A Wide-Scanning Metasurface Antenna Array for 5G Millimeter-Wave Communication Devices

In this paper we present a high-performance compact phased array antenna which is easy to integrate into mobile devices for 5G-and-beyond wireless telecommunications. The proposed design features high efficiency and wide-scan capabilities. The linear array consists of eight elements realized using substrate integrated waveguide technology in combination with two rows of metasurfaces that are used to optimize the transition towards free space for enhanced impedance matching characteristics. The integrated metasurface structure also enables a larger half-power beamwidth and wide-angle scanning at array level. A prototype has been realized using a dielectric substrate of Rogers RO4003C with relative permittivity of 3.55. The array is designed with an inter-element spacing of half-wavelength at 29.5 GHz and is characterized using dedicated millimeter-wave anechoic and reverberation chambers. The measurement results show that the proposed antenna array can scan from $\phi = -55^{\circ }$ to $\phi = 55^{\circ }$ with a gain fluctuation less than 3 dB in the frequency band of operation from 27 GHz to 29.5 GHz, and a measured total efficiency above 70 % with an uncertainty of 10% (95% confidence interval). Furthermore, when compared to the state-of-the-art, the proposed antenna provides a much wider scanning range while occupying a significantly smaller and compact volume.

the same system and the limited space typically available on 27 printed circuit boards (PCBs) embedded in mobile devices 28 makes the integration of multiple phased arrays challenging, 29 though this is necessary to perform adaptive beamforming to 30 guarantee full radio coverage. 31 In this context, one should notice that the integration of 32 a large number of antenna elements in a limited portion of 33 space poses problems in terms of power consumption and 34 thermal management [1]. In this regard, an important role is 35 played by the array beamforming network which could lead 36 to extra insertion and radiation losses. Therefore, embedding 37 the antenna array directly in the PCB of a 5G cellular device 38 is useful to minimize the insertion loss between the antenna 39 elements and the radio-frequency integrated circuits. 95 %, the realized antenna array efficiency [7]. As a matter of 80 fact, one can notice that, in available scientific publications, 81 the reported antenna efficiency levels at mm-wave frequen-82 cies are often based on numerical estimations only.

83
The paper is divided in three sections. In Section II, the 84 geometry of the developed antenna element is presented. 85 In Section III, the simulated array performance and the rele-86 vant measurement results are discussed. Finally, the conclud-87 ing remarks are summarized in Section IV.

89
The geometry of the proposed metasurface-based end-fire 90 antenna is shown in Fig. 2.The bottom layer is hidden since 91 it is identical to the top layer. Two rows of metal patches, 92 on both the top and bottom metal layers, are integrated in such 93 a way to optimize the impedance transition from the SIW-94 based open-ended waveguide structure to free space [8].

95
In the design of an open ended SIW, the main challenge 96 is to match the characteristic impedance Z s of the structure 97 to the free-space impedance Z air . Typically, Z air 377 98 is much larger than Z s . Therefore, strong reflections occur 99 at the antenna aperture unless suitable design solutions are 100 implemented.

101
The easiest way to improve impedance matching is to load 102 the antenna with a dielectric material by extending the inner 103 core of the substrate across the array aperture. In this way, the 104 impedance level at the aperture can be modified by changing 105 the realized effective permittivity eff , as one can infer from 106 the general relation between impedance and permittivity Z = 107 √ µ 0 / 0 eff , assuming that non-magnetic materials are being 108 used.

109
In [9], the transition at array aperture is further improved by 110 realizing perforations in the dielectric substrate. The diameter 111 of such perforations increases with the distance from the 112 aperture to synthesize smaller values of eff and, thereby, 113 a higher impedance. A more extensive review of end-fire 114 antennas in SIW technology can be found in [10]. 115 In this study, to enhance the impedance matching at the 116 transition, we propose the integration of a suitable meta-117 surface in combination with the basic SIW structure. Such 118 a metasurface consists of metal patches printed on the 119 extension of the substrate core at the array aperture (see 120 VOLUME 10, 2022   (1) where c 0 is the speed of the light in free space.

132
More rows of metasurface elements can be used to make 133 the transition to free space smoother. In literature, metasur-134 faces with multiple rows of elements are often adopted in 135 such a way to enhance the antenna gain [13]. In our study, 136 however, we make use of two rows only to mitigate the peak 137  gain of the individual array element while increasing the 138 HPBW. Fig. 6 shows the normalized radiation pattern of the 139 central antenna element embedded in the array structure with 140 and without metasurfaces. One can notice that the integration 141 of the metasurfaces significantly enhances (nearly doubling) 142 the HPBW of the basic antenna element. At the same time, 143 a considerable improvement is achieved in reducing the cou-144 pling with the adjacent elements, as shown in Fig. 6c, with a 145 similar decoupling mechanism as shown in [14]. This, in turn, 146 is useful to enable wide-angle scanning capabilities at array 147 level.

148
The dimensions of the metal patches forming the metasur-149 face can be tuned to control the impedance matching charac-150 teristics of the antenna and the center frequency of operation. 151 The working principle of the considered metasurface is based 152 on the observation that the current flow in the y− direction, 153 along the metal patches and across the relevant gaps, can 154 be modeled, at circuit level, as a series of inductors and 155 capacitors, respectively [5]. Therefore, the length of the metal 156 patches and the gap between them have a tremendous impact 157 on the achievable impedance matching characteristics of the 158 antenna structure.

159
The proposed antenna element is fed by a mini-SMP con-160 nector type 18S101-5H0E4. The mini-SMP connector is only 161 used for passive testing, but it is not required when the anten-162 nas are directly connected to beamforming integrated circuits 163 (ICs). The feeding pin of the mini-SMP connector is inserted 164 in an unplated via hole with a diameter equal to 0.4 mm, while 165 the outer metal surface of the connector is soldered on top 166 of the SIW structure. To prevent risk of short-circuiting, the 167 bottom and top layers of the SIW include a circular clearance 168 with a diameter of 1.1 mm around the feeding pin.

169
To determine the optimal dimensions of the metasurface 170 for enhanced impedance matching, a dedicated parameter 171 study has been carried out using CST Microwave Studio .

188
The characterization of the array performance has been car-  more accurate results as compared to an anechoic chamber 199 when it comes to power-based metrics such as efficiency [15]. 200 Besides that, the reverberation chamber allows for filtering 201 out unwanted radiation sources from other parts of the system 202 and, in this way, evaluate the actual radiation characteristics 203 of the antenna array under test, so overcoming an intrinsic 204 limitation of anechoic chambers [17]. 205 The measured and simulated S-parameter results are shown 206 in Fig. 8. To improve readability, only the S-parameters 207 relevant to the central antenna elements have been plot-208 ted. In spite of the narrower bandwidth (with respect to 209 the return-loss level of 10 dB), the measured reflection 210 coefficient is in good agreement with the simulated one. 211 Please note that the S-parameters have been evaluated after 212 de-embedding of the effect of the passive components 213 between the VNA cable and the antenna port using the 214 method described in [16]. The coupling between adjacent 215 elements is below −20 dB.

216
The radiation pattern measurement setup is shown in Fig. 5 217 where the antenna array is placed at the center of a mm-wave 218 anechoic chamber which uses a horn antenna as a probe [6], 219 that can be oriented to measure the co-or cross-polarized 220 component of the radiated electromagnetic field (in cross-221 polarization in Fig. 5). A spherical scan was performed 222 in the far-field, where the radiation patterns in both the 223 E-and H -plane were measured in co-polarization and 224    by combining the different contributions from each port in 246 post-processing. A peak gain of 13.1 dBi has been simulated 247 with a side lobe level below −10 dB.

248
For a more extensive characterization of the scanning capa-249 bilities of the array antenna, the active reflection coefficient is 250 determined using the de-embedded S-parameters. The results 251 are shown in Fig. 11, only for the scan range from φ = 0 • to 252      Table 2, which are, 279 in this case, all Type A uncertainties [19]. In this measure-280 ment, the RF beamformer quality is the most significant 281 uncertainty contributor, since it has some unwanted radiation, 282 and the PCB losses vary across the different traces. The 283 latter is extracted from the variation across insertion loss 284 between and is included as 'PCB Loss Splitter' in Table 2. 285 The variation in mismatch between the ports is extracted from 286 the measured reflection coefficients in the same measure-287 ments.The measurement repeatability was estimated by car-288 rying out the measurements nine times for different antenna 289 positions and calibrations, and the same approach was used 290 to estimate the uncertainty due to unwanted radiation [20]. 291 The expanded uncertainty was estimated by using the root-292 sum-square approach as dictated by the CTIA standardized 293 test plan [21], expanded with a coverage factor (k p ) of 1.96. 294 We state that the measurements are in good agreement with 295 the simulations when the simulation results lie within the 296 uncertainty bounds of the measurements, meaning they are 297 statistically not significantly different.

298
The simulated and measured results for the total efficiency 299 are shown in Fig. 12, the latter reporting the best estimate 300 taken from the mean of the nine measurements. The simulated 301 efficiency values are slightly larger than the measured ones, 302 but always within the 95 % confidence interval of the rever-303 beration chamber. While the efficiency was not measured 304 for different scanning angles, it should be noted that the 305 simulated total efficiency is above 70 % for the entire scan 306 range.

307
The array design presented in this manuscript is compared 308 to state-of-the-art solutions already proposed in the scientific 309 literature for 5G mobile communication and based on dif-310 ferent antenna technologies with end-fire radiation pattern. 311 The benchmarking is performed on the basis of the figures of 312 merit (FoM) reported in Table 3. [26] are characterized by a scanning range which is limited 315 to ±40 • or less. Conversely, the array solution detailed in 316 this research study features measurably wider scanning capa-317 bilities which, in turn, can enable a full 3D radio coverage 318 by integration of a reduced number of modules. Further-319 more, the proposed array is more compact. Note that the 320 long transmission lines embedded in the physical prototype 321 shown in Fig. 4 are needed only for testing purposes. That 322 being said, such transmission lines in SIW technology are 323 completely shielded from interference and the top and bottom 324 metal layers can be used for accommodating, for example, 325 batteries or other suitable components. Most of the antenna 326 arrays listed in Table 3 are instead considerably larger than 327 the one presented in this work, especially when considering 328 the mobile-device application. Furthermore, the proposed 329 solution is easily scalable to a larger or a smaller number of 330 antenna elements, based on specific requirements.

332
In this paper, a metasurface-based linear array of end-fire 333 like antennas with wide-angle scanning capabilities has been 334 presented. The radiating structure is implemented in SIW 335 technology to enable an easy integration into mainstream 336 PCBs, for use in smartphones or other mobile devices. Fur-337 thermore, a metasurface is used to optimize the impedance 338 transformation from the open-ended dielectric-loaded SIW 339   Professor with the Group of Electromagnetics in Wireless Telecommunica-524 tions, Eindhoven University of Technology, Eindhoven, since 2020. He has 525 authored or coauthored more than 200 publications in international peer-526 reviewed journals, book chapters, and conference proceedings. He holds 527 22 families of patents in antenna-related technologies and advanced com-528 putational techniques. His current research interests include the full-wave 529 analysis and design of passive devices and antennas for satellite, wireless, 530 and radar applications, the development of analytically based numerical 531 techniques devoted to the modeling of wave propagation and diffraction 532 processes, the theory of special functions for electromagnetics, the determin-533 istic synthesis of sparse antenna arrays, and the solution of boundary-value 534 problems for partial differential equations of mathematical physics. He is 535 currently a member of the Applied Computational Electromagnetics Society 536 (ACES), the Institution of Engineering and Technology (IET), the Interna-537 tional Union of Radio Science (URSI), and the Italian Electromagnetic Soci-538 ety (SIEm). He was a recipient of the Young Antenna Engineer Prize at the 539