Millimeter Wave Power Transmission for Compact and Large-Area Wearable IoT Devices based on a Higher-Order Mode Wearable Antenna

,

IoT BANs have various applications in healthcare monitoring, fitness tracking, defense and wearable sensing [13], [14].Powering such systems using microwave and mmWave rectennas has attracted significant research interest, from Ultra-High Frequency (UHF) to mmWave 5G bands [15]- [17], where wireless power transmission is increasingly seen as a reliable and scalable method for powering the IoT [18].At mmWave bands, higher end-to-end WPT efficiencies can be achieved due to the improvement in antennas' gain [4], [9], [12].Furthermore, the wide spectrum available enables the co-existence of several applications such as WPT and communications, enabling a massive mmWave-connected IoT [17].However, the free space path loss and the additional Non-Line-of-Sight (LoS) losses at mmWave bands, for example due to attenuation by water molecules at 28 GHz, may imply that WPT to wearables at mmWave bands is less efficient compared to UHF bands.In addition, the antennas need to maintain a broad bandwidth to benefit from the available spectrum.At an antenna-design level, existing flexible and wearable antenna implementations do not fully address both requirements, where the efficiency and bandwidth are improved at the expense of a thicker antenna [17].mmWave-powered devices can be broadly classified into compact and large-area receivers.For a compact receiver, several analytical studies have shown that a higher antenna efficiency can be achieved at mmWave bands [10], [11].Nevertheless, channel gain or s-parameter measurements have not been used to compare mmWave WPT to area-constrained devices.For large-area systems, beamforming rectennas [12], as well as large transmitting arrays have been investigated [9], [19], showing potential performance gains.However, the performance of a large-area mmWave rectenna array has not been compared, analytically or experimentally, to its similarlysized UHF counterpart.Therefore, this work aims to present a wearable antenna design suitable for mmWave WPT, as well as present an extensive evaluation of its performance as a compact and a large-area power receiver.

A. mmWave Antenna & System Requirements for WPT:
To enable mmWave-powered BANs, the following antenna design criteria need to be met: 1) High-efficiency antennas with broad angular coverage for increased energy harvesting probability [20].
IEEE INTERNET OF THINGS JOURNAL, VOL.X, NO.X 2) Conformable and low antenna profile for improved integration in clothing and on thinner substrates [21].
3) Broad bandwidth, to enable the antennas to be used in the full 5G spectrum [12], [17].
While several theoretical and practical studies have shown the benefits of mmWave WPT [4], [9], [12], the performance of a mmWave-powered network based on a real antenna, meeting the above criteria, remains unclear.To explain, the benefits of mmWave WPT compared to its UHF counterpart need to be evaluated.Therefore, a realistic investigation on the feasibility of mmWave WPT needs to account for: 1) Antenna radiation properties, based on real measured antenna radiation gain and patterns.
2) Realistic non-linear rectifier model, based on real commercially-available devices.
3) Empirical LoS and N-LoS propagation models, to reflect the higher losses at mmWave bands.

B. Summary of Contributions
In this paper, a higher-order dual-mode broadband microstrip patch antenna is proposed on textiles for wearable WPT.The proposed antenna achieves: 1) The highest thickness-normalized efficiency of a wearable antenna.
2) Wider bandwidth compared to existing microstrip patch antennas.
3) The lowest profile of a broadside textile antenna with full body-isolation.

4) Wider angular coverage compared to wearable antennas
with broadside off-body radiation.
The DC power output of a mmWave-powered BAN based on the proposed antenna is then investigated numerically and experimentally demonstrating: 1) 11 dB measured basestation-to-body path (channel) gain improvement for an area-constrained receiver.
2) Up to 6.3× higher DC power output compared to a similar-sized UHF microstrip patch antenna in LoS and 2× higher power in N-LoS.
The proposed antenna design is introduced in Section II with the antenna measurements in Section III.The performance of the proposed antenna as a wireless receiver is then evaluated in Section IV, along with potential IoT applications of the proposed antenna, and mmWave WPT.

A. Broadband Microstrip Antenna Design
To design an antenna for wearable energy harvesting at mmWave bands the antenna needs to maintain a wide input bandwidth (S 11 ).As mmWaves have very poor tissue penetration capabilities beyond the skin [22], the antenna's patterns need to be predominantly off-body.In addition, due to the wide variation in the angle of incidence of power, the antenna needs to cover a wide beamwidth.Previously, microstrip antennas with multiple radiative modes were used to obtain several resonances at nearby frequencies, improving the bandwidth compared to a conventional patch.To explain, the majority of reported flexible and wearable patch antennas resonate 135 at a single-mode transverse magnetic (TM) mode, which is 136 typically the first-order TM 10 mode [23].Nevertheless, to 137 excite multiple modes simultaneously, a complex antenna 138 structure is often needed [23], [24]. 139 The dimensions of a rectangular microstrip patch antenna can be calculated analytically for a given TM resonant mode based on the length and width [25].For a planar (i.e.t<<λ) microstrip cavity, the resonance frequency f mn of a TM mn mode is given by where c is the speed of light, and f mn and k mn are the resonant 140 frequency and wavenumber for the mn mode, respectively 141 [25].a and b are the patch dimensions on the m and n 142 respective axes.

143
To achieve a wide S 11 bandwidth covering the mmWave 144 5G bands ), the proposed antenna is 145 designed to match 50 Ω for two resonant TM modes.Firstly, 146 a TM 02 patch is designed with a=9 mm as shown in Fig. 147 1-a.Given r =1.95 (1.9< r <2 due to the non-uniform gaps 148 between the fabric and the polyimide), the TM 02 patch has a 149 calculated theoretical resonance at 24.3 GHz using (1).Fig. 1     b, where the −10 dB beamwidth of the proposed antenna is 50 • (38%) wider than that of a CG patch.
To explain, in spite of the observed bore-sight "null" of the TM 20 mode, the broadside angular coverage of the proposed antenna will result in better energy harvesting coverage from arbitrarily-directed radiation, which was previously found to improve the DC power harvested compared to narrower-beam higher-gain antennas [20].

B. Bending and Wearable Operation
As the antenna is designed for wearable applications, it is essential to simulate the antenna in proximity with a human tissue model.Due to the short wavelength and to simplify the electromagnetic problem reducing the solver's time, a simplified layered tissue model is used.The skin layer, shown in Fig. 6, is based on the measured skin properties above 26.5 GHz, reported in [22].It was previously shown that mmWaves do not penetrate deeper than the skin layer [27].Fig. 6-a shows the model of the patch operating at 3 mm separation from the human model, as well as the dielectric properties and thickness of each layer.The simulated reflection coefficient on the tissue model shows less than 2% shift in the antenna's resonance when placed on the body, this is attributed to the additional capacitance introduced by the higher r of the tissue compared to air.The impact of bending has been investigated by bending the antenna across a 15 mm radius, as shown in Fig. 4-b, which results in less than 4% shift in the resonance, as shown in Fig. 5, and does not alter the S 11 < −10 dB bandwidth from 25 to 28.5 GHz.
The tissue model has been used to calculate the antenna's Specific Absorption Rate (SAR) at 28 GHz, shown in Fig. 6b, to showcase the antenna's compliance with the IEEE C95.1 The antenna has been fabricated using photolithography on 233 flexible polyimide laminates, using the method detailed in 234 [17], which was previously used to realize antennas resilient 235 to bending [21] and machine washing [28].Two antenna with the measured resonances of both prototypes.From 26 to 28 GHz, a discrepancy can be observed between the two prototypes.This can be attributed to the length of the connector's unshielded pin, approximately λ/4 at 27 GHz, which can act as an impedance transformer around 27 GHz causing additional reflection [29], as observed in the S 11 of prototype-2 in Fig. 7.For both measured prototypes, the antenna maintains an impedance bandwidth between 24.9 and 31.1 GHz, with an S 11 under −8 and −10 dB on prototype 1 and 2, respectively.
In wearable operation, the antenna needs to withstand bending and maintain its bandwidth in human proximity.The antenna's S 11 was measured on-body and under bending.The antenna was bent both in space and on-body, as shown in the inset in Fig. 8, while measuring the S 11 .As observed in Fig. 8, the antenna's S 11 response remains mostly unaffected in human proximity and under bending.

B. Antenna Radiation Patterns
The 3D polarimetric radiation patterns of the proposed antenna were measured in an anechoic chamber at 24 and 26 GHz; the measurements were limited by the available power amplifier's operation frequency and the WR-42 horn cut-off frequency.To improve the repeatability of the measurements, the antenna was mounted on a large (radius>5λ 0 ) circular ground plane, as in [17].Fig. 9 shows the 3D measured directivity D, normalized to 10 dBi, and the polarization of the measured D. The polarization of the textile antenna has been normalized using a standard 20 dBi WR-42 horn.
The peak measured D is 13.2 and 11.2 dBi, at 24 and 26 GHz respectively.The measured results are indicative of the antenna's performance over a continuous ground plane, such as a fully-textile shielding between the antenna and the human body, which was shown to reduce the SAR of mmWave wearable antennas [17].The simulated and measured relative gain patterns of the antenna, with the connector and the large ground plane, are shown in Fig. 10.The underlined plane in Fig. 10 indicates the principal plane of the respective TM mode at each frequency.It can be observed that the simulated and insensitive harvesting [31].On the other hand, while circular polarization can be preferred in directional WPT [32], it is still prone to mismatch between right and left-hand circularly polarized incident radiation [33].Finally, in the more general wearable BAN antenna case, a high polarization purity is not required due to the mobility and rotation of the antenna, for example when the antenna is used on the wrist or arm [34].
The 3D total radiated power was used to measure the efficiency (inclusive of mismatch) of the proposed antenna, with respect to a reference antenna.A WR-42 horn antenna was used as a reference using the method detailed in [35].The radiated power of the textile antenna and the reference horn were measured at 24 and 26 GHz.Three measurements of both the antenna-under-test and the reference horn were performed to minimize the uncertainty; a 1.3% standard deviation was observed.The measured S 11 of the connectorized antenna was used to calculate the radiation efficiency excluding mismatch.
Using the measured efficiency, the antenna's measured gain is 8.2 dBi at 26 GHz.its dual-resonant structures.To compare the efficiency with 330 reported mmWave antennas, the measured efficiency was 331 normalized to the substrate's electrical thickness using where t is the antenna's thickness and η rad. is the measured 333 antenna efficiency.This metric allows the evaluation of the 334 antenna's efficiency with respect to its volume, whereas the 335 aperture efficiency only considers the area [37].

336
Compared to the textile antennas in [38] and [5], operating 337 at 60 GHz, the proposed antenna achieves a higher radiation 338 efficiency while having a lower profile (compared to λ 0 ).339 While [17] achieves a higher efficiency using a reflector-340 backed textile antipodal antenna, which increases the thickness 341 and complexity of the antenna.However, the proposed antenna 342 compares favorably when η norm. is considered, as visualized 343 in Fig. 11.

344
Compared to the multi-mode patch in [24], the proposed 345 antenna compares favorably when the antenna efficiency is 346 normalized to the thickness, despite being implemented on 347 textiles and not RF laminates.In addition, the wider bandwidth 348   This work [7] [24] [17] Space in [24] was achieved using a Substrate-Integrated Waveguide 349 (SIW) cavity and a very thick (0.11λ 0 ) substrate.While broadband "wire-type" antennas such as [36] achieve higher   In section IV-B, the wireless power harvesting capabilities of a large-area rectenna system based on the proposed antenna is evaluated and compared to a UHF off-body rectenna.

A. Compact Single-Receiver
Higher power can be delivered to a compact receiver in mmWave bands due to the comparable physical aperture size of the antenna to the wavelength.In [4], it was analytically shown that for antennas of a fixed area, the power received increases with the frequency.This section demonstrates the benefits of using the proposed antenna for LoS WPT for a transmitter and receiver with area-constrained antennas.
Two symmetric textile patch antennas were connected to a Vector Network Analyzer (VNA)'s ports to measure the forward transmission between them in space, as shown in Fig. 12.As the proposed antenna is only matched for mmWave bands and will have a very high reflection coefficient for UHF bands, the measured forward transmission has been corrected post-measurement to exclude the impedance mismatch losses at 2.4 GHz.By positioning the antennas at d=50 cm, d is equal to 4×λ 2.4GHz ensuring the antennas are in the far-field.Fig. 13 shows the measured forward S 21 from 0.5 to 4 GHz and from 22 to 34 GHz.It can be seen that due to the antenna's improved gain and efficiency in mmWave bands, it is possible to achieve over 11 dB higher S 21 at 26.5 GHz compared to 2.4 GHz, despite the increased path loss, even after excluding the mismatch losses at 2.4 GHz.The additional advantage of using patch antennas for WPT in mmWave bands is the ability to implement the off-body patch on a lowprofile substrate, as shown in Fig. 11.Moreover, while the impedance mismatch losses were excluded at 2.4 GHz for a fair comparison, achieving a low S 11 at 2.4 GHz with such a compact antenna (0.06×0.07λ 2 2.4GHz ) will require a matching generation [42].The analytical diode PCE formulas proposed 418 in [42] are used as a starting point for the PCE evaluation.

419
However, as discussed in [43], such analytical closed form Balance (HB) simulation is more suited.Therefore, in our comparison between mmWave and UHF WPT performance, HB results are used for a more conservative estimate.
The diode considered in this work is the VDI W-ZBD GaAs Schottky diode, used for low-power rectification up to 100 GHz for its low forward voltage and parasitic capacitance [44].Fig. 14-c shows the calculated PCE for the rectifier using the analytical formulas from [42] and HB simulation.The peak PCE of 77% observed in the HB model is within 10% of the reported PCE at 36 GHz in [45], based on commercial Schottky diodes [4].
Two transmitters are considered: a 28 GHz 53 dBm Equivalent Isotropically Radiated Power (EIRP) and an 892 MHz (the center resonant frequency of [46]) 36 dBm EIRP.These are the maximum permissible EIRP levels for both bands.In mmWave bands, the maximum EIRP is higher than sub-5 GHz bands.For example, an EIRP of 75 dBm is permitted for a 5G basestation, where the higher EIRP is achieved using high-gain phased arrays [47].The 53 dBm EIRP could be realized using a 1 W transmitter and a 23 dBi antenna, which is permissible for license-free bands above 5 GHz [48].Below 5 GHz, the EIRP limit is capped at 4 W (approximately 36 dBm).From a practical implementation perspective, the antenna's theoretical minimum physical aperture area A Phys. can be calculated from the aperture efficiency e a using for a fixed e a and a target antenna gain G (dimensionless) [49].Assuming a 90% aperture efficiency, typical for radiating aperture antennas [49], a 23 dB antenna at 28 GHz would occupy an area of 2000 mm 2 .This represents 95% less area than a 6 dBi antenna operating at 900 MHz.Therefore, utilizing mmWave wireless charging base-stations promises reduced base-station antenna size allowing more pervasive deployment in future IoT micro-cells.
Multiple array sizes based on the proposed antenna are considered, with a 5.4 mm (λ/2 at 28 GHz) spacing between the array elements.The power harvested by the rectennas is calculated using an empirical propagation model.The UHF  of the patch, it is an accurate representation of the antenna's performance as a wearable wireless power receiver.
Despite the increased path losses, the lower PCE, and the empirical path loss exponent, for both the LoS and N-LoS cases, higher power can be delivered to the wearable receivers at 28 GHz compared to 892 MHz.This demonstrates that body-centric WPT and Radio Frequency Energy Harvesting (RFEH) from an off-body source can achieve improved endto-end efficiency due to the smaller size of the antennas.Table II compares the power received by the mmWave and UHF antennas calculated using (5).The received DC power density is calculated using P RX /A Phys. .Owing to the higher permissive EIRP and the compactness of the mmWave patch, the received power density is higher than that at UHF.Therefore, the large-area implementation can harvest up to 6.3× more power than the UHF rectenna occupying the same area.In addition, it is observed that despite the higher path-loss exponent in mmWave N-LoS propagation, higher power can still be harvested up to 3.5 m at 28 GHz, based on the proposed textile antenna.

C. mmWave WPT in IoT Applications
For a compact device, it was found that the proposed mmWave textile antenna will outperform its counterpart operating at 2.4 GHz by 11 dB in terms of the channel gain, implying a 10-fold improvement in the received power.In the large-area scenario, a similar sized receiver can harvest p to 6× higher power than its similar-sized sub-1 GHz counterpart.Several recent WPT applications can benefit from the proposed

Fig. 5 .
Fig. 5. Comparison of the proposed patch with a common geometry (CG) TM 10 patch: (a) bandwidth; (b) radiation pattern, where the shaded red area indicates the angular coverage improvement.

4 IEEEFig. 6 .Fig. 7 .
Fig. 6.Simulation of the patch antenna on the body model: (a) body phantom structure, (b) simulated SAR at 28 GHz for a 0.5 W input.

Fig. 9 .
Fig. 9. Measured 3D directivity D and polarization of the proposed antenna, on a large ground plane, at 26 GHz.
This article has been accepted for publication in IEEE Internet of Things Journal.This is the author's version which has not been fully edited and content may change prior to final publication.Citation information: DOI 10.1109/JIOT.2021.3107594

Fig. 11 .
Fig. 11.Comparison of recent microstrip antennas based on the thicknessnormalized efficiency ηnorm., and the fractional bandwidth.

351
efficiency and bandwidth on low-loss substrates, the antenna 352 requires double-sided etching and the efficiency is only achiev-353 able with tanδ<0.001substrates.Other printed TM 10 patches, 354 [39], [40], implemented on low-cost substrates or using printed 355 conductors achieve a narrower bandwidth and a lower radiation 356 efficiency than the proposed antenna.
357 IV.MMWAVE WIRELESS-POWERED BAN EVALUATION 358 The textile patch antenna's radiation properties are uti-359 lized to analytically evaluate the performance of a mmWave-360 powered BAN.In this section, we compare the WPT efficiency 361 based on the proposed patch to a similar sized off-body an-362 tenna based on the same area, in Section IV-A, demonstrating 363 that a higher-efficiency WPT link can be achieved at mmWave 364 bands compared to UHF, when the antennas' area is restricted. 365

420Fig. 14 .
Fig. 14.The large-area mmWave power receiver: (a) textile-based rectenna array with directional transmitter and ideal DC combining, (b) a single rectenna cell, (c) analytically calculated and simulated PCE of the rectifier at 28 GHz.

Fig. 15 .P RX has been calculated using 478 PFig. 16 .
Fig. 15.DC power harvested by different-sized arrays, based on the measured antenna gain and empirical path loss model, at 28 GHz: (a) LoS, (b) N-LoS.
-150 a shows the layout of the first TM 02 patch iteration.The 151 simulated S 11 , obtained from CST Microwave Studio based 152 on the substrate's measured properties ( r =1.95, tanδ=0.026),153 is shown in Fig. 1.

TABLE I COMPARISON
OF THE PROPOSED TEXTILE ANTENNA WITH REPORTED PLANAR MMWAVE MICROSTRIP ANTENNAS.
* Simulated result; † right/left-hand circular polarization isolation (estimated from the patterns).

TABLE II TEXTILE
-BASED LOS MMWAVE AND UHF WPT SUMMARY.