*Communication* **Characterization of Tiled Architecture for C-Band 1-Bit Beam-Steering Transmitarray**

**Dmitry Kozlov 1, Irina Munina 2, Pavel Turalchuk 2, Vitalii Kirillov 2, Alexey Shitvov <sup>3</sup> and Dmitry Zelenchuk 4,\***

<sup>1</sup> Nokia Bell-Labs, D15Y6NT Dublin, Ireland; dmitry.1.kozlov@nokia.com

<sup>2</sup> Department of Microelectronics and Radio Engineering, St. Petersburg Electrotechnical University LETI, 197376 St. Petersburg, Russia; ivmunina@etu.ru (I.M.); paturalchuk@etu.ru (P.T.); vvkirillov@etu.ru (V.K.)


**Abstract:** A new implementation of a beam-steering transmitarray is proposed based on the tiled array architecture. Each pixel of the transmitarray is manufactured as a standalone unit which can be hard-wired for specific transmission characteristics. A set of complementary units, providing reciprocal phase-shifts, can be assembled in a prescribed spatial phase-modulation pattern to perform beam steering and beam forming in a broad spatial range. A compact circuit model of the tiled unit cell is proposed and characterized with full-wave electromagnetic simulations. Waveguide measurements of a prototype unit cell have been carried out. A design example of a tiled 10 × 10-element 1-bit beam-steering transmitarray is presented and its performance benchmarked against the conventional single-panel, i.e., unibody, counterpart. Prototypes of the tiled and single-panel C-band transmitarrays have been fabricated and tested, demonstrating their close performance, good agreement with simulations and a weak effect of fabrication tolerances. The proposed transmitarray antenna configuration has great potential for fifth-generation (5G) communication systems.

**Keywords:** antenna array; antenna measurements; beam pattern; beam steering; equivalent circuit modelling; transmitarray

#### **1. Introduction**

Emerging architectures of the fifth-generation (5G) new radio communication systems employ complementary use of both sub-6 GHz and beyond 24 GHz spectrum regions, whereby, in outdoor scenarios, the low-frequency bands are envisioned to provide wide uniform coverage, whereas the millimetre-wave radio would allow directed ultra-high throughput within the wide sub-6 GHz coverage area. Moreover, although millimeterwave propagation channels exhibit many peculiar features, which may even call for the use of quasi-optical analysis and design techniques, some advanced communication principles and system architectures, primarily aimed at millimetre-wave frequencies, can be implemented and verified with the aid of low-frequency proof-of-concept prototypes.

Multiple-antenna millimeter-wave radio systems, commonly referred to as multipleinput-multiple-output (MIMO) architecture with a large number of antenna elements at the radio access nodes and user terminals enable spatial multiplexing and diversity by means of intelligent beamforming. The latter feature seems to be an indispensable attribute of the 5G communication and radar systems, alongside the exploitation of unconventional degrees of freedom in radio propagation.

Although fully digital beamforming in massive MIMO systems can, in theory, achieve optimal performance, the current state of the digital hardware makes this approach unfeasible for millimeter-wave radio, due to prohibitively high cost and as yet insufficient

**Citation:** Kozlov, D.; Munina, I.; Turalchuk, P.; Kirillov, V.; Shitvov, A.; Zelenchuk, D. Characterization of Tiled Architecture for C-Band 1-Bit Beam-Steering Transmitarray. *Sensors* **2021**, *21*, 1259. https://doi.org/ 10.3390/s21041259

Academic Editor: Naser Ojaroudi Parchin Received: 31 December 2020 Accepted: 6 February 2021 Published: 10 February 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

resolution of the analog/digital-to-digital/analog converters, [1]. On the other hand, fully analog beamforming does not provide essential flexibility in design. In the course of previous studies, it appeared that millimeter-wave channels typically have much less degrees of freedom than can be achieved with fully digital beamforming, thus making the latter redundant. Therefore, many hybrid architectures have emerged recently, aimed to efficiently exploit the sparsity of millimeter-wave channels by combining the key features of both beamforming approaches to achieve optimal performance in applications at reduced complexity and cost.

In particular, the use of refractive dielectric lenses and focusing arrays proved to be technologically advantageous and economically efficient. The use of intelligent reflecting and transmitting surfaces, [1,2], including multi-beam transmitarrays, flat and hybrid lenses, impedance-modulated holographic surfaces, programmable metasurfaces with arbitrary control of the propagated wavefronts, all of which can be realized in conventional planar multi-layer technology using either non-linear materials or surface-mount RF components, opened new avenues in the design of millimeter-wave communication and sensing systems.

Recently, the feasibility of low-bit beam-steering and phase-only beamforming has been demonstrated as a means of further cost-reduction, [3,4]. Beam-switching at the focal-plane array has also been found to be a useful feature for millimeter-wave compact small-cell architectures, [5,6]. A number of different electronically controlled transmitarray architectures have reported recently for applications from C-band to V-band, with various performance functional from merely beam collimation to wide-angle beam-steering, beamforming and complete wavefront and polarization control. A 28-GHz circularly-polarized reconfigurable transmitarray comprising 400 binary phase unit-cells of receiver-transmitter type with an integrated phase-switch network was experimentally demonstrated in [7] as an attractive solution for many applications operating in Ka-band, such as satellite communications, point-to-point links and heterogeneous wireless networks. The use of a co-designed slot-array focal source antenna enabled a significant reduction of the antenna profile. An X-band electronically reconfigurable transmitarray with enhanced transmission bandwidth and efficiency achieved by using new contactless probe-feeding of the antenna patch was demonstrated in [8], aiming at advanced communication applications. A successful attempt to extend the application of low-cost transmitarrays to V-band was experimentally demonstrated in [9], although no electronic control was available at the time for two-dimensional beam-steering. Most of the above concepts have been demonstrated using integrated transmitarrays fabricated in planar printed-circuit technology. However, fabricating large single-panel transmitarrays raises the cost of proof-of-concept prototyping and makes the technology unaffordable for teaching laboratories.

Our research is aimed at adopting the transmitarray architecture for MIMO communications in C-band. In our previous publications [10,11], we reported on a low-frequency prototype of novel 1-bit dual-polarized tiled transmitarray, whereby the required phase distribution across the array aperture was built from standalone unit cells manufactured individually and assembled in the required pattern using a rectangular latticed plastic frame, Figure 1. Some preliminary simulation and measurement results were presented, and it appeared that the tiled architecture can be a viable solution for fast prototyping and teaching experiments, without significant performance deterioration, as compared with a similar single-panel transmitarray. Moreover, the possibility of replacing and adding individual elements in the tiled array makes it both repairable and adjustable for a specific focal distance and feed type. This paper revisits previous simulations and provides new results of modelling and experimental characterization of the tiled transmitarray.

**Figure 1.** One-bit dual-polarized tiled transmitarray architecture: (**a**) design of the array tile (vertically exploded view), comprising two identical proximity-coupled square-ring radiators on the opposite sides of the tile connected via two U-shaped feed loops; (**b**,**c**) schematic view of the surface currents on the proximity coupled feed loops and square-ring patches in two phase states, respectively (the ground plane is not shown); (**d**) a section of the tiled transmitarray partially assembled; (**e**) a section of the integral single-panel transmitarray; (**f**) proposed device architecture, [10], including individual unit cells (different colors indicate one of the two phase states) to be mounted in the plastic grid frame and spatially fed by a focal-source patch antenna.

#### **2. Transmitarray Model**

The model of a transmitarray, first presented by the authors in [10], is given below for consistency. In a spatially phase-modulated transmitarray, the normalized wave amplitude received by a unit cell from the focal source reads:

$$a\_{mn} = \frac{\lambda e^{jkR\_{mn}}}{4\pi r R\_{mn}} \mathbf{F}^{fs}\_{mn} \cdot \mathbf{F}^{ncr}\_{mn} \tag{1}$$

where *m* = 1, 2, ... , *M* and *n* = 1, 2, ... , *N* are the row and column indexes of the array which define the position of the unit cell with respect to the reference one, *k* and *λ* are free-space wavenumber and wavelength, **F***f s mn* is the complex vector field pattern of the focal-plane source transmitting in the direction of the unit cell defined by the corresponding polar and azimuthal angles of the local coordinate system (CS) with the origin at the focal point, **F***ucr mn* is that of the unit cell on receive in the direction of the focal plane source defined by the respective angles of the local CS with the origin at the center of the unit cell (note that the antenna pattern on receive is conjugate of that on transmit due to the reciprocity), the dot symbol denotes the Hermitian inner product of the two complex vector patterns, and *Rmn* is the distance between the focal-source and unit-cell CSs. The unit cells are assumed to be matched to the incoming wave at all angles of incidence determined by the angular aperture of the transmitarray. The effects of the element coupling and finite aperture of the transmitarray can, in principle, be accounted for in the unit-cell antenna patterns by infinite array analysis, [12], or embedded element technique, [13].

The focal-source and unit-cell antenna patterns, in the case of linear polarization, reduce to scalar-valued functions. After sampling and retardation of the incident spherical wavefront, the complex amplitude antenna pattern of the transmitarray, *F*(*θ*, *φ*), can be calculated by the pattern multiplication principle, as follows:

$$F(\theta,\phi) = \sum\_{m=1}^{M} \sum\_{n=1}^{N} b\_{mn} F\_{mn}^{uc}(\theta,\phi) e^{l\Psi\_{mn}^{uc}(\theta,\phi)} e^{lkl(m\sin\theta\cos\phi + n\sin\theta\sin\phi)}\tag{2}$$

where *Fuc mn*(*θ*, *φ*) and *ψuc mn*(*θ*, *φ*) are, respectively, the unit-cell amplitude and phase patterns on transmit, *θ* and *φ* are azimuthal and polar angles in the spherical CS with the origin at the center of the transmitarray aperture and the polar direction aligned with the transmitarray optical axis, *bmn* = *Tmnamn*-complex amplitudes of the waves radiated by each unit cell, and *Tmn*-the corresponding complex transmission coefficients. Equation (2) enables accounting for the effect of the finite array on the standalone pattern of the element, [14]. Also, the unit-cell radiation pattern is assumed to be independent of the transmission coefficient, i.e., of the specific phase shift for the phase-modulated transmitarray.

The above model can be adopted for the design of the proposed tiled transmitarray by suitably adjusting the unit-cell transmission coefficients for given focal-source and unitcell antenna patterns. In transmitarray antennas, beam steering is achieved by spatially modulating the phase distribution of the emitted wavefront across the array aperture, as follows:

$$\arg(b\_{mn}) = -k \operatorname{\boldsymbol{\tau}}\_{\text{s}} \cdot \operatorname{\boldsymbol{\tau}}\_{mm\text{\textquotedblleft}} \tag{3}$$

where *<sup>r</sup>s*(*θs*, *<sup>φ</sup>s*) is the unit vector in the beam-steering direction (*θs*, *<sup>φ</sup>s*), while the array vector *<sup>r</sup>mn* <sup>=</sup> (*xmn*, *ymn*, 0) comprises the coordinates of the unit cell. For symmetrical unit cells, the required continuous local phase shift follows from (1) and (2) as:

$$\arg(T\_{mm}) = \arg(b\_{mm}) - \psi\_{mm}^{fs} + kR\_{mm} - \psi\_{mm}^{uc} \tag{4}$$

where *ψf s mn* is the focal-source phase pattern in the direction of the unit cell (NB: typically, the phase pattern, with respect to the phase center of the antenna, is nearly flat within the angular range of the main lobe). In the proposed 1-bit transmitarray, the phase distribution (4) is discretized according to the following recipe (shown for the wrapped phase):

$$\arg \left( T\_{mn}^d \right) = \begin{cases} \begin{array}{c} 0^\circ \,\forall |\text{arg}(T\_{mn})| \le 90^\circ\\ 180^\circ \,\text{otherwise} \end{array} \,, \tag{5} $$

It is important to note that the effect of the 1-bit phase quantization on radiation characteristics was analyzed in [11,15]. It was shown that a 1-bit resolution results in the antenna gain reduction of up to 4 dB, higher sidelobe level and noticeable beam squint.

#### **3. Unit Cell Design and Characterization**

The detailed description of the unit cell design and preliminary results of the measurements inside the rectangular waveguide were reported in [10]. This unit cell structure has been employed in the current study. It is noteworthy that the proposed unit cell structure was conceived as a blank of a reconfigurable pixel of single-panel transmitarrays, using surface-mount solid-state switches to add functionality. However, in the context of the tiled transmitarray, power routing is much more challenging and thus is not addressed in this work.

The unit cell design, first reported in [10], was implemented in a stacked 6-layer structure, Figure 1a. The receiving and transmitting antennas were represented by squarering microstrip elements with electromagnetic (proximity coupled) feeds in the form of open-ended half-wavelength semi-annular (U-shaped) microstrip loops in the layer beneath the square-ring antenna. The proximity coupling allowed a wider bandwidth when the feed loop and ring were properly aligned, [16]. The track widths of the square ring and

feed loops were numerically optimized for the maximum return-loss bandwidth and low insertion loss, using CST Microwave Studio simulations with Floquet periodic boundary conditions (FPBCs) and assuming infinite ground plane. The pair of loop resonators were connected to each other by a buried via hole.

The receiving and transmitting sides of the tiled unit cell were separated by two ground plane electrodes bonded together using a 0.2 mm layer of Rogers RO4350B and protruded by the buried vias. The redundancy of the two ground planes was imposed by the manufacturing process. The metallic patterns of the square-ring radiators and feed loops were formed on and between dielectric layers of 0.51 mm thick Rogers RO4003 material (dielectric constant Dk = 3.5, dissipation factor Df = 0.0018). The layers of 0.1 mm bonding film Rogers RO4003C (Dk = 3.38) were used to stack the RO4003 layers. The lateral size of the unit cell of the single-panel transmitarray was 24 mm × 24 mm (~0.46 *λ* at the design frequency of 5.75 GHz) and its thickness was <0.045 *λ*. The tiled unit cells were trimmed by 0.5 mm around the edge in order to keep the same array period in both single-panel and tiled transmitarrays.

In the proposed unit-cell design, a 180◦ phase shift is implemented by switching the feed point of the U-shaped resonator on the receiving side of the transmitarray, Figure 1b,c. The state when the resonators at the receiving and transmitting sides are connected such that the currents flowing in the patches are codirectional is referred to as the phase state I (or 0◦ state). In the reciprocal phase state II, the resonators are connected at the opposite ends, so that the currents flow in the opposite directions thus imparting a 180◦ phase shift with respect to the phase state I. Two pairs of feed loops are used on each side of the structure, placed orthogonal to each other so that the unit cell can support two orthogonal linear polarizations for each phase state. The transmission and reflection coefficients measured in the waveguide were similar in both phase states and for both polarizations. The 10 dB return-loss bandwidth spanned 160 MHz from 5.67 to 5.83 GHz. The differential phase error did not exceed ±6◦ across the operating band.

The transmitarray design approach adopted in our study is based upon the unit cell characterization in terms of insertion loss and differential phase shift (i.e., the phase shift in one phase state with respect to the other)-numerical with full-wave electromagnetic simulations (CST Microwave Studio), as well as experimental inside a rectangular waveguide. The tiled transmitarray has slotted dielectric substrate and ground plane, as well as additional dielectric frame 3D-printed in ABS (acrylonitrile butadiene styrene, Dk = 2.35 as measured), see Figure 1d, necessary to arrange the tiles in desired planar phase pattern. Thus, the effect of the discontinuity, i.e., the width of the gap between the adjacent unit cells, on the tile radiation performance is inherent to the design of the tiled transmitarray and we aimed to minimize its impact within the operating band.

The effect can be elucidated with the aid of the compact circuit model of the tiled unit cell shown in Figure 2a. It is noteworthy that the circuit model is loosely related to the actual geometry of the unit cell and it is derived essentially by emulating the bandpass response of the unit cell in the two phase-states. Nevertheless, the compact model provides useful insights on the interactions of different parts of the unit cell structure.

**Figure 2.** Equivalent-circuit modelling of the unit cells of the single-panel (*g* = 0 mm) and tiled (*g* = 1 or 2 mm) transmitarrays: (**a**) network topology (left) and compact electrical circuit model (right); (**b**) comparison of the circuit model with full-wave electromagnetic simulations (CST Microwave Studio) for the phase state I; (**c**) same as (**b**) but for the phase state II; (**d**) effect of the edge gap width, *g*, on the differential phase shift.

The circuit model topology constitutes a canonical parallel–parallel connection of the cascaded two-ports, Figure 2a. Being reduced to equivalent elements, the circuit comprises two parallel RLC-circuits (*Rp*, *Lp*, and *Cp*) associated with the receiving and emitting square-ring patches loaded by the respective U-shaped resonators and coupled via the two ideal admittance inverters, *J*<sup>0</sup> and *Jg*, with characteristic admittances *Y*<sup>0</sup> and *Yg*, respectively. The circuit model differs for the two phase states, due to the opposite direction of the current flowing on the receiving patch and this difference is implemented by changing the sign of the inverter admittance *Y*0, with its positive value corresponding to the phase state I and negative to the phase state II. The *J*0-inverter represents the primary coupling of the patches by the via connection, see Figure 1a. The effects of the edge gap are modelled by the additional inverter with a characteristic admittance *Yg*, which can accurately model the out-of-band transmission zeros, see Figure 2c.

Putting *Rp* to zero, it can be shown that the transmission zeros appear at frequencies where the following condition fulfils:

$$\Upsilon\_0 = -\Upsilon\_\emptyset / \left(1 - \Upsilon\_\emptyset^2 Z\_p^2\right) ,\tag{6}$$

where *Zp* is the complex impedance of the parallel LC circuit.

The resonance nulls in (6) can appear only when *Y*<sup>0</sup> and *Yg* are in phase, due to the negative sign of the denominator in the vicinity of the resonant frequency of the LC circuit. This condition determines the out-of-band 180◦ steps of the differential phase shifts, demonstrated in Figure 2d when the differential circuit mode with the opposite direction of the currents on the receiving and emitting patches is superseded by the common mode that is driven by the floating ground plane.

The model parameters are shown in Table 1. The parameters were extracted by bestfitting to the full-wave electromagnetic simulations, as follows. Firstly, the initial 'patch' circuit parameters *Rp*, *Lp*, and *Cp* were fitted using the full-wave simulation of the reflection (S11) for the rectangular patch over an infinite ground plane and simplified circuit model without the inverters. In the second step, characteristic admittance of the *Y*0-inverter is extracted by fitting to the full-wave simulations of transmission (S21) of the single-panel transmitarray unit cell (i.e., *g* = 0 mm). Finally, characteristic admittance of the *Jg*-inverter is obtained by fitting the model to the full-wave simulations of S21 of the two square-ring patches coupled only through the slotted ground plane, i.e., in the absence of the *J*0-inverter. It appeared that the absolute value of the characteristic admittance *Yg* decreases for the wider gap.

**Table 1.** Parameters of the compact circuit model of the transmitarray unit cell extracted by fitting to the full-wave simulations (see Figure 2b,c).


Prototype unit cells emulating the structure of the tiled (1 mm gap width) and singlepanel (0 mm gap width) unit cells were fabricated and measured inside the waveguide, see Figure 3a,b respectively. The results in both cases demonstrate noticeable downshift of the central frequency with respect to the design value, c.f., Figure 2b,c, as well as expected degradation of the differential phase shift for the tiled structure, c.f., Figure 2d. The observed shift of the central frequency has been attributed primarily to the specifics of the measurement setup, i.e., different boundary conditions for the unit cell in the waveguide, as compared with the FPBCs in the simulations.

**Figure 3.** Simulated and measured transmission and differential phase shift of the single-panel without gap (**a**) and tiled with 1 mm gap (**b**) transmitarray unit cells. The results were obtained inside a rectangular waveguide (**c**).

Concluding on the results of characterization of the single-panel and tiled unit cells, it can be noticed that both structures demonstrate similar performance within the operating band. Moreover, two orthogonal polarizations demonstrated close performance, according to the full-wave simulations with FPBCs in [10]. The effects of the gap width can be modelled with a reasonably good accuracy using the compact circuit model in Figure 2a. The beam collimating and steering performances of the transmitarray with specific binary phase distributions are discussed in the next section.

#### **4. Beam Steering by the Tiled Transmitarray**

Although one may preemptively conclude from the results of the preceding section that the array performance of the tiled architecture should be commensurable with the single-panel transmitarray, there are still important factors yet unaccounted for. Here we shall apply the analytical model (2), alongside the full-wave simulations and antenna measurements, to evaluate the performance of the tiled architecture against the conventional single-panel transmitarray.

Two sets of prototype 10 × 10-element transmitarrays, viz., a set of 120 tiles hardwired for the two phase states and arranged in specific aperture pattern and a set of three single-panel transmitarrays routed for different beam scan angles (0◦, 15◦, and 30◦), were fabricated in multi-layer printed circuit board technology by two manufacturers using similar materials, but different fabrication processes. The design of the unit cell of the single-panel transmitarrays had to be adjusted to comply with the company-specific fabrication process. That included slightly (10%) decreasing the via diameter and the width of the straight section connecting the annular track of the U-shaped feed to the via, but nevertheless, according to our simulations these changes were not expected to have a prominent effect on the transmitarray performance.

The measured and simulated boresight gains (H-plane) versus frequency of the tiled and single-panel transmitarrays illuminated by a patch-antenna feed are shown in Figure 4. The results demonstrate a 3 dB gain bandwidth of 140 MHz from 5.66 to 5.8 GHz for both transmitarrays.

**Figure 4.** Measured and simulated (with the patch-antenna feed and with the plastic frame in case of the tiled array) boresight gains (H-plane) of the tiled and single-panel transmitarrays versus operating frequency.

The measured gains in Figure 4 are up to 3 dB lower than the simulated values in the operating band for both transmitarrays, which can be attributed to the simulation accuracy, particularly in estimation of the conductor and dielectric losses, as well as to the fabrication tolerances. Non-uniform amplitude distribution across the transmitarray aperture, due to slightly different transmittance of the unit cells in the two phase-states, might have been another contributing factor. This can be inferred from the measurement results in Figure 3. The measured results also indicate a noticeable (<40 MHz) upshift of the peak-gain frequency of the tiled array with respect to that of the single-panel array. Nevertheless, both arrays demonstrate adequate performance within the operating band.

The measured and simulated H-plane and E-plane beampatterns of the single-panel and tiled transmitarrays are shown in Figure 5 at the operating frequency of 5.75 GHz. It appears that the tiled array in measurements exhibits a lower gain and a higher beampointing error against the simulations, as compared with the single-panel transmitarray. It is noteworthy that the measured 15◦ beampattern of the single-panel transmitarray and simulated 15◦ beampattern of the tiled transmitarray feature a higher main lobe as

compared with the corresponding central beampatterns. With the aid of the beampattern model (2) we have attributed this feature to the quantization error inherent to the 1-bit phase-shift design, which leads to sub-optimal radiating power combining at boresight of the transmitarray. Deviation of the differential phase shift from 180◦ causes decreasing peak gain of the steered beams with respect to the central beams in both transmitarrays.

**Figure 5.** Measured (solid lines) and simulated with the patch antenna feed (dashed lines) beampatterns for different beam-scanning angles (viz., 0◦, 15◦, and 30◦): (**a**) assembled transmitarray with the patch-antenna focal source visible; (**b**) binary phase distribution for different beam-scanning angles; (**c**) H-plane single-panel transmitarray beampatterns; (**d**) H-plane tiled transmitarray beampatterns; (**e**) E-plane tiled transmitarray beampatterns.

Figure 6 shows the measured beam patterns in orthogonal polarization (cross-polarization) in the principal E and H planes and diagonal D-plane. All patterns exhibit a prominent peak at boresight with respective cross-polarization ratio (CPR) ~14.5 dB. The shape of the cross-polarization beampattern is typical for dual-polarization transmitarrays, c.f., [17], and indicates polarization leakage due to coupling between the orthogonal feeds and between the patches on the opposite sides of the unit cell, as indicated by our equivalent-circuit characterisation. The measured figure agrees well with the data reported elsewhere, c.f. [17].

**Figure 6.** Measured cross-polarization patterns of the tiled transmitarray radiating at boresight plotted in the principal (E and H) and diagonal (D) planes.

Table 2 shows the performance comparison of the reference single-panel transmitarray discussed in this paper against a selection of published C-band transmitarray implementations, including the theoretical ('theor.'), measured ('meas.') and simulated ('sim.') data, [17–20]. Apart from one passive two-layer frequency-selective surface (FSS) lens, [20], the other transmitarrays adopt the conventional receiver-transmitter architecture with electronic control of the array functional (i.e., beam-steering, beam-forming or polarization conversion). As our design advances, it will integrate electronic control and provide wider bandwidth and better beam pointing accuracy.



#### **5. Conclusions**

A comprehensive characterization of the tiled transmitarray architecture first proposed in [10] has been carried out in this paper. A new compact circuit model has been devised to analyze the broadband transmission and differential phase shift characteristics of the tiled transmitarray unit cells.

The unit cell characterization at normal incidence has been carried out using the proposed circuit model and full-wave electromagnetic simulations. It appeared that the tiled and single-panel unit cells demonstrate commensurable performance within the operating frequency band, although the tiled unit cell exhibits a higher differential phase error.

The antenna gain and radiation patterns of the fabricated tiled and single-panel transmitarrays have been measured for different beam-scan angles, as well as compared with full-wave electromagnetic simulations. The tiled transmitarray demonstrated slightly lower gain and higher beam-pointing error as compared with the single-panel transmitarray. The measured results are in a good quantitative agreement with simulations.

In conclusion, it has been shown that the tiled transmitarrays can be effectively designed, modelled and fabricated to demonstrate the antenna performance commensurate with conventional single-panel transmitarrays. Considering the cost of manufacture and flexibility in configuring the transmitarray for various applications, the proposed tiled transmitarray architecture proves to be a feasible and economically effective solution for 5G communication systems. The future work will be carried out on advancing the analytical model by taking into account essential effects due to spillover, [21], coupling and array nonuniformity, adopting the tiled architecture for millimeter-wave applications, investigating the heterogeneous and conformal transmitarrays enabled by the tiled architecture, as well as developing hybrid approaches to beam-scanning and beam-forming by combining tiled transmitarrays with focal plane antenna arrays.

**Author Contributions:** Conceptualization, D.K., P.T., I.M. and A.S.; methodology, V.K.; validation, P.T., I.M., V.K. and D.K.; investigation, D.K., I.M., P.T., V.K. and D.Z.; Writing—original draft preparation, D.K. and A.S.; Writing—review and editing, D.K., A.S., I.M., P.T. and D.Z.; project administration, I.M. and D.Z.; resources, D.K., I.M., P.T. and D.Z.; funding acquisition, I.M., A.S. and D.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Russian Science Foundation under Project 17-79-20374 and A.S. was supported by The Royal Society under Grant IE160128 and D.Z. was supported by The Royal Society under Grant IES\R1\191236236. The APC was funded by MDPI.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** The authors would like to thank Kieran Rainey and Adrian McKernan for technical support with the antenna measurements.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


### *Article* **Wideband Multiport Antennas**

#### **Mehdi Seyyedesfahlan 1,\*, Abdulkadir Uzun 2,3, Anja K. Skrivervik <sup>1</sup> and Ibrahim Tekin 2,4,\***


Received: 27 October 2020; Accepted: 3 December 2020; Published: 5 December 2020

**Abstract:** In this paper, a wideband four port 2–6 GHz antenna is proposed. One-, two-, and four-port antennas are implemented and characterized between 2 and 6 GHz. The isolation between the ports is improved by connecting and optimizing the ground plane sections. The results show that the antennas' reflection coefficients are better than 10 dB in the frequency band. The measured isolation between the ports is greater than 15 dB (between 2.3 and 6 GHz) and 10 dB in the whole band for two- and four-port antennas, respectively, however, it is more than 20 dB around 2.4 and 5–6 GHz for both antennas. The calculated correlation coefficient between ports is below −30 dB (>2.14 GHz) and −15 dB for the two- and four-port antennas, respectively. The measured gain and efficiency scale are 3.1–6.75 dBi and 62–98%, respectively. To the best of our knowledge, an antenna both being wideband from 2 to 6 GHz and having independent four ports is only addressed in this work. The four-port antenna can be used for MIMO systems or smartphones operating on many wireless systems simultaneously such as 3G/4G/5G Sub-6 GHz and WLAN including the next generation WiFi7 with full-duplex operation.

**Keywords:** wideband antenna; MIMO antenna; four-port wideband antenna

#### **1. Introduction**

Recently, there is an increasing demand for higher throughput and more reliable transceiver systems with an application on 4G wireless systems and mobile communication. Multiple Input Multiple Output (MIMO) technology could be a promising candidate for this purpose [1]. The MIMO technique is based on using multiple antennas to increase the data rate by means of uncorrelated signals.

For the MIMO system to function as expected, the mutual coupling between antenna elements should be as low as possible. A standard approach to achieve MIMO operation is to develop multiple antennas that are sufficiently separated to achieve the desired level of signal independence and port-to-port isolation. However, this will make the transceiver system bulky and result in increased assembly costs. Additionally, ease of integration and miniaturization are two major challenges ahead of MIMO antennas. Thus, the design of the MIMO antenna is the first important thing to be addressed to improve the overall system performance. Planar type antennas are preferred for MIMO applications due to ease of integration and low cost. For miniaturization purposes, there are no options but to space antenna elements closer or designing multiport, single-element antenna. Various studies have been carried out aiming to design such compact antenna systems [2,3]; they are commonly based on the planar antenna prototype [4,5]. The first approach is to decrease the spacing between antennas and keep the mutual coupling at an acceptable level by applying isolation improvement techniques. Examples of this approach can be found in various literature, such as adding a ground wall with connecting line and shorting pins [6], T-shaped ground plane [7], the corrugated ground plane with λ/4 slot [8], modified PIFA with a small local ground plane [9], techniques based on dispersion engineering called

negative group delay (NGD) technique [10], use of external lumped element decoupling networks between the feed ports to allow matching of even and odd modes to a common impedance and thereby producing small cross-correlation and maximum gain over a limited frequency range [11], and other compact designs of MIMO antennas [2–5,12]. These methods can reduce the overall size of antennas and insulator regardless of the difficulty in insulator design.

The second approach takes advantage of multiport, single-element antennas to propose a more compact solution. A novel design of dual-feed, single-element antennas for 4G MIMO terminals is proposed and analyzed in [13]. The antenna consists of a radiating patch which is fed by two input ports. The idea is to use an isolated mode antenna (iMAT) [14] to reduce the antenna size and mutually couple the ports. The iMAT works based on exciting and different propagating modes of antenna for different ports. The iMAT antenna idea is also used in [15] to design a novel u-shaped single-element antenna with better performance, compared with two separate monopole antennas, in [16,17] to design a multiple compact multimode patch antenna, and in [18–20] for a multimode antenna that is not based on a patch antenna.

An important factor in MIMO systems is its bandwidth, which is determined by the bandwidth of the antenna element. Thus, a wideband single-element antenna with multiple ports could be very useful for a wideband MIMO system. In general, MIMO systems use many antennas to obtain multiport systems. Note that a multiport wideband antenna can also be used in smartphones that use many different wireless protocols at different frequency bands at the same time. There exist antennas which either are wideband or are multiport with narrow bandwidth. Nevertheless, combining multiport with wide bandwidth operation forms our antenna's novelty, which has four ports and can operate between 2 and 6 GHz. We propose a structure to increase the number of radiating element feeding/receiving ports only by rotating the main single port monopole antenna. Of course, monopole antenna is well known and there are many reports on how to make it wideband; however, increasing the number of ports while matching the ports and decreasing coupling between the ports requires many attempts. Moreover, when all the ports use common radiating elements, it needs a smart method to mitigate coupling between the ports. In this paper, we use a unit structure and bridge between the ground planes of ports to alleviate coupling between ports. We designed and optimized the antenna for frequency band between 2 and 6 GHz and achieved minimum isolation of 10 dB between the four ports. The aim is to introduce a multi-purpose (multiport and wideband) structure; however, for the desired application/band, the isolation between ports can be increased only by optimizing the ground plane and connection between ports.

The four-port antenna reported in this paper can be used for a multi-frequency system requiring many antennas. The 4 × 4 MIMO implemented for a WLAN on 2.4 and 5.2 GHz band is one example. The four-port antenna can also be used for the sub-6 GHz band 5G system. For a multiple radio system currently used in smartphones, let us assume there are four radios and these radios are 3G (2 GHz band), WLAN (2.4 GHz), 1–6 GHz 5G (3.6 GHz band), and WLAN (5.2 GHz). One can directly connect these four radios to the proposed four-port antenna without any switches and duplexers. RF filters can be deployed for each radio band to provide enough selectivity. However, with our antenna, all these radios can operate simultaneously. The key focus is on new mobile 5G bands including spectrum in the 3.5 GHz range that has been assigned in numerous countries. However, several countries including China and Japan plan to use spectrum in the 4.4–4.9 GHz range for 5G in addition to a growing number of countries considering the 3.5–4.2 GHz range, as well as the 2.3 and 2.5/2.6 GHz bands for 5G NR [21].

To have a wideband multiport antenna, a wideband planar structure should be selected. In this work, a printed monopole disk antenna [22] is selected for multiport use, due to its wide bandwidth operation. Figure 1 shows the monopole disk antenna with a single port. The disk monopole antenna is modified to two- and four-port versions for different frequency ranges. The geometrical symmetry of the antenna shape not only makes the design easy but also gives the versatility of adding and increasing the number of ports.

**Figure 1.** Schematic for the proposed single port wideband disk antenna: (**a**) cross-sectional view; (**b**) top view; and (**c**) etched ground dimensions.

In this paper, two metrics are used for the assessment of the isolation between antenna ports: the S parameter and the correlation coefficient. The correlation coefficient expresses antenna pattern independence to the S parameter, which is necessary for a MIMO antenna. This paper is organized as follows. Section 2 demonstrates the design of single-, dual-, and quad-feed disk monopole antenna with wideband operations. Section 3 presents and compares the simulated and measured results for the S parameter as well as the radiation patterns of the antennas. Section 4 summarizes and concludes the paper.

#### **2. Multiport Antennas Design**

In this section, the design process for the one-, two-, and four-port antennas are introduced. The antennas contain a radiating disk and microstrip transmission line as the antenna feed. Both twoand four-port antennas have structures similar to the single-port antenna, and the various dimensions shown in Figure 1 are optimized for each antenna to match each port to 50 Ω and decrease the mutual coupling between ports of each antenna, over the frequency band of 2–6 GHz. The scheme for increasing the number of the ports is to exploit the single-port antenna geometry (Figure 1) as the basis of n-port antennas, and then rotate/add the structure by 90◦ (with respect to disk center) to form the new port. The advantage of this procedure is that the ports (in multiport types) would be similar, and the design parameters in Figure 1 are optimized for all ports, simultaneously. The optimization is performed to approach the specified reflection coefficient and isolation between the ports over the desired frequency bandwidth. The antennas were simulated and prepared for fabrication on d = 0.787 mm thick (copper cladding tc = 35 μm) Rogers RT/duroid 5880 laminate with a dielectric constant of 2.2 and tangent loss of 0.0009.

#### *2.1. Single-Port Antenna*

The schematic of the single-port antenna and the parameters for which optimizations are performed are shown in Figure 1. The antenna can be divided into two major parts: the radiating disk and the transmission line that feeds the disk. The fabricated antenna with the dimensions of 6.8 cm × 4.4 cm is shown in Figure 2. In the bottom layer, an incomplete triangular shape ground plane supports the signal line in the top layer and can have coupling with the radiating disk.

**Figure 2.** Fabricated single-port antenna: (**a**) top view; and (**b**) bottom view.

The dimension of the disk (*r*) adjusts the antenna operating frequency, while Δ*r* is the spacing between the disk (top layer) and ground plane (bottom layer) edge. Since the antenna is similar to a monopole antenna, the disk will resonate with a quarter-wavelength diameter (2*r* = λ/4). The radius of the disk for resonating at 2 GHz in the free space is calculated as 18.75 mm, which is used as the initial value for r. To match the antenna to 50 Ω in the band of 2–6 GHz, other parameters (shown in Figure 1) are utilized to tune the antenna over the entire desired frequency band or in some specific frequencies. θ and *hp* control the dimensions of the ground plane. The gap specified by the dimensions of *ga*/*gb*, as well as the location (*hd*) and dimensions of the dumbbell-shaped etching, affect the antenna reflection coefficient by changing the inductance/capacitance of the transmission line and improving the feed line S11 magnitude.

The optimized parameters for fabricating the antenna (Figure 2) are reported in Table 1. Different parameters of the single-port antenna are swept around the optimized values to show their effect on the antenna reflection coefficient.


**Table 1.** Geometrical dimensions for different antennas.

Values are in mm.

Figure 3a shows that the optimum value for a disk radius of 21 or 22 mm can give the best reflection coefficient values at less than −15 dB. As the spacing between the disk and ground plane is increased up to 3 mm, the antenna matching is improved, while greater values deteriorate the antenna performance, due to the decoupling between the microstrip line and the disk antenna, as shown in Figure 3b.

**Figure 3.** Single-port antenna simulated reflection coefficient for different: (**a**) disk radii; and (**b**) the gap length between disk and the ground plane.

When the ground plane angle (θ) is increased, the antenna |S11| is improved for higher frequencies, while the impedance matching worsens in the middle of the band (Figure 4a). As shown in Figure 4b, the height variation of the truncated triangle ground causes a frequency shift in the antenna reflection coefficient.

**Figure 4.** Single-port antenna simulated reflection coefficient in terms of triangle parameters: (**a**) angle; and (**b**) height.

The effect of the dumbbell-shaped etched ground plane on matching the antenna between 2 and 6 GHz is demonstrated in Figure 5a. Note that the legend with the word "No" in Figure 5a points to the full ground (without dumbbell-shaped etching). Changing the dimensions of the etched area in Figure 5b,c shows its major effect on the antenna reflection coefficient for frequencies greater than 3 GHz. Although the effect of some parameters is not that significant in the single-port antenna, they play a drastic role in tuning the multiport antennas in the wideband operation.

*Sensors* **2020**, *20*, 6960

**Figure 5.** Single-port antenna simulated reflection coefficient for different (**a**) location; (**b**) width; and (**c**) height of the dumbbell-shaped etched ground.

#### *2.2. Two-Port Antenna*

The two-port antenna is obtained by rotating the single-port antenna by 90◦, with respect to the center of the disk, and adding another port. As shown in Figure 6, the ground planes of the two ports are connected via a circular ring sector.

**Figure 6.** Fabricated two-port antenna: (**a**) top view; and (**b**) bottom view.

The angle of the sector is 90◦−θ, while its inner and outer radius are *Rd* and *hp*, respectively. By connecting the grounds of the two ports, better isolation between the ports is obtained. When the first port of the antenna is fed, the received signal in the second port includes two components: (a) the signal that passes over the disk; and (b) the signal that flows from the connected ground of the ports. Therefore, these two components can cancel each other, if the phase difference of 180◦ is kept when these two components arrive at the second port. Out of phase condition between the mentioned two current trajectories improves the ports' isolation significantly and can be achieved by optimizing some of the antenna parameters. The working mechanism of the connection is shown (see Figure 7) in the simulated current distribution on the antenna at 2.4 GHz and at different phases.

**Figure 7.** The simulated magnitude of current distribution on top and bottom layers at 2.4 GHz and different 0-, 45-, 90-, and 135-degree phases.

Note that the power that is dissipated in the vicinity of port two (due to cancellation) can decrease the radiation efficiency of the antenna, whereas the used PCB board is chosen to have a very small tangent loss. The dimensions of the fabricated two-port antenna are 7.6 cm × 7.6 cm and the rest of the parameters are given in Table 1.

Changing the disk radius (*r*) and Δ*r* can both affect the accepted/reflected power by the first port on the disk side as well as the coupled power to the second port through the disk. The influence of this complicated process on the insertion loss between the ports, for various *r* and Δ*r*, is shown in Figure 8.

**Figure 8.** Simulated insertion loss between the ports of two-port antenna for various: (**a**) disk radius; and (**b**) disk-ground spacing.

As shown in Figure 8, the most significant effect of the parameters *r* and Δ*r* on |S21| is between 2.7 and 4.5 GHz. At these frequencies, insertion loss can be adjusted to be below −20 dB, by the disk size and the gap spacing. The ground plane angle also affects insertion loss between ports in a limited frequency band of 4–5 GHz (Figure 9a). Increasing the height of the ground plane shifts the |S21| to lower frequencies in the 3–6 GHz band (Figure 9b).

**Figure 9.** Simulated insertion loss between the ports of two-port antenna for various: (**a**) angle; and (**b**) height of the truncated triangular-shaped ground plane.

As discussed above, the connection between the ports' ground plays an important role in improving the insertion loss between them. The important factor *Rd*, which controls the dimension of the connected part, and corresponding isolation between the ports, is swept around its optimum value *Rd* = 36 mm, as shown in Figure 10. This results in increasing *Rd*, and hence decreases the thickness of the connected section and causes the S21 curve to shift to lower frequencies.

**Figure 10.** Effect of the inner radius of the ground plane sector on simulated insertion loss.

#### *2.3. Four-Port Antenna*

Another multiport antenna is a four-port antenna that is formed by rotating/adding a single-port antenna with respect to the disk center. In this type of antenna, the spacing between the ports is 90◦, as shown in Figure 11.

θ

θ

θ

θ

θ

θ

**Figure 11.** Fabricated four-port antenna: (**a**) top view; and (**b**) bottom view.

For this antenna, the basic geometry is similar to that of the single port, while the edge of the disk at the ports is etched (in top view) and the grounds are connected using a metal bar with a thickness of *Ct*. The etched areas on the disk are a rectangular section with dimensions of *ea* and *eb* (Figure 11), which alleviate the ports' reflection coefficient to be below −10 dB within the band of 2–6 GHz. The ground connection also controls the insertion loss between the ports, the same technique as used in the two-port antenna. The dimensions of the fabricated four-port antenna are 12.4 cm × 12.4 cm, with parameters given in Table 1. Note that, for an application on a mobile phone, antenna size can be made smaller by bending from the microstrip line sections. The proposed four-port antenna is designed for a wide frequency band, starting from 2 GHz. By excluding WiFi 2.4 GHz frequency, while shifting start frequency to 3 GHz, which means, if only 5G systems are chosen, the disk size and hence the overall antenna size will be smaller by a factor of 1.5 times to achieve an 8 cm × 8 cm antenna. Moreover, for mobile applications, part of the feeding network can also be placed in a different PCB layer, or a flexible board may be folded/wrapped and antenna size can be further made smaller. Moreover, the antenna can be optimized/improved by separating 5G or WiFi system and having two antennas. For example, for a 5G sub-6 GHz system, the antenna size can be optimized to 3 GHz, and for WiFi it could be around 5 GHz band.

As shown in Figure 12, etched areas on the disk (geometrical parameters *ea* and *eb* in Figure 11) and the ground plane (geometrical parameters *ga* and *gb* in Figure 1) play a very crucial role in adjusting each port's reflection coefficient below −10 dB. The legend entry "No" indicates no etching is performed on the copper.

**Figure 12.** Effect of etched area's simulated reflection coefficient of four-port antenna: (**a**) disk; and (**b**) edge of the ground plane.

Due to the wide bandwidth of 2–6 GHz and the number of ports, developing the antenna for |S11| < −10 dB and desired isolation between the port is challenging or maybe impossible for some geometries. Consequently, it is proposed to match the ports to 50 Ω with |S11| < −10 dB and improve ports isolation for some specific frequencies, while it exceeds 10 dB for whole the bandwidth. Therefore, isolation between the ports is optimized to target the higher values around the frequency of 2.4 GHz and bandwidth of 5–6 GHz that are used by WLAN.

Since the ports are symmetric, S21 = S41 = S32 = S43, and S31 = S42, only S21, and S31 is plotted. As shown in Figure 13, as the ground plane angle (θ) is increased, the magnitudes of the S21 and S31 shift to higher frequencies.

**Figure 13.** Effect of ground plane angle θ on the simulated isolation (a) S21 and (b) S31 between different ports of four-port antenna.

The increasing θ causes an increase in the ground plane size, and, as a result, the length of the connected part between ground planes of the ports is decreased. Moreover, increasing the height of the ground plane (*hp*) increases the length of the connected part and shifts S21 and S31 to lower frequencies (see Figure 14).

**Figure 14.** Effect of the ground planes height on the different ports' isolation (**a**) S21 and (**b**) S31 in a four-port antenna.

When the thickness of the connected part (*Ct*) is increased, as in Figure 11b, since the lower side of the connection part is limited/fixed by the triangular-shaped plane (*hp*), the upper edge is extended toward the disk. Therefore, by increasing the thickness of the connected part (*Ct*), its average length is decreased, shifting S21 and S31 to higher frequencies (see Figure 15).

**Figure 15.** Influence of the ground joint thickness on the different ports simulated isolation (a) S21 and (b) S31 in four-port antenna.

#### **3. Measurements and Simulations**

The antennas were simulated, fabricated, and measured using the geometrical parameters in Table 1. The antennas were characterized for S parameters and 3D cross-polar and co-polar gain at some specific frequencies. The correlation coefficient between ports i and j of N-port antennas is calculated using the simulated and measured S parameters using (1). Equation (1) is an approximation to calculate the pattern independence between the ports using the S parameter. Its precision is increased as the radiation efficiency of the antenna is increased [23]

$$\rho\_{\ell}(i,j,N) = \frac{\left|\mathbb{C}(i,j,N)\right|^2}{\prod\_{k=i,j} [1 - \mathbb{C}(i,j,N)]}, \quad \mathbb{C}(i,j,N) = \sum\_{n=1}^{N} S\_{i,n}^\* S\_{n,k} \tag{1}$$

The measurements were performed at Sabanci University Anechoic Chamber that is suitable for the frequency range from 700 MHz to 50 GHz and is equipped with a PNA5245A vector network analyzer (working up to 50 GHz).

#### *3.1. Reflection Coe*ffi*cient*

Simulated and measured reflection coefficient results from 1 to 6 GHz of the single-port antenna are shown in Figure 16. The measured |S11| has some shifts for frequencies greater than 4 GHz. This shift can result from the PCB dielectric constant variation in different frequencies or the effect of the measurement setup. Although the antennas are measured inside the anechoic chamber, due to their isotropic radiation pattern (which is discussed in the next section), the absorbers, feeding cable, and setup in close distance to the antenna can affect its performance. The measurement shows that the antenna reflection coefficient is below −10 dB for the whole 1–6 GHz frequency band.

**Figure 16.** Measured and simulated reflection coefficient for the single-port antenna.

The measured and simulated S parameters and calculated correlation coefficient using (1) for the two-port antenna are shown in Figure 17. Some frequency shift around 200 MHz is also seen in the measured S parameters. The measured results comply with the simulated ones and the antenna is matched to 50 Ω for frequencies greater than 1.1 GHz. Isolation between the ports is better than 15 dB for higher frequencies (>2.3 GHz). After a frequency of 2.14 GHz, a correlation coefficient of better than −30 dB is obtained from the measured/calculated ρ21.

**Figure 17.** Measured and simulated (**a**) S-parameters and (**b**) correlation coefficient for two-port antenna.

The four-port antenna is characterized for S11, S21, and S31, as shown in Figure 18. The measured and simulated results agree well, and only some frequency shift is seen in S21. As mentioned for the two-port antenna, the frequency shifts between simulation and measurement can be the effect of setup (such as cables) that reflect back the radiated field from the antenna and change the antenna performance. When cables are used for measuring two close ports such as S21, their influence is more pronounced than for the other ports such as S31. In addition, as the radiating disk is surrounded by the metal ground plane (the number of ports is increased), the antenna radiation on the ground plane direction is reduced. Thus, the effect of any cables, which are extended in the same plane as the ground plane, is decreased by increasing the number of ports.

ρ

ρ

ρ

**Figure 18.** Measured and simulated (**a**) S-parameters and (**b**) correlation coefficient for four-port antenna.

The measurements show that the S parameters are below −10 dB over the desired frequency band of 2–6 GHz. Isolation between the ports is better than 20 dB at around 2.4 GHz and between 5 and 6 GHz. In addition to the S parameters, good agreement between the measured and simulated correlation coefficient is also obtained (see Figure 18b).

For MIMO applications, although there are no specific requirements on isolation values, lower values of isolation will ease the work done by the baseband processor. Two- and four-port antennas can operate at frequencies below 6 GHz with good isolation values. For the two-port antenna, isolation is lower than 15 dB in the overall band mainly around 20 dB, which may be sufficient for MIMO applications. For the four-port antenna, isolation is lower than 10 dB in the overall band, more than 20 dB around 2.4 and 5–6 GHz, and more than 10 dB in the whole band. Further, for a 5G MIMO system, the four-port antenna can also be used for some portions of the bandwidth. The least isolation is 10 dB; however, the 5G sub-6 GHz system will not use the whole 4 GHz available, but a few hundred MHz bandwidth from the spectrum. When we consider a realizable 5G massive MIMO system with a few hundred MHz bandwidth operations, the four-port antenna may achieve more than 20 dB isolation, which may be sufficient for a MIMO system such as in 5–6 GHz band.

There is also interest in how these antennas will perform in a real environment. Most of the time, measured results are performed in a controlled environment such as an anechoic chamber. When these antennas are placed in a real environment, the isolation, as well as the return loss of the antenna, may change. However, most of these isolation and reflections are due to the antenna itself, the so-called self-interference signal. Isolation and reflection will not degrade significantly if an object is not placed in the vicinity of the antenna. Specifically, for the 5G 3 GHz frequency, the free space path loss around 3 GHz at 1 m is 42 dB, if an object is placed at 1 m from the antenna. The two-way path loss will be 84 dB lower, which will worsen the isolation and reflection. However, it will not be that significant if the isolation is around 20–30 dB range. If a very close object is placed by the antenna, this may cause a few dB change in the isolation; however, if the objects are placed far away from the antennas, similar performance should be expected. As measured in [24], the performance of high isolation antennas in a real environment will definitely change, but the normal operation of the antenna will remain stable.

#### *3.2. 2D and 3D Gains*

The 3D gain of antennas was measured at frequencies of 2, 2.4, 3.4, 4.4, 5.2, and 5.8 GHz. The origin of the Cartesian coordinate system, which describes the gain of antennas, is the disk center and is shown in Figures 2, 6 and 11 for one-, two-, and four-port antennas, respectively. In all systems, the feeding port is on the z-axis (Port 1), and the x-axis is perpendicular to the disk. Since the radiation pattern of the antennas is similar to a dipole antenna, the antennas' co- and cross-poles are indicated by G<sup>θ</sup> and G<sup>φ</sup> components. Therefore, it is expected that the antennas co-pole gain (Gθ) will significantly dominate the cross-pole gain (Gφ). Furthermore, the one- and four-port antennas are symmetric with respect to the xz-plane; thus, the patterns are measured only on the hemisphere on y > 0 space.

The simulated and measured 2D gains for the one- (Figure 19), two- (Figure 20), and four-port (Figure 21) antennas were obtained at different frequencies (2.4, 4.4, and 5.8 GHz) and on xz-, yz-, and xy-planes. As shown in Figures 19–21, the measured (dashed lines) and simulated (solid lines) gains in the θ direction (G<sup>θ</sup> in the red curve) dominate the gains in the φ direction (G<sup>φ</sup> in the blue curve). The simulated gains for G<sup>φ</sup> are smaller than the measured values, which could be due to the antennas' imprecise alignment in the measurement setup or the AUT (antenna under test) tilt during the measurement. When the antenna gain at one pole is much smaller than the other pole, some tilt in AUT can cause a significant increase in the value of the cross-pole. The effect of the feed cable is seen in the measured G<sup>θ</sup> gain at around θ = 180◦ and on the xz- and yz-planes. The radiation patterns show an isotropic antenna characteristic on the xy-plane. The electric field component of G<sup>θ</sup> gain on the xz-plane is perpendicular to disk at θ = 0◦ and 180◦ and results in null at these angles. Moreover, the fact that the antennas radiation at θ = 0◦ and 180◦ is lower (smaller gain) than at other angles shows the antennas behave very similar to a dipole antenna.

The 3D gain of the antenna at 4.4 GHz for total gain, θ polarization, and φ polarization are shown in Figure 22. As discussed, the level of G<sup>θ</sup> is higher than G<sup>φ</sup> for all of the antennas. For all ports of the antenna, nulls are seen around the antenna feeding ports. One can note that the antenna almost radiates in the available space, making an ideal antenna for MIMO wireless systems.

#### *3.3. Gain and E*ffi*ciency versus Frequency*

The gains of the antennas are measured in some specific frequencies including 2, 2.4, 3.4, 4.4, 5.2, and 5.8 GHz, as shown in Figure 23a. The agreement between the measured and simulated gains of the antennas is decreased as the number of the ports is reduced. Since the single-port antenna radiates in all directions, the absorbers near the antenna in the anechoic chamber change the antenna performance and specifically in the lower frequencies. Note that the ground plane around the two- and four-port antennas reduces the antenna radiation on the antenna plane (xz-plane), in the direction that the near absorbers to the AUT are positioned. Consequently, the destructive effect of these absorbers, near the two- and four-port antennas, are partially canceled and the measurements get closer to simulated results.

The antennas' measured gain varies from 3.08 dBi at 3.4 GHz for a single port to 6.74 dBi at 5.8 GHz for four-port antennas.

The radiation efficiency of the antennas is calculated using the measured (with 2◦ angular spacing) average 3D gain technique. The calculated efficiency plot is presented in Figure 23b. Due to the big structure of the antennas, the accuracy of the simulated radiation efficiencies is low. As the frequency is increased, each antennas' radiation efficiency is increased. The values change from 50% to 100 %.

Table 2 presents the comparison of proposed antennas with the reported sub-6 GHz MIMO antennas in [25–30]. The presented two- and four-port antennas in this work outperform previously published sub-6 GHz antennas for 5G applications in [25–30] with the larger bandwidth and single radiating element, for which the antenna size, gain, efficiency, and performance in MIMO applications are comparable to existing studies where multiple radiating elements are used.

**Figure 19.** Measured (dashed lines) and simulated (solid lines) gain for the single-port antenna on three different planes and frequencies.

**Figure 20.** Measured (dashed lines) and simulated (solid lines) gain for the two-port antenna on three different planes and frequencies.

**Figure 21.** Measured (dashed lines) and simulated (solid lines) gain for the four-port antenna on three different planes and frequencies.

**Figure 22.** Measured 3D total, θ-polarized and φ-polarized gains at 4.4 GHz for: (**a**) single-port antenna; (**b**) two-port antenna; and (**c**) four-port antenna.

**Figure 23.** (**a**) Measured and simulated peak gain; and (**b**) measured antenna efficiency values.


**Table 2.** Comparison of proposed antennas in this work and antennas in [25–30].

Abbreviations: Ref., References; BW, Bandwidth; BW Def., Bandwidth Definition; ECC, Envelope Coefficient Coefficient.

#### **4. Conclusions**

A wideband single-port antenna, with λ/4 diameter of the radiating disk (monopole-like) and dipole-like radiation pattern, was designed and manufactured. The geometry of the single-port antenna was utilized as the prototype for the two- and four-port antennas, by rotating (90◦) the single-port geometry with respect to the disk center and adding a new port. The ground planes of the ports (other than the single port) were connected to improve/increase the isolation between ports. The measured and simulated data are in good agreement. The acceptable correlation coefficient during the bandwidth makes the antenna suitable for the MIMO application for the 5G NR sub-6 GHz band. Finally, the design challenge of two or four separate antennas being near each other and any potential coupling between them can be solved by these monolithic compact antennas that contain good matching, proper isolation between the ports and omnidirectional-like radiation pattern. The antennas are not only very compatible, but their reflection coefficient/isolation between the ports can be further improved, to achieve even better values (for a limited bandwidth or single operation between 2 and 6 GHz), by optimizing the dimensions of the introduced parameters (assuming prior knowledge of their influence on Sii and Sij).

**Author Contributions:** Conceptualization, I.T., and M.S.; funding acquisition, I.T; software, M.S.; investigation, M.S.; writing—original draft preparation, M.S. and A.U.; writing—review and editing, I.T and A.K.S.; and project administration, I.T. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by The Scientific and Technological Research Council of Turkey (TUBITAK) under Grant 114E494.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
