Multi-Mode Antennas for Ultra-Wide-Angle Scanning Millimeter-Wave Arrays

In this paper, a novel multi-mode millimeter-wave antenna array with enhanced scan range and reduced scan losses is presented. The individual array element consists of a differentially fed microstrip patch on top of which a cylindrical dielectric resonator is integrated. The radiation pattern of the antenna element can be reconfigured by changing the phase offset between the feeding ports of the patch and the dielectric resonator to excite two distinct radiating modes. With such a feature, the field of view can be divided into two different subspaces, with the first one covering the angular range from −75° to 0° and the second one from 0° to +75°. In a <inline-formula> <tex-math notation="LaTeX">$1\times16$ </tex-math></inline-formula> linear array configuration, the achieved scan range extends from −75° to 75° along the horizontal plane with a maximal gain loss of 3 dB, which is better than the ideal <inline-formula> <tex-math notation="LaTeX">$\cos \theta _{0}$ </tex-math></inline-formula> behavior. The proposed design operates in the frequency range between 27 GHz and 29.5 GHz and, thanks to its wide-scan capabilities, constitutes an effective solution for upcoming 5G/6G millimeter-wave wireless communications.


I. INTRODUCTION
P HASED array antennas constitute key components of millimeter-wave (mm-wave) wireless systems thanks to their electronic beam-steering capabilities.They are employed in a variety of applications such as satellite communications, radar, space exploration, and remote sensing, as well as in emerging 5G/6G technologies which can benefit from high antenna gain to compensate for the larger propagation losses experienced at mm-wave frequencies.
The design of phased arrays poses a number of challenges that need to be addressed.One of the most significant drawbacks which this class of radiating structures is typically subject to is the limited scan range.According to theory, the scan capabilities of an array are limited by the beamwidth of the embedded element pattern.In the ideal case, the array gain decays following a cos θ 0 trend, where θ 0 denotes the scan angle.The most common solutions adopted in the design of wide-angle scanning arrays are microstrip patch antennas which typically feature a beamwidth of 110 • and can therefore enable a scan range of ±55 • both in azimuth and elevation with a typical maximum gain loss of 4-5 dB when integrated into an array configuration [1].
Furthermore, when scanning toward the end-fire region, the increase of mutual coupling between array elements causes a fast degradation of efficiency and gain antenna characteristics.Many studies were carried out to reduce the mutual coupling between elements, for example, by introducing a metal cavity [2] between the adjacent antenna elements or metal vias [4] as a separating structure.In [3], instead, wide-angle scanning performance is achieved by using highimpedance periodic structures that suppress surface waves while broadening the embedded element radiation pattern.
Another approach to improve the scan performance of antenna arrays relies on the enhancement of the beamwidth of the embedded element pattern.Different design concepts are available in the scientific literature [5], [6], [7], [8], [9].In [5], two metal walls are introduced on the side of the basic patch antenna element to increase the relevant beamwidth up to 120 • .In [6], a large beamwidth is achieved with a Ushaped microstrip antenna enclosed with two metal walls integrated in the plane opposite to the scan direction of the linear array.
Another approach to achieve wide scanning is to use pattern-reconfigurable antennas.The working principle is based on the segmentation of the scan range in sub-spaces, where each sub-space is covered by steering the embedded element pattern based on different modal configurations.Usually, pattern reconfiguration is realized by introducing pin diodes to switch between radiating modes [7].This approach removes the need for broadening the beamwidth of the embedded element pattern and can benefit from more directive radiating modes, thus enabling an additional advantage in filtering out undesired grating lobes.
Different from the reconfigurable antenna technology, metamaterial-based lens design methodologies can result in broader scan ranges as compared to traditional arrays thanks to the integration of metamaterial structures [8], [9].As a matter of fact, metamaterials can enable unconventional electromagnetic wave propagation properties by controlling permittivity and permeability independently in a complex way [10].However, metamaterial lenses are typically very complex and large, which is undesired for onboard applications.
When considering the mm-wave frequency range, most of the listed methods above are not easy to realize because of manufacturing limitations or extra losses introduced by passive components.
In this paper, we propose a wide-scanning array topology based on a novel two-port multi-mode antenna element realized by integrating a pin-fed dielectric resonator (DR) with a capacitively fed microstrip patch.In our design, the mode switching is realized by selecting different phase offsets at the antenna feeding ports.The proposed antenna solution has been implemented in a linear array that can switch between two radiating modes: the first mode covering the sub-space from −75 • to +20 • when the two ports of the individual array element are fed with the same phase and the second mode covering the sub-space from −20 • to +75 • when the ports are fed with an offset of 180 • .In the presented design, no switches are required to modify the current paths along the antenna structure; to this end, use is made of multichannel beamforming RF integrated circuits (RFICs) with integrated vector modulators, which are readily available on the market nowadays [11].In this way, a 1 × 16 linear array of the developed antenna can scan in the angular range from −75 • to +75 • with a maximal scan loss of only 3 dB.This paper is organized as follows: Section II concerns the antenna element design and discusses the working principle with the related theory and the geometry of the radiating structure.Section III details the simulation and measurement results of the 1 × 16 linear array.In Section IV, the proposed solutions is benchmarked against the state of the art, and Section V concludes with some final remarks.

II. PRINCIPLE AND ANALYSIS OF THE ANTENNA ELEMENT
For point-to-multipoint wireless communications, phased arrays with wide-angle scanning capabilities are usually preferred.The scanning capability of an antenna array is mostly determined by the embedded element pattern.Consider a linear array of N elements.The array pattern S(θ 0 ) can be expressed as a multiplication of the ideal array factor and the element factor f (θ 0 ): upon adopting a suitable spherical coordinate reference system and assuming a θ 0 -only dependence.In (1), d is the distance between the elements, k 0 is the wave number in free space, and ψ n denotes the phase offset applied to each array port.As the beam is steered towards larger angles, the array gain is attenuated proportionally to the element pattern.For this reason, antenna elements with a wide halfpower beamwidth (HPBW) are often employed to realize wide-angle scanning arrays [1].
In order to extend the scan range above the physical limit of cos θ 0 , in this work, we will investigate the use of antenna elements with two radiating modes.By multi-mode antenna, we mean a single physical radiator with separate feeding ports that excite different characteristic modes.In this way, the field of view can be divided into two sub-spaces, where each sub-space is covered by one of two radiating modes.
The radiation pattern scanning capabilities of an array based on a multi-mode antenna concept are illustrated in Fig. 1 in the particular case of two radiating modes, both characterized by a cos 2 (θ 0 ) element pattern behavior as an example.Each radiating mode identifies a different sub-space where the array beam is scanned.As one can notice, the overall scan range is broadened as compared to a traditional array implementation while attenuating the impact of unwanted grating lobes.Such a feature of multi-mode antennas allows for the integration in an array configuration without the need to keep an inter-element spacing of λ 0 /2 or smaller, with λ 0 being the wavelength in free space.This enables additional degrees of freedom in the design process that can be favorably exploited to enhance array performance in terms of scan loss mitigation.

A. THEORY
To realize a multi-mode antenna based on a phase control principle, we make use of the combination of an electric and a magnetic dipole with orthogonal orientation.
The total far-field of a pair of orthogonal electric and magnetic dipoles can be obtained by superimposing their individual far-fields because these are orthogonal to each other [12], [13].Therefore, the total E θ and E φ components of a z-directed electric dipole (length l e , current amplitude I e ) and a y-directed magnetic dipole (length l m , current amplitude I m ) are given by [12]: e −jkr jηI e l e sin θ − e jδ jI m l m cos φ (2) where we assume time-harmonic fields and where k = ω √ 0 μ 0 is the wave number, and δ is the phase offset between the two ports of the same element.Upon assuming ηl e I e = l m I m = lI, the expression of the total electric field, given by: can be simplified as: when δ = 180 • , and: when δ = 0 • .According to the equations above, the maximum and minimum of |E T | are located along the ±x-direction.One can notice that E T reaches its maximum intensity along θ = 90 • for δ = 0 • , and along θ = −90 • for δ = 180 • as pointed out in [12], where the objective was to synthesize an end-fire-like radiation pattern.
The simplified basic principle is illustrated in Figs.2a  and 2b, where the effect of the superposition of an electric dipole along the z-axis and a magnetic dipole along the yaxis is shown in case of the same feeding phase and in case of a phase offset of 180 • , respectively.However, it is worth mentioning here that the simplified illustration above makes use of ideal electric and magnetic dipoles but, in reality, the actual radiation patterns will be different, depending on the specific antenna design implementation.For example, when we integrate such an antenna in an array configuration, the direction of maximum radiation will be slightly steered towards the direction of θ = 45 • due to the presence of the ground plane and the adjacent elements.It is important to mention that the design proposed in [12] cannot support two independent radiating modes and, therefore, cannot be used to enhance the scan range of an antenna array using the system architecture and principle described in this research study.To achieve such a goal, one should use a suitable multi-port antenna solution, such as the one detailed in the following sub-sections.

B. MULTI-MODE ANTENNA GEOMETRY
The novel multi-mode antenna element that we propose in this study consists of a differentially fed loaded microstrip patch integrated with a pin-fed cylindrical DR.Such an antenna can support multiple radiating modes by selecting different phase offsets between the two ports.The antenna structure is sketched in Fig. 3a and Fig. 3b, where the cylindrical DR is placed on top of the differentially fed patch element.In the aforementioned figures, the prepreg layers are made transparent in such a way as to allow the view of the rectangular cavity that is realized by grounding the top and the bottom layers with through-via fences surrounding the entire structure.Such a cavity is useful to reduce the parasitic coupling level with nearby antenna elements when integrated in an array.The antenna ports are fitted with mini-SMP 18S142-40ML5 connectors to enable the passive characterization of the radiating structure.
The proposed antenna is built using 8 metal layers as shown in the printed circuit board (PCB) stack-up of Fig. 4. As one can notice, the Rogers RO4350B cores with relative permittivity r = 3.66 and loss tangent tan δ = 0.0037 at 10 GHz are interleaved with Rogers RO4450F prepreg layers with a thickness of 0.1 mm.From the bottom to the top, the four PCB layers have different thicknesses, equal to 0.168 mm, 0.168 mm, 0.508 mm, and 0.422 mm, respectively.The dielectric rod is instead realized with Rogers RO4360G2 material having r = 6.3 and tan δ = 0.0023; its height is 3.2 mm while the relevant diameter is 2.5 mm.The patch structure is differentially fed by two metal pads located on metal layer 6 (M6), which are connected to the mini-SMP connector through a via feed.The size of the patch is 1.65 mm × 1.65 mm; a hole with a diameter of 0.9 mm is realized at its center to allow for the feeding pin of the DR to be inserted through it.The capacitively coupled metal pads have dimensions of 0.9 mm × 0.63 mm and are located at a distance of 0.125 mm from the patch.To implement the 180 • phase offset between the two pads, the via feeds are connected to a rat-race coupler, as shown in Fig. 5a, located on metal layer 2 (M2).The optimized diameter of the ring forming the rat-race is 2.57 mm to enforce out-ofphase signals at the two feeding probes, whereas the width of the transmission line is 0.1 mm to ensure a 70 characteristic impedance level.The width of the 50 transmission lines directly connected to the feeding vias is 0.2 mm.The relevant dimensions of the antenna structure are summarized in Table 1.
The central transmission line, corresponding to the decoupling port of the rat-race coupler, is terminated through a via pin with a 50 resistor soldered on the bottom layer of the board.The simulated RF performance of the developed power divider is shown in Fig. 5b.As it can be noticed, excellent impedance matching characteristics have been achieved in the frequency range from 26 GHz to 31 GHz, in combination with the desired phase offset of 180 • between the two output ports.
The DR is fed by a vertical via pin having diameter of 0.55 mm, and going through the entire PCB as illustrated in Fig. 4, and is directly connected to the mini-SMP connector.

C. SIMULATION RESULTS
The design and optimization of the antenna element and the relevant linear array have been carried out using the commercial electromagnetic field solver CST Microwave Studio c .
The simulated electric and magnetic field distributions excited by the embedded antenna element are shown in Figs.6a and 6b, respectively.As it appears in Fig. 6, the Hfield forms basically a loop when the DR pin is fed, whereas the E-field is directed along the z-direction.As a matter of fact, the resonant mode excited by the vertical feeding pin in the DR is the fundamental one, TM z 011 .Conversely, when the patch port is excited, the E-field forms a loop, while the H-field is directed along the y-direction.This distribution corresponds to the TE y 111 fundamental mode.The first mode is equivalent to an electric dipole, and the second one to a magnetic dipole.Because of the orthogonality between the two dipoles, the coupling between the two ports is very low, typically below the level of −35 dB, in the frequency range between 26 GHz and 30 GHz, as it can be noticed in Fig. 7.
From Fig. 7, it can also be seen that the antenna element shows very broadband properties, with an impedancematching bandwidth of 6 GHz around the center frequency of 27 GHz.

FIGURE 7. Simulated reflection coefficient of the single element and coupling between the two ports.
The simulated multi-mode characteristics of the proposed antenna element are illustrated in Fig. 8.In particular, we can notice how the element radiation pattern is steered towards the −x direction when the relevant input ports are fed with the same phase and towards the +x direction when a phase offset of 180 • is enforced.

III. LINEAR ANTENNA ARRAY
The developed multi-mode antenna element has been integrated into a 1 × 16 linear array.The overall size of such an array is 80.55 mm × 13 mm × 4.88 mm, with the antenna  elements being separated by a distance of 5 mm.The simulation model of the linear antenna array is shown in Fig. 9, where the upper layers of the first radiating element have been hidden in such a way as to allow the view of the relevant feeding network.
The simulated beam steering characteristics of the array are illustrated in Fig. 10, where the enlarged scan range from −75 • to +75 • along the elevation plane can be readily noticed.In order to highlight the radiation properties of the array in terms of scan losses, two dashed lines following the cos θ 0 and the cos 0.5 θ 0 law, respectively, have been added.From basic theory, it is well known that conventional phased antenna arrays are subject to a gain degradation which follows a cos θ 0 law in the ideal case [1].In the presented case, we want to highlight that the scan capabilities of the

FIGURE 11. Improvement of SLL and peak gain by varying the phase difference between the two ports ( φ).
developed antenna array can outperform the ideal case without making use of complex and/or lossy solutions in the design, by only adjusting the phase offset between different antenna ports.
As mentioned above, in order to reach wide scanning capabilities, two radiating modes are excited by controlling the phase offset between the two feeding ports of the individual antenna array element.As it appears from Fig. 11, an optimal phase difference can be identified for each steering direction.Tailored phase offsets can lead not only to better performance in terms of realized gain but also in terms of side lobe level and, eventually, grating lobes.
In order to validate the simulation results, a physical prototype of the linear array has been realized.The manufactured demonstrator is shown in Fig. 12, where one can notice the white cylindrical DRs mounted on the top of the PCB and the mini-SMP connectors soldered on the bottom.The PCB is instead covered by a black solder mask.
The gain pattern measurement setup is shown in Fig. 13 where the antenna array is placed at the center of a mm-wave anechoic chamber which uses an open-ended waveguide as a probe [14].A spherical scan in the far-field region of the device under test is performed so to measure the co-and   cross-polarized components of the gain along both the Eand H-planes.To this end, a suitable mini-SMP adapter was used to characterize each antenna port, while the other ports were terminated with 50 loads.For the sake of conciseness, only the measured characteristics of the central array element at 28 GHz are plotted and compared to the simulated results.Fig. 14 shows the normalized gain pattern of the 8-th element embedded in the linear array along the xz-plane (see Fig. 14a) and the yz-plane (see Fig. 14b) when the DR port is fed, while Fig. 15 shows the normalized gain pattern when the patch port is fed.In both cases, simulation and measurement results are in excellent agreement.
To evaluate the total embedded radiation pattern of each individual array element, the two relevant patterns are combined by post-processing through the following equations: where f 1 θ and f 2 θ are the θ -components of the element pattern when the DR-port and the patch-port are fed, respectively; in the same way, f 1 φ and f 2 φ are the φ-components of the radiation pattern when the DR-port and the patch-port are fed, respectively.Furthermore, a θ and a φ are the unit vectors in the θ and φ direction, respectively.The combined radiation pattern coverage is shown in Fig. 16.By controlling the phase offset at the input ports, it is possible to excite a different radiating mode and, in this way, illuminate a different angular region.In particular,  By combining the two radiating modes, one can achieve an overall HPBW of 150 • that is useful to enable wide-angle scanning at the array level.
Fig. 17 shows the combined radiation patterns of the two corner elements and the central element along the yz-plane (17a) and xz-plane (17b), respectively.The gain pattern of element 1 shows a larger ripple level which is caused by the proximity to the truncation of the ground plane at the corner of the PCB structure.
The measured reflection coefficients of the central element are shown in Fig. 18.In order to perform the S-matrix measurement, suitable adapters, and cable jumpers were adopted.The scattering parameters have been evaluated upon de-embedding the effect of the aforementioned transitions connected between the Vector Network Analyser (VNA) cable and the antenna ports using the method described in [23].This is illustrated in Fig. 18, where the simulated reflection coefficient is compared to the measured one using the cable and the adapter only.
The measurement results in both cases follow the trend of the simulation results, with the only exception of a shift of the resonance around 29 GHz towards higher frequencies (see Fig. 18).This effect results in a reflection coefficient with a magnitude above −10 dB in the frequency range between 26.5 GHz and 28.5 GHz (see Fig. 18b).This is caused by manufacturing tolerances as it has been verified by numerical simulations.
With the same measurement setup and by using the same de-embedding methodology, the isolation between the central antenna element and the adjacent ones has been characterized.The results are plotted in Fig. 19a and Fig. 19b.As one can notice, the coupling level measured in the frequency range from 26.5 GHz to 30 GHz is below −15 dB when the DR port is fed and below -14.2 dB when the patch port is fed.This performance has been achieved while optimizing the input reflection coefficient in such a way as to keep its magnitude below −11 dB (see Fig. 18b).It is worth noting here that the largest coupling level when feeding the DR port is observed towards the DR ports of the adjacent array elements.Such a parasitic process can be reduced by controlling the inter-element spacing.
For a more complete characterization of the array structure, the active reflection coefficient was simulated and compared to the measurement results.The active reflection coefficient ( A ) at the individual antenna port is evaluated through the following expression [24]: with ψ m and ψ n denoting the feeding phase at the m−th and n−th port, respectively.To evaluate the total active reflection coefficient of the individual antenna element ( A E ) we then combine the A relevant to the two ports of each element as follows [25]: The total active reflection coefficient of the individual array element is shown in Fig. 20 for different scan angles from 0 • to 75 • from 25 GHz to 30 GHz.Because of the symmetry of the array structure, only the results related to one sub-space are reported in Fig. 20a and Fig. 20b.As can be noticed, the measured results are somehow worse than the simulated ones.The suboptimal performance of the active reflection coefficients is attributed to manufacturing tolerances.These tolerances can introduce variations in the physical properties of the components, leading to deviations from the desired specifications.While we made efforts to minimize these tolerances during the manufacturing process, some discrepancies may still arise, resulting in non-ideal active reflection coefficients.However, it is worth highlighting that despite these limitations, the prototype implementation successfully allowed us to validate our design concept and evaluate the realized gain of the proposed antenna array.
The measured realized gain, as demonstrated in our research study (see Fig. 14 and Fig. 15), exhibits characteristics that are closely in line with our expectations and numerically simulated predictions.This indicates that while the active reflection coefficients may not be optimal, the overall system performance, as reflected by the measured radiation pattern properties, confirms the benefits of our design strategy.In this context, one can notice that the magnitude of the active reflection coefficient remains below −7 dB (equivalent to a VSWR of approximately 2.6:1) throughout the entire scan range and operating frequency band of the array.
The measured scan performance of the array is documented in Fig. 21.Here the field of view is divided into two sub-spaces, as anticipated in the previous sections.One can notice that thanks to the low gain of the radiating mode in the opposite direction to the main beam, the secondary radiation lobes, as well as any possible undesired grating lobes, are drastically attenuated.This favorable characteristic is even more visible in the color map shown in Fig. 22, where the side-lobe level is plotted as a function of frequency and main-lobe direction.It is worth mentioning here that the proposed antenna array design is not subject to grating lobes.

IV. DISCUSSION
The design solution detailed in this research study has been benchmarked against the current state of the art for phased array antennas with wide scanning capabilities, as documented in the scientific literature.As shown in Table 2, the majority of existing solutions are limited to a scan range of up to ±60 • , whereas the proposed array solution extends the scan range to ±75 • .
While Table 2 indicates comparable overall performance, we would like to emphasize that our design achieves a significantly wider beam scan range.
In comparison to previous works, such as [22], our design incorporates a multi-port antenna solution that enables fine control of the radiation pattern at the individual element level through phase shifting, in such a way to support two radiating modes.
Each mode features maximum radiation at approximately +/-40 • , in contrast to the ideal case, where the maximum radiation is typically observed along the boresight.The latter limits the beam-steering capabilities at the antenna-array level, resulting in a narrower scan range.With our approach instead, by combining these modes, we are able to generate multiple lobes that can be steered independently, thus allowing for a significantly wider scan range at array level.This unique feature enables the proposed antenna array to surpass the performance of the ideal case without the need for complex or expensive technical solutions in the design.
Furthermore, as shown in [1], the degradation of the realized gain featured by a conventional phased array outside the relevant scan range is very sharp and severe, whereas our array design is characterized by a limited gain loss of 3 dB over a very wide scan range from −75 • to +75 • , and of about 6 dB in the even broader angular range from −85 • to +85 • .
The increased scan range is a key feature of our technical proposal that enables greater coverage and flexibility in practical applications.We believe that the enhanced performance sets our design apart from existing approaches and offers distinct advantages in terms of system capabilities.

V. CONCLUSION
In this paper, a novel wide-scanning linear antenna array with multi-mode characteristics is presented.Such an array is based on a novel dual-fed antenna element that can support two radiating modes by controlling the phase difference between the two relevant input ports.The proposed antenna is realized by integrating, in the same physical structure, a capacitively fed patch with a cylindrical dielectric resonator mounted on top of it.By introducing a phase offset of 180 • between the antenna ports, the embedded element pattern can be steered in such a way as to cover two complementary sub-spaces, the first one from −75 • to +20 • and the second one from −20 • to +75 • .Thanks to these characteristics, the developed antenna array is able to scan from −75 • to +75 • with a gain loss smaller than 3 dB (or from −85 • to +85 • with a loss below 6 dB), while most of the alternative array solutions available in the scientific literature cannot scan beyond ±60 • .Furthermore, the proposed design is characterized by a peak side lobe level always below −9.5 dB.Measurement results showed a wide operating frequency band from 27 GHz to 29.5 GHz with an active return loss above 7 dB across the entire scan range.

FIGURE 1 .
FIGURE 1. Array scanning when two radiating modes are used in the array pattern multiplication formula with the array factor (in dashed line).

FIGURE 2 .
FIGURE 2. Basic principle of two complementary unilateral antennas with E-Field of a z-directed electric dipole and with H-Field of a y-directed magnetic dipole (a) in case of same feeding phase between the two ports and (b) in case of 180 • phase offset between the two ports.

FIGURE 3 .
FIGURE 3. (a) Exploded view of the antenna structure with mini-SMP connectors and transparent prepeg layers and (b) Top view of the single element with transparent cylindrical DR.

FIGURE 5 .
FIGURE 5. (a) Feeding network of the capacitively fed patch antenna realized with a rat-race power divider and (b) relevant scattering parameters.

FIGURE 6 .
FIGURE 6.(a) E-Field and H-Field of the TM z 011 mode when the DR port of the embedded central element is activated and (b) E-Field and H-Field of the TE y 111 mode when the patch port of the embedded central element is activated.

FIGURE 8 .
FIGURE 8. (a) 3D realized antenna gain of the two radiating modes of the single antenna element when both ports are fed with the same phase (on the left) and with a phase offset of 180 • (on the right) and (b) 2D cut in the xz-plane.

FIGURE 10 .
FIGURE 10.Simulated scan range of the linear array at 28 GHz.

FIGURE 12 .
FIGURE 12. Top and bottom views of the manufactured linear array.

FIGURE 13 .
FIGURE 13.Antenna under test (AUT) at the center of the mm-wave anechoic chamber.

FIGURE 14 .
FIGURE 14. Measured (black) and simulated (orange) normalized gain pattern of the central element at 28 GHz, in the (a) φ = 90 • -plane and (b) in the φ = 0 • -plane when port 15 is fed and all the other ports terminated on 50 loads.

FIGURE 15 .FIGURE 16 .
FIGURE 15.Measured (black) and simulated (orange) normalized gain pattern of the central element at 28 GHz, in the (a) φ = 90 • -plane and (b) in the φ = 0 • -plane when port 16 is fed and all the other ports terminated on 50 loads.

FIGURE 17 .
FIGURE 17. Normalized gain pattern of the two corner elements and the central element at 28 GHz, (a) in the V-plane and (b) in the H-plane with the two ports fed without phase offset.

FIGURE 19 .
FIGURE 19.Coupling coefficients of (a) port 15 and (b) port 16 with the adjacent elements.

FIGURE 20 .
FIGURE 20.Active reflection coefficient of the central element when radiating (a) in broadside and (b) while scanning, with simulations in solid lines and measurements in dashed lines.

FIGURE 21 .
FIGURE 21.Measured scan range of the linear array at 28 GHz.

FIGURE 22 .
FIGURE 22. Side lobe levels of the linear array while scanning versus frequency as a function of the main-lobe direction.