Rectiﬁers’ Design and Optimization for a Dual-Channel RF Energy Harvester

: This paper presents the design and implementation of two front-ends for RF (Radio Frequency) energy harvesting, comparing them with the commercial one—P2110 by Powercast Co. (Pittsburgh, PA, USA) Both devices are implemented on a discrete element board with microstrip lines combined with lumped elements and are optimized for two di ﬀ erent input power levels ( − 10 dBm and 10 dBm, respectively), at the GSM900 frequencies. The load has been ﬁxed at 5k Ω , after a load-pull analysis on systems. The rectiﬁers stages implement two di ﬀ erent Schottky diodes in two di ﬀ erent topologies: a single diode and a 2-stage Dickson’s charge pump. The second one is compared with the P2110 by generating RF ﬁelds at 915 MHz with the Powercast Powerspot. The main aim of this work is to design simple and e ﬃ cient low-cost devices, which can be used as a power supply for low-power autonomous sensors, with better performances than the current solutions of state-of-the-art equipment, providing an acceptable voltage level on the load. Measurements have been conducted for input power range − 20 dBm up to 10 dBm; the best power conversion e ﬃ ciency (PCE) is obtained with the second design, which reaches a value of 70% at 915 MHz. In particular, the proposed device exhibited better performance compared to the P2110 commercial device, allowing a maximum distance of operation of up to 22 meters from the dedicated RF power source, making it suitable even for IoT (Internet of Things) applications.


Introduction
In recent years, the great use of low-power autonomous systems and sensors [1][2][3][4][5][6] increased the need for self-sustainable devices, which are capable of harvesting and using energy from the environment, particularly for those that need a continuous power supply (such as human health monitoring systems) [7][8][9][10][11][12][13]. These devices, such as low-voltage front-ends for photomultiplier [4] or monitoring systems for buildings with a low power consumption [6], can use energy scavenged from the environment, which is typically poor but enough to ensure the system functionality. Since this energy is available in many different forms (thermal, vibrational, . . . ), some works focused on multi-source energy harvesters, combining techniques to compensate for this lack of retrievable energy [11]. In this perspective, the development of wireless communication with smartphones and RF transmitters provided a steady availability of electromagnetic waves, which corresponds to free RF energy [13]. Energy harvesting's aim is to exploit the generated field as a source of power, collecting waves in the environment and converting them in a DC electrical signal. Good efficiency levels are reached with solutions proposed in other papers: some of them use a nonvolatile memory structure [12], while others propose IC structures with CMOS technology [7]. Moreover, a double-band design has been proposed, which uses a diplexer to harvest RF power at two frequencies [5]. Despite their efficiency level (that usually reaches 50%), these have a complex design and are not simple to realize.
In this work, we focused on the possibility to design a simple circuit configuration with easily available components in order to provide a simple energy harvesting system with a good power conversion efficiency. The most commonly available frequencies of RF signals cover a spectrum range from about 400 MHz to 2.5 GHz [10], which includes mobile phones and Wi-Fi devices' bands. Despite this, the amount of harvested energy is strongly influenced by the surroundings, such as the distance from RF sources. Several designs capable of converting RF into DC energy have already been proposed, usually working from −30 to 30 dBm, with a peak efficiency of 80% [14,15]. As a drawback, a standard urban environment has an available power maximum peak of about −30 dBm, distributed across multiple frequency spectra. In this perspective, some commercial devices able to work with their own generated RF source have been designed, such as the Powerspot, a 3 W RF transmitter. This commercial device is sold for working with Powercast harvesters [16].
A generic RF energy harvesting (EH) system is composed of four main blocks (see Figure 1): the antenna, which captures the electromagnetic waves and generates an electric RF signal; the matching network, needed to transfer the maximum amount of input power from the antenna to the rectifier, also avoiding energy reflection; the rectifier, which converts the RF signal to a DC voltage on the load; and, finally, the energy storage section, for storing harvested energy and for filtering spikes (this can simply consist in a shunt capacitor) [17]. In this work, the aim is to propose an effective, optimized, and repeatable design method, as a result of previously conducted works [18][19][20], in order to improve the performance of RF EH systems with standard architecture, as shown in Figure 1. To prove the validity of the proposed technique, a prototype board has been implemented and a performance comparison with a commercial RF energy harvester, the PB2110 from Powercast Co. [21], has been conducted and reported.
All harvesters used the same antenna, sold with the evaluation board of the Powercast P2110 [22], to ensure a comparison between commercial and designed harvesters. Matching networks are made by mixed lumped and distributed elements, in order to have more degrees of freedom in designing them. Since a good network has not to be dissipative, resistive elements are not recommended for this kind of system. Since the rectifier section needs diodes with a low junction voltage and fast switching times, Schottky diodes are the best choice for it, because of their low junction capacitance and their small amount of required power, to minimize losses and maximize available power [23][24][25][26]. Using a more complex topology, such as a full-bridge rectifier or a voltage multiplier instead of a single diode, brings a greater voltage value on the load; however, the greater the number of diodes used, the greater the power absorption of the device, significantly reducing its In this work, the aim is to propose an effective, optimized, and repeatable design method, as a result of previously conducted works [18][19][20], in order to improve the performance of RF EH systems with standard architecture, as shown in Figure 1. To prove the validity of the proposed technique, a prototype board has been implemented and a performance comparison with a commercial RF energy harvester, the PB2110 from Powercast Co. [21], has been conducted and reported.
All harvesters used the same antenna, sold with the evaluation board of the Powercast P2110 [22], to ensure a comparison between commercial and designed harvesters. Matching networks are made by mixed lumped and distributed elements, in order to have more degrees of freedom in designing them. Since a good network has not to be dissipative, resistive elements are not recommended for this kind of system. Since the rectifier section needs diodes with a low junction voltage and fast switching times, Schottky diodes are the best choice for it, because of their low junction capacitance and their small amount of required power, to minimize losses and maximize available power [23][24][25][26]. Using a more complex topology, such as a full-bridge rectifier or a voltage multiplier instead of a single diode, brings a greater voltage value on the load; however, the greater the number of diodes used, the greater the power absorption of the device, significantly reducing its overall efficiency. Moreover, diodes are non-linear devices that work as a variable impedance that changes with input power itself, making it more difficult to design an efficient matching network.
Concerning the harvester efficiency, several definitions, depending on the considered section where it is calculated, are reported in the literature. In this study, the definition of the harvester power conversion efficiency (PCE) is as follows [27]: where V L and I L are the output voltage and current measured on the load, respectively, while P in is the available input power that the antenna delivers towards the matching network and, therefore, to the harvesting circuitry. This work is organized as follows: in Section 2, the design method and optimization strategy of the designed two harvesters are presented and discussed; in Section 3, the measured results are reported; and, finally, Section 4 shows the conclusions.

Design Method and Optimization
Two devices have been designed for different input power levels: an HPD (high-power device), optimized for 10 dBm, in order to make a comparison with the commercial device P2110, and an LPD (low-power device) optimized for a −10 dBm level to work with a lower power availability, which is below the Powercast input power range of operation. The LPD rectifier topology is based on a single SMS7630 diode by Skyworks Solution Inc. [28], which has a junction voltage of 0.34 V. The HPD implements a 2-stage voltage multiplier [29], using two HSMS-2852 by Avago Technologies [30], a couple of series diodes in a single package. Both circuits are implemented on a TLX8 substrate by Taconic [31] with microstrip transmission lines. The lumped elements used for matching networks are from the GJM1555 series capacitors and LQW15AN inductors by Murata (except for the LPD shunt inductor, which is a 0805CS by Coilcraft). For both the circuits, simulated and measured load-pull analyses have been conducted to find the best value of impedance that maximizes efficiency; then, a study has been conducted for different input power levels with fixed loads. From this investigation, it comes out that the best efficiency is obtained for a load of about 5 kΩ, as reported in Figure 2. The high-level schematics of both designed systems are shown in Figure 3.
J. Low Power Electron. Appl. 2020, 10, 11 3 of 11 overall efficiency. Moreover, diodes are non-linear devices that work as a variable impedance that changes with input power itself, making it more difficult to design an efficient matching network. Concerning the harvester efficiency, several definitions, depending on the considered section where it is calculated, are reported in the literature. In this study, the definition of the harvester power conversion efficiency (PCE) is as follows [27]: % ⋅ 100 (1) where and are the output voltage and current measured on the load, respectively, while is the available input power that the antenna delivers towards the matching network and, therefore, to the harvesting circuitry.
This work is organized as follows: in Section 2, the design method and optimization strategy of the designed two harvesters are presented and discussed; in Section 3, the measured results are reported; and, finally, Section 4 shows the conclusions.

Design Method and Optimization
Two devices have been designed for different input power levels: an HPD (high-power device), optimized for 10 dBm, in order to make a comparison with the commercial device P2110, and an LPD (low-power device) optimized for a −10 dBm level to work with a lower power availability, which is below the Powercast input power range of operation. The LPD rectifier topology is based on a single SMS7630 diode by Skyworks Solution Inc. [28], which has a junction voltage of 0.34 V. The HPD implements a 2-stage voltage multiplier [29], using two HSMS-2852 by Avago Technologies [30], a couple of series diodes in a single package. Both circuits are implemented on a TLX8 substrate by Taconic [31] with microstrip transmission lines. The lumped elements used for matching networks are from the GJM1555 series capacitors and LQW15AN inductors by Murata (except for the LPD shunt inductor, which is a 0805CS by Coilcraft). For both the circuits, simulated and measured loadpull analyses have been conducted to find the best value of impedance that maximizes efficiency; then, a study has been conducted for different input power levels with fixed loads. From this investigation, it comes out that the best efficiency is obtained for a load of about 5 kΩ, as reported in Figure 2. The high-level schematics of both designed systems are shown in Figure 3.   Design and simulations have been conducted within the AWR Environment Design software by Cadence Design Systems, Inc. (San Jose, CA, USA) The diodes were simulated using their non-linear SPICE-model parameters. As a first design, ideal lines and lumped elements were used. The LPD matching network is made by 3 lumped elements (shunt inductor, series inductor, and shunt capacitor), while HPD's one is a 5 elements network (series inductor and two inductor-capacitor Ltype networks). Since diodes are non-linear devices, the impedance of the rectifier is affected by the input power level [32,33]; thus, the rectifiers have been matched to the antenna for the selected input power levels Pin (−10 dBm for the LPD and 10 dBm for the HPD). Then, the values of both lumped elements and microstrips have been optimized using the Smith chart, as shown for HPD in Figure 4, for different values of Pin; to get a good matching, the impedance of the rectifier circuit must be equal to the antenna impedance at the working frequency and the considered power level. . HPD input impedance (real and imaginary part) plotted on a Smith chart for input power levels from −25 to +20 dBm at 1 dBm steps (impedances are normalized to 50 Ω, which is the same as the input antenna) at a frequency of 915 MHz; it has been matched for an input power level of 10 dBm in order to get maximum efficiency for that value. Design and simulations have been conducted within the AWR Environment Design software by Cadence Design Systems, Inc. (San Jose, CA, USA) The diodes were simulated using their non-linear SPICE-model parameters. As a first design, ideal lines and lumped elements were used. The LPD matching network is made by 3 lumped elements (shunt inductor, series inductor, and shunt capacitor), while HPD's one is a 5 elements network (series inductor and two inductor-capacitor L-type networks). Since diodes are non-linear devices, the impedance of the rectifier is affected by the input power level [32,33]; thus, the rectifiers have been matched to the antenna for the selected input power levels Pin (−10 dBm for the LPD and 10 dBm for the HPD). Then, the values of both lumped elements and microstrips have been optimized using the Smith chart, as shown for HPD in Figure 4, for different values of Pin; to get a good matching, the impedance of the rectifier circuit must be equal to the antenna impedance at the working frequency and the considered power level. Design and simulations have been conducted within the AWR Environment Design software by Cadence Design Systems, Inc. (San Jose, CA, USA) The diodes were simulated using their non-linear SPICE-model parameters. As a first design, ideal lines and lumped elements were used. The LPD matching network is made by 3 lumped elements (shunt inductor, series inductor, and shunt capacitor), while HPD's one is a 5 elements network (series inductor and two inductor-capacitor Ltype networks). Since diodes are non-linear devices, the impedance of the rectifier is affected by the input power level [32,33]; thus, the rectifiers have been matched to the antenna for the selected input power levels Pin (−10 dBm for the LPD and 10 dBm for the HPD). Then, the values of both lumped elements and microstrips have been optimized using the Smith chart, as shown for HPD in Figure 4, for different values of Pin; to get a good matching, the impedance of the rectifier circuit must be equal to the antenna impedance at the working frequency and the considered power level. . HPD input impedance (real and imaginary part) plotted on a Smith chart for input power levels from −25 to +20 dBm at 1 dBm steps (impedances are normalized to 50 Ω, which is the same as the input antenna) at a frequency of 915 MHz; it has been matched for an input power level of 10 dBm in order to get maximum efficiency for that value. . HPD input impedance (real and imaginary part) plotted on a Smith chart for input power levels from −25 to +20 dBm at 1 dBm steps (impedances are normalized to 50 Ω, which is the same as the input antenna) at a frequency of 915 MHz; it has been matched for an input power level of 10 dBm in order to get maximum efficiency for that value. Tables 1 and 2 report the used components in both definitive layouts, while in Figure 5a,b the implemented prototype boards for the LPD and HPD EH circuits are depicted, respectively; simulations have been conducted using the available 2-port S-parameters of commercial components. L5 was used to represents the parasitic inductive effect of the input port microstrip. Substituting ideal lines with microstrips introduced an inductive effect that has been compensated by removing the lumped element, in order to maintain a good impedance matching. Tables 1 and 2 report the used components in both definitive layouts, while in Figure 5a,b the implemented prototype boards for the LPD and HPD EH circuits are depicted, respectively; simulations have been conducted using the available 2-port S-parameters of commercial components. L5 was used to represents the parasitic inductive effect of the input port microstrip. Substituting ideal lines with microstrips introduced an inductive effect that has been compensated by removing the lumped element, in order to maintain a good impedance matching.

Simulations and Experimental Results
As mentioned, both circuits have been preliminarily simulated on AWR. The simulated plot of both circuit efficiency is shown in Figure 6. The best PCE is reached for input power levels near the chosen one for the optimization, exceeding 65%.
As shown, the discrepancy between ideal and commercial components is more evident in LPD than in HPD. This is because of the different diode model used, which is more precise for the second one. Moreover, due to the lower power levels, LPD is much more sensitive to the parasitic effects of the lumped elements.
After simulation sessions, the two designed circuits have been physically realized and tested. The transferred power on the load has been measured and compared with the P2110, by connecting them to the same RF power generator. The chosen reference load is always 5 kΩ for the three systems. Since Powercast implements a DC/DC voltage regulator, it has been excluded in order to perform a fair comparison between the commercial harvester and the implemented prototypes by testing only the RF to DC section as it represents the core of the RF energy harvesting circuitry. In Figure 7, the functional block diagram of P2110 is shown [21]: the test reference load has been connected before the DC/DC boost converter, at V CAP pin, leaving the D SET pin unconnected, so that V CAP is connected to D OUT , which is the harvester analog output that provides a voltage proportional to the harvested power. In order to guarantee the boost converter not to influence the RF harvester characterization process, the RESET pin has been driven at a high logic level by means of an external voltage generator in order to disable V OUT and the voltage monitor.

Simulations and Experimental Results
As mentioned, both circuits have been preliminarily simulated on AWR. The simulated plot of both circuit efficiency is shown in Figure 6. The best PCE is reached for input power levels near the chosen one for the optimization, exceeding 65%.  As shown, the discrepancy between ideal and commercial components is more evident in LPD than in HPD. This is because of the different diode model used, which is more precise for the second one. Moreover, due to the lower power levels, LPD is much more sensitive to the parasitic effects of the lumped elements.
After simulation sessions, the two designed circuits have been physically realized and tested. The transferred power on the load has been measured and compared with the P2110, by connecting them to the same RF power generator. The chosen reference load is always 5 kΩ for the three systems. functional block diagram of P2110 is shown [21]: the test reference load has been connected before the DC/DC boost converter, at VCAP pin, leaving the DSET pin unconnected, so that VCAP is connected to DOUT, which is the harvester analog output that provides a voltage proportional to the harvested power. In order to guarantee the boost converter not to influence the RF harvester characterization process, the RESET pin has been driven at a high logic level by means of an external voltage generator in order to disable VOUT and the voltage monitor. For better accuracy, both output load DC voltage and DC current have been measured simultaneously and the output harvested power has been calculated as the product of them. As shown in Figure 8, the HPD circuit reaches its maximum efficiency for an input signal frequency of 915 MHz, exceeding 70% above 5 dBm input power, and has a good efficiency at 850 MHz too. LPD showed a small overall efficiency at 915 MHz remaining similar down to 850 MHz, which is the circuit's best matching frequency. On the other hand, the Powercast P2110 turned off for input power levels lower than −5 dBm, at each frequency. In Figure 8 shows the comparison between LPD, HPD, and P2110 efficiency. For better accuracy, both output load DC voltage and DC current have been measured simultaneously and the output harvested power has been calculated as the product of them. As shown in Figure 8, the HPD circuit reaches its maximum efficiency for an input signal frequency of 915 MHz, exceeding 70% above 5 dBm input power, and has a good efficiency at 850 MHz too. LPD showed a small overall efficiency at 915 MHz remaining similar down to 850 MHz, which is the circuit's best matching frequency. On the other hand, the Powercast P2110 turned off for input power levels lower than −5 dBm, at each frequency. In Figure 8 shows the comparison between LPD, HPD, and P2110 efficiency.
As a further comparison, P2110 and HPD have been measured in an outdoor, free-space environment too, connecting a Powercast patch antenna to them and then generating an electromagnetic source signal with the Powercast Powerspot transmitter, which radiates a 3W RF power at 915 MHz (see Figure 9), specifically provided for Powercast harvesting devices. The LPD has not been measured outdoors because, due to its lower working power levels, it is not suitable for this kind of application and it is not directly comparable with the other two devices, while an overall test bench comparison of the three harvesters is reported in Figure 8. The P2110 barely worked 20 m away from the source, while the HPD still worked at 22 m. Voltage and current have been measured as before, while input power has been estimated by Friis' transmission equation. Of course, a better analysis could be conducted in an anechoic chamber, avoiding field reflections and external sources to have greater accuracy. Table 3 summarizes the measurement results. The drop at a distance of 4 m is related to outdoor environment problems, such as signal interferences and multiple paths caused by the external source and the surrounding.  As a further comparison, P2110 and HPD have been measured in an outdoor, free-space environment too, connecting a Powercast patch antenna to them and then generating an electromagnetic source signal with the Powercast Powerspot transmitter, which radiates a 3W RF power at 915 MHz (see Figure 9), specifically provided for Powercast harvesting devices. The LPD has not been measured outdoors because, due to its lower working power levels, it is not suitable for this kind of application and it is not directly comparable with the other two devices, while an overall test bench comparison of the three harvesters is reported in Figure 8. The P2110 barely worked 20 meters away from the source, while the HPD still worked at 22 meters. Voltage and current have been measured as before, while input power has been estimated by Friis' transmission equation. Of course, a better analysis could be conducted in an anechoic chamber, avoiding field reflections and external sources to have greater accuracy. Table 3 summarizes the measurement results. The drop at a distance of 4 m is related to outdoor environment problems, such as signal interferences and multiple paths caused by the external source and the surrounding.

Conclusions
In this work, we have presented the design of two RF energy harvesting circuits, compared with the commercial one-P2110 by Powercast Co. Designed devices used commercial components and Schottky diode and were realized on a lossy substrate. All the used elements are low-cost components. Moreover, both the designed devices have a simple topology and are easily feasible; this is especially needed for IoT systems, where a great number of low-power sensors and devices are used, so a continuous power supply is a significant part of the application. Due to this, the designed system is suitable for a high-scale production with minimal costs and still maintaining good performance that is comparable (and superior, as is the case of HPD) with current commercial solutions. Due to the small power density commonly available in an urban environment, it is also helpful to design devices that can work with a lower power availability, maintaining an acceptable efficiency. In outdoor measurements, HPD managed to provide a good voltage level on its load without any DC voltage regulators, even at a distance of several meters. Good overall efficiency levels are reached for either LPD, HPD, or P2110, while the proposed harvesters are able to work a superior input signal power range with greater efficiency, becoming a very good candidate even for low-voltage, low-power RF energy harvesting suitable for IoT devices and other low-power applications.