Design of a Compact and Highly Efficient Energy Harvester System Suitable for Battery-Less Low Cost On-Board Unit Applications

Abstract

This study addresses the general problem regarding the power supply in specific on-board unit (OBUs) solutions. In detail, this paper refers to a subset of the so-called electronic toll collection (ETC) applications such as assets control and vehicle identification, where simplicity, low costs, and maximum compactness represent the most important features. In this context, the next generation of OBUs, developed specifically with reference to such applications, will involve energy harvester-based battery-less techniques. Previous studies have mainly concentrated on performance optimization by achieving maximum energy transmission to the OBUs. This study discusses a technique suitable for both maximizing performance and minimizing the dimensions of transponder energy harvesters suitable for assets control and vehicle identification operating at 5.8 GHz. The technique assumes that an optimal source impedance exists that maximizes the energy transfer to the transponder, thus enabling its power supply in a battery-less configuration. We discuss a solution based on a compact patch antenna designed to exhibit this optimal source impedance to the RF-to-DC rectifier. This approach avoids the use of a lossy matching network. For the sake of comparison, the same function is compared with an equivalent development, which includes the interstage matching network between the antenna and the RF-to-DC rectifier. We introduce experimental results demonstrating that the ultracompact energy harvester optimized at −5 dBm of impinging power is capable of increasing both the charge current and energy efficiency from 340 to 450 μA and from 37% to 47%, respectively.

1. Introduction

In recent years, many concepts developed in the field of intelligent transportation systems (ITS) have started to merge with the Internet of Things (IoT) to become part of this paradigm [1]. It is well known that, until now, limitations in the present communications technologies have reduced the full exploitation of the IoT vision in terms of mobile applications. This translated into the poor development of IoT-based approaches oriented toward the infomobility ecosystem. As a consequence, it is expected that full development of the 5G communications technology will enable wide implementation of the IoT paradigm in the mobile world, including ITS/IoT systems [2,3,4,5,6].

Nevertheless, regardless of the technology, three central classes of applications may be identified in such a system, i.e., road safety, traffic efficiency, and a wide-ranging category of other utility applications. Up to now, one of the most relevant applications in this third class has been related to toll road systems or electronic toll collection (ETC) [7,8]. Typically, these systems are mainly linked with electronic payment for highways or main roads, although the same systems or solutions may also be used for assets control and vehicle identification (access control or parking payment). The common approach followed in such a system is an architecture based on roadside units (RSUs) and on-board units (OBUs) [7,8,9].

It is a fact that in specific applications, such as assets control and vehicle identification, cost reduction, low complexity, and high reliability represent fundamental requirements for the whole system. These severe requirements make the replacement of the current approach based on RSUs and OBUs with 5G-based applications very challenging.

The technological evolution in such identification systems will be focused on OBU cost and complexity reduction. One of the most relevant contributions in this direction will be the development of battery-less solutions that translate into simplified architectures, reduced costs, and reduced logistic chains, thereby resulting in increased diffusion of the system. It is worth noting that the battery-less attribute is maximally exploited in conjunction with electronic components capable of operating at very low power supply [10]. As exemplified in

Table 1. Compatibility of on-board unit (OBU) applications with respect to battery-less solutions.

Application	OBU Performance	OBU Power Consumption	OBU Complexity and Cost	OBU Battery-Less Compatibility
ETC	High	High	High	Low
Asset control	Medium	Medium	Low	Medium/low
Vehicle identification	Low	Low	Low	High

, systems oriented to ETC applications involve OBU with performances, power consumption requirements, levels of complexity, and costs that are not compliant with a battery-less approach. Consequently, only applications that match with lower performances, as well as reduced complexity and power requirement, can be implemented using a battery-less approach.

In this context, the key function for implementing a battery-less system has become the development of advanced radio-based energy harvesting solutions, a technological approach commonly used by a wide range of applications, including the IoT [11], smart cities [12], and mobile healthcare [13], and, in particular, for completely battery-less systems [14,15,16].

Concerning the specific class of applications, the harvester system has to ensure a highly efficient reduced charge time in order to enable the OBU functionality, as illustrated in Figure 1. The OBU must be charged by the RSU transmitted signal during the so-called “TX data and power” time interval. Consequently, the implemented harvesting system has to charge and activate the OBU in a time that does not exceed the order of ten seconds.

The aim of the present study are to illustrate an efficient and compact energy harvester solution, compatible with battery-less OBU design, based on the approach described in [17] but characterized by the use of an antenna showing a specific impedance. The key parameters for improving the harvester performance in this approach consist of maximizing the input current of the DC/DC converter. The solution proposed in [17] suggests that, by making use of an RF power generator that shows quite a peculiar impedance to the rectifier, it is possible to obtain optimum charge performances. This impedance is almost never purely resistive, and this target is reached by means of a specific matching network that transforms the input impedance of a 50 Ω antenna into such an optimum impedance [17].

In our approach, the results are obtained by means of a proper patch antenna which has been designed to directly show the optimum impedance to the harvester. The use of such an antenna offers the advantage of minimizing the feed line complexity, avoiding any need for additional circuitry, and thus decreasing the system dimensions and complexity. However, the real advantage consists of the power loss reduction because of the matching network removal. The reduction in losses translates into an increased DC/DC input current and, consequently, an enhanced efficiency.

A comparative analysis between some state-of-the-art harvesting solutions, specifically designed for low power RF applications, and the proposed application is carried out, showing the advantage of the present approach with respect to the reference applications [18,19,20].

The use of this novel harvesting approach in the battery-less OBU design enables the development of systems with a very simple architecture, high compactness, reduced costs, and increased performance for assets control and vehicle identification applications.

2. Harvester Architecture and Operation

The reference application for this study is the development of a compact harvester system for a low-cost battery-less OBU working at 5.8 GHz [7,8] and optimized for assets control and vehicle identification applications. This means that the energy available for system activation is the energy supplied by the RSU during the wake up and negotiation time account, as illustrated in Figure 1. Consequently, the harvester must possess a maximum conversion efficiency and minimum charge time and size. Additionally, the operating range between RSU and OBU has to be limited to about 1.5 to 2 m. This leads to optimization of the harvester’s performances with respect to the power impinging to the rectifier, related to the required distance. In our reference application, as explained in Section 5, a power level of −5 dBm has been chosen for the optimization procedure.

The proposed harvester block is based on the design principle presented in [17]. The top diagram in Figure 2 illustrates the harvester as designed in [17], where the so-called optimum matching network is used to show the desired impedance to the RF-to-DC conversion section. The bottom diagram in Figure 2 describes the modified architecture with the matching network removed. The proposed system consists of a chain composed of the optimum matched antenna, the corresponding RF-to-DC rectifier, the specific DC/DC boost, and a tank capacitor, representing the energy storage for the following sections.

In order to achieve the objective of this study, the key sub-system is the cascade of the RF-to-DC rectifier and the DC/DC converter blocks. The cascade of these two blocks may be addressed as the RF-to-DC rectifier, which acts as a frequency power converter that transfers RF energy to DC. The intrinsically nonlinear behavior of this power converter requires the harvester block to be analyzed by applying nonlinear techniques.

In this application, the actual goal is to maximize the harvester performance that translates into increasing the RF-DC output current $I_{CGH}$ as much as possible, which feeds the DC/DC boost, thus, reducing the DC/DC “charge” time [17,21]. To this aim, the DC/DC boost, which is implemented by a BQ25570 evaluation board [22], is modeled with an equivalent circuit showing a variable input termination.

With reference to Figure 3 and following the reasoning carried out in [17], the relation between current and voltage at the DC/DC input is given by:

$$I_{CHG}=\{\begin{array}{c}\frac{V_{RFDC}-V_{TH}}{R_{RFDC}}when{V}_{IN}>V_{TH}\\ \\ \\ 0otherwise\end{array}$$

(1)

for the hardware involved, the minimum input voltage is the threshold $V_{TH}=$ 0.34 V.

As stated in [17], the DC/DC block is described by its variable input termination within the simulation. In particular, the boost converter inside the DC/DC block operates draining as much current as possible. Since the input source shows a non-zero output resistance R_RFDC, when the current absorption increases, the input voltage reduces. Consequently, this leads to a variation of DC/DC state parameters.

Benefitting from this equivalent circuit model, an effective source-pull analysis may be performed. The source-pull algorithm consists of varying the $Z_{src}$ of the ideal generator across any feasible value. For each of the sets, the DC/DC charge process is simulated and an estimated $I_{CGH}$ is computed. Comparing all the collected $I_{CGH}$ values, the optimal $Z_{src}$ is formally identified as the one that leads to the maximum charge current.

Figure 4 depicts the results of the source-pull procedure carried out with respect to the chosen reference power impinging the rectifier of −5 dBm. The resultant optimal impedance $Z_{opt}$ has a very small real part, namely, 12.1 Ω. This is expected since it must maximize the output current. Instead, quite a high imaginary part, namely, −j22.3 Ω, is found as compared with the real part of 12.1 Ω. The corresponding best current is ~145 μA. This value, as well as $Z_{opt}$, strongly depends on the diode chosen for the voltage tripler.

3. Antenna Design

In most automotive applications, the OBU is typically placed below the front window of the vehicle pointing toward the RSU when the car is moving [23]. Hence, the most suitable antenna to be implemented in an OBU is a patch antenna, with a major lobe that is almost hemispherical. This feature permits the link with the RSU. Many solutions have been proposed for the patch, depending also on the polarization on the communication. Here, we refer to the case of linear polarization operating at 5.8 GHz, as prescribed by the European Telecommunications Standards Institute (ETSI) standard [7].

The key to the success of the proposed approach consists of the ability to implement the necessary optimal input impedance $Z_{opt}$ with a minimum waste of power. Therefore, the antenna becomes a critical component for system implementation. The goal of the antenna design procedure is to provide a reasonable match with the computed $Z_{opt}$, while pursuing the best gain.

As previously discussed, a relevant improvement in OBU design is given by the architecture simplification, from which a low-cost and small size system can be obtained. On this basis, the use of a patch antenna, which is a planar technology component, allows us to build the entire harvesting system in a single step. Therefore, high integration and maximum compactness can be achieved. In addition, the patch antenna, as an unbalanced structure, can be directly connected with the network without the use of a balun and a matching network, which would increase the system size.

With reference to Figure 5, the proposed antenna consists of a $L_{0}x{W}_0$ rectangular patch, coupled with a pair of identical $L_{1}x{W}_1$ rectangular patch parasites, one for each side. The parasites are placed at a small distance $g$, and thus are excited by proximity coupling from the fringing field of the main patch. This is a common strategy to enhance the antenna performance, either in terms of gain or bandwidth [24,25].

It is well known that this kind of antenna can be described by the cavity model, which represents the field under the patch as a superimposition of resonant modes. These modes depend on the working frequency, and thus the patch sizes and the position of the feeding network, which allows us to only excite some modes. The proposed patch antenna is designed to excite the first resonant mode, from which a radiated far field with a maximum gain in the normal direction is obtained. This first resonant mode is related to a certain input patch impedance trend and is equal to zero at the center of the patch and the maximum on the edge. Therefore, different values of $Z_{patch}$ can be obtained by changing $W_{slit}$ and $L_{slit}$.

As discussed, for the sake of miniaturization and efficiency, a complex matching network must be avoided to prevent energy waste. This requirement spurs the idea of synthesizing at the edge of the antenna an impedance $Z_{patch}$, which is compatible with the required $Z_{opt}$, after the simplest impedance transformations.

The rectangular patch is designed to use a low-cost commercial substrate (${\epsilon }_r=3.75$, $tan\delta =0.02$ and $h$ = 1.575 mm) and, as previously mentioned, to show a maximum gain of ~6.5 dB in the broadside direction.

With reference to Figure 6, the input impedance of the synthesized patch, evaluated at 5.8 GHz, is compared to the desired $Z_{opt}$. This evaluation is carried out graphically by means of various Smith charts. The idea is to pursue the nominal matching to $Z_{opt}$ exploiting only a section of transmission line of length $\mathit{l}$. Accordingly, the effective normalized values of $Z_{patch}$ and $Z_{opt}$ change as well, and so does the length of the rotation needed to transform one into the other. In particular, Figure 6a shows the case of a transmission line of impedance $Z_{0A}$ = 50 Ω. The impedance seen by the patch after the rotation corresponding to $l_A$, which is the best possible in this case, does not accurately synthesize the real part of $Z_{opt}$. In the case of $Z_{0B}$ = 100 Ω, described by Figure 6b, the optimal impedance $Z_{opt}$ is almost exactly hit after a rotation corresponding to the length $l_B$. Finally, the case depicted in Figure 6c shows the rotation corresponding to $l_C$ around the reference impedance $Z_{0C}$ = 200 Ω. In this case, the real part of the input impedance $Z_{in}$ is slightly larger than the real part of $Z_{opt}$.

Apparently, the best solution is the case where $Z_0=Z_{0B}$. However, with reference to the source-pull result in Figure 4, it can be seen how the area around the best impedance $Z_{opt}$ is a plateau. Hence, the third solution is only marginally worse than $Z_{0B}$ in terms of impedance, but it is reasonably better for the loss and compactness, considering how much smaller $l_C$ is than $l_B$. Furthermore, higher transmission line impedance implies a narrower width, resulting in a positive effect on spurious radiation reduction, and hence antenna efficiency.

In view of this consideration, a final optimization was carried out starting from this heuristically good condition. The best gain and match are pursued by exploiting all the available degrees of freedom and the ones given by the parasitic patches. The best design exhibits the gain in Figure 7, with a maximum of 6.75 dB at 5.8 GHz. The corresponding $Z_{in}$ is (15–j25) Ω, in agreement with the nominal value of (12.1–j 22.3) Ω.

Figure 8 shows the simulated reflection coefficient of the proposed antenna. The optimal condition at the nominal frequency of 5.8 GHz is clearly highlighted. It is worthwhile to remark that the proposed harvester operates in a narrowband around the center frequency. Nevertheless, the 10 dB return loss bandwidth extends from 5.7 GHz to 5.9 GHz, confirming the effectiveness of the proposed solution implemented without a matching network.

Figure 7 shows the simulated radiation pattern. The maximum gain of 6.6 dB is achieved in the boresight direction. This good performance proves that the design conditions imposed for the optimal impedance do not affect the radiative performance of the antenna. The absence of an external matching network is a strong aid in keeping a high level of efficiency.

The proposed antenna was designed to show the optimum impedance to the harvester. However, in order to minimize the OBU dimensions, the designed optimum impedance antenna must also be used for OBU communications. The complete block diagram of the system is illustrated in Figure 9. Note, the introduction of a control unit that drives the modulator. The modulator block is directly connected to the antenna.

The description of the full system exceeds the aims of the present study but the reference system configuration with the optimum-matching network makes use of a modulator block based on a cold FET architecture, directly connected to the 50 Ω interconnection between the antenna and the optimum matching network.

In a system based on the proposed harvester configuration, the modulator block must work on a different impedance. It has been demonstrated that such a block may be connected directly to the antenna and harvester even if they show an impedance different from 50 Ω.

As illustrated in [20] it is possible to evaluate the condition under which the present harvester design is compatible with a backscattering ASK modulation scheme based on a two-symbol modulation (S_i with i = 0, 1). In particular, with reference to Figure 10 and according to the results showed in [20], is possible to make explicit the impedance condition for modulator implementation as:

$$\{\begin{array}{c}\left|Z_{MOD}\left(s_0\right)\right|\gg max\left(\left|Z_{HARV}\right|,\left|Z_{RF}\right|\right)\\ m_{ASK}=1-\left|{\Gamma }_{BO}\right|\left|2\frac{Z_{MOD}\left(s_1\right)}{Z_{RF}}+1\right|\end{array}$$

(2)

Under this condition, the modulator does not significantly affect the matching between the antenna and the harvester during the “TX Data and Power” time interval.

4. Assembly

With reference to Figure 11a, the design showing the best performance in the simulation was assembled with the harvester directly connected to the patch, and then fabricated on a commercial dielectric substrate ISOLA FR408.

The calculated final values for the parameters are reported in

Table 2. List of the geometric parameters of the antenna design. All dimensions are in mm.

${\mathit{L}}_{\mathbf{0}}$	${\mathit{W}}_{\mathbf{0}}$	${\mathit{L}}_{\mathbf{1}}$	${\mathit{W}}_{\mathbf{1}}$	g	${\mathit{W}}_{\mathbf{slit}}$	${\mathit{W}}_{\mathbf{slit}}$	${\mathit{W}}_{\mathbf{in}}$	$\mathit{l}$
15.93	12.5	9.53	2.33	1.58	0.266	0.125	1.23	2.21
${\mathit{L}}_{\mathbf{2}}$	${\mathit{W}}_{\mathbf{2}}$	${\mathit{L}}_{\mathbf{3}}$	${\mathit{W}}_{\mathbf{3}}$	${\mathit{W}}_{\mathit{s}\mathit{u}\mathit{b}}$	${\mathit{L}}_{\mathit{s}\mathit{u}\mathit{b}}$	$\mathit{L}$	$\mathit{C}$	$\mathit{D}$
45	0.7	2.8	3.1	0.67	30	1 nH	100 pF	BAT1503

.

It is noteworthy that the occupied area is only 30 × 45 mm² as compared the solution based on a 50 Ω antenna, shown in Figure 11b, which is significantly much larger with an area of 35 × 62 mm². The latter is employed as a reference in the following section to comparatively validate the result of the proposed solution based on an antenna directly showing the desired impedance.

It can be observed that, without the need of a matching network, the dimensions of the device are very compact. Indeed, $L_m$ = 13.5 mm is the length of the input section of the reference antenna, while $L_n$ = 18.5 mm is the length of the matching network needed for the transformation of 50 Ω in $Z_{opt}$. It is noteworthy that the gain of the reference antenna, measured at the 50 Ω section, exceeds 7 dB.

5. Experimental Validation

The performance of the proposed harvester solution is compared with the performance of a system based on [17] (the reference system). The prototype of the proposed solution (Figure 12a) and the reference system were evaluated, experimentally, in operative conditions as illustrated in Figure 12b.

In order to reproduce the effective operative condition, a specific experimental setup was prepared. The RSU, compliant with ETC applications, was implemented making use of a signal generator supplying the CW signal at 5.8 GHz. The transmitting antenna, visible in Figure 12b, together with the receiving antenna, is fully compliant with the antenna used in RSU for ETC. The impinging power to the antenna section was measured making use of a calibrated power meter connected to a certified antenna. Finally, the $I_{CGH}$ current supplied by the rectifier to the BQ25570 circuit was evaluated using a precision multimeter.

Figure 13 shows the power impinging on the device under test (DUT) in the operative condition, meaning a nominal EIRP of 39 dBm. The DUT is progressively moved away from the RSU keeping the boresight direction. The power at the rectifier section is estimated from the measurement at the antenna section, considering a gain of 6.6 dB, as previously estimated in Section 3.

As anticipated in Section 3, the value of $Z_{opt}$ was estimated as the result of the source-pull procedure considering the impinging power at the −5 dBm at the rectifier section. With reference to Figure 13, this power level is chosen because it corresponds to the distance of 1.75 m, the center of the range of interest for application proposed in this paper, as anticipated in Section 2.

For the sake of repeatability, subsequent experiments were performed by varying the transmission power rather than the distance. Figure 14 shows the measured current $I_{CGH}$ for the two prototypes, as well as the difference of the resulting currents corresponding to the two devices. The current is plotted against the impinging power at the rectifier input. It can be highlighted that the proposed solution starts to perform better than the reference solution starting from the impinging power of −11 dBm, which correspond to 2.8 m, according to Figure 14.

At the target optimizing impinging power, the $I_{CGH}$ value of the proposed solution is more than 110 μA greater than in the reference solution. The best improvement is observed at −4 dBm, corresponding to 560 μA for the proposed solution and 430 μA for the reference solution.

Moreover, Figure 15 describes the comparison of conversion efficiency shown by both the harvester prototype and the reference prototype with respect to the impinging power. The measured results confirm a very similar behavior for lower impinging power levels. At the optimized impinging power reference level (−5 dBm), the proposed solution exhibits a significant improvement of ~12.0% in conversion efficiency.

The efficiency was evaluated by comparing the power in DC, calculated on the basis of the DC/DC charge current $I_{CGH}$ and the fixed input voltage with the same DC/DC, i.e., 0.34V, with respect to the impinging power. A peak exceeding 50% is observed at −7.3 dBm.

At the same time, a research concerning state-of-the-art equivalent systems was carried out. The aim was to assess the performances obtained by the proposed solution with respect to the state-of-the-art approaches illustrated in the literature.

Table 3. Harvesting performances in comparison with the state-of-the-art approaches.

Design	Freq.	Input Power	η	ICHG	Load and Condition
[18]	5.8 GHz	−5 dBm	16%	N.A.	1500 Ω, fixed load
[19]	5.8 GHz	−5 dBm	23%	145 μA	3000 Ω, fixed load
[20]	5.8 GHz	−5 dBm	16%	135 μA	DC/DC, current sink
Present work	5.8 GHz	−5 dBm	45%	450 μA	DC/DC, current sink

summarizes the comparison with the state-of-the-art at 5.8 GHz with respect to charging current and conversion energy efficiency. The data show that for an input power of −5 dBm, the charge current increases from 135 μA to 450 μA, while in terms of efficiency, there is an increase from 16% to 45%. In two of the illustrated solutions, the results were obtained making use of fixed load conditions that represented a far simpler constraint with respect to the measurement conditions implemented in the present work.

6. Conclusions

We demonstrated a novel energy harvesting solution based on an optimum antenna design suitable for implementation in an on-board unit for vehicle communications. Experimental results, carried out on the bases of a comparison between the present development and the harvesting prototype based on the integration of a 50 Ω antenna showing similar gain combined with the matching network, confirm the better performance of the proposed approach.

As compared with the reference harvester, the proposed compact prototype optimized at −5 dBm of impinging power provides an improvement in both charging current ($I_{CGH}$) and efficiency of about 100 μA and 15%, respectively. The two prototypes show almost equivalent performances up to −11 dBm of impinging power, while for higher value the compact prototype shows a significantly better behavior.

Furthermore, on the basis of results illustrated in Table 3, it is possible to state that the harvesting approach illustrated in the paper shows the best performance as compared with those of all the other approaches. In particular, Table 3 highlights an efficiency that is double with respect to the efficiency obtained in [19] and more than three times the charge current measured in [19,20].