Single-monopole UHF rectenna

The adopted design strategies for achieving long-distance scavenging capabilities are first described through the optimization of a standard single-antenna rectenna layout able to rectify ambient energy at 868 MHz: the corresponding performance represents the lower limit the final two-monopole solution must go beyond.

The envisaged next-generation RFID tag, described by the scheme of Fig. 1, has to perform energy-consuming operations: in particular during the localization phase, when the activation of a microcontroller unit (MCU) and the consequent UWB switch modulation of the received UWB signal for the backscattering are needed. Indeed, the MCU represents the most power-hungry part of the system and strategies to keep it in an idle state for most of the time, except during the active phases, play a crucial role. To this extent, the usage of a low-power and low-latency wake-up radio (WUR) [Reference Del Prete, Costanzo, Magno, Masotti and Benini34] is adopted in this project: its purpose is to wake up the active components only when required.

Figure 3 shows the photo of the single-antenna prototype, realized on 0.5 mm-thick Rogers RO4350B substrate (ε _r = 3.48, tan(δ) = 0.0037 at 10 GHz): it consists of a meandered-monopole (in order to fulfill the footprint constraint) with a zoomed view of the lumped-element matching network. The corresponding circuit schematic is reported in Fig. 4, where the two-path rectifier (consisting of a couple of voltage doublers) for both the rectifying (RECT) and the WUR operations is visible. It is worth noticing the replacement of the standard lossy resistive divider with a novel fully reactive power divider: it properly splits the received RF power into two unequal portions, in order to guarantee activation for low-power levels and, most of all, without deteriorating the main rectifying action (as verified later on), but it results an almost lossless component, except for those due to the parasitic of the non-ideal capacitors [Reference Costanzo, Dardari, Aleksandravicius, Decarli, Del Prete, Fabbri, Fantuzzi, Guerra, Masotti, Pizzotti and Romani6].

Fig. 3. Photo of a single-monopole rectenna, with a detailed view of the matching/rectifying sections.

Fig. 4. Circuit schematic representation of the rectenna of Fig. 3, with components value.

As the last design choice, the ultra-low-power commercially available PMU TI BQ25570 [35], well suited for ultra-low-power applications, is placed as the load of the rectifier branch. As already stated in the Introduction, a clear distinction between the energy requirements of the “cold start” and “already active” states of this PMU must be made: to comply with its “cold start”, a dc voltage of 330 mV and a dc power of 15 µW are needed. After the critical start-up phase, this PMU is able to continue the charging operation with less strict input requirements: voltages as low as 100 mV and consumption of only 7.5 µW are sufficient. The first couple of values are adopted as a reference in the design of the presented rectifying units: as will be clearer in the following, this choice has a strong impact on the maximum operative distance of the system.

The circuital/electromagnetic co-design approach has demonstrated its effectiveness and relevance when non-linear radiating circuits are involved [Reference Costanzo, Donzelli, Masotti and Rizzoli36, Reference Masotti, Francia, Costanzo and Rizzoli37]: this strategy is adopted in the optimization of the single-monopole rectenna of Figs 3 and 4, where harmonic-balance analyses are repeated to maximize the RF-to-dc conversion efficiencies when low UHF power levels are available in the rectenna location (i.e. −20÷−10 dBm) [Reference Masotti, Costanzo, Del Prete and Rizzoli13, Reference Costanzo, Donzelli, Masotti and Rizzoli36]. The results of the optimization procedure are shown, in solid lines, in Fig. 5 in terms of open-circuit (OC) voltage at both rectifier and WUR outputs (Fig. 5(a)) and of RF-to-dc conversion efficiency and rectified voltage on the optimum load (10 kΩ in the present case) for a range of available RF power at the rectenna input wider than that one adopted in the optimization. In the same figure, the corresponding measured values are superimposed: very good agreement between measured and simulated results is obtained in all cases, except at the higher power levels, where the breakdown voltage of the Schottky diode is not providing accurate model prediction inside the circuit simulator. Figure 5(a) inspection allows a first encouraging comment on the previously mentioned design choices: the capacitive divider works in a proper manner, providing a V _WUR around 20–25% of rectifier one (V _RECT).

Fig. 5. (a) Open-circuit output voltage and (b) rectified voltage with RF-to-dc conversion efficiency as a function of available RF power.

Before verifying the most stringent design constraint in terms of reference node-tag distance, the validity of the capacitive divider solution, which is able to provide a WUR implementation with no need for a dedicated antenna, has to be proven further. In order to assess the effectiveness of this solution, it is important to evaluate how the two signals might affect each other. Indeed, since the WUR is obtained from a direct derivation of the RF voltage which feeds the rectifier, a variation of the loading conditions of one signal could translate in a consequent alteration of the second one. Though the WUR alteration due to different rectifier conditions is an obvious consequence of the adopted capacitive divider implementation, the opposite case would result particularly detrimental, since the WUR should perform its operations in a seamless fashion, without affecting the main functionalities of the harvester.

The results of these proofs are shown in Fig. 6: in Fig. 6(a) the dc voltages at the RECT output port (rect Vdc) and at the WUR output port (WUR Vdc) are given as a function of the RECT load resistance. In this way, all the possible RECT operating regions (from almost OC up to almost short-circuit) are considered. The same content pertains to Fig. 6(b), where the operating region of the WUR is varied, instead. As can be evinced from Fig. 6(a) inspection, variations of the rectifier voltages reflect in direct variations of the wake-up ones, as expected, being these directly derived from the main voltage signal. If we think to the final tag usage, this is not a big concern because it does not result in fake tag activation: indeed, for almost 95% of the time, the PMU provides a constant value at the rectifier output, corresponding to the optimum loading conditions, and even though the maximum power point (MPP) tracking mechanism adopted by the PMU [Reference Costanzo, Romani, Masotti, Arbizzani and Rizzoli38] periodically refreshes its state, the refresh rate is significantly lower (few Hz) with respect to the frequency of the UHF CW modulation employed for tag addressing (in the order of kHz), therefore it is not affecting the WUR functionalities.

Fig. 6. Comparison between measured and modeled dc voltage at the rectifier and WUR outputs for different values of (a) the rectifier load and (b) the WUR load.

The more significant evaluation of the dual case is described in Fig. 6(b) by sweeping the WUR load value and comparing how this translates respectively to the rectifier and the wake-up outputs. Obviously, the voltage at the wake-up output is changed accordingly, but this variation exerts only a minor influence on the rectifier signal, as desired: in fact, the WUR will experience a variation of its operating condition during the tag usage, but must not compromise the main rectifier efficiency. The graph reported in Fig. 6 has been obtained for a fixed available power level at the input port of the circuit of Fig. 4 of −13 dBm, but equivalent conclusions are obtained all over the expected incident power range. Therefore, these results prove the effectiveness of the proposed seamless implementation of the WUR.

As regards the maximum operating distance between the UHF source and the rectenna, we refer to the optimum RF link conditions: perfect polarization matching between the linearly polarized transmitter and the receiving antenna is assumed. When 2 W ERP are provided at 10 m distance, the RF power received by the 1.8 dBi-gain monopole and made available at the harvester input is −14.3 dBm (~37.2 µW). With this received power level, the measured dc voltage is approximately 360 mV and the RF-to-dc conversion efficiency is equal to 35%, which results in a total available dc power of 13 µW. This means that, as usual, the dc voltage goal is less critical than the power one: in fact, the obtained dc voltage is able to satisfy the PMU requirements, although the dc power is still lower than the required value. Note that these measurements refer to the most favorable link conditions of perfect polarization matching. The limits of the single-monopole rectenna under examination are further emphasized by the practical need of a circularly polarized UHF transmitter for orientation independency of the RFID tag: this means a further ideal 3 dB reduction of the available RF power, resulting in measured dc voltage and power outputs of 220 mV and 5 µW.

Dual-monopole UHF rectenna with decoupling network

Because of the expected unsatisfactory results of the single-monopole solution, a more sophisticated solution is needed for the complex scenario under test.

Hence, a two-monopole layout (one vertical and one horizontal), with series dc-combination of the two rectennas output, is investigated. One can note that this solution is ideally more versatile and more efficient than the single linearly polarized antenna previously examined; but its choice is convenient with respect to a circularly polarized rectenna, too: in fact, the latter could be highly mismatched with the incoming UHF signal if inverse rotation is imposed by the tag position or by the harsh scenario.

However, the footprint constraint imposed by the indoor application (10 cm × 10 cm) forces to have highly-coupled antennas (with ports distance lower than λ/20), resulting in degradation of their radiation properties as well as signals interference due to the coupling between the antenna ports. These uncontrolled interferences could result in undesired re-radiation of a portion of the coupled signal instead of being converted to dc, as well as destructive signals power combinations: but this cannot be a priori known.

For this reason, the architecture of the two closely-spaced rectennas with dc-combined outputs of Fig. 7 has been first studied. An initial full-wave simulation of the sole linear portion of the rectenna allows to retrieve both the near-field (scattering parameters and surface currents) and far-field performance of the two-monopole assembly.

Fig. 7. Circuit schematic representation of a two-element rectenna without decoupling network.

A poor transmittance value of around −5 dB is evaluated (see solid red line in Fig. 8) in the frequency band of interest. This result is confirmed by the surface current plot of Fig. 9(a), where only the metallization of the whole structure is made visible, for the sake of visualization, and port 1 (of the horizontal monopole) is excited: a strong electromagnetically induced current onto the parasitic vertical monopole is clearly visible. Finally, the radiation surface (in the same previous excitation conditions) is shown in Fig. 9(b): the presence of the unexcited vertical monopole significantly distorts the radiation mechanism, as expected by antenna theory.

Fig. 8. Modeled electromagnetic port coupling for a couple of neighboring monopoles (one vertical and one horizontal) with (dashed blue line) and without (solid red line) defected ground plane.

Fig. 9. Full-wave simulation output at 868 MHz: surface currents (left) and radiation surface (right) for two closely-spaced monopoles, when the horizontal one is fed, only.

However, one cannot a priori say if the effects of the aforementioned behavior are destructive or not, in terms of available power at the antennas port: for this reason, a more effective analysis has been carried out through the accurate evaluation of the Norton equivalent current generators at the antenna ports (see Fig. 7) (for both the single-monopole and the new two-monopole layout). The following formula [Reference Rizzoli, Costanzo and Monti39], derived by the rigorous application of the electromagnetic theory (i.e. the reciprocity theorem [Reference Harrington40]), can be adopted for this purpose:

(1)$$J_{eq,k}(\,f_{RF})\approx j\;\lsqb {1 + R_0Y_{kk}(\,f_{RF})} \rsqb \,\;\displaystyle{{2{\kern 1pt} \lambda _{RF}\,{\kern 1pt} r\,e^{\,j\beta r}} \over \eta} \,\;\displaystyle{{{\bi E}_{\bi i}\; \bullet {\bi E}_{\bi R}^{(k)} (\theta _L,\;\varphi _L)} \over U},$$

where E _i is the incident electric field (considered as a plane wave coming from the far RF source and impinging on the phase center of the receiving antenna along the link direction (θ _L,ϕ _L)), E^(k)_R represents the field radiated by the k-th port of the receiving antenna, when it works in transmitting mode, and is driven by a sinusoidal voltage source of frequency f _RF, electro-motive force U, and internal impedance R ₀. Y _kk(f _RF) is the k-th admittance value (derived by full-wave analysis) of the multi-port receiving antenna: hence, this term automatically takes into account the effects of the transadmittance contributions (Y_kp(f _RF), k ≠ p) on the k-th port of interest.

The calculations carried out for both the circuits of Figs 4 and 7, for a vertically polarized incident field with 1 V/m of intensity, at a distance r = 1 m, provide the following values for the magnitude of (1): 3.8 mA for the single monopole, <1 mA at the vertical monopole input port, for the strongly electromagnetically coupled two-antenna case.

These last results suggest not to investigate further this simple two-monopole layout without any decoupling mechanism.

As a first attempt of reduction of the ports coupling, we resort to the defected ground structure reported in [Reference Chiu, Cheng, Murch and Rowell20], consisting of the removal of slits of copper from the ground plane in the region below the ports to be decoupled, with the additional need for lumped inductors in between the slits: this turns out to be a resonator whose resonant frequency is related to the distributed capacitance values among the faced slits and the inductance. Figure 10 contains both the overall view of the two monopoles with defected ground (the substrate is again made invisible) and the zoomed view of the region below the two-monopole ports: slits width is identical to slits distance and equal to 0.8 mm, whereas the (ideal) inductance value is 22 pF for all the lumped indictors. The output of the corresponding full-wave simulation in terms of transmittance magnitude is reported in Fig. 8, superimposed to that one of the previous, strongly-coupled solution: an encouraging −21 dB of the |S ₂₁| parameter is guaranteed. Note that a de-embedding of the ports (of Fig. 10(a)) up to the slits is performed in the full-wave simulation in order to reproduce the more realistic and critical situation.

Fig. 10. (a) Layout of two closely-spaced monopoles with defected ground plane (only the metallization is made visible); (b) detail of the ground plane with slits and lumped inductors.

The limits of the previous solution are multiple: first of all, there is also the need of a matching network (as indicated in Fig. 7); the positioning of the inductances represents a delicate task; finally, the etching of the ground is an add-on to be included in the prototyping phase. Moreover, the replacement of the ideal inductors with real ones brings to a not negligible worsening of the decoupling behavior: the |S ₂₁| value raises to almost −17 dB. For these reasons, we look for a simpler decoupling solution involving a limited number of lumped components, only: it needs for the proper design of a lumped-element decoupling network (similar to the one proposed in [Reference Chen, Wang and Chung32]) and working in the antennas operating bandwidth. It is worth noticing that, differently from the explicative scheme of Fig. 2, the adopted harvester layout, chosen in the present work and shown in the photo of Fig. 11 exploits a more effective unique network combining goals of both decoupling and matching. This is an additional advantage with respect to the aforementioned defected-ground architecture. The corresponding circuit schematic representation of Fig. 12 highlights the decoupling/matching network topology: it results simply composed of two parallel lumped elements and two couples of series ones, realized with SMD components.

Fig. 11. Photo of the realized two-monopole prototype. The bottom-left part of the tag, containing a green solder mask, is designed for integration of the electronic components.

Fig. 12. Circuit schematic representation of a two-element rectenna with decoupling/matching network.

Of course, the output of the electromagnetic simulation of the two monopoles is adopted in the design of the circuit of Fig. 12, thus taking into account the reactive near-field frequency behavior of the unwanted array [Reference Craeye and González-Ovejero14]. To obtain the four-port decoupling network design, a two-step procedure is adopted. First, a linear analysis is carried out with the network terminated as follows: on the left-end side, the network terminations consist of the two-port, frequency-dependent Z-matrix, derived by EM-simulation, representing the two-port monopoles layout; on the right-end side, the rectifying sections input impedance of 9–j150 Ω are used, that are the capacitive divider input impedance when loaded by the rectifiers, for a receiving reference power of −13 dBm.

The results of this first approximate optimization are accounted for in the EM simulation of the resulting subsystem and are encouraging with respect to the performance without decoupling network (see Fig. 9). The corresponding results are reported in Fig. 13, where the surface currents and the radiation surface obtained by exciting port 1 (horizontal dipole) clearly show that the behavior of the horizontal dipole is almost independent of the vertical one.

Fig. 13. Full-wave simulation output at 868 MHz: surface currents (left) and radiation surface (right) for two decoupled closely-spaced monopoles, when the horizontal one is fed, only.

The equivalent Norton current generators can now be evaluated for the accurate active rectenna simulation by resorting again to (1): under the same conditions as before, the magnitude of (1) for the vertical monopole is around 4 mA (very close to the case of the single monopole), thus confirming again the results of Fig. 13 and the achievement of the design goal.

As can be evinced from Fig. 12, another design choice is worth mentioning: there are two WUR paths, one for each rectenna (both the vertical and the horizontal one), still exploiting the purely reactive divider solution. This solution is again imposed by the harsh electromagnetic environment where the tag will operate: in fact, the duplication of the double path guarantees the powering of both the WURs for any polarization of the incoming signals, thus assuring a stronger rectification (as demonstrated by the following results) starting from the initial activity. As before, the series dc connection has been realized for both the rectifier and the WUR.

A second non-linear/EM optimization is then carried out in a more rigorous way, by considering the active rectenna of Fig. 12 as a whole, for a wide range of incoming RF powers (from −20 to −10 dBm) [Reference Masotti, Costanzo, Del Prete and Rizzoli13, Reference Costanzo, Donzelli, Masotti and Rizzoli36], as for the single-monopole rectenna. The final layout of the harvester, shown in Fig. 11, foresees the inclusion of a miniaturized UWB chip antenna, according to the scheme of Fig. 1, for future UWB-based localization of the system (a proper soldering pattern is realized on the backside of the PCB).

The modeled scattering parameters of Fig. 14 (solid lines), normalized again to 9–j150 Ω, provide at a glance the matching and decoupling behavior of the final topology of the network (whose element values are listed in Fig. 12): port 1 and port 2 are the additional test ports (both shown in Fig. 12 and in the prototype photo of Fig. 11) included in the final layout exactly for a separate check of the designed lumped-element network. The matching and decoupling conditions at the UHF 868 MHz frequency are satisfactory: the simulated value of the S ₂₁ magnitude is around −20 dB, but taking into consideration the real model of the lumped components while simulating. The test lines (indicated with arrows in Fig. 12) are then exploited for the evaluation of the equivalent measured performance: the measured scattering parameters magnitudes are superimposed (with dashed lines) in Fig. 14, after a proper de-embedding of the test lines [Reference Crupi and Schreurs41] and a re-normalization of the S parameters. A good agreement is obtained between simulated and measured data, except for a small reduction of the measured decoupling, which is still below −10 dB.

Fig. 14. Comparison between the measured and modeled insertion and return loss (with sign reversal) at the two test ports, in correspondence of the interface between the decoupling network and the two rectifying sections.

The single- and the two-monopole rectennas performance is compared through a twofold test. In both cases, a circularly polarized transmitting UHF patch antenna is adopted: this choice allows a less stringent system operation, since tags should be able to rotate freely, without being restricted to predefined orientation positions. In the first part of the test, the transmitting patch antenna is kept fixed and provides an available RF power in the tag location of −13 dBm, hence each monopole (both in the single- and in the double-antenna case) receives around −16 dBm. Figure 15 shows the comparison between the two solutions in terms of rectified OC voltage and rectified power: the latter is not a direct measurement, but it is evaluated by following the operating rule of modern dc–dc converters, i.e. tracking the MPP as V _RECT = V _OC/2 [Reference Costanzo, Romani, Masotti, Arbizzani and Rizzoli38]; from the knowledge of this value and that one of the optimum load (10 and 15 kΩ for the single- and double-monopole rectennas, respectively), the dc power is extrapolated. This measurement/calculation is repeated for different orientation of the rectennas. As expected, each rectenna rectifies almost the same amount of power for all the orientations; however, through the two-monopole architecture, more than double rectification is achieved whatever orientation is considered.

Fig. 15. Single- and double-monopole performance for different orientations with respect to a fixed circularly polarized UHF source.

In the second phase of the test, the position of the UHF circularly polarized patch source with respect to the fixed rectennas is varied while emitting the maximum value of 2 W ERP. The measured rectified outputs are finally shown in Fig. 16: the limits imposed by the “cold start” of the adopted PMU (P _DC = 15 µW, V _DC = 330 mV) are reported with dashed red lines to better appreciate the fulfillment of the design goals by the advanced RF-uncoupled closely-spaced two-monopole rectenna. Hence, the effectiveness of the new rectenna array architecture and, most of all, of the achieved decoupling between the array elements is demonstrated: it ensures the stringent PMU requirements for the highest operation distance foreseen by the presented system.

Fig. 16. Comparison between single- and double-monopole rectenna for different distance from a 2 W ERP circularly polarized transmitter, in terms of (a) output dc power and (b) rectified voltage.