself to effectiveeffective experimental experimental verification verification of theof th actuale actual behavior behavior of the of entire the entire rectenna, rectenna, i.e., the i.e., antenna the antennaacting as acting the power as the source power of source the rectifier. of the rectifier. In order In to order create to a referencecreate a reference comparison comparison with the with expected the expectedsimulated simulated results, the results, link adopted the link was adopted first modeled was first by modeled a Friis equation by a Friis and theequation actual and rectified the actual power rectified(PRECT) waspower straight (PRECT forwardly) was straight computed forwardly by the computed following by equation:the following equation:  2 = ∙ ∙ ∙ λ ∙ (2) P = P ·G ·G · 4π ∙ ·η − (2) RECT TX TX RX 4π·D RF dc where PTX is the power transmitted at the remote location (the power exciting the transmitting antenna),where PTX GisTX the and power GRX are transmitted the gains atof the the remote receiving location and transmitting (the power exciting antennas, the transmittingrespectively,antenna), D is the linkGTX distance,and GRX areλ is the the gains wavelength of the receiving at the operating and transmitting frequenc antennas,y. It is noteworthy respectively, thatD isexperimental the link distance, and theoreticalλ is the wavelength analyses atmay the be operating carried frequency.out consider Iting is noteworthy the maximum that gain experimental direction, and for theoretical both the transmittinganalyses may and be carriedreceiving out consideringantennas. As the the maximum transmitting gain direction,antenna, forwe both adopted the transmitting a commercial and logperiodicreceiving antennas. antenna As(PCB the VA5JVB transmitting), with antenna, GTX = 6 we dBi. adopted As mentioned a commercial before, logperiodic the receiving antenna antenna (PCB gainVA5JVB is G),RX with = 1.4G TXdBi.= 6Figure dBi. As 5 reports mentioned the before,rectified the dc receiving power versus antenna distance, gain is GforRX two= 1.4 different dBi. Figure ERP5 levels.reports Considering the rectified a dc mini powermum versus required distance, power for twoof 6 differentμW for ERPthe dc-dc levels. converter Considering [27], a minimumit can be concludedrequired power that the of rectenna 6 µW for is the able dc-dc to operate converter at 6-m [27 distance], it can befor concludedthe 0.5-W-ERP that experiment, the rectenna and is able up toto operate12-m distance at 6-m distancefor the for2-W-ERP. the 0.5-W-ERP It is note experiment,worthy that and upthe tomodels 12-m distance used during for the rectenna 2-W-ERP. optimizationIt is noteworthy for both that thethe modelsdispersive used linear during sub- rectennacircuit and optimization the nonlinear for both devices the dispersiveallowed one linear to obtainsub-circuit predicted and the performances nonlinear devices that were allowed in very one good to obtain agreement predicted with performances the experimental that were one inover very a widegood range agreement of transmitted with the RF experimental power, as can one be over obse arved wide in range the plots of transmitted of Figure 5 where RF power, the measured as can be andobserved predicted in the rectenna plots of Figureoutput5 dcwhere power the measuredare compared and predictedover a number rectenna of outputRF source-rectenna dc power are distances.compared over a number of RF source-rectenna distances.

SensorsSensors 20172017,, 1717,, 17321732 77 ofof 2121 Sensors 2017, 17, 1732 7 of 21

Figure 5. Measured and simulated rectifiedrectified power forfor differentdifferent linklink distances.distances. Figure 5. Measured and simulated rectified power for different link distances. 2.1.2.2.1.2. DC/DC Converter Converter 2.1.2. DC/DC Converter TheThe DC/DCDC/DC converter converter used used in the system features an ultra-lowultra-low powerpower buck-boostbuck-boost converterconverter The DC/DC converter used in the system features an ultra-low power buck-boost converter designeddesigned inin STMicroelectronicsSTMicroelectronics 0.320.32 µμmm CMOSCMOS microelectronicmicroelectronic technologytechnology [14[14].]. The overalloverall designed in STMicroelectronics 0.32 μm CMOS microelectronic technology [14]. The overall architecturearchitecture can bebe divideddivided intointo twotwo mainmain blocks:blocks: a start-up circuit, which allows for ICIC bootstrapbootstrap architecture can be divided into two main blocks: a start-up circuit, which allows for IC bootstrap withwith RFRF inputinput sourcessources typicallytypically providingproviding lowlow voltages,voltages, andand thethe mainmain DC/DCDC/DC converter which alsoalso with RF input sources typically providing low voltages, and the main DC/DC converter which also providesprovides aa fractional fractional open-circuit open-circuit voltage voltage (FOCV) (FOCV) MPPT MPPT algorithm algorithm in order in order to adapt to adapt to the bestto the power best provides a fractional open-circuit voltage (FOCV) MPPT algorithm in order to adapt to the best transferpower transfer condition. condition. The IC The dynamically IC dynamically decides decid whetheres whether to route to powerroute power to the to load the or load to aor small to a power transfer condition. The IC dynamically decides whether to route power to the load or to a self-supplysmall self-supply capacitor capacitor Cconv: C thisconv achieves: this achieves very fast very activation fast activation times eventimes in even the presencein the presence of large of buffer large small self-supply capacitor Cconv: this achieves very fast activation times even in the presence of large capacitorsbuffer capacitors at the load at the output load port.output When port. C Whenconv is sufficientlyCconv is sufficiently charged, charged, all power all is power routed is to routed the load. to buffer capacitors at the load output port. When Cconv is sufficiently charged, all power is routed to Shouldthe load. C convShouldget C excessivelyconv get excessively discharged, discharged, all power all is power routed is here routed to replenishhere to replenish it before it the before IC fails. the the load. Should Cconv get excessively discharged, all power is routed here to replenish it before the AIC block fails. diagramA block diagram of the circuit of the IC ci isrcuit reported IC is reported in Figure in6. Figure 6. IC fails. A block diagram of the circuit IC is reported in Figure 6.

Figure 6. DC/DC low-power converter block diagram. FigureFigure 6. 6.DC/DC DC/DClow-power low-powerconverter converter block block diagram. diagram. The start-up module consists of a 16-stage charge pump implemented with low-threshold The start-up module consists of a 16-stage charge pump implemented with low-threshold MOSFETs and driven by an internal oscillator. A minimum voltage of approximately 250 mV from MOSFETs and driven by an internal oscillator. A minimum voltage of approximately 250 mV from Vrect is required to keep it operating. During the start-up phase, when the input voltage (i.e., the Vrect is required to keep it operating. During the start-up phase, when the input voltage (i.e., the rectenna output voltage) approaches 250 mV, the charge pump circuit starts working, and the rectenna output voltage) approaches 250 mV, the charge pump circuit starts working, and the output voltage on the self-supply capacitor Cconv is boosted. However, internal devices are not fully output voltage on the self-supply capacitor Cconv is boosted. However, internal devices are not fully switched on until the output voltage gets to 600 mV, so that the output charging rate is initially switched on until the output voltage gets to 600 mV, so that the output charging rate is initially limited by the sub-threshold state of the system. As soon as the generated voltage reaches 0.6 V the limited by the sub-threshold state of the system. As soon as the generated voltage reaches 0.6 V the

Sensors 2017, 17, 1732 8 of 21

The start-up module consists of a 16-stage charge pump implemented with low-threshold MOSFETs and driven by an internal oscillator. A minimum voltage of approximately 250 mV from Vrect is required to keep it operating. During the start-up phase, when the input voltage (i.e., the rectenna output voltage) approaches 250 mV, the charge pump circuit starts working, and the output voltage on the self-supply capacitor Cconv is boosted. However, internal devices are not fully switched on until the output voltage gets to 600 mV, so that the output charging rate is initially limited by the sub-threshold Sensors 2017, 17, 1732 8 of 21 state of the system. As soon as the generated voltage reaches 0.6 V the start-up circuit becomes fully operationalstart-up circuit and the becomes charge fully pump operational improves and its the charging charge ratepump until improves the output its charging exceeds rate 1.36 until V, which the is theoutput minimum exceeds operating 1.36 V, which voltage is the of theminimum main DC/DCoperating converter. voltage of At the this main point DC/DC the charge converter. pump At is disabledthis point and the power charge conversion pump is occursdisabled through and powe ther buck-boost conversion DC/DCoccurs through converter. the Althoughbuck-boost the DC/DC overall efficiencyconverter. of theAlthough start-up the stage overall settles efficiency between of 5%the andstart-up 15%, stage its sole settles purpose between is to 5% initially and 15%, bootstrap its sole the mainpurpose DC/DC is to converter, initially bootstrap so that its the impact main DC/DC on operative converter, efficiency so that can its impact be considered on operative negligible. efficiency canOnce be considered the 1.36 V negligible. threshold is reached, an in-rush current of about 11 µA is absorbed from the energy sourceOnce the by 1.36 the moduleV threshold for ais short reached, time an to in-rush complete current bootstrapping of about 11 the μA converter is absorbed functionalities. from the Afterenergy this, thesource module by canthebe module sustained for with a anshort input time power to ofcomplete just 935 nWbootstrapping showing a quiescentthe converter current of 121functionalities. nA. The different After this, functional the module modes can arebe sustained summarized with withan input the correspondingpower of just 935 supply nW showing voltages in Tablea quiescent1. It is worthcurrent recallingof 121 thatnA. theseThe different values refer functional to the voltagemodes onare thesummarized self-supply with capacitor the corresponding supply voltages in Table 1. It is worth recalling that these values refer to the voltage Vconv. Another aspect worth consideration is the high output resistance of the rectenna, which causes significanton the self-supply voltage drops capacitor on its V outputconv. Another node asaspect the currentworth consideration increases. is the high output resistance of the rectenna, which causes significant voltage drops on its output node as the current increases. Table 1. DC/DC converter functional modes. Table 1. DC/DC converter functional modes.

Vrect 0 V–0.250 V 0.250 V–1.6 V Vconv 0 V–0.600 V 0.600 V–1.36 V >1.36 V mode Switched Charge pump Charge pump DC/DC converter fully functional off (limited efficiency) (fully functional)  121 nA quiescent current  935 nW minimum input power 11 μA in-rush current from source needed to bootstrap the DC/DC converter

In order to extract the maximum power from the source, the IC adopts an FOCV MPPT In order to extract the maximum power from the source, the IC adopts an FOCV MPPT technique. technique. The input source is kept at 50% of the open-circuit voltage of the rectenna, which actually The input source is kept at 50% of the open-circuit voltage of the rectenna, which actually represents represents the maximum power transfer condition for the system. The open-circuit voltage is the maximum power transfer condition for the system. The open-circuit voltage is sampled for 2 µs sampled for 2 μs every 8 conversion cycles of the DC/DC converter. The buck-boost DC/DC everyconverter 8 conversion operates cycles in discontinuous of the DC/DC current converter. conduction The buck-boostmode, and DC/DCis switched converter when the operates source in discontinuousvoltage crosses current the conductionreference MPPT mode, voltage. and is switched The overall when efficiency the source is voltagehighly affected crosses theby referencesource MPPTimpedance voltage. and The open-circuit overall efficiency voltage. is highly affected by source impedance and open-circuit voltage. Thereafter,Thereafter, a full a analysisfull analysis was performed was performed with the with aim ofthe obtaining aim of an obtaining exhaustive an characterization exhaustive of thecharacterization DC/DC converter of the DC/DC in the specificconverter running in the spec conditionsific running of the conditions system proposed.of the system An proposed. equivalent modelAn equivalent of the rectenna model describedof the rectenna in Section described 2.1.1 inwas Section extracted 2.1.1 fromwas extracted the static from voltage-current the static characteristicsvoltage-current and characteristics used for characterization. and used for Thischaracterization. consists in a DCThis voltage consists source in a DC (V OCvoltage) with source a 22 k Ω series(VOC resistance.) with a 22 The kΩ outputseries resistance. voltage V harvThe ofoutput the DC/DC voltage isVharv set of at 2.3the V,DC/DC which is is set the at chosen 2.3 V, which maximum is operatingthe chosen voltage maximum of the operating sensing node voltage as described of the sens ining Section node 2.3as .described During operation, in Section the2.3. variations During onoperation, Vharv will the be limitedvariations to on a fewVharv hundred will be limited mV as to the a few worst hundred case. AmV 10 as M theΩ loadworst was case. connected A 10 MΩ to Vharvload. Results was connected show that to when Vharv. theResults open-circuit show that voltage when of the the open-c rectennaircuit V OCvoltageranges offrom the rectenna 0.32 V to V 1.4OC V, theranges related from efficiency 0.32 V growsto 1.4 V, from the related 35% to efficiency 73%, with grows a plateau from reached35% to 73%, early with on ata plateau 0.8 V. The reached whole characterizationearly on at 0.8 isV. reportedThe whole in characterization Figure7. is reported in Figure 7.

Sensors 2017, 17, 1732 9 of 21 Sensors 2017, 17, 1732 9 of 21

Sensors 2017, 17, 1732 9 of 21

FigureFigure 7.7.DC/DC DC/DC efficiency efficiency vs. open circuit input voltage. Figure 7. DC/DC efficiency vs. open circuit input voltage.

WithWithWith respectrespect respect toto FigureFigureto Figure4 4b,b, 4b, VRECT VRECT VRECT will will will be bebe kept kept by byby thethe the IC IC at at at approximately approximately approximately half half half of VOC, of of VOC, VOC, as the as as the the DC/DCDC/DCDC/DC converter converter converter works works works to tomaintain maintain a a statestate of maximum power power transfer transfer by byproviding providingproviding an input anan inputinput impedanceimpedanceimpedance equalequal equal toto thetheto the sourcesource source impedanceimpedance impedance (22 (22(22 kkΩΩ inin this thisthis case). case).case). Moreover, Moreover,Moreover, the thethe extracted extractedextracted model modelmodel was waswas consideredconsideredconsidered validvalid valid throughoutthroughout throughout thethe the operativeoperative operative range rangerange ofof the the system. system.system. Figure FigureFigure 8 8shows 8 shows shows the the thededicated dedicated dedicated board board board containingcontainingcontaining thethe DC/DCtheDC/DC DC/DC converter converter along along with with its its subsidiary components. components.

FigureFigure 8. 8.Harvesting Harvesting module module board. board.

2.2. The2.2. ActiveThe Active Node Node Figure 8. Harvesting module board. The sensingThe sensing and communicationand communication node, node, whose whose internal internal architecture architecture is shown is shown inFigure in Figure9, includes 9, 2.2. Theincludes Active a Nodetemperature and relative humidity sensor, a low-power microcontroller and a sub-GHz a temperature and relative humidity sensor, a low-power microcontroller and a sub-GHz radio radio device for data communication. STMicroelectronics HTS221 is an ultra-compact sensor based deviceThe for sensing data communication. and communication STMicroelectronics node, whose HTS221internal isarchitecture an ultra-compact is shown sensor in basedFigure on9, on a planar capacitance technology that integrates humidity and temperature sensing with a mixed aincludes planarsignal capacitancea temperatureASIC to provide technology and data relative measured that humidity integrates through sensor, humidity standard a low-power digital and temperature serial microcontroller interfaces. sensing The and withultra-low a sub-GHz a mixed signalradiopower device ASIC STMicroelectronics to for provide data communication. data measured STM32L1 throughmicrocontroller,STMicroelect standardronics based digital HTS221 on the serial ARM is interfaces.an Cortex-M3 ultra-compact The core, ultra-low interfaces sensorpower based STMicroelectronicson a withplanar the capacitance sensor STM32L1 for datatechnology acqu microcontroller,isition that andintegrates manages based humidity on communication the ARM and Cortex-M3 temperature with the core, STMicroelectronics sensing interfaces with witha mixed the sensorsignalSPIRIT1 forASIC data module,to acquisition provide a low-power data and managesmeasured sub-GHz communication RFthrough transceiver. standard with thedigital STMicroelectronics serial interfaces. SPIRIT1 The ultra-low module, apower low-power STMicroelectronics sub-GHz RF transceiver. STM32L1 microcontroller, based on the ARM Cortex-M3 core, interfaces with the sensor for data acquisition and manages communication with the STMicroelectronics SPIRIT1 module, a low-power sub-GHz RF transceiver.

Sensors 2017, 17, 1732 10 of 21 Sensors 2017, 17, 1732 10 of 21

FigureFigure 9. 9. BlockBlock diagram diagram of of the the sensing sensing node. node.

Wireless communication is based on DASH7 [28], an open source Wireless Sensor and Actuator Wireless communication is based on DASH7 [28], an open source Wireless Sensor and Actuator Network (WSAN) protocol; the microcontroller implements OpenTag, a DASH7 protocol stack and Network (WSAN) protocol; the microcontroller implements OpenTag, a DASH7 protocol stack and a a minimal Real-Time (RTOS) designed to be light and compact and targeted to minimal Real-Time Operating System (RTOS) designed to be light and compact and targeted to run on run on resource-constrained . resource-constrained microcontrollers. The DASH7 network architecture has a star structure where all the nodes, which are typically The DASH7 network architecture has a star structure where all the nodes, which are typically low-power devices able to transmit and receive data, communicate only with a gateway that is never low-power devices able to transmit and receive data, communicate only with a gateway that is never offline and connects the DASH7 network to other networks and to the web. offline and connects the DASH7 network to other networks and to the web. DASH7 supports two communication models: pull and push [29]. The pull model consists of a DASH7 supports two communication models: pull and push [29]. The pull model consists of a request-response mechanism initiated by the gateway; it uses an advertising protocol for rapid request-response mechanism initiated by the gateway; it uses an advertising protocol for rapid ad-hoc ad-hoc node synchronization before sending an addressed request to a node and waiting for the node synchronization before sending an addressed request to a node and waiting for the response [29]. response [29]. The data transfer to the gateway initiated by the nodes is based on the push model The data transfer to the gateway initiated by the nodes is based on the push model (e.g., beaconing); (e.g., beaconing); this approach is implemented as an automated message or beacon that is sent at this approach is implemented as an automated message or beacon that is sent at specific time intervals. specific time intervals. In this work, the beaconing approach is used to send the temperature and humidity data from the In this work, the beaconing approach is used to send the temperature and humidity data from sensor node to the gateway; this method is the least power hungry and the node can send the message the sensor node to the gateway; this method is the least power hungry and the node can send the as soon as the harvesting module provides enough power to power-up the node or wake up it from a message as soon as the harvesting module provides enough power to power-up the node or wake deep sleep state. up it from a deep sleep state. The node can be supplied with a voltage ranging from 1.8 V to 3.6 V, and is programmed to The node can be supplied with a voltage ranging from 1.8 V to 3.6 V, and is programmed to perform two different stop policies: off-mode and standby-mode. The policy is selected by the harvesting perform two different stop policies: off-mode and standby-mode. The policy is selected by the module, as will be explained in the following section. harvesting module, as will be explained in the following section. As can be seen in Figures 10 and 11, the two modes behave in the same way at power on, when, As can be seen in Figures 10 and 11, the two modes behave in the same way at power on, when, after the start-up phase, the node acquires data from the sensor and transmits them to the gateway. after the start-up phase, the node acquires data from the sensor and transmits them to the gateway. Afterwards, if off-mode is selected, the node generates a signal for the harvesting module to inform it Afterwards, if off-mode is selected, the node generates a signal for the harvesting module to inform it that the power supply can be switched-off. When the power supply is once more provided, the node that the power supply can be switched-off. When the power supply is once more provided, the node will perform a new start-up phase, data acquisition and message transmission. This phase is associated will perform a new start-up phase, data acquisition and message transmission. This phase is with a significant energy overhead. Figure 10 shows a schematic of the current consumption of the associated with a significant energy overhead. Figure 10 shows a schematic of the current node and its behavior in off-mode indicating the average current in each of the three phases (start-up, consumption of the node and its behavior in off-mode indicating the average current in each of the data acquisition and transmission); in all each activation phase lasts 680 ms and the average current three phases (start-up, data acquisition and transmission); in all each activation phase lasts 680 ms consumption over this time is 4.7 mA at 1.9 V. and the average current consumption over this time is 4.7 mA at 1.9 V. On the contrary, if the standby-mode is selected, after the first start-up phase and message On the contrary, if the standby-mode is selected, after the first start-up phase and message transmission, the node informs the harvesting module that transmission is completed and that it transmission, the node informs the harvesting module that transmission is completed and that it is is entering a deep sleep state in which its current consumption is 3 µA. The node then waits for an entering a deep sleep state in which its current consumption is 3 μA. The node then waits for an external interrupt, generated by the harvesting module, to wake up and perform a new data acquisition external interrupt, generated by the harvesting module, to wake up and perform a new data and message transmission. In standby-mode, therefore, the power supply is never switched-off, and the acquisition and message transmission. In standby-mode, therefore, the power supply is never overhead consists in constant power consumption. The schematic of the node’s current consumption switched-off, and the overhead consists in constant power consumption. The schematic of the node’s and its behavior in standby-mode is shown in Figure 11. In this configuration, with the exception current consumption and its behavior in standby-mode is shown in Figure 11. In this configuration, of the first power-on in which the current consumption is the same as each activation in off-mode, with the exception of the first power-on in which the current consumption is the same as each the start-up phase does not have to be repeated for each activation phase because the power supply activation in off-mode, the start-up phase does not have to be repeated for each activation phase because the power supply is never switched-off. As can be seen in Figure 11, in standby-mode each

Sensors 2017, 17, 1732 11 of 21

Sensors 2017, 17, 1732 11 of 21 is neverSensors 2017 switched-off., 17, 1732 As can be seen in Figure 11, in standby-mode each activation phase following11 of 21 theactivation first one phase lasts 38 following ms, and the the first average one lasts current 38 ms consumption, and the average during current this consumption time is 3.81 during mA at this 1.9 V. In activationstandby-modetime is 3.81 phase mA, the following at charge 1.9 V. consumption theIn standby-modefirst one lasts for, each38the ms charge data, and acquisition theconsumption average and current for message each consumption data transmission acquisition during is thisand much lesstimemessage than is in3.81 transmissionoff-mode mA at, but1.9 isV. a much currentIn standby-mode less of than 3 µA in, has theoff-mode tocharge be, guaranteedbut consumption a current to of keep for3 μ eachA the has nodedata to be acquisition alive guaranteed in the and deepto sleepmessagekeep state. the transmissionnode alive in isthe much deep lesssleep than state. in off-mode, but a current of 3 μA has to be guaranteed to keep the node alive in the deep sleep state.

FigureFigure 10. 10.Schematic, Schematic, not not to to scale, scale, of of the the node node currentcurrent consumptionconsumption in in off-mode; off-mode; the the average average charge charge consumptionFigureconsumption 10. Schematic, in in each each activation activationnot to scale, phase phase of the is isQ node= =I ∗current ∗T = = 4.7 mA4.7 consumption mA ∗∗ 680 ms680 ms in = =off-mode; 3.196 mC3.196 mC. .the average charge consumption in each activation phase is = ∗ = 4.7 mA ∗ 680 ms = 3.196 mC. TheThe selection selection between betweenoff-mode off-modeand andstandby-mode standby-mode has a a significant significant impact impact on on the the power power budget. budget. The selection between off-mode and standby-mode has a significant impact on the power budget. TheTheoff-mode off-modeis is associated associated with with a a constant constant energyenergy overhead,overhead, consisting in in the the energy energy consumed consumed for for Thestarting off-mode the issystem. associated By contrast, with a constant the standby-mode energy over is head,associated consisting with ain constant the energy power consumed overhead. for starting the system. By contrast, the standby-mode is associated with a constant power overhead. Hence, startingHence, whichthe system. of the Bytwo contrast, modes coststhe standby-modeless will dep endis associated on the frequency with a constant of activation. power For overhead. frequent which of the two modes costs less will depend on the frequency of activation. For frequent activation, Hence,activation, which the of stand-by the two modesmode consumescosts less willless depenerendgy. on Furthermore, the frequency the of frequency activation. of For activations frequent the stand-by mode consumes less energy. Furthermore, the frequency of activations strictly depends activation,strictly depends the stand-by on the harvested mode consumes power. less energy. Furthermore, the frequency of activations on the harvested power. strictlyFigure depends 12 shows on the the harvested sensor node power. connected to the harvesting module describer in Section 2.1. FigureFigure 12 12 shows shows the the sensor sensor node node connectedconnected to the harvesting module module describer describer in in Section Section 2.1. 2.1 .

Figure 11. Schematic, not to scale, of the node current consumption in standby-mode; the average FigureFigurecharge 11. 11.consumption Schematic,Schematic, innot noteach to toscale, activation scale, of the of phasenode the nodecurrentfollowing current consumption the consumptionpower in on standby-mode phase in standby-modeis ; = ∗ = the average; the averagecharge 3.81 mA chargeconsumption ∗ 38 ms ~ 145 μC. consumption in each activation in each phase activation following phase the following power on thephase power is = ∗ = on phase is Q 3.81 mA = I ∗ T = ∗ 38 ms ~ 145 μC.3.81 mA ∗ 38 ms ∼ 145 µC.

Sensors 2017, 17, 1732 12 of 21

Sensors 2017, 17, 1732 12 of 21 Sensors 2017, 17, 1732 12 of 21

Figure 12. Sensing node and harvesting module boards. Figure 12. Sensing node and harvesting module boards. Figure 12. Sensing node and harvesting module boards. 2.3. Policies for Power Optimization 2.3. Policies for Power Optimization 2.3. PoliciesIn this forsection, Power the Optimization proposed adaptive feature is described in detail, as well as the architectural choicesIn thisIn this section,and section, dimensioning the the proposed proposed of control adaptive adaptive circuits featurefeature shown is is described described in Figure in indetail, 13. detail, The as well asinterface well as the as betweenarchitectural the architectural the choicesharvestingchoices and and dimensioning module dimensioning and the ofactive of control control node circuitsiscircuits implemen shownshownted through in in Figure Figure four 13. control13 .The The signalsinterface interface connected between between to the the harvestingmicrocontrollerharvesting module module (Policy and and the, theReset_V activeDD node, nodeStart is and implemen is implementedReset_Startted through), besides through four the control fourground signals control and connectedpositive signals supply connectedto the (microcontrollerVDD). (Policy, Reset_VDD, Start and Reset_Start), besides the ground and positive supply to the microcontroller (Policy, Reset_VDD, Start and Reset_Start), besides the ground and positive (VDD).The Policy signal selects one of the two stop policies of the active node (off-mode and supply (VDD). standby-modeThe Policy). signal selects one of the two stop policies of the active node (off-mode and The Policy signal selects one of the two stop policies of the active node (off-mode and standby-mode). standby-mode).

Figure 13. Schematic of the harvesting module and its interface with the active node. FigureFigure 13. 13.Schematic Schematic of of the the harvesting harvesting module an andd its its interface interface with with the theactive active node. node. The first factor to be analyzed is the buffer capacitance Cbuff, which plays the role of energy

storageThe needed first factor by the to systembe analyz to offseted is the totalbuffer po capacitancewer request C ofbuff the, which node plays during the its role active of energyphase. SuchstorageThe firstrequests needed factor are by tosummarized the be system analyzed toin offsetTable is the the2, whichbuffer total po in capacitancewertegrates request the currentofCbuff the, node which absorbed during plays by its the the active node role phase. when of energy storageswitchingSuch needed requests into by areactive the summarized systemmode from to in offset each Table of the 2,the which total other power intwotegrates modes. request the current of the absorbed node during by the itsnode active when phase. Suchswitching requests into are active summarized mode from in Tableeach of2, the which other integrates two modes. the current absorbed by the node when switching into activeTable mode 2. Charge from each requests of theof node other during two active modes. states according to mode. Table 2. Charge requests of node during active states according to mode. Mode Charge (μC) Operating Voltage (V) Table 2. ChargeModeOFF requests Charge of3196 node ( μ duringC) Operating active 2.3–1.9 states Voltage according (V) to mode. STANDBYOFF 3196145 2.3–1.9 ~2 Mode Charge (µC) Operating Voltage (V) STANDBY 145 ~2 OFF 3196 2.3–1.9 STANDBY 145 ~2 Sensors 2017, 17, 1732 13 of 21

In order to have a voltage drop of about 0.4 V, the resulting Cbuff must be at least

Charge(OFF) 3196 µC C = = = 7990 µF =∼ 8 mF (3) bu f f ∆V 0.4 V where the voltage drop was chosen as a trade-off between system power consumption and wake-up time. In this instance, a lower bound of 1.9 V was chosen by considering a 100 mV margin to the minimum 1.8 V supply voltage required by the node to operate, while the upper bound was chosen by considering that a lower voltage implies lower power consumption, though at the same time it requires a larger buffer capacitance which causes a longer start-up time when the system has to be booted for the first time. Hence a maximum operating voltage of 2.3 V was chosen and, as a result, an 8 mF Cbuff must be considered. Three control circuits are necessary to generate the correct control signals for the micro-controller (Figure 13). The first one is the Control circuit for OFF mode. This block has the task of monitoring the voltage across Cbuff capacitance voltage and consequently to provide power supply to the sensing node only when voltage is in the acceptable range (1.9 V–2.3 V). In this way, the sensing module is only switched on when VDD exceeds 2.3 V, i.e., when the buffer capacitance has sufficient energy to sustain at least one complete data transmission. Likewise, power supply is taken off as soon as VDD drops below 1.9 V, as the sensing module cannot work at lower voltages. During off-mode, a reset signal (Reset_VDD) must be issued by the microcontroller at the end of each transmission stage, so that power supply is simultaneously turned off. This behavior ensures that the next transmission will only be held when the buffer capacitance is fully recharged, i.e., when it reaches 2.3 V again. A NCP303 low-power voltage supervisor was used to implement the block, along with two AS11P2 analog switches and two resistors in order to obtain the desired hysteresis, respectively 2.4 MΩ and 500 kΩ. The total quiescent current of the block is 500 nA, mostly due to the NCP303 supervisor. The second one is the Control circuit for STANDBY mode. This block has the task of monitoring the capacitance voltage and consequently providing the start signal only when voltage is above 2.0 V. During standby-mode, the sensing node is always powered, and subsequent transmissions are regulated by the start signal, so that it must only be issued when the buffer capacitance is storing enough energy to sustain transmission. Actually, the expected voltage drop due to a single transmission in standby-node is a mere 18 mV as explained in following equation

Charge(STANDBY) 145 µC ∆V(STANDBY) = = = 18 mV (4) Cbu f f 7990 µF so a threshold voltage of 1.9 V + ∆V = 1.918 V could be enough to sustain standby-mode, but a safer margin of 2.0 V − 1.9 V = 100 mV was chosen for this condition. A reset signal (Reset_Start) from the microcontroller is also requested at the end of each transmission in order to force the start signal to a low level, even if the related capacitance voltage drop has not been sufficient to trigger the voltage supervisor. This happens because the microcontroller is sensitive to a positive start signal edge, so that a low-to-high transaction is always required to awaken the sensing node. An NCP303 low-power voltage supervisor was used to implement the block, along with one AS11P2 analog switch and a pull-up resistor of 2.4 MΩ. The total quiescent current of the block is 400 nA, mostly due to the NCP303 supervisor. The ability to dynamically select whether to operate in off-mode or standby-mode is a crucial aspect of the solution presented, as it enables the system to fully exploit the available environmental energy under any circumstance without human intervention. In both cases the minimum transmission period is obtained, as the microcontroller is either supplied (off-mode) or awoken (standby-mode) as soon as the buffer capacitance has recovered the energy lost, but depending on the variable amount of extracted energy, one solution can be less advantageous than the other, if not altogether unfeasible. Sensors 2017, 17, 1732 14 of 21

Transmission Period (TP) is defined as the time interval between two consecutive transmissions and can be calculated as: Q TP = needed (5) (I − Iq − Iq ) Sensors 2017, 17, 1732 extracted harv node 14 of 21 where Qneeded is the charge lost during each transmission, Iextracted is the current provided by the where Qneeded is the charge lost during each transmission, Iextracted is the current provided by the DC/DC DC/DC converter, Iqharv is the quiescent current of the harvesting module that is directly supplied by converter, Iqharv is the quiescent current of the harvesting module that is directly supplied by VHARV VHARV while Iqnode is the quiescent current of the rest of the system supplied by VDD. These values differwhile between Iqnode is thethe twoquiescent operative current modes, of the and rest are of summarized the system supplied in Table3 .by VDD. These values differ between the two operative modes, and are summarized in Table 3. Sensors 2017, 17, 1732 14 of 21 Table 3. Quiescent currents of harvesting module and sensor node during recharge phase. Table 3. Quiescent currents of harvesting module and sensor node during recharge phase. where Qneeded is the charge lost during each transmission, Iextracted is the current provided by the DC/DC

converter, Iqharv is the quiescentI qharvcurrent (µA) of the harvesting module thatIqnode is (µA) directly supplied by VHARV

while Iqnode is theQneeded quiescent (µC) currentControl circuit of the rest ofControl the circuitsystem suppliedControl by circuit VDD .for These values differ Sensor node between the two operative modes,OFF-mode and are summarizedSTANDBY-mode in Table 3. Policy detection switched-off switched-off switched-off 0.5 OFF Table 3.3196 Quiescent currents of harvesting module and sensor nodeswitched-off during recharge phase. 0.5 Iqharv(μA) Iqnode(μA) 0.4 0.6 3 Qneeded(μC) Control0.5 circuit Control circuit Control circuit for STANDBY 145 4 Sensor node OFF-mode STANDBY-mode Policy detection 4.5 switched-off switched-off switched-off 0.5 OFF 3196 switched-off The most significant difference is that in off-mode the charge needed is significantly higher than The most significant difference is that in off-mode the charge0.5 needed is significantly higher than in in standby-mode, but the total quiescent current is only 500 nA, since all the other modules are standby-mode, but the total quiescent current is only 5000.4 nA, since all the other0.6 modules are switched3 off switched off during recharge periods.0.5 As a consequence, in off-mode nearly all the current from the duringSTANDBY recharge periods.145 As a consequence, in off-mode nearly all the current4 from the power converter power converter can be used to recharge the buffer capacitance. can be used to recharge the buffer capacitance. 4.5 Thus, this last mode proves more suitable when the extracted current is relatively low, whereas the standby-modeThe most significantbecomes moredifference cost effectiveis that in off-mode at higher the extracted charge needed currents, is significantly when its lower higher value than of in standby-mode, but the total quiescent current is only 500 nA, since all the other modules are activation charge becomes predominant. A detailed graph of transmission data rates versus extracted switched off during recharge periods. As a consequence, in off-mode nearly all the current from the currents is shown in Figure 14, analytically obtained through Equation (5). power converter can be used to recharge the buffer capacitance.

Figure 14. Transmission period vs. Extracted currents in different operating modes.

Thus, this last mode proves more suitable when the extracted current is relatively low, whereas the standby-mode becomes more cost effective at higher extracted currents, when its lower value of activation charge becomes predominant. A detailed graph of transmission data rates versus extracted currents is shown in Figure 14, analytically obtained through Equation (5). FigureFigure 14. 14. Transmission period vs. Extracted currents in different operating modes. The available currentTransmission is defined period as the vs. effective Extracted current currents that in differentcan be used operating by themodes. active part of the systemThus, or, inthis other last modewords, proves the whol moree extracted suitable when current the subtracted extracted currentby Iqharv ,is which relatively is always low, whereas drawn The available current is definedavailable asextracted the effectiveqharv current that can be used by the active part of the fromthe standby-mode the DC/DC converterbecomes more (I cost = I effective − I at ).higher As shown extracted in the currents, graphs, whenas the itsavailable lower valuecurrent of system or, in other words, the whole extracted current subtracted by I , which is always drawn from risesactivation above charge4.190 µA, becomes standby-mode predominant. becomes theA detailbest choiceed graph in terms of transmissionqharvof minimum datadata rates,rates and versus the relatedextracted threshold currents value is shown is about in Figure 13 min. 14, analytically obtained through Equation (5). WithThe available this assumption, current is the defined Control as circuit the effective for policy current detection that has can the be usedtask ofby correctly the active identifying part of the whether the current available is either above or below the threshold value of 4.190 µA, and system or, in other words, the whole extracted current subtracted by Iqharv, which is always drawn communicating this to the microcontroller through the policy signal. Since the correct relationship from the DC/DC converter (Iavailable = Iextracted − Iqharv). As shown in the graphs, as the available current betweenrises above the 4.190 power μA, received standby-mode by the becomesrectenna the and best the choicepower in extracted terms of by minimum the DC/DC data has rates, been and fully the

related threshold value is about 13 min. With this assumption, the Control circuit for policy detection has the task of correctly identifying whether the current available is either above or below the threshold value of 4.190 μA, and communicating this to the microcontroller through the policy signal. Since the correct relationship between the power received by the rectenna and the power extracted by the DC/DC has been fully

Sensors 2017, 17, 1732 15 of 21

the DC/DC converter (Iavailable = Iextracted − Iqharv). As shown in the graphs, as the available current rises above 4.190 µA, standby-mode becomes the best choice in terms of minimum data rates, and the related threshold value is about 13 min. With this assumption, the Control circuit for policy detection has the task of correctly identifying whether the current available is either above or below the threshold value of 4.190 µA, and communicating this to the microcontroller through the policy signal. Since the correct relationship between the power received by the rectenna and the power extracted by the DC/DC has been fully investigated (see details in Results section), the least power-consuming solution is to simply monitor the average voltage at the rectenna output. This value is solely related to the extracted current through the efficiencies of the interposed stages, and experimental results show that a value ≥585 mV for VRECT is required to obtain an available current of 4.190 µA. Since the low-power voltage supervisor 2 V NCP303 has a threshold value of 2 V, a dedicated low-power voltage amplifier with a 0.585 V = 3.42 gain was added, composed of an LPV521 op-amp and two resistors of 20 MΩ and 4 MΩ. Moreover, a simple RC filter was inserted with τ = 100 ms, in order to reject the periodic fluctuations caused by the FOCV algorithm of the DC/DC converter, which actually disconnect the rectenna for about 4 µs every 8 extraction cycles. The overall quiescent current of the block is 600 nA, as already reported in Table3.

3. Results The fully autonomous DASH7 IoT sensor node with RF harvesting capability was characterized experimentally. Moreover, we tested the capabilities of implementing the adaptive behavior. The available energy is mainly affected by two factors: the transmission power and the distance between transmission and receiving antennas. For the former, two different transmitter powers were selected to be compliant with the RFID standard [26], respectively 0.5 W ERP and 2.0 W ERP. Then, a full analysis with respect to values of distance was performed. Starting from experimental results for the rectenna (Figure4) and DC/DC (Figure7), a precise analytic model of the harvesting module was extracted, and this provides the power extracted by the DC/DC converted as a function of RF source power and node distance. The related results are presented in Section 3.1. Furthermore, model reliability was proved by two experiments accomplished with a regulated RF source placed at a defined distance from the harvesting module (logperiodic directive antenna “PCB VA5JVB”, with GTX = 6 dBi). Combining the harvesting analytic model described in Section 3.1 with Equation (5) and the measured currents reported in Table3, yielded a detailed estimate of startup times and transmission periods as reported in Section 3.2. The foregoing results are summarized in Section 3.3. Section 3.4 presents the experimental results of the sensor node, while Section 3.5 describes the tests performed on voltage supervisors.

3.1. Harvesting Module Results First of all, a specific graph reporting the available power versus node distance is presented in Figure 15. The available power is the total power extracted from the DC/DC converter. It also accounts for losses on the capacitor and on instrumentation. The graph shows two different curves for 0.5 W ERP and 2.0 W ERP, which intersect at threshold values. The value of 1.373 µW is the threshold value at which the system presented can be considered self-sustained, as it can successfully supply the minimum actual load of the system, which is the control circuit for off-mode described in Figure 13. The far from inconsiderable result obtained is that with a transmitted power of just 0.5 W ERP the system presented can start at 8.4 m, while with 2.0 W ERP the distance can rise up to 16.8 m. Minimum distance of 8.4 m with 0.5 W ERP transmitted power was verified through a dedicated experiment. Sensors 2017, 17, 1732 16 of 21 SensorsSensors 2017 2017, ,17 17, ,1732 1732 1616 of of 21 21

Figure 15. Power extracted vs. Node distance. FigureFigure 15. 15. Power Power extracted extracted vs. vs. Node Node distance. distance.

3.2.3.2. Startup/Transmission Startup/Transmission Period PeriodPeriod Results Results WhenWhen thethethe system systemsystem starts startsstarts from fromfrom a completely aa completelycompletely discharged discdischargedharged situation, situation,situation, an initial anan startupinitialinitial startup timestartup is requiredtimetime isis requiredtorequired charge to to the charge charge 8-mF the bufferthe 8-mF 8-mF capacitance buffer buffer capacitanc capacitanc whose valueee whose whose depends value value depends ondepends the harvested on on the the harvested harvested current. current. current. OnceOnce booted, booted, the the node node is is automatically automatically configurconfigur configurededed toto operate operate inin the the best best operating operating mode, mode, providingprovidingproviding the the shortestshortest transmissiontransmission period.period. period. AA A detaidetai detailedledled graphgraph graph ofof of startupstartup startup andand and transmissiontransmission transmission timestimes times isis reportedisreported reported in in inFigureFigure Figure 16. 16. 16 .

Figure 16. Transmission data rate and policy relationship with node distance. FigureFigure 16. 16. Transmission Transmission data data rate rate and and policy policy relationship relationship with with node node distance. distance.

AnalyticAnalytic results results show show show that that that the the the minimum minimum minimum transmissi transmissi transmissiononon period period period is is about isabout about 21 21 21s,s, corresponding corresponding s, corresponding to to an an to extractedanextracted extracted powerpower power ofof 26.526.5 of 26.5 μμWW µ obtainedobtainedW obtained withwith with anan antennaantenna an antenna openopen open circuitcircuit circuit voltagevoltage voltage ofof 1.6501.650 of 1.650 V.V. Similarly,Similarly, V. Similarly, thethe operatingoperating mode mode is is switched switched from from off-mode off-mode to to standby-mode standby-mode according according to to Figure Figure 13 13 when when the the extracted extracted powerpower is is 11.011 11.011 μ μWW (open (open circuit circuit voltage voltage 1.170 1.170 V), V), with with a a period period of of 12 12′43′43″″. .

Sensors 2017, 17, 1732 17 of 21 the operating mode is switched from off-mode to standby-mode according to Figure 13 when the extracted 0 Sensorspower 2017 is 11.011, 17, 1732µ W (open circuit voltage 1.170 V), with a period of 12 43”. 17 of 21

3.3. Power/Transmission Power/Transmission Period Combined Results Figure 17 shows the relationshiprelationship betweenbetween nodenode distance,distance, powerpower extractedextracted fromfrom DC/DCDC/DC and obtainable transmission periods. St Startingarting from from distance (bottom horizontal axis), the single blue line represents extractedextracted powerpower when when the the source source is 0.5is 0.5 W ERP,W ERP, the triplethe triple pink pink line isline extracted is extracted power power when whenthe source the source is 2.0 W is ERP. 2.0 OnceW ERP. the Once available the poweravailabl ise known, power theis known, obtainable the periodobtainable can be period found can on thebe founddotted on green the line,dotted whose green values line, arewhose reported values along are reported the top horizontal along the axis.top horizontal Moreover, axis. three Moreover, operative threeregions operative of extracted regions power of extracted are identifiable: power are a turn identifiable: off region a where turn off the region node where cannot the be switchednode cannot on, bean switchedoff-mode operating on, an off-mode region operating and a standby-mode region andone. a standby-mode one.

Figure 17. CombinedCombined graph graph showing showing the the relationship relationship between between distance, extracted power and transmission periods.

Table 4 recaps the threshold values shown in Figure 17. Table4 recaps the threshold values shown in Figure 17.

Table 4. Minimum distances in the three operating conditions. Table 4. Minimum distances in the three operating conditions.

Antenna Voc Transmission Distance (m) Extracted Power (μW) (V)Antenna Voc Extracted TransmissionPeriod Distance0.5 W ERP (m) 2.0 W ERP Switch-on 0.490 (V) Power 1.373 (µ W) Period→∞ 0.5 W ERP8.4 2.0 W ERP 16.8 Policy ThresholdSwitch-on 1.170 0.490 11.011 1.373 → ∞ 0:12:43 8.4 4.6 16.8 9.1 Min. TransPolicy Period Threshold 1.650 1.17026.531 11.011 0:12:430:00:21 4.63.5 9.1 7.0 Min. Trans Period 1.650 26.531 0:00:21 3.5 7.0 3.4. Sensor Node Results 3.4. Sensor Node Results The effective operation of the DASH7 node was tested, in terms of power consumption, transmissionThe effective capability operation and adaptive of the DASH7data rate. node Figure was 18a tested, shows in waveforms terms of powerof signals consumption, related to off-modetransmission operation capability. Supply and VDD adaptivedecreases data as expected rate. Figure from 2.318a V shows to 1.9 V waveforms over a time of interval signals of related about 680to off-mode ms, which operation correspond. Supply toV DDa singledecreases transmission. as expected Wi fromth reference 2.3 V to 1.9 to VFigure over a10, time this interval time accounts of about for680 the ms, microcontroller which correspond start-up to a single phase, transmission. the data acquisition With reference and finally to Figure data 10 transmission., this time accounts The reset for signalthe microcontroller (Reset_VDD) is start-up activated phase, at the the end data of acquisition each transmis andsion. finally Similarly, data transmission. in Figure The18b the reset control signal signals(Reset_V forDD standby-mode) is activated atare the shown. end of Voltage each transmission. drop is about Similarly, 20 mV as in expected, Figure 18 bwhile the control in this signalsmode the for Reset_Startstandby-mode signalareshown. is activated Voltage instead drop of is Reset_V about 20DD mVat the as end expected, of transmission. while in this mode the Reset_Start signal is activated instead of Reset_VDD at the end of transmission.

Sensors 2017, 17, 1732 18 of 21 Sensors 2017, 17, 1732 18 of 21

Sensors 2017, 17, 1732 18 of 21

(a) (b)

Figure 18. Voltage drop(a) and control signals of micro-controller in off-mode( b(a)) and standby-mode (b). Figure 18. Voltage drop and control signals of micro-controller in off-mode (a) and standby-mode (b). Figure 18. Voltage drop and control signals of micro-controller in off-mode (a) and standby-mode (b). Transmitted data are then received by a dedicated gateway, which displays data directly onto a PC terminal.Transmitted Figure data 19 areshows then the received received by data a dedicated bytes with gateway, different which sensor displays values. data directly onto a PC terminal.Transmitted Figure data 19 are shows then the received received by dataa dedicated bytes with gateway, different which sensor displays values. data directly onto a PC terminal. Figure 19 shows the received data bytes with different sensor values.

Figure 19. Data received format. Figure 19. Data received format. 3.5. Voltage Supervisor and Policy SelectionFigure Results 19. Data received format. 3.5. VoltageFinally, Supervisor power optimization and Policy Selection circuits Resultsfor policy selection were tested. In Figure 20 control signals 3.5. Voltage Supervisor and Policy Selection Results behavior can be observed with respect to DC/DC output voltage VHARV. Considering signals reported Finally, power optimization circuits for policy selection were tested. In Figure 20 control signals in FigureFinally, 13, power VHARV optimization is initially below circuits the for activation policy selection threshold were of tested. 2.3 V, In Figureso all other20 control signals signals are behavior can be observed with respect to DC/DC output voltage VHARV. Considering signals reported disabled.behavior canAs besoon observed as VHARV with exceeds respect 2.3 to V, DC/DC the node output can be voltage enabled VHARV so V. ConsideringDD (=VHARV) is signals provided, reported and in Figure 13, VHARV is initially below the activation threshold of 2.3 V, so all other signals are otherin Figure supervisors 13,VHARV areis supplied initially below through the V activationDD too. Start threshold signal ofis 2.3high, V, soas allVHARV other is signalsabove are2.0 disabled.V, while disabled. As soon as VHARV exceeds 2.3 V, the node can be enabled so VDD (=VHARV) is provided, and Aspolicy soon is asalso V HARVhigh exceedsas VOC is 2.3kept V, above the node 1.170 can V bein enabledthis stage so (sVeeDD Table(=VHARV 4). As) isRF provided, source is andswitched other other supervisors are supplied through VDD too. Start signal is high, as VHARV is above 2.0 V, while off,supervisors the buffer are voltage supplied VHARV through beginsVDD to too.decrease, Start signaland policy is high, is asimmediately VHARV is above forced 2.0 low. V, while Then policy start policy is also high as VOC is kept above 1.170 V in this stage (see Table 4). As RF source is switched signalis also goes high low as V whenOC is V keptHARV above falls below 1.170 2.0 V inV thisas expected, stage (see while Table supply4). As voltage RF source VDD isis switcheddisconnected off, off, the buffer voltage VHARV begins to decrease, and policy is immediately forced low. Then start whenthe buffer VHARV voltage goes below VHARV 1.9begins V. to decrease, and policy is immediately forced low. Then start signal signal goes low when VHARV falls below 2.0 V as expected, while supply voltage VDD is disconnected goes low when VHARV falls below 2.0 V as expected, while supply voltage VDD is disconnected when when VHARV goes below 1.9 V. VHARV goes below 1.9 V.

Sensors 2017, 17, 1732 19 of 21 Sensors 2017, 17, 1732 19 of 21

FigureFigure 20. 20. PowerPower optimization optimization circuits circuits and and policy policy detection detection waveforms. waveforms.

4.4. Conclusions Conclusions InIn this workwork we we proposed proposed an autonomousan autonomous battery-less battery-less sensor sensor node withnode RF with harvesting RF harvesting capability capabilityin the 868 in MHz the 868 UHF MHz band. UHF Temperature band. Temperature and humidity and humidity data are data acquired are acquired and transmitted and transmitted using a usingstandard a standard DASH7 DASH7 protocol, protocol, which allowswhich theallows node the to node be compatible to be compatible with IoT with infrastructure. IoT infrastructure. AnAn optimized optimized harvesting harvesting module, module, combining combining a a dedicated dedicated rectifying rectifying antenna antenna and and an an ultra-low ultra-low powerpower DC/DC DC/DC converter converter with with MPPT, MPPT, is is used used to to obtain obtain considerable considerable results results at at long long distances. distances. Besides Besides that,that, innovative innovative solutions solutions for for adaptive adaptive behavior behavior are are introduced. introduced. In In particular, particular, by by monitoring monitoring the the rectifyingrectifying antenna antenna open-circuit open-circuit voltage, voltage, a a dedicated dedicated low-power low-power management management circuitry circuitry can can select select two two differentdifferent operatingoperating modesmodes of theof sensingthe sensing node. node. Furthermore, Furthermore, in both cases,in both the minimumcases, the transmission minimum transmissionperiod is achieved period accordingly is achieved to theaccordingly available RFto powerthe available through RF the power managementthrough the module.power managementThe node achieves module. operation The node with achieves standard operation wireless wi protocolth standard in micropower wireless protocol condition. in micropower condition.As a result, the node is able to operate with an antenna open-circuit voltage of solely 520 mV, correspondingAs a result, to the 12.3 nodeµW is input able power. to operate Consequently, with an antenna with a RFopen-circuit source of 2.0voltage W ERP, of solely the node 520 can mV, be correspondingturned on up to to 16.8 12.3 m μW distance. input power. Consequently, with a RF source of 2.0 W ERP, the node can be turned on up to 16.8 m distance. Acknowledgments: This work was performed within the activities of the joint research laboratory of University Acknowledgments:of Bologna and STMicroelectronics This work was performed at the Advanced within Research the activities Center of the on Electronicjoint research Systems laboratory of the of university. University ofAuthor Bologna Contributions: and STMicroelectronicsMatteo Pizzotti at the and Advanced Davide Fabbri Research designed, Center implemented on Electronic and Systems tested theof the electronic university. section of the “harvesting and power management module”, and wrote the related sections of the paper. Luca Perilli Author Contributions: Matteo Pizzotti and Davide Fabbri designed, implemented and tested the electronic designed, implemented and tested the “active sensing node”, and wrote the related sections of the paper. sectionMassimo of the del “harvesting Prete designed, and power implemented management and tested module”, the “rectenna”,and wrote the and related wrote sections the related of the sections paper. ofLuca the Perillipaper. designed, Michele Diniimplemented designed and tested provided the the“active “DC/DC sensing converter”. node”, and Eleonora wrote the Franchi related Scarselli, sections Aldo of the Romani, paper. MassimoRoberto Canegallo,del Prete designed, Alessandra implemented Costanzo and and Diego tested Masotti the “rectenna”, supervised and the wrote development the related of thesections system of andthe paper.of individual Michele electronic Dini designed and RF and blocks, provided and the “DC/DC writing of converter”. the paper. Eleonora Matteo Pizzotti Franchi wrote Scarselli, the commonAldo Romani, parts, merged and coordinated the whole paper. Roberto Canegallo, Alessandra Costanzo and Diego Masotti supervised the development of the system and of individualConflicts ofelectronic Interest: andThe RF authors blocks, declare and nothe conflictwriting of of interest. the paper. Matteo Pizzotti wrote the common parts, merged and coordinated the whole paper. References Conflicts of Interest: The authors declare no conflict of interest. 1. Yick, J.; Mukherjee, B.; Ghosal, D. survey. Comput. Netw. 2008, 52, 2292–2330. References[CrossRef ] 2. Sudevalayam, S.; Kulkarni, P. Energy harvesting sensor nodes: Survey and implications. IEEE Commun. 1. Yick, J.; Mukherjee, B.; Ghosal, D. Wireless sensor network survey. Comput. Netw. 2008, 52, 2292–2330. Surv. Tutor. 2011, 13, 443–461. [CrossRef] 2. Sudevalayam, S.; Kulkarni, P. Energy harvesting sensor nodes: Survey and implications. IEEE Commun. Surv. Tutor. 2011, 13, 443–461.

Sensors 2017, 17, 1732 20 of 21

3. Del Prete, M.; Masotti, D.; Costanzo, A.; Magno, M.; Benini, L. A 2.4 GHz–868 MHz dual-band wake-up radio for wireless sensor network and IoT. In Proceedings of the IEEE 11th International Conference on Wireless and , Networking and Communications (WiMob), Abu Dhabi, UAE, 19–21 October 2015; pp. 322–328. 4. Vullers, R.J.M.; van Schaijk, R.; Doms, I.; van Hoof, C.; Mertens, R. Micropower energy harvesting. Solid-State Electron. 2009, 53, 684–693. [CrossRef] 5. Lu, X.; Wang, P.; Niyato, D.; Kim, D.I.; Han, Z. Wireless networks with RF energy harvesting: A contemporary survey. IEEE Commun. Surv. Tutor. 2015, 17, 757–789. [CrossRef] 6. Gollakota, S.; Reynolds, M.S.; Smith, J.R.; Wetherall, D.J. The emergence of RF-powered computing. Computer 2014, 47, 32–39. [CrossRef] 7. Oh, S.; Wentzloff, D.D. A −32 dBm sensitivity RF power harvester in 130 nm CMOS. In Proceedings of the 2012 IEEE Radio Frequency Integrated Circuits Symposium, Montreal, QC, Canada, 17–19 June 2012; pp. 483–486. 8. Carvalho, N.; Georgiadis, A.; Costanzo, A.; Stevens, N.; Kracek, J.; Pessoa, L.; Roselli, L.; Dualibe, F.; Schreurs, D.; Mutlu, S.; et al. Europe and the future for WPT: European contributions to wireless power transfer technology. IEEE Microw. Mag. 2017, 18, 56–87. 9. Chaour, I.; Fakhfakh, A.; Kanoun, O. Enhanced passive RF-DC converter circuit efficiency for low RF energy harvesting. Sensors 2017, 17, 546. [CrossRef][PubMed] 10. Eid, A.; Costantine, J.; Tawk, Y.; Ramadan, A.H.; Abdallah, M.; ElHajj, R.; Awad, R.; Kasbah, I.B. An efficient RF energy harvesting system. In Proceedings of the 2017 11th European Conference on Antennas and Propagation (EUCAP), Paris, France, 19–24 March 2017; pp. 896–899. 11. Collado, A.; Georgiadis, A. Conformal hybrid solar and electromagnetic (EM) energy harvesting rectenna. IEEE Trans. Circuits Syst. I Regul. Pap. 2013, 60, 2225–2234. [CrossRef] 12. Ouda, M.H.; Khalil, W.; Salama, K.N. Wide-range adaptive RF-to-DC power converter for UHF RFIDs. IEEE Microw. Wirel. Compon. Lett. 2013, 26, 634–636. [CrossRef] 13. Stoopman, M.; Keyrouz, S.; Visser, H.J.; Serdijn, K.P.W.A. Co-design of a CMOS rectifier and small loop antenna for highly sensitive RF energy harvesters. IEEE J. Solid-State Circuits 2014, 49, 622–634. [CrossRef] 14. Dini, M.; Romani, A.; Filippi, M.; Tartagni, M. A nano-current power management IC for low voltage energy harvesting. IEEE Trans. Power Electron. 2015, 31, 4292–4304. [CrossRef] 15. Romani, A.; Tartagni, M.; Sangiorgi, E. Doing a lot with a little: Micropower conversion and management for ambient-powered electronics. Computer 2017, 50, 41–49. [CrossRef] 16. Syed, Y.; Hegde, B.G.; Prabhakar, T.V.; Manjunath, M.; Vinoy, K.J. RF energy harvesting chip powered sensor node. In Proceedings of the 2016 IEEE International Conference on Electronics, Circuits and Systems, Monte Carlo, Monaco, 11–14 December 2016; pp. 748–751. 17. Choi, K.W.; Ginting, L.; Rosyady, P.A.; Aziz, A.A.; Kim, D.I. Wireless-powered sensor networks: How to realize. IEEE Trans. Wirel. Commun. 2017, 16, 221–234. [CrossRef] 18. Li, T.; Fan, P.Y.; Chen, Z.C.; Letaief, K.B. Optimum transmission policies for energy harvesting sensor networks powered by a mobile control center. IEEE Trans. Wirel. Commun. 2016, 15, 6132–6145. [CrossRef] 19. Kamalinejad, P.; Mahapatra, C.; Sheng, Z.; Mirabbasi, S.; Leung, V.C.M.; Guan, Y.L. Wireless energy harvesting for the . IEEE Commun. Mag. 2015, 53, 102–108. [CrossRef] 20. Mason, A.; Al-Shamma’a, A.I.; Shaw, A. An antenna for wireless sensor applications. In Proceedings of the 2009 Loughborough Antennas Propagation Conference, Leicestershire, UK, 2 February 2009. 21. D’Elia, A.; Perilli, L.; Viola, F.; Roffia, L.; Antoniazzi, F.; Canegallo, R.; Cinotti, T.S. A self-powered WSAN for energy efficient heat distribution. In Proceedings of the 2016 IEEE Sensors Applications Symposium (SAS), Catania, Italy, 2016; pp. 1–6. 22. Moon, J.-I.; Sim, D.-U.; Park, S.-O. Compact PIFA for 2.4/5 GHz dual ISM-band applications. Electron. Lett. 2004, 40, 844–846. [CrossRef] 23. De Carli, L.G.; Juppa, Y.; Cardoso, A.J.; Galup-Montoro, C.; Schneider, M.C. Maximizing the power conversion efficiency of ultra-low-voltage CMOS multi-stage rectifiers. IEEE Trans. Circuits Syst. I Regul. Pap. 2015, 62, 967–975. [CrossRef] 24. Essel, J.; Brenk, D.; Heidrich, J.; Weigel, R. A highly efficient UHF RFID frontend approach. In Proceedings of the 2009 IEEE MTT-S International Microwave Workshop on Wireless Sensing, Local Positioning, and RFID, Cavtat, Croatia, 24–25 September 2009. Sensors 2017, 17, 1732 21 of 21

25. Analog Devices. Ultralow Power Energy Harvester PMU with MPPT and Charge Management; ADP5091; Analog Devices: Norwood, MA, USA, 2016. 26. Report ITU-R SM.2255-0(10/2012). Technical Characteristics, Standards, and Frequency Bands of Operation for Radio-Frequency Identification (RFID) and Potential Harmonization Opportunities; International Union: Geneva, Switzerland, 2015. 27. Masotti, D.; Costanzo, A.; del Prete, M.; Rizzoli, V. Genetic-based design of a tetra-band high-efficiency radio-frequency energy harvesting system. IET Microw. Antennas Propag. 2013, 7, 1254–1263. [CrossRef] 28. DASH7 Alliance Protocol 1.0: Low-Power, Mid-Range Sensor and Actuator Communication. Available online: http://www.dash7-alliance.org/ (accessed on 1 April 2017). 29. Ergeerts, G.; Nikodem, M.; Subotic, D.; Surmacz, T.; Wojciechowski, B.; de Meulenaere, P.; Maarten, W. DASH7 alliance protocol in monitoring applications. In Proceedings of the 2015 10th International Conference on P2P, Parallel, Grid, Cloud and Internet Computing (3PGCIC), Krakow, Poland, 4–6 November 2015; pp. 623–628.

© 2017 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).