WIRELESS powering system implementation and design

Transcription

1 1046 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 4, APRIL 2014 Scalable RF Energy Harvesting Zoya Popović, Fellow, IEEE, Sean Korhummel, Student Member, IEEE, Steven Dunbar, Student Member, IEEE, Robert Scheeler, Student Member, IEEE, Arseny Dolgov, Member, IEEE, Regan Zane, Senior Member, IEEE, Erez Falkenstein, Member, IEEE, and Joseph Hagerty, Member,IEEE (Invited Paper) Abstract This paper discusses harvesting of low-power density incident plane waves for electronic devices in environments where it is difficult or impossible to change batteries and where the exact locations of the energy sources are not known. As the incident power densities vary over time and space, distributed arrays of antennas with optimized power-management circuits are introduced to increase harvested power and efficiency. Scaling in array size,power,dcload,frequency, and gain is discussed through three example arrays: a dual industrial scientific medical band Yagi-Uda array with a low-power startup circuit; a narrowband 1.96-GHz dual-polarized patch rectenna array with a reconfigurable dc output network designed for harvesting base-station power; and a broadband dual-polarized 2 18-GHz array with multi-tone performance. The efficiency of rectification and power management is investigated for incident power densities in the W/cm range. Index Terms Power management, rectifier, RF, wireless powering, wireless sensors. I. BACKGROUND AND INTRODUCTION WIRELESS powering system implementation and design differs significantly for inductive or resonant near-field powering [1], far-field directive power beaming [2] [5], and nondirective low-power far-field harvesting [6] [9]. The latter is differentiated from RF identification (RFID) in that the powering is independent of signal transmission. Additionally, the power transfer is not as sensitive to device orientation as in the case of near-field power transfer with resonant inductors or far-field directive beaming. In this paper, we address far-field power harvesting of low incident power densities in the range of W/cm using arrays of antennas integrated with rectifiers and power-management circuits. The design, implementation, and characterization of narrowband, multi-frequency, and Manuscript received October 10, 2013; revised December 17, 2013; accepted December 18, Date of publication February 05, 2014; date of current version April 02, This work was supported by the Rehabilitation Engineering Research Center, U.S. Department of Education NIDRR Grant H133E040019, and the National Science Foundation, Grant ECCS Z. Popović, S. Korhummel, S. Dunbar, R. Scheeler, A. Dolgov, E. Falkenstein, and J. Hagerty are with the Department of Electrical, Computer and Energy Engineering, University of Colorado at Boulder, Boulder, CO USA ( zoya@colorado.edu). R. Zane was with the Department of Electrical, Computer and Energy Engineering, University of Colorado at Boulder, Boulder, CO USA. He is now with the Department of Electrical and Computer Engineering, Utah State University, Logan, UT USA. Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /TMTT Fig. 1. High-level block diagram of scalable RF energy harvesting approach based on rectenna arrays. Incident waves from arbitrary angles and at multiple frequencies, power levels, and polarizations produce rectified output power at each rectenna element of the array. The reconfigurable dc combining network chooses series and parallel combinations for the most efficient dc impedance given the power-management circuit and storage element. broadband arrays are overviewed with a focus on scaling in power and frequency, as well as efficiency optimization of both the rectenna and power-management circuit at very low power levels. A general block diagram of a scalable approach to RF power harvesting is shown in Fig. 1. Incident waves at various frequencies, power densities, polarizations, and incidence angles are received by a rectenna array. Each antenna element is integrated with a simple rectifier, with RF-isolated dc collection lines that are connected reconfigurably in series and/or parallel depending on the resistance required by the power-management circuit. Since, in general, the received power varies due to multipath, the power-management circuit needs to perform power tracking. Co-design of the power reception and power-management circuits is required to achieve the highest total system efficiency. A. Efficiency of Wireless Harvesting The efficiency of the wireless power reception includes the efficiency of the integrated antenna rectifier (rectenna),,and the converter efficiency,, and can be written as (1) IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See for more information.

2 POPOVIĆ et al.: SCALABLE RF ENERGY HARVESTING 1047 where is the dc power output of the rectenna, is the dc power delivered to the storage element, and is the incident power on the rectenna of geometric area given by where is the angle-dependent incident power density of one or more plane electromagnetic waves. To obtain the total incident power, in general the total power density is integrated over a sphere. The array element rectenna efficiency defined in this way is determined by measuring the dc voltage across a known load and by calibrating the incident power density: (2) (3) where is the power density at normal plane-wave incidence, assuming a single transmitter in the far field and an electrically small antenna with constant across its aperture. Note that this is the most conservative definition since the geometric area of the antenna is always smaller than its effective area, and (3) thus takes into account the aperture efficiency and losses of the antenna, the impedance mismatch between the antenna and the rectifier, and the losses in the rectifier circuit [7]. In the case of rectenna arrays larger than a free-space wavelength at the lowest frequency of operation, the power density is nonuniform across the array. Since the efficiency is a nonlinear function of incident power density, the array rectification efficiency can be expressed as where the sum is performed over the array elements. Multiple transmitters can be taken into account by adding the dc powers resulting from a superposition of plane waves. B. RF Harvesting Environment A number of recent publications review measurements of available RF power densities in various locations, e.g., Atlanta, GA, USA, Tokyo, Japan [10], and London, U.K. [11]. The measured power densities are in the range of W/cm in the UHF and various industrial scientific medical (ISM) bands. Another example of harvesting is a wireless sensor for monitoring cell-phone base-station communications. Such sensors can be placed near or on a base-station tower, and harvest antenna sidelobe or reflected power sufficient to enable monitoring of the base-station activity [12]. Multiple studies [13], [14] show that power densities within 50 m of a cell tower exceed 1 W/cm, but vary widely with location and time of day, and are also different for different types of base-stations. The sketch of a base-station tower and specified antenna radiation pattern in one sector is shown in Fig. 2(a) with calculated line-of-sight normalized received power density shown in Fig. 2(b) as a function of distance from the base-station. (4) Fig. 2. Cellular base-station tower with specified antenna radiation patterns: (a) which are used to calculate normalized power density received within angle as a function of distance from the tower [17]. The incident powering wave will, in general, reach the sensor with a polarization that varies in time, depending on the radiating source and propagation environment. Since the orientation of the sensor is not always known, the dependence on alignment of the source relative to the receiver should be minimized. In the case of a linearly polarized source, a linearly polarized receiving antenna will receive a power where the cosine term takes in to account the polarization mismatch between the transmit and receive antennas, described by the angle. In a Rayleigh multipath propagation environment, for a vertically polarized transmitting antenna, the electric field at the receiving antenna will on average contain equal power in the two orthogonal polarizations. Therefore, a dual-orthogonal polarization rectenna, which rectifies the power contained in orthogonally polarized waves independently and adds the dc output, will on average receive the most power with the least variation over time and with increased overall efficiency over a linearly polarized rectenna. This property has been recognized by many of the rectenna designs in the literature, e.g., [15] and [16], and is discussed in detail in [9].

3 1048 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 4, APRIL 2014 TABLE I SUMMARY OF RECTENNA ARRAYS DISCUSSED IN THE PAPER. :NUMBER OF ANTENNA ELEMENTS; :NUMBER OF RECTIFIERS; :NUMBER OF DC COLLECTION CIRCUITS. ALL ARRAYS USE THE SKYWORKS SMS7630 ZERO-BIAS GaAs SCHOTTKY DIODE 1) An extremely low-mass and inexpensive dual-frequency Yagi-Uda array for power densities in the range of 1 W/cm. In this example, the two scaling factors are antenna gain and frequency since the rectenna is designed for harvesting power in two ISM wireless communication bands of 915 MHz and 2.45 GHz. 2) A narrowband dual-polarized patch antenna array designed to harvest power and monitor operation of a 1.96-GHz cellular base-station. In this example, the scaling factors are array size, harvested power, dc load, and power-management efficiency scaling. 3) An ultra-broadband array tested between 2 18 GHz is reviewed and a comparison given. In this example, scaling is performed in size, frequency range harvested, and dc collection circuit load. Table I gives an overview of the scaling parameter space for the specific examples presented here. Fig. 3. Photographs of the three arrays discussed in this paper. (a) Dual- ISM band Yagi-Uda array. (b) Narrowband 20-element dual-linearly polarized patch array for 1.96 GHz, showing front and isolated ground plane back side. (c) 64-element dual-circularly polarized broadband spiral array [6]. Although there are environments with substantial RF ambient power, in order to reliably power devices, it is useful to additionally have controlled compliant low-power transmitters that operate at frequencies that overlap with the harvesting frequency range. The allowable power levels are published by the Federal Communications Commission (FCC) for ISM bands ( MHz and , , , and GHz). Additional regulations allow short period transmission of modulated higher power pulses (FCC CFR47, part ), which is useful for startup conditions discussed later in this paper. C. Arrays and Scaling This paper discusses three specific array designs, shown in Fig. 3, each focusing on a different set of challenges associated with harvesting, and with different types of scaling. II. RECTENNAS AND DC COLLECTION CIRCUITS Time-domain Fourier analysis of ideal rectennas has been presented by a number of authors, e.g., [7], [18] and gives insight into the waveforms of the voltage and current across the diode. The design of a rectenna requires nonlinear rectifier characterization, based on experimental or simulated source pull [7]. After the rectifier impedance over a range of power and frequency is determined, an optimal antenna matched over the design parameters is designed. The integrated rectenna and dc collection network is then characterized in terms of its dc equivalent circuit for various RF input powers, allowing for optimization of the power-management circuit. A. Dual-Frequency Yagi-Uda Rectenna Array The requirements for the energy harvester presented in this paper are available online 1 : smallest possible mass; incident power density of 1 W/cm at 2.45 GHz and 915 MHz; linear vertical polarization; known source direction; fixed dc power load, k. For minimal size with largest possible harvested power, the design is optimized for 2.45-GHz operation, while ensuring that the rectenna provides some output at 915 MHz. The approximately known incident wave direction indicates that scaling the antenna gain with minimal antenna size will provide higher rectified power. 1 [Online]. Available:

4 POPOVIĆ et al.: SCALABLE RF ENERGY HARVESTING 1049 TABLE II SUMMARY OF DIODE SOURCE PULL SIMULATIONS Fig. 5. Measured Yagi-Uda rectenna efficiency obtained from (3) for three incident power levels as a function of the dc load. Note that the efficiency drops for higher input power since it is optimized for the low incident power diode impedance. Fig. 4. (a) Dual-frequency Yagi-Uda rectenna array layout. CPS shorted stubs are long at 2.45 GHz, and inductive at 915 MHz. The capacitor is chosen to tune out the inductance. (b) Simulated antenna input impedance versus frequency from 900 MHz to 2.6 GHz for the antenna with (solid blue line in online version) and without (dashed red line in online version) the shunt capacitor, superimposed with source pull data for the diode at 915 MHz and 2.45 GHz, showing contours of constant rectified dc power into a 2.2-k load. The highest power is 13 dbm, and the contours are given with 2-dB increments. Source pull simulations for a Skyworks GaAs zero-bias Schottky diode SMS are performed using AWR/NI Microwave Office s harmonic balance nonlinear simulator with a nonlinear SPICE diode model to determine the optimum impedance for highest dc voltage across the fixed 2.2-k- load. The input power levels are calculated by assuming a 8-dBi gain antenna at 2.45 GHz and W/cm, and the results are shownintableii. Usually, a Yagi-Uda antenna array is designed to be matched to 50. From the source pull data in Table II, it is seen that an inductive reactance is needed to match the diode at 2.45 GHz, as is commonly done in RFID tags [19]. The printed antenna layout with the -long reflector, -long driven element, and -long director element, all 4 mm in width, is shown in Fig. 4(a). The distances of the reflector and director from the driven element are and, resulting in a match directly at the diode rectifier terminals where the source pull reference plane is defined. A dc collection network is designed to isolate the dc load from the rectenna, using two quarter-wave shorted co-planar strip (CPS) lines. Isolation is enabled by the capacitor that acts as an RF short. The dc collection network and load do not affect the RF match. The shorted stubs are open circuits at the diode feed point, but provide a path to deliver power to the dc load. The three-element Yagi-Uda array achieves 6dBi gain when matched at 2.45 GHz. The gain is reduced to 3dBiat 915 MHz, but this is still sufficient to provide 13.5 dbm of power to the antenna feed point, which turns on the diode provided the antenna is matched. At 915 MHz, the shorted CPS stub is electrically short and inductive, with a reactance given by. The value of the bypass capacitor is not critical at 2.45 GHz, but at 915 MHz, the capacitance can be used to tune the inductive reactance of the shorted stubs. The antenna complex input impedance is simulated using the 3-D planar method of moments solver Axiem (AWR MWO). The simulated performance is combined with measured data provided by ATC for 600-L 0402 package capacitors and an ideal 2.2-k dc load to determine the input impedance seen at the feed point of the antenna. Capacitances between 1 27 pf have no effect on the 2.45-GHz performance, and a value of 15 pf provides a good match at 915 MHz, as shown in Fig. 4(b). The rectenna is fabricated on a flexible substrate (Rogers Ultralam 3850, 1-mil thick, ) with a final mass after metal etching of approximately 0.5 g. The measured data in our laboratory, corresponding to Table II, is shown in Fig. 5. Notice that an increase in incident power reduces performance because the rectenna is designed for low input power. B. Narrowband Dual-Polarization Patch Rectenna Array The effective area that collects RF power can be increased with multiple rectenna elements that are not RF connected so that the dc power is summed after rectification. A 20-element 22 cm 28 cm patch rectenna array [see Fig. 3(b)] is designed to receive and rectify power densities found near cell tower basestations, which are on the order of W/cm. The rectennas are

5 1050 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 4, APRIL 2014 Fig. 6. Dual-linearly polarized patch rectenna with two rectifying diodes, each connected to a linearly polarized antenna feed. Impedance matching microstrip lines match the high radiating-edge patch impedance to the diode impedance obtained by source pull (Fig. 7) at 1.96 GHz. The RF null of the patch is used as a dc-isolated terminal of the dc collection circuit. Fig. 8. Measured rectified power at 1.96 GHz for a single patch antenna element and various incident power density levels in one polarization. The response to the orthogonal polarization is within a few percent [9]. Fig. 7. Measured rectified power contours in dbm, for 0-dBm input power for the SMS7630 diode. The optimal load is around at 1.96 GHz. The Smith chart is normalized to 120. narrowband, centered at the expected transmitter frequency of 1.96 GHz. Each dual-polarized antenna (Fig. 6) isprintedona 0.75-mm-thick substrate with. Two Skyworks SMS7630 Schottky diodes are connected to the two feeds corresponding to the two orthogonal linear polarizations. The matching circuit are designed to match the radiating-edge high impedance of the patch to the optimal diode impedance, obtained by source pull as shown in Fig. 7 for. The RF null in the center of the patch is used for a dc connection since it is automatically RF isolated. Fig. 8 shows measured rectified power in one polarization as a function of incident power density and dc load. It is seen that the optimal load for which the rectified power is the highest is around 500.Whenmanysuchelementsare arrayed, the power will increase, but the load will vary depending on the inter-element connections. A 4 5elementarray of identical patch antennas such as the one in Fig. 7 is fabricated with dc-isolated ground planes and connected to a reconfigurable dc network, described in Section II-C. C. Broadband Spiral Rectenna Array The challenge in broadband rectenna design lies in the nature of the antenna and diode frequency-dependent impedance. For Fig. 9. Simulated -parameters of the spiral antenna shown in inset, along with optimal SMS7630 diode source impedance from 2 to 8 GHz shown in black and obtained by source pull simulations. maximal power transfer, the antenna impedance would match the optimal diode impedance at all frequencies. Since this is difficult to accomplish with low loss and in a compact way, our suboptimal, but simple approach is to present a constant impedance to the diode by using a frequency-independent antenna element. Broadband right- and left-hand circularly polarized spiral antennas designed to cover 2 18-GHz bandwidth are presented in detail in [6]. A layout of the rectenna element is shown in Fig. 9, along with load pull diode data, shown in the black area on the Smith chart. The impedance of the diode for low incident power levels is measured and simulated from 2 to 8 GHz for this plot and is shown together with the antenna impedance over this range. The impedance of the antenna can only be approximately matched over the entire frequency range so the efficiency of harvesting is not expected to be constant over frequency. In the array in Fig. 3(c), each element independently rectifies, and the element spacing is compact. At lower frequencies, the increased mutual coupling between elements and the presence of the bias lines increases the power delivered to the diodes. Alternating left and right circular polarization between neighboring elements, and an additional rotation by 90 from element to element implies that the array suffers an average 3-dB input polarization loss for every possible polarization of the incident energy, ensuring a flat polarization response. The ambient RF power levels vary by several orders of magnitude, implying a varying dc load as reflected in the simulated I V characteristics

6 POPOVIĆ et al.: SCALABLE RF ENERGY HARVESTING 1051 Fig. 10. Simulated I V curves of a single spiral rectenna element as a function of load resistance and input RF power to the diode. The peaks in rectified power are indicated by circles with the corresponding optimal load resistance (100 to 63 k ). Fig. 11. Reconfigurable dc connection network for one row of the 20-element patch array (layout and circuit diagram) allows for combinations shown in the table. indicates rectennas connected in series, and such rows connected in parallel. The measured optimal dc load and maximal received power are given in each case. of a single rectenna element, shown in Fig. 10. The results are plotted in analogy with photovoltaic curves. III. ARRAY DC LOAD SCALING In Fig. 10, the optimal load impedance ranges from 100 to 63 k over 50 db of input power variation. For maximum dc power generation, it is necessary to either match the output characteristics of the source with the load, or to insert an intermediate dc dc converter with peak power tracking. In either case, the rectenna array presents some equivalent dc load, which can be reconfigured in the case of the array by series and parallel connections between the array elements. This has been addressed with a discussion of dc power output combining in [20] and dc load dependence in [21]. The nonlinear performance of the diodes will have an effect on the performance for various dc connections. Increased current or voltage in a predominately series or parallel-connected array can lead to over-biasing of the individual diodes such that the rectification process is uniformly degraded over the array, which is a common phenomenon in solar arrays. Furthermore, the dc power level is very sensitive to under-illumination. A. 20-Element Narrowband Patch Antenna Array The 20-element patch antenna array is fabricated in a way that all patch elements are dc separated. This is accomplished by dc breaks in the ground plane, with capacitive shorts to ensure a common RF ground for all the patches and matching networks, as seen in the photograph of Fig. 3(b). The reconfigurable dc connection circuit diagram for one row of the 20-element patch array is shown in Fig. 11, where the switches S1 are double-pole double-throw (DPDT) and S2 are single-pole single-throw (SPST). For all switches in the horizontal position, the rectenna elements Ri,j are connected in series. For S2 open and S1 in position p, they are in parallel. The different rows are connected similarly. The switched network allows for all combinations shown in the table. The measured optimal dc load varies from 10 to Fig. 12. Example measured rectified power as a function of dc load for various incident power densities for the 20-element arrays. (top) When all elements are dc-connected in parallel, the optimal dc load is very low. (bottom) Whenall elements are connected in series, the optimal load is high. In both cases, a CP wave is incident. 5k, while the maximal received power remains roughly constant. Example measured rectified power as a function of dc load for various incident power densities for the 20-element arrays are shown in Fig. 12. B. Broadband 64-Element Spiral Array The 64-element array was connected using 2 1 parallel pairs connected in series to form a 2 2 sub-subarray; four of these are connected in parallel to create a 4 4 subarray. Four such 16-element subarrays are combined in the 8 8arraywith reconfigurable connectivity.

7 1052 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 4, APRIL 2014 TABLE III RECTIFICATION EFFICIENCY FOR 64-ELEMENT SPIRAL ARRAY WITH THREE DC-CONNECTION SCHEMES FOR TWO INCIDENT POWER DENSITIES AND A RANDOMLY POLARIZED ILLUMINATION AT 3GHz For different dc connections, the diodes across the array are biased at different quiescent points on the diode I V curve, resulting in different efficiency. Table III shows efficiency measured across a 600- load for three dc-connection schemes for a low and high incident power density. For this array, the parallel configuration of the subarrays gave the highest efficiency over the range of incident power densities. IV. STARTUP AND POWER-MANAGEMENT SCALING The purpose of the power-management circuit is to act as a buffer between the rectenna power source and the energy storage device (Fig. 1). As an ideal buffer in the harvesting application, the converter performs three functions, which are: 1) creates optimal impedance match to maximize the rectenna efficiency,, over the full range of incident power densities; 2) transfers the harvested energy with minimal loss to the energy storage element over the full range of rectenna output voltages; and 3) monitors the energy storage and provide charge control and protection as appropriate for the energy storage used (battery or capacitor). Since the efficiency of the rectenna depends on the matching behavior of the converter, and the efficiency of the converter depends on the operating conditions of the rectenna and the energy storage device, it is best to codesign these blocks for the given application and expected conditions. A. Yagi-Uda Array Startup Circuit For low incident power densities, it is conceivable that the storage element can get depleted and the microcontroller and power-management circuit, as well as the intended application, cannot continue to function. The startup problem is addressed in the case of the Yagi-Uda array and can be applied to all three arrays. We assume that we can illuminate the rectenna with a short pulse of compliant increased power level. If the energy storage element is depleted, some method is required to initially start up the system. Doing so requires a minimum amount of dc power from the rectenna, as well as: 1) boosting the typically very low output voltage from the rectenna to a value that can be consumed by readily available electronics; 2) applying a sharp reset of the resident microprocessor to mitigate the effects of slow supply slew rates; and 3) minimizing overall power consumption to conserve as much for the intended application as possible. Fig. 13(a) indicates a simplified startup circuit that can be used to prime a larger system. Once the system starts, this startup circuit may be partially or completely disabled. With the rectenna illuminated by an RF wave, some dc current will flow into the startup circuit. At low current levels the exponential characteristics of the diode,, will allow the capacitor,, Fig. 13. (a) Combined boost converter and charge pump startup circuit. Low-voltage boot oscillator drives transistor. Boosted output rectified by andstoredon continues powering the oscillator. Charge pump roughly doubles the output voltage to power the application. (b) Rectenna open-circuit voltage prior to connection to startup circuit; Boosted output level from startup circuit after charge-up time. is never high enough to directly power a programmable processor necessitating a boost stage. to charge up to nearly that of the rectenna output voltage. As it continues to charge, the low-voltage boot oscillator begins to drive the boost transistor,, and the rectenna voltage is boosted through the inductor,. As the boot voltage,, increases the operation of the boot oscillator is reinforced and the system self-starts. The lower limit of operation of this startup circuit is determined by the lowest operating voltage of the boot oscillator and boost transistor,. This particular boot oscillator begins oscillating at 800 mv. Other oscillators are available that operate to lower voltages [22]. Due to the small currents involved, this boost converter will typically operate in discontinuous current mode (DCM), and its input impedance is proportional to both the inductance and frequency of the boot oscillator [23]. It is best to co-design this startup impedance to match the rectenna output under expected operating conditions to maximize startup capability. When operating at low power densities, it may be impossible to reach the desired operating voltage,, with just a single boost stage. An additional charge pump constructed from a capacitor and two diodes [see Fig. 13(a)] could nearly double the output voltage of the boost converter. Connecting the output of the dual-frequency Yagi of Fig. 4 to the startup circuit of Fig. 13(a) and tuning the startup circuit s impedance to 2k mh khz yields the performance shown in Fig. 13(b). A range of power densities illuminates the rectenna and its open circuit voltage is measured. Upon connection of the startup circuit, the output

8 POPOVIĆ et al.: SCALABLE RF ENERGY HARVESTING 1053 Fig. 15. Rectenna array power-management system. A low-power Texas Instruments MSP430 microcontroller (MCU) generates power stage control signals and runs optimization and MPPT algorithms. The rectenna voltage is sensed to calculate input power and the output is used for battery charge control. Fig. 14. (a) Measured I V characteristics of patch rectenna for several incident RF power densities compared to measured rectenna emulator I V curve based on dc model. The inset shows the emulator circuit. (b) Photograph of the array emulator circuit with multiple emulator circuits. is allowed to stabilize and is then measured. Even at the lowest activation power density tested, 2.5 W/cm, the startup circuit provides ample bias to operate a microcontroller. The variation of power density, as well as power variations across a larger array such as the 20-element patch array from Fig. 3(b), could cause dramatic differences in the dc rectenna model required for power-management circuit co-design. A four-quadrant model is developed in [23] for the 1.96-GHz patch rectenna, and a circuit constructed to emulate the 20-element array with various dc connections. The simple model assumes a narrowband antenna equivalent source, and ideal filtering, along with the SPICE diode model. The emulator circuit allows design and testing of the power-management circuit with controlled input power, and is shown in Fig. 14 along with the agreement with measured rectenna curves. An adjustable-gain amplifier is used to implement the ac source of the model. The isolation transformers ideally do not significantly change the I V characteristics of the cells, but allow the cell output to float so that only one ground-referenced power supply is needed for the entire array and the individual cells can be connected together in series and parallel combinations. The component values of the model can be selected over a wide range since they are only meant to represent the low-frequency behavior of the rectenna. The rectenna array power-management system architecture is shown in Fig. 15. A low-power Texas Instruments MSP430 microcontroller generates power stage control signals and runs optimization and maximum power point tracking (MPPT) algorithms, which have been demonstrated in a number of previous publications, e.g., [24] [27]. The rectenna voltage is sensed to calculate input power and the output used to charge a 3-V lithium battery. The power-management circuit is a synchronous boost power stage, a comparator and analog switch for high-side gate drive, and a few low-power discrete logic gates. The control logic is designed to be powered directly from the battery so that no additional power conversion is required. The circuit is designed using low-cost readily available commercial hardware in order to achieve a practical solution for small-scale commercial or scientific applications. At low input power levels, the converter operates in asynchronous mode (using a Schottky diode as the high-side switch) to prevent the losses caused by the high-side gate drive. At higher input power levels, the MCU activates the comparator across the high-side Schottky diode. This enables the synchronous rectification operating mode, where the high-side field-effect transistor (FET) clamps the diode during the conduction interval to reduce conduction losses. Traditional synchronous designs usually rely on a high-frequency high-power clock in the range of several megahertz to generate a short precise dead-time interval between the turn-off and turn-on transitions of the two FETs to prevent shoot-through current. The use of a comparator to generate the high-side gate drive, however, effectively enables soft switching of the FET and only a clock on the order of the switching frequency needs to be generated for the low-side FET. The power-management approach is compared to direct connection between a rectenna array and a battery. A 24-h segment of cell tower data is used to update the rectenna emulator at 1-min intervals [28] using the setup in Fig. 15. The output voltage and current are recorded every few seconds. Four test cases are considered, simulating different average power levels and array configurations: a series connection of rectenna elements resulting in a 20-k Thevenin equivalent resistance, and a series/parallel connection resulting in 5 k, each at two power levels (125- and 250- W average rectified power). The measured output power for each test case is integrated and normalized to theoretical maximum available at the maximum power point (MPP). Fig. 16 shows a plot of the fraction of

9 1054 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 4, APRIL 2014 Fig. 16. Measured fraction of energy harvested over a 24-h period with simulated cell-phone base-station representative power density data and using the rectenna array emulator for two dc configuration cases. A comparison with a direct connection (no power-management circuit) shows good scaling of efficiency with a scaled power-management approach. TABLE IV SUMMARY OF MEASURED HARVESTING PERFORMANCE OF DUAL-FREQUENCY YAGI-UDA RECTENNA energy harvested for the higher power level over the 24-h simulated period and in comparison with a direct connection with no power-management circuit. It is seen that the converter closely tracks theoretical maximum power and that the rectenna-battery mismatch is significant at lower power. Additionally, the converter harvesting efficiency scales well with power. The fraction of harvested energy is defined as Fig. 17. Comparison of total recitified power for independent and simultaneous dual-frequency illumination of the 64-element spiral array. (a) Scatter plot of paired measurements showing power of combined independent illumination,, versus simultaneous illumination, -. (b) Power increase of two-tone over individual tone illumination - ). The two axes are frequencies of the two tones from (a) and the incident power density is kept constant. V. SPECTRAL RESPONSE AND FREQUENCY SCALING Although we consider the entire powering system with a lowpower FCC-compliant transmitter, multiple or modulated transmitters can also be taken into account as in [29] where the increase over single-tone waves for a chaotic (spread-spectrum) modulated signal is quantified. In [30], multi-band powering for sensors is discussed. For the Yagi-Uda array, in an uncontrolled environment 1 the rectified power was measured as shown in Table IV, where the expected values are obtained by nonlinear simulations. The increase in rectified power at the higher frequency is attributed to additional sources that were present in the area in this ISM band, which increase the overall efficiency when waves at both design frequencies are incident on the array, with possible additional sources at different frequencies present in the environment. For true multi-tone characterization, with a range of tone spacings and amplitudes, the broadband CP spiral array is con- sidered between 2 8 GHz [6]. The array is tested in a controlled multi-tone environment in order to quantify the dc response improvement when more waves of more than one frequency are incident simultaneously. Two signal sources, separated by 20 in incidence angle from the rectenna and 45 in polarization, are used as multiple inputs with variable frequency and power. In order to ensure uncorrelated input waves, paired frequencies and power levels are generated randomly with a uniform distribution. The experiment is run by first generating one randomized frequency between 2 8 GHz and one incident power level between approximately 0.1 W/cm 0.1 mw/cm. Each source is turned on separately to record the independent dc rectified powers. Both sources are then turned on simultaneously and the rectified power - recorded for this case and compared to the sum of the two independent rectified powers. This process is repeated for randomized input pairs. The result shown in Fig. 17(a) reveals that the

10 POPOVIĆ et al.: SCALABLE RF ENERGY HARVESTING 1055 independent sources combine nonlinearly when simultaneously incident to produce more rectified power in every instance. Also shown is the mean increase in power for three portions of the curve. The amount of power increase, - ), when both sources are on correlates roughly with the amount of rectified power when both sources are off. This correlation is much stronger than the frequency or incident power correlation. For this reason, Fig. 17(b) is plotted according to. It should also be noted that, in general, the power increase diminishes as the two frequencies approach one another. Furthermore, local power-increase minima occur for certain difference frequencies. This is due to a spiral input impedance, which efficiently radiates the difference frequency rather than further rectifying the energy. A map of simultaneous illumination power increase is shown in Fig. 17(b) for frequencies of 2 8 GHz with constant incident power density. VI. CONCLUSIONS In this paper, several rectenna arrays have been presented for low power density power reception. The scaling in antenna gain, array size and received power, dc load, frequency of operation, and integrated power-management circuits have been discussed. Such scalable harvesting devices are envisioned to be useful for various radio-wave rich environments, although previous work has shown that available power levels in the microwave range are very low, e.g., [10] and [11]. Therefore, it has been recommended to include an efficient low-power compliant transmitter to ensure sufficient power for operating the microcontroller of the power-management circuit, and enable useful electronic functions such as wireless sensors [31], [32]. The startup circuit described in Section IV-A is enabled by ISM band regulations, which allow short-period transmission of modulated higher power pulses (FCC CFR47, Part ). In some applications, both near- and far-field powering could be applicable, as discussed in [33] and [34]. Different applications might call for different rectifier and antenna topologies. Although an advantage of the patch antennas such as those in the array in Fig. 3(b) is that the circuitry can be placed behind the antenna ground and mounted on any object, antennas with no ground plane, such as the Yagi-Uda and spiral, can be used when less directional coverage is required. Antennas can be fabricated on a variety of substrates, including flexible substrates with ink-jet printing [35]. Other circuit topologies, such as full-wave rectifiers or charge pumps, can be employed for increased efficiency at higher input power levels. The hardware results shown here and in [23] [25] are based on low-cost off-the-shelf components and demonstrated record efficiencies for power levels down to approximately 100 W. Below this level, the quiescent losses of available control hardware become the limiting factor. With the significant advancements in ultra-low-power wireless sensors, a range of custom solutions have been developed for power conversion at these low power levels, e.g., [36] and [37], and can be integrated with the scalable approaches demonstrated in this paper. REFERENCES [1] G. A. Covic and J. T. Boys, Inductive power transfer, Proc. IEEE, vol. 101, no. 6, pp , Jun [2] W. Brown, The history of power transmission by radio waves, IEEE Trans. Microw. Theory Techn., vol. MTT-32, no. 9, pp , Sep [3] N. Shinohara and H. Matsumoto, Experimental study of large rectenna array for microwave energy transmission, IEEE Trans. Microw. Theory Techn., vol. 46, no. 3, pp , Mar [4] L. Epp, A. Khan, H. Smith, and R. Smith, A compact dual-polarized 8.51-GHz rectenna for high-voltage (50 V) actuator applications, IEEE Trans. Microw. Theory Techn., vol. 48, no. 1, pp , Jan [5] B.StrassnerandK.Chang, Microwave power transmission: Historical milestones and system components, Proc. IEEE, vol. 101, no. 6, pp , Jun [6] J. A. Hagerty, F. Helmbrecht, W.McCalpin,R.Zane,andZ.Popović, Recycling ambient microwave energy with broadband antenna arrays, IEEE Trans. Microw. Theory Techn., vol. 52, no. 3, pp , Mar [7] E. Falkenstein, M. Roberg, and Z. Popović, Low-power wireless power delivery, IEEE Trans. Microw. Theory Techn., vol.60,no.7, pp , Jul [8] G. Vera, A. Georgiadis, A. Collado, and S. Via, Design of a 2.45 GHz rectenna for electromagnetic (EM) energy scavenging, in IEEE Radio Wireless Symp., 2010, pp [9] Z. Popović, E. Falkenstein, D. Constinett, and R. Zane, Low-power far-field wireless powering for wireless sensors, Proc. IEEE (Special Issue), vol. 101, no. 6, pp , Jun [10] R. Vias, H. Nishimoto, M. Tentzeris, Y. Kawahara, and T. Asami, A battery-less, energy harvesting device for long range scavenging of wireless power from terrestrial tv broadcasts, in IEEEMTT-SInt.Microw. Symp. Dig., Montreal, QC, Canada, Jun [11] M. Piñuela, P. D. Mitcheson, and S. Lucyszyn, Ambient RF energy harvesting in urban and semi-urban environments, IEEE Trans. Microw. Theory Techn., vol. 61, no. 6, pp , Jun [12] T. Le, K. Mayaram, and T. Fiez, Efficient far-field radio frequency energy harvesting for passively powered sensor networks, IEEE J. Solid-State Circuits, vol. 43, no. 5, pp , May [13] T. Haumann, U. Munzenberg, W. Maes, and P. Sierck, HF-radiation levels of GSM cellular phone towers in residential areas, in Proc. 2nd Int. Biol. Effects of EMFS Workshop, Oct. 2002, pp [14] D. Willkomm, S. Machiraju, J. Bolot, and A. Wolisz, Primary users in cellular networks: A large-scale measurement study, in Proc. 3rd IEEE New Frontiers in Dynam. Spectr. Access Networks Symp.,,Oct. 2008, pp [15] Z. Harouni, L. Cirio, L. Osman, A. Gharsallah, and O. Picon, A dual circularly polarized 2.45-GHz rectenna for wireless power transmission, IEEE Antennas Wireless Propag. Lett., vol. 10, pp , [16] D.-G. Youn, K.-H. Kim, Y.-C. Rhee, S.-T. Kim, and C.-C. Shin, Experimental development of 2.45 GHz rectenna using fss and dual-polarization, in 30th Eur. Microw. Conf., Oct. 2000, pp [17] A. Dolgov, Power management system for online low power RF energy harvesting optimization, Masters Thesis, Dept. Elect. Comput. Eng., Univ. Colorado at Boulder, Boulder, CO, USA, [18] J. O. M. McSpadden et al., Design and experiments of a high conversion efficiency 5.8-GHz rectenna, in IEEE MTT-S Int. Microw. Symp. Dig., 1998, pp [19] G. Marrocco, The art of UHF RFID antenna design: impedancematching and size-reduction techniques, IEEE Antennas Propag. Mag., vol. 50, no. 1, pp , Feb [20] R. J. Gutmann and J. M. Borrego, Power combining in an array of microwave power rectifiers, IEEE Trans. Microw. Theory Techn., vol. MTT-27, no. 12, pp , Dec [21] N. Shinohara and H. Matsumoto, Dependence of DC output of a rectenna array on the method of interconnection of its array elements, Elect. Eng. Jpn., vol. 125, no. 1, pp. 9 17, [22] J. Bohan, Low voltage driven oscillator circuit, U.S. Patent A, Mar. 29, [23] A. Dolgov, R. Zane, and Z. Popović, Power management system for online low power RF energy harvesting optimization, IEEE Trans. Circuits Syst., vol. 57, no. 7, pp , Jul [24] T. Paing, A. Dolgov, J. Shin, J. Morroni, J. Brannan, R. Zane, and Z. Popović, Wirelessly powered wireless sensor platform, in Eur. Microw. Conf. Dig., Munich, Germany, Oct. 2007, pp

11 1056 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 62, NO. 4, APRIL 2014 [25] T. Paing, J. Shin, R. Zane, and Z. Popović, Resistor emulation approach to low-power RF energy harvesting, IEEE Trans. Power Electron., vol. 23, no. 3, pp , May [26] D. Costinett, E. Falkenstein, R. Zane, and Z. Popović, RF-powered variable duty cycle wireless sensor, in Eur. Microw. Conf., 2010, pp [27] E. Falkenstein, D. Costinett, R. Zane, and Z. Popović, Far-field RF-powered variable duty cycle wireless sensor platform, IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 58, no. 12, pp , Dec [28] D. Willkomm, S. Machiraju, J. Bolot, and A. Wolisz, Primary users in cellular networks: A large-scale measurement study, in Proc. 3rd IEEE New Frontiers Dynam. Spectr. Access Networks Symp., Oct , 2008, pp [29] A. Collado et al., Improving wireless power transmission efficiency using chaotic waveforms, in IEEE MTT-S Int. Microw. Symp. Dig., Montreal, QC, Canada, Jun [30] A. Costanzo et al., RF/baseband co-design of switching receivers for multiband microwave energy harvesting, Sens. Actuators A, Phys., vol. 179, pp , [31] J.Bernhard,K.Hietpas,E.George, D. Kuchima, and H. Reis, An interdisciplinary effort to develop a wireless embedded sensor system to monitor and assess corrosion in the tendons of prestressed concrete girders, in IEEE Top. Wireless Commun. Conf., 2003, pp [32] T.H.Teo,G.K.Lim,D.Sutomo,K.H.Tan,P.K.Gopalakrishnan,and R. Singh, Ultra low-power sensor node for wireless health monitoring system, in Proc. IEEE Int. Circuits Syst. Symp., New Orleans, LA, USA, May 2007, pp [33] V. Rizzoli, D. Masotti, N. Arbizzani, and A. Costanzo, Cad procedure for predicting the energy received by wireless scavenging systems in the near and far-field regions, in IEEEMTT-SInt.Microw.Symp.Dig., May 2010, pp [34] J. Garnica, R. A. Chinga, and J. Lin, Wireless power transmission:fromfarfield to near field, Proc. IEEE, vol. 101, no. 6, pp , Jun [35] G. Orecchini, L. Yang, M. M. Tentzeris, and L. Roselli, Wearable battery-free active paper-printed RFID tag with human energy scavenger, in IEEEMTT-SInt.Microw.Symp.Dig., Baltimore, MD, USA, Jun [36] H.-K. Chiou and I.-S. Chen, High-efficiency dual-band on-chip rectenna for 35- and 94-GHz wireless power transmission in m CMOS technology, IEEE Trans. Microw. Theory Techn., vol. 58, no. 12, pp , Dec [37] T. Paing, J. Shin, R. Zane, and Z. Popović, Custom IC for ultralow power rf energy scavenging, IEEE Trans. Power Electron. Lett., vol. 26, no. 6, pp , Jun Zoya Popović (S 86 M 90 SM 99 F 02) received the Dipl.Ing. degree from the University of Belgrade, Belgrade, Serbia, in 1985, and the Ph.D. degree from the California Institute of Technology, Pasadena, CA, USA, in Since 1990, she has been with the University of Colorado at Boulder, where she is currently a Distinguished Professor and holds the Hudson Moore Jr. Chair in Electrical, Computer and Energy Engineering. In 2001, she was a Visiting Professor with the Technical University of Munich, Munich, Germany. Since 1991, she has graduated 50 Ph.D. students. Her research interests include high-efficiency, low-noise, and broadband microwave and millimeterwave circuits, active antennas, applications of microwaves in medicine, and wireless powering for batteryless sensors. Prof. Popović was elected a Foreign Member of the Serbian Academy of Sciences and Arts in She was the recipient of the 1993 and 2006 Microwave Prizes presented by the IEEE Microwave Theory and Techniques Society (IEEE MTT-S) for best journal papers. She was the recipient of the 1996 URSI Issac Koga Gold Medal and was named an NSF White House Presidential Faculty Fellow in In 2000, she was the recipient of the German Humboldt Research Award for Senior U.S. Scientists. She was alos the recipient of the IEEE MTT-S Distinguished Educator Award in Sean Korhummel (S 05) received the Bachelors Degree of Science in electrical engineering from the University of California at Santa Cruz, Santa Cruz, CA,USA,in2008,andiscurrentlyworkingtoward the Ph.D. degree in the area of cavity resonators and rectifiers for power harvesting applications at the University of Colorado at Boulder, Boulder, CO, USA. Mr. Korhummel was the recipient of the 2013 IEEE Microwave Theory and Techniques Society (IEEE MTT-S) International Microwave Symposium (IMS) Wireless Power Harvesting Student Design Competition. Steven Dunbar (S 12) received the B.S. degree in electrical and computer engineering from the University of Colorado at Boulder, Boulder, CO, USA, in 1992, the M.S.E.E. degree from the University of Texas,Arlington,TX,USA,in1996,andiscurrently working toward the Ph.D. degree at the University of Colorado at Boulder, Boulder, CO, USA. He was with Motorola, the Inovonics Wireless Corpoation, and RF Micro Devices. He is currently with Texas Instruments Incorporated, as a Field Applications Engineer. Mr. Dunbar is a licensed Professional Engineer (PE) in the State of Colorado. Robert Scheeler (S 12) received the B.S. degree in electrical engineering from North Dakota State University, Fargo, ND, USA, in 2008, and the M.S. and Ph.D. degrees in electrical engineering from the University of Colorado at Boulder, Boulder, CO, USA, in 2011 and 2013, respectively. His research interests include low-noise receivers, on-chip noise calibration standards, monolithic microwave integrated circuit design, wireless energy harvesting, and noninvasive core body temperature measurement using microwave radiometers. Regan Zane (SM 07) received the Ph.D. degree in electrical engineering from the University of Colorado at Boulder, Boulder, CO, USA, in He is currently a USTAR Professor with the Department of Electrical and Computer Engineering, Utah State University, Logan, UT, USA. From 2001 to 2012, he was a faculty member with the University of Colorado at Boulder. From 1999 to 2001, he was a Research Engineer with the GE Global Research Center, Niskayuna, NY, USA. His research concerns bidirectional converters for dc and ac micro-grids, high step-down power converters for dc distribution systems such as high-efficiency data centers, advanced battery management systems (BMSs) for electric vehicles, LED drivers for lighting systems, and low-power energy harvesting for wireless sensors. Dr. Zane is currently an Associate Editor for the IEEE TRANSACTIONS ON POWER ELECTRONICS and IEEE POWER ELECTRONICS LETTERS. He is the publicity chair for the IEEE Power Electronics Society. He was the recipient of the National Science Foundation (NSF) CAREER Award in 2004, the 2005 IEEE Microwave Prize, the 2007 and 2009 IEEE Power Electronics Society Transactions Prize Letter Awards, and the 2008 IEEE Power Electronics Society Richard M. Bass Outstanding Young Power Electronics Engineer Award. He was also the recipient of the 2006 Inventor of the Year Award, the 2006 Provost Faculty Achievement Award, the 2008 John and Mercedes Peebles Innovation in Teaching Award, and the 2011 Holland Teaching Award of the University of Colorado at Boulder. Arseny Dolgov, photograph and biography not available at time of publication. Erez Falkenstein, photograph and biography not available at time of publication. Joseph Hagerty, photograph and biography not available at time of publication.