A Wide-Band High-Efﬁciency Hybrid-Feed Antenna Array for mm-Wave Wireless Systems

: A wide-band and high-efﬁciency planar antenna array with a novel hybrid-feed structure is proposed in this article. By combing the coaxial-line feed magneto-electric (ME) dipoles with the aperture coupled dielectric cavity, a hybrid-feed 2 × 2-unit ME-dipole sub-array is invented. The low-loss ridge gap waveguide (RGWG) corporate-feed network is used to replace the high-loss substrate-based feed networks and high-cost RWG feed networks. New forms of RGWG H-plane divider are designed to build the RGWG feed network. An 8 × 8-unit ME-dipole antenna array is designed and fabricated to verify the validity of the array. The radiation part consists of two layers of a low-cost printed circuit board (PCB), and the feeding part consists of two copper plates manufactured by computer numerical control (CNC) milling. Measured results show that a relative bandwidth of 16.4% with |S 11 | < − 10 dB is achieved, with a maximum radiation efﬁciency of 85%. The stable symmetric radiation patterns are observed in both the E-plane and H-plane, covering the operation band. Based on the measured results, a 16 × 16-unit ME-dipole antenna array is simulated. Results indicate that the proposed array has wide-band and high-efﬁciency features, which is suitable for large-scale array design in mm-wave wireless systems.


Introduction
Due to the advantages of a high data rate and high detection accuracy, mm-wave wireless systems have been widely used in 5G base stations, auto-motive radar, and blast furnace radar [1][2][3]. A wide-band and high-efficiency mm-wave antenna array is critical to the mm-wave wireless system. In some specific applications, large-scale mm-wave antenna arrays are required to achieve high-gain, narrow beam-width, or multi-direction beams features [3][4][5]. In these applications, both the efficiency, bandwidth, and fabrication cost must be considered in the antenna array design.
The performance of an mm-wave antenna array is determined by the antenna unit and feed network. Various mm-wave planar antenna arrays based on low-cost printed circuit board (PCB) technology have been presented [6][7][8]. The slots array are based on a substrate-integrated waveguide (SIW) in [6,7] and substrate-integrated coaxial line (SICL) in [8]; these arrays have the features of a low-profile and low fabrication cost but suffer from a narrow bandwidth. The ME-dipole is a complementary antenna unit with the advantages of a wide impedance matching band and symmetric radiation patterns [9]. Different kinds of ME-dipoles on the mm-wave band have been reported [10][11][12][13]. These mm-wave arrays exhibit wide-band and low-profile features. However, the feed-networks of these arrays all use the substrate-based transmission line while the substrate loss of the array cannot be ignored with the increase in array dimension. Figure 1 shows the geometry of the proposed 2 × 2 ME-dipole sub-array with a hybrid feed. Figure 2 gives the details of the proposed 2 × 2 ME-dipole sub-array by dividing it into a radiation part and feeding part. The radiation part consists of two Teflon layers with a dielectric constant of 2.1 and tangent loss of 0.001. The two layers have different thicknesses. The four coaxial-feed ME-dipoles on the top layer are modified from the units in the article [13]; a pair of short-end patches and the three pairs of metallics, plated via holes, forms the ME-dipole, and the coaxial-feed T-probe and the gaps between two patches forms a GSG transmission structure, which can excite the ME-dipole. The coaxial-feed line on the top layer is connected with the second layer through a metallic pad. A ring pad is added in the feeding point of the T-probe to realize wide-band impedance matching of the ME-dipole. The second layer is the dielectric cavity with four metallic probes. The four pins with depth of hp are concentric with the feed-pin on the top layer; the metallic pads between the feed-pins and coaxial probes act as the impedance-matching section. The feeding part is on the lower two layers; the third layer and forth layer is the air-filled coupling slot and the short-end RGWG section, which are both made of copper.
Electronics 2021, 10, x FOR PEER REVIEW 3 of 16 feeding part is on the lower two layers; the third layer and forth layer is the air-filled coupling slot and the short-end RGWG section, which are both made of copper.

Operation Mechanism
In order to analyze the working principle of the coaxial-feed ME-dipole, the full-wave simulation software ANSYS Electronics Desktop was used. The simulated surface current distributions on the 2 × 2-ME-dipole sub-array are shown in Figure 3. T represents a period of time; it can be seen that at t = 0 and t = T/2, the current on the inner GSG feed (T-shape probe and the gaps between two patches) are dominant, which indicates that the quarterwavelength apertures are excited, equivalent to x-direction magnetic-dipoles. On the other hand, when t = T/4 and t = 3T/4, the currents are mainly concentrated on both sides of the quarter-wavelength short-end patches along the y-direction, which indicates that the x-direction electric dipoles are excited. The electric dipoles and the magnetic dipoles are excited alternatively, which forms the ME-dipole sub-array.  feeding part is on the lower two layers; the third layer and forth layer is the air-filled coupling slot and the short-end RGWG section, which are both made of copper.

Operation Mechanism
In order to analyze the working principle of the coaxial-feed ME-dipole, the full-wave simulation software ANSYS Electronics Desktop was used. The simulated surface current distributions on the 2 × 2-ME-dipole sub-array are shown in Figure 3. T represents a period of time; it can be seen that at t = 0 and t = T/2, the current on the inner GSG feed (T-shape probe and the gaps between two patches) are dominant, which indicates that the quarterwavelength apertures are excited, equivalent to x-direction magnetic-dipoles. On the other hand, when t = T/4 and t = 3T/4, the currents are mainly concentrated on both sides of the quarter-wavelength short-end patches along the y-direction, which indicates that the x-direction electric dipoles are excited. The electric dipoles and the magnetic dipoles are excited alternatively, which forms the ME-dipole sub-array.

Operation Mechanism
In order to analyze the working principle of the coaxial-feed ME-dipole, the full-wave simulation software ANSYS Electronics Desktop was used. The simulated surface current distributions on the 2 × 2-ME-dipole sub-array are shown in Figure 3. T represents a period of time; it can be seen that at t = 0 and t = T/2, the current on the inner GSG feed (T-shape probe and the gaps between two patches) are dominant, which indicates that the quarter-wavelength apertures are excited, equivalent to x-direction magnetic-dipoles. On the other hand, when t = T/4 and t = 3T/4, the currents are mainly concentrated on both sides of the quarter-wavelength short-end patches along the y-direction, which indicates that the x-direction electric dipoles are excited. The electric dipoles and the magnetic dipoles are excited alternatively, which forms the ME-dipole sub-array.
The main function of the feeding part is to excite the 2 × 2-unit ME-dipole through the 1-to-4 power divider, which is composed of the dielectric cavity and coaxial probes. Figure 4 exhibits the simulated electric-field distributions in the dielectric cavity. The cavity is excited by the offset slot etched on the top of the short-end RGWG. The coupling slot is on the central part of the cavity and the TE 230 mode is excited in the dielectric cavity. The EM-wave feed from the RGWG is evenly intercepted by the coaxial probe through the TE 230 mode, and then fed into the ME-dipole sub-array. At each quarter in the period, the electric-field amplitude around the four coaxial probes are nearly equal; the phase is opposite to the y-axis. Therefore, by placing the radiation units on both sides of the y-axis symmetrically, the equal amplitude in-phase radiation of the elements in the sub-array can be realized. The main function of the feeding part is to excite the 2 × 2-unit ME-dipole through the 1-to-4 power divider, which is composed of the dielectric cavity and coaxial probes. Figure 4 exhibits the simulated electric-field distributions in the dielectric cavity. The cavity is excited by the offset slot etched on the top of the short-end RGWG. The coupling slot is on the central part of the cavity and the TE230 mode is excited in the dielectric cavity. The EM-wave feed from the RGWG is evenly intercepted by the coaxial probe through the TE230 mode, and then fed into the ME-dipole sub-array. At each quarter in the period, the electric-field amplitude around the four coaxial probes are nearly equal; the phase is opposite to the y-axis. Therefore, by placing the radiation units on both sides of the y-axis symmetrically, the equal amplitude in-phase radiation of the elements in the sub-array can be realized.

Performance and Parametric Analysis
The operation mechanism of the proposed 2 × 2-unit ME-dipole sub-array has been studied in the previous section. Performance of the radiation part and the feeding part of the ME-dipole sub-array are discussed, respectively, in this section. In addition, several key parameters of the ME-dipole sub-array will be analyzed to give the design guidelines of the antenna. The original dimensions of the single ME-dipole are as follows: the height of the substrate is nearly λg/4, where the λg is the guide wavelength of the substrate. The length of the single ME-dipole (lp) is set to 0.5 λg, the width of each short-end patch (wp) is set to 0.2 λg and the gap (wd) between the two patches is set to 0.1 λg, which forms an  The main function of the feeding part is to excite the 2 × 2-unit ME-dipole through the 1-to-4 power divider, which is composed of the dielectric cavity and coaxial probes. Figure 4 exhibits the simulated electric-field distributions in the dielectric cavity. The cavity is excited by the offset slot etched on the top of the short-end RGWG. The coupling slot is on the central part of the cavity and the TE230 mode is excited in the dielectric cavity. The EM-wave feed from the RGWG is evenly intercepted by the coaxial probe through the TE230 mode, and then fed into the ME-dipole sub-array. At each quarter in the period, the electric-field amplitude around the four coaxial probes are nearly equal; the phase is opposite to the y-axis. Therefore, by placing the radiation units on both sides of the y-axis symmetrically, the equal amplitude in-phase radiation of the elements in the sub-array can be realized.

Performance and Parametric Analysis
The operation mechanism of the proposed 2 × 2-unit ME-dipole sub-array has been studied in the previous section. Performance of the radiation part and the feeding part of the ME-dipole sub-array are discussed, respectively, in this section. In addition, several key parameters of the ME-dipole sub-array will be analyzed to give the design guidelines of the antenna. The original dimensions of the single ME-dipole are as follows: the height of the substrate is nearly λg/4, where the λg is the guide wavelength of the substrate. The length of the single ME-dipole (lp) is set to 0.5 λg, the width of each short-end patch (wp) is set to 0.2 λg and the gap (wd) between the two patches is set to 0.1 λg, which forms an

Performance and Parametric Analysis
The operation mechanism of the proposed 2 × 2-unit ME-dipole sub-array has been studied in the previous section. Performance of the radiation part and the feeding part of the ME-dipole sub-array are discussed, respectively, in this section. In addition, several key parameters of the ME-dipole sub-array will be analyzed to give the design guidelines of the antenna. The original dimensions of the single ME-dipole are as follows: the height of the substrate is nearly λ g /4, where the λ g is the guide wavelength of the substrate. The length of the single ME-dipole (l p ) is set to 0.5 λ g , the width of each short-end patch (w p ) is set to 0.2 λ g and the gap (w d ) between the two patches is set to 0.1 λ g , which forms an electric dipole. The diameter of the metallic vi as was set to 0.6 mm; accordingly, the diameter of the hole on the ground plane was set to 5.6 mm, forming a 50-ohm coaxial feed line. A pad structure is introduced into the transition between the T-shape MS-line probe and the coaxial-line to replace the direct MS-line transition. The widths of the MS-line T-shape probes are set to 0.28 mm and 0.2 mm.
As illustrated in Figure 5a-c, the resonant frequencies are mainly controlled by the dimension of the short-end patches (l p , w p , w d ). Besides, the diameter of the pads also affects the impedance matching bandwidth, which is shown in Figure 5d. By tuning these key parameters, good performance can be achieved for the ME-dipole. Figure 6 gives the radiation patterns and |S11| curves of the optimized ME-dipole, in this paper and [13]; it can be seen that the modified ME-dipole in this paper has a better impedance matching bandwidth, which is beneficial for a sub-array design. line. A pad structure is introduced into the transition between the T-shape MS-line probe and the coaxial-line to replace the direct MS-line transition. The widths of the MS-line Tshape probes are set to 0.28 mm and 0.2 mm.
As illustrated in Figure 5a-c, the resonant frequencies are mainly controlled by the dimension of the short-end patches (lp, wp, wd). Besides, the diameter of the pads also affects the impedance matching bandwidth, which is shown in Figure 5d. By tuning these key parameters, good performance can be achieved for the ME-dipole. Figure 6 gives the radiation patterns and |S11| curves of the optimized ME-dipole, in this paper and [13]; it can be seen that the modified ME-dipole in this paper has a better impedance matching bandwidth, which is beneficial for a sub-array design. line. A pad structure is introduced into the transition between the T-shape MS-line probe and the coaxial-line to replace the direct MS-line transition. The widths of the MS-line Tshape probes are set to 0.28 mm and 0.2 mm. As illustrated in Figure 5a-c, the resonant frequencies are mainly controlled by the dimension of the short-end patches (lp, wp, wd). Besides, the diameter of the pads also affects the impedance matching bandwidth, which is shown in Figure 5d. By tuning these key parameters, good performance can be achieved for the ME-dipole. Figure 6 gives the radiation patterns and |S11| curves of the optimized ME-dipole, in this paper and [13]; it can be seen that the modified ME-dipole in this paper has a better impedance matching bandwidth, which is beneficial for a sub-array design. The main function of the hybrid-feed structure is to divide the EM-wave that couples from the offset slot to the coaxial probes through the resonant mode in the dielectric cavity. The first step is to choose the appropriate resonant mode. When the coupling slot is in the center of the cavity, the TE 230 mode is the lowest resonant mode that can realize the Electronics 2021, 10, 2383 6 of 15 uniform excitation of the four coaxial probes. The relationship between the dimension of the cavity and the resonant mode can be summarized as follow: where w c , l c and h c is the width, length, and height of the cavity; the original value of the three parameters are set to 12 mm, 11.3 mm, and 1.2 mm. According to (1), choosing the dielectric cavity instead of the air-filled cavity can reduce the size of the cavity, which is helpful to reduce the spacing between the adjacent sub-arrays in a large-scale array design.
Considering the influence of the metallic vi as, the length and width of the dielectric cavity were analyzed. The results are shown in Figure 7a. The parameter analysis of the height of the dielectric cavity (h c ) and the depth of the coaxial-probe (h p ) are depicted in Figure 7b.
Results show that the value of h c determines the resonant frequency. In addition, the bandwidth is the widest when the ratio of h p to h c is 2/3 and h c is 1.2 mm. Figure 6. Performance of the original ME-dipole in [13] and modified ME-dipole in this paper. (a) |S11| curves. (b) Radiation patterns in the E-plane. (c) Radiation patterns in the H-plane.
The main function of the hybrid-feed structure is to divide the EM-wave that couples from the offset slot to the coaxial probes through the resonant mode in the dielectric cavity. The first step is to choose the appropriate resonant mode. When the coupling slot is in the center of the cavity, the TE230 mode is the lowest resonant mode that can realize the uniform excitation of the four coaxial probes. The relationship between the dimension of the cavity and the resonant mode can be summarized as follow: where wc, lc and hc is the width, length, and height of the cavity; the original value of the three parameters are set to 12 mm, 11.3 mm, and 1.2 mm. According to (1), choosing the dielectric cavity instead of the air-filled cavity can reduce the size of the cavity, which is helpful to reduce the spacing between the adjacent sub-arrays in a large-scale array design. Considering the influence of the metallic vi as, the length and width of the dielectric cavity were analyzed. The results are shown in Figure 7a. The parameter analysis of the height of the dielectric cavity (hc) and the depth of the coaxial-probe (hp) are depicted in Figure 7b. Results show that the value of hc determines the resonant frequency. In addition, the bandwidth is the widest when the ratio of hp to hc is 2/3 and hc is 1.2 mm.  The second step is to set the spacing between the adjacent ME-dipoles in the x-and ydirection. Figure 7c shows the simulated radiation patterns of the sub-array under different x d and y d values. With the increase in the spacing, the gain of the sub-array gradually increases, and the side-lobe level also rises. When x d = y d = 3.5 mm, the peak gain is achieved. The last step is to choose the dimension of the coupling slot; the slot length (l s ), width (w s ), height (h s ), and offset (x m ) values affect the impedance matching bandwidth. The parameter adjustment process is similar to that in [15,21]. Following the above design guidelines, a good impedance matching bandwidth of the sub-array can be achieved. The optimized details of the sub-array are listed in Table 1. The simulated feeding and radiation performance of the optimized sub-array are depicted in Figure 7d. Results show that the |S 11 |< −10 dB bandwidth is from 26.9 GHz to 32.1 GHz, with a relative bandwidth of 17.9%. The realized gain of the sub-array at 29 GHz is 13.58 dBi, with a radiation efficiency of 90%. The side-lobe level is −13.2 dB and the cross-polarization discrimination (XPD) is 45.4 dB. The simulation results indicate that the proposed hybridfeed ME-dipole sub-array has a wide impedance-matching bandwidth with good radiation performance. In addition, the hybrid-fed sub-array is the key part to combine the wideband coaxial-feed ME-dipoles with the low-loss and low-cost gap waveguide feed-network, which is suitable for a large-scale mm-wave antenna array design. Table 1. Details of the proposed hybrid-feed 2 × 2-unit ME-dipole sub-array (unit: mm).

Parameter
Value Parameter Value

RGWG Feed Network Design and Array Configuration
Based on the proposed 2 × 2-ME-dipole sub-array, an 8 × 8-unit ME-dipole antenna array was designed and fabricated. The proposed 8 × 8 ME-dipole array is fed by a 1-to-16 RGWG corporate-feed network. The distance between the neighboring sub-arrays are 12.8 mm (1.28 λ 0 ) in both the x-direction and y-direction. The basic unit of the feed network is an equal power RGWG H-plane T-junction, as shown in Figure 8a. Three RGWG sections are introduced into the RGWG H-T as the impedance transformers. The power splitting ratio of the output ports and the impedance matching bandwidth of the input port can be tuned by adjusting the ridge height and length of the three RGWG sections. It is worth mentioning that by adjusting the power splitting ration based on this method, the phase difference between the output ports can be controlled within a small range. Figure 8b exhibits the simulated S-parameters of the equal power RGWG H-T and unequal power RGWG H-T-junction (power-splitting ration = 3.5 dB). Results show that the phase difference between the output ports is within ±5 • and the |S 11 | is less than −15 dB over the frequency band from 26 GHz to 36 GHz for the two RGWG H-T-junctions.
The total 1-to-16 feed-network is shown in Figure 9a, and the simulated performance of the feed-network is shown in Figure 9b; the phase and amplitude differences among all the output ports are within ±5 • and ±0.2 dB; and the |S11| of the input port is less than −17.5 dB, covering the bandwidth of 26 GHz~32 GHz. Tolerance of the gap height is also simulated and verified, and the results are depicted in Figure 6. Results shows that the designed feed network still works normally within the gap height error range of ±0.04 mm. RGWG H-T-junction (power-splitting ration = 3.5 dB). Results show that the phase difference between the output ports is within ±5° and the |S11| is less than −15 dB over the frequency band from 26 GHz to 36 GHz for the two RGWG H-T-junctions.
(a) (b) The total 1-to-16 feed-network is shown in Figure 9a, and the simulated performance of the feed-network is shown in Figure 9b; the phase and amplitude differences among all the output ports are within ±5° and ±0.2 dB; and the |S11| of the input port is less than −17.5 dB, covering the bandwidth of 26 GHz~32 GHz. Tolerance of the gap height is also simulated and verified, and the results are depicted in Figure 6. Results shows that the designed feed network still works normally within the gap height error range of ±0.04 mm.

Measurement of the 8 × 8-Unit ME-Dipole Antenna Array
In order to verify the validity of the proposed antenna array, a prototype of the 8 × 8unit hybrid-feed ME-dipole antenna array prototype was fabricated and tested. The configuration and photograph of the 8 × 8-unit hybrid-feed ME-dipole antenna array is shown in Figure 10. The radiation part is made of two Teflon layers with a thickness of 1.6 mm and 1.2 mm. The feeding part is located in lower two layers, which are made of copper with a thickness of 1.2 mm and 2.1 mm. A ridged gap waveguide to WR-30 rectangular waveguide vertical transition was invented as the feed port, which is shown in Figure 11. The optimized structure dimensions are shown in Table 2. The S-parameters of the vertical  The total 1-to-16 feed-network is shown in Figure 9a, and the simulated performance of the feed-network is shown in Figure 9b; the phase and amplitude differences among all the output ports are within ±5° and ±0.2 dB; and the |S11| of the input port is less than −17.5 dB, covering the bandwidth of 26 GHz~32 GHz. Tolerance of the gap height is also simulated and verified, and the results are depicted in Figure 6. Results shows that the designed feed network still works normally within the gap height error range of ±0.04 mm.

Measurement of the 8 × 8-Unit ME-Dipole Antenna Array
In order to verify the validity of the proposed antenna array, a prototype of the 8 × 8unit hybrid-feed ME-dipole antenna array prototype was fabricated and tested. The configuration and photograph of the 8 × 8-unit hybrid-feed ME-dipole antenna array is shown in Figure 10. The radiation part is made of two Teflon layers with a thickness of 1.6 mm and 1.2 mm. The feeding part is located in lower two layers, which are made of copper with a thickness of 1.2 mm and 2.1 mm. A ridged gap waveguide to WR-30 rectangular waveguide vertical transition was invented as the feed port, which is shown in Figure 11. The optimized structure dimensions are shown in Table 2. The S-parameters of the vertical

Measurement of the 8 × 8-Unit ME-Dipole Antenna Array
In order to verify the validity of the proposed antenna array, a prototype of the 8 × 8-unit hybrid-feed ME-dipole antenna array prototype was fabricated and tested. The configuration and photograph of the 8 × 8-unit hybrid-feed ME-dipole antenna array is shown in Figure 10. The radiation part is made of two Teflon layers with a thickness of 1.6 mm and 1.2 mm. The feeding part is located in lower two layers, which are made of copper with a thickness of 1.2 mm and 2.1 mm. A ridged gap waveguide to WR-30 rectangular waveguide vertical transition was invented as the feed port, which is shown in Figure 11. The optimized structure dimensions are shown in Table 2. The S-parameters of the vertical transition is shown in Figure 12. Results indicate that the transition structure has a reflection coefficient of less than −15 dB in the frequency range of 26.2 to 31.4 GHz, and the insertion loss of the transition is less than 0.05 dB in the same frequency range. The total size of the array is 67.5 mm × 67.5 mm × 6.25 mm, and the effective radiation size is 56 mm × 56 mm.
transition is shown in Figure 12. Results indicate that the transition structure has a reflection coefficient of less than −15 dB in the frequency range of 26.2 to 31.4 GHz, and the insertion loss of the transition is less than 0.05 dB in the same frequency range. The total size of the array is 67.5 mm × 67.5 mm × 6.25 mm, and the effective radiation size is 56 mm × 56 mm.    transition is shown in Figure 12. Results indicate that the transition structure has a reflection coefficient of less than −15 dB in the frequency range of 26.2 to 31.4 GHz, and the insertion loss of the transition is less than 0.05 dB in the same frequency range. The total size of the array is 67.5 mm × 67.5 mm × 6.25 mm, and the effective radiation size is 56 mm × 56 mm.      The fabricated prototype of the 8 × 8-unit hybrid-feed ME-dipole array is shown in Figure 13. The radiation part is made up of low-cost PCB technique and the two layers are bonded by conductive adhesive process. Metallic air-filled holes are used to replace the metallic pins in the array. The feed network of the array was made up of copper and machined by CNC milling; the smallest size was 0.5 mm. The radiation part and the feeding part was assembled by eight M2 screws. Two 1.6 mm in diameter locating pins were used to align the four layers, so as to ensure the alignment error can be controlled within ±0.02 The fabricated prototype of the 8 × 8-unit hybrid-feed ME-dipole array is shown in Figure 13. The radiation part is made up of low-cost PCB technique and the two layers are bonded by conductive adhesive process. Metallic air-filled holes are used to replace the metallic pins in the array. The feed network of the array was made up of copper and machined by CNC milling; the smallest size was 0.5 mm. The radiation part and the feeding part was assembled by eight M2 screws. Two 1.6 mm in diameter locating pins were used to align the four layers, so as to ensure the alignment error can be controlled within ±0.02 mm. Tolerance of the array was analyzed by the full-wave simulation software HFSS, to ensure a better performance of the array. The fabricated prototype of the 8 × 8-unit hybrid-feed ME-dipole array is shown in Figure 13. The radiation part is made up of low-cost PCB technique and the two layers are bonded by conductive adhesive process. Metallic air-filled holes are used to replace the metallic pins in the array. The feed network of the array was made up of copper and machined by CNC milling; the smallest size was 0.5 mm. The radiation part and the feeding part was assembled by eight M2 screws. Two 1.6 mm in diameter locating pins were used to align the four layers, so as to ensure the alignment error can be controlled within ±0.02 mm. Tolerance of the array was analyzed by the full-wave simulation software HFSS, to ensure a better performance of the array. The reflection coefficient of the fabricated array was measured by a Keysight E8361A VNA and the radiation performance was measured by a mm-wave far-field antenna measurement system, which is shown in Figure 14. The measured and simulated reflection coefficient are plotted in Figure 15. The measured bandwidth with |S11| < −10 dB is from 26.05 GHz to 31.15 GHz. The trend of the measured curve is in good agreement with the simulated results. The measured and simulated gain vs. frequency curves are also given in Figure 15. The measured maximum gain is 25.15 dBi and the maximum radiation efficiency is 89% over the operation bandwidth. A 0.4 dB gain difference can be observed between the measured and simulated ideal gain, which may mainly be caused by the surface roughness of the feed network and the fabrication error. In order to verify the gain difference, the metal surface roughness and fabrication error are all calculated into the electric conductivity of the copper that is used in feed network; the value is 4.2 × 10 7 S/m. The simulated gain vs. frequency curve under this situation is also given in Figure 15. It has good consistency with the measured curve, which indicates that the estimation result can be used in a large-scale array design. The reflection coefficient of the fabricated array was measured by a Keysight E8361A VNA and the radiation performance was measured by a mm-wave far-field antenna measurement system, which is shown in Figure 14. The measured and simulated reflection coefficient are plotted in Figure 15. The measured bandwidth with |S 11 | < −10 dB is from 26.05 GHz to 31.15 GHz. The trend of the measured curve is in good agreement with the simulated results. The measured and simulated gain vs. frequency curves are also given in Figure 15. The measured maximum gain is 25.15 dBi and the maximum radiation efficiency is 89% over the operation bandwidth. A 0.4 dB gain difference can be observed between the measured and simulated ideal gain, which may mainly be caused by the surface roughness of the feed network and the fabrication error. In order to verify the gain difference, the metal surface roughness and fabrication error are all calculated into the electric conductivity of the copper that is used in feed network; the value is 4.2 × 10 7 S/m. The simulated gain vs. frequency curve under this situation is also given in Figure 15. It has good consistency with the measured curve, which indicates that the estimation result can be used in a large-scale array design.  The measured radiation patterns in the E-plane and H-plane are shown in Figure 16. The stable radiation patterns can be observed over the operation band from 27 GHz to 31 GHz. The measured first side-lobe levels are less than −12.5 dB in both the E-plane and H-plane, respectively, with the cross-polarization level better than 35 dB in both planes.  The measured radiation patterns in the E-plane and H-plane are shown in Figure 16. The stable radiation patterns can be observed over the operation band from 27 GHz to 31 GHz. The measured first side-lobe levels are less than −12.5 dB in both the E-plane and Hplane, respectively, with the cross-polarization level better than 35 dB in both planes.  The measured radiation patterns in the E-plane and H-plane are shown in Figure 16. The stable radiation patterns can be observed over the operation band from 27 GHz to 31 GHz. The measured first side-lobe levels are less than −12.5 dB in both the E-plane and Hplane, respectively, with the cross-polarization level better than 35 dB in both planes.

Analysis and Comparison of the Hybrid-Feed ME-Dipole Antenna Array
Based on the measurement results of the 8 × 8-unit hybrid-feed ME-dipole array, a 16 × 16-unit hybrid-feed ME-dipole array was further designed and simulated. The configuration of the 16 × 16-unit ME-dipole antenna array is shown in Figure 17.

Analysis and Comparison of the Hybrid-Feed ME-Dipole Antenna Array
Based on the measurement results of the 8 × 8-unit hybrid-feed ME-dipole array, a 16 × 16-unit hybrid-feed ME-dipole array was further designed and simulated. The configuration of the 16 × 16-unit ME-dipole antenna array is shown in Figure 17.

Analysis and Comparison of the Hybrid-Feed ME-Dipole Antenna Array
Based on the measurement results of the 8 × 8-unit hybrid-feed ME-dipole array, a 16 × 16-unit hybrid-feed ME-dipole array was further designed and simulated. The configuration of the 16 × 16-unit ME-dipole antenna array is shown in Figure 17. The simulated feeding and radiation performance of the 16 × 16-unit ME-dipole antenna array are shown in Figure 18. The influence of machining error and metal surface roughness on array performance can be summarized as the decrease in metal conductivity The simulated feeding and radiation performance of the 16 × 16-unit ME-dipole antenna array are shown in Figure 18. The influence of machining error and metal surface roughness on array performance can be summarized as the decrease in metal conductivity used in feed-network. The conductivity of the copper used in feed-network is set to 4.2 × 10 7 S/m, which is equal to the value derived from the measured results of the 8 × 8unit hybrid-feed ME-dipole array. Simulated results shows that the relative bandwidth with |S 11 | < −15 dB is 12%. The simulated gain of the 16 × 16 array is higher than 30.13 dBi over the band of 27 GHz to 31 GHz, with the radiation efficiency better than 76%. The simulated E-plane and H-plane radiation patterns are given in Figure 19a-c. The first side-lobe level of both planes at 27 GHz, 29 GHz, and 31 GHz are lower than −13 dB and the XPD are all better than 35 dB. The simulated results indicate that the proposed planar array has wide-band and high-efficiency features.
Electronics 2021, 10, x FOR PEER REVIEW 13 of 16 used in feed-network. The conductivity of the copper used in feed-network is set to 4.2 × 10 7 S/m, which is equal to the value derived from the measured results of the 8 × 8-unit hybrid-feed ME-dipole array. Simulated results shows that the relative bandwidth with |S11| < −15 dB is 12%. The simulated gain of the 16 × 16 array is higher than 30.13 dBi over the band of 27 GHz to 31 GHz, with the radiation efficiency better than 76%. The simulated E-plane and H-plane radiation patterns are given in Figure 19a-c. The first side-lobe level of both planes at 27 GHz, 29 GHz, and 31 GHz are lower than −13 dB and the XPD are all better than 35 dB. The simulated results indicate that the proposed planar array has wideband and high-efficiency features.    Table 3 lists the performance comparison of the planar mm-wave antenna arrays, including the fabrication technology, impedance matching bandwidth, maximum gain, radiation efficiency, XPD, array dimensions, and fabrication cost. By comparison with the substrate-integrated coaxial line slot array in [7] and printed GWG feed slot array in [6], the proposed antenna array in this work has a wider bandwidth and higher radiation efficiency. The ME-dipole array fed by gap waveguide or SIW are fabricated by PCB technology in [10,11,19]; the three antennas have a good impedance-matching bandwidth with a low fabrication cost. However, the insertion loss caused by the substrate-based feed network increases with the expansion of the array scale, which reduces the radiation efficiency of the large-scale array antennas. Furthermore, the LTCC process used in [10] also has a high fabrication cost. The full corporate-feed slot array in [15] has a better radiation efficiency with a wide-band feature but the diffusion bonding process costs high. The pure GWG slot antenna array in [18] has a high radiation efficiency with reduced fabrication cost, but the bandwidth is narrow. In addition, the antenna array in [21] combine the MEdipole with the GWG feed network, but the aperture efficiency is limited since the feed network occupies much of the radiation aperture. By comparison with the aforementioned antenna arrays, the proposed hybrid-feed ME-dipoles antenna array in this paper gives an approach for a wide-band, high-efficiency, large-scale antenna array with moderate fabrication cost.  Table 3 lists the performance comparison of the planar mm-wave antenna arrays, including the fabrication technology, impedance matching bandwidth, maximum gain, radiation efficiency, XPD, array dimensions, and fabrication cost. By comparison with the substrate-integrated coaxial line slot array in [7] and printed GWG feed slot array in [6], the proposed antenna array in this work has a wider bandwidth and higher radiation efficiency. The ME-dipole array fed by gap waveguide or SIW are fabricated by PCB technology in [10,11,19]; the three antennas have a good impedance-matching bandwidth with a low fabrication cost. However, the insertion loss caused by the substrate-based feed network increases with the expansion of the array scale, which reduces the radiation efficiency of the large-scale array antennas. Furthermore, the LTCC process used in [10] also has a high fabrication cost. The full corporate-feed slot array in [15] has a better radiation efficiency with a wide-band feature but the diffusion bonding process costs high. The pure GWG slot antenna array in [18] has a high radiation efficiency with reduced fabrication cost, but the bandwidth is narrow. In addition, the antenna array in [21] combine the ME-dipole with the GWG feed network, but the aperture efficiency is limited since the feed network occupies much of the radiation aperture. By comparison with the aforementioned antenna arrays, the proposed hybrid-feed ME-dipoles antenna array in this paper gives an approach for a wide-band, high-efficiency, large-scale antenna array with moderate fabrication cost. * The aperture efficiency can be calculated as η = Gλ 0 /(4π Ae), where G is the gain, λ 0 is the wavelength at the central frequency, and A e is the physical aperture area of the array [7]. ** Simulated results.

Conclusions
A low fabrication cost hybrid-feed ME-dipole antenna array with wide-band highefficiency features for a mm-wave large-scale array design is presented in this paper. A novel hybrid-feed 2 × 2-unit ME-dipole sub-array combine the coaxial-feed ME-dipoles with an aperture-coupled dielectric cavity is proposed as the basic unit of the array. A 16 × 16-unit and an 8 × 8-unit antenna array were designed, and the latter array was fabricated. The corporate-feed network built by the new RGWG H-T-junction was adopted to feed the arrays. The total array can be fabricated by low-cost PCB technology, CNC milling, and assembled without any expensive bonding process. The measured results show that the proposed array can achieve a relative bandwidth of 18% with |S 11 | < −10 dB and a maximum radiation efficiency of 80%. The proposed hybrid-feed ME-dipole antenna array has potential in mm-wave wireless system applications that require large-scale arrays.

Conflicts of Interest:
The authors declare no conflict of interest.