A Broadband Differential-Fed Dual-Polarized Hollow Cylindrical Dielectric Resonator Antenna for 5G Communications

A broadband differential-fed dual-polarized hollow cylindrical dielectric resonator antenna (DRA) is proposed in this article. It makes use of the HEM111, HEM113, and HEM115 modes of the cylindrical hollow DRA. The proposed DRA is simply fed by two pairs of conducting strips and each pair of strips is provided with the out-of-phase signals. After introducing four disconnected air holes into the DRA, a broadband characteristic is achieved, with little effect on the antenna gain of its higher-order modes. To verify this idea, frosted K9-glass is applied to fabricate the hollow cylindrical DRA. The differential S-parameters, radiation patterns, and antenna gain of the DRA are studied. It is found that the proposed differential-fed dual-polarized DRA is able to provide a broad differential impedance bandwidth of ~68% and a high differential-port isolation better than ~46 dB. Moreover, symmetrical broadside radiation patterns are observed across the whole operating band. The proposed DRA covers the frequency bands including the 5G-n77 (3.4–4.2 GHz), 5G-n79 (4.4–5.0 GHz), WLAN-5.2 GHz (5.15–5.35 GHz), and WLAN-5.8 GHz (5.725–5.825 GHz), which can be used for 5G communications.


Introduction
The dielectric resonator antenna (DRA) [1][2][3] has been widely investigated in the past three decades. It has lots of attractive characteristic such as light weight, ease of excitation, and being bandwidth-controllable. In recent years, the dual-polarized antenna has received more and more attention. Its main advantage is to save the number of antennas in the base-station system. In addition, it can improve the channel capacity and reduce co-channel interference. As a result, many studies on the dual-polarized DRA have been reported [4][5][6][7][8][9][10]. For example, in [4], a dual-polarized DRA fed by a coplanar waveguide is proposed. In this design, even and odd modes of the coplanar waveguide structure are utilized, with a relatively narrow impedance bandwidth of~7% obtained. The bandwidth of the dual-polarized DRA can be enlarged by using suitable excitation. For instance, in [5], two pairs of balanced probes are used to excite the two orthogonal mode of the HEM 111 mode of the cylindrical DRA, with an impedance bandwidth of~20% obtained. In addition, the dual-mode method can be used to widen the bandwidth. For instance, in [6], a dual-polarized cylindrical DRA is designed using its HEM 111 and HEM 113 modes, with an impedance bandwidth of~30% reported. The disadvantage is

S Parameters of the Dual-Polarized Antennas Using the Single-Ended and Differential Ports
In designing the dual-polarized antennas, the port matching and port isolation are needed to considered. Table 1 compares the S parameters of the dual-polarized antenna with two single-ended ports and differential ports. When the ports are single-ended, the port matching (S 11 and S 22 ) and port isolation (S 12 and S 21 ) are considered. However, it is different as the ports are differential ones. In this condition, the port matching is defined as S dd11 and S dd22 , while the port isolation is defined as S dd12 and S dd21 . In [38][39][40], a dual-polarized antenna with two differential-ports is considered as a single-ended four-ports network. Its differential S parameters is given as: where S ij (i = 1, 2, 3, 4; j = 1, 2, 3, 4) denotes the single-ended S-parameters. As can be seen from the above formulas, the matching of the differential ports (S dd11 and S dd22 ) are related to the matching (S 11 and S 22 ) and isolation (S 12 and S 21 ) of the single-ended ports. where Sij (i = 1, 2, 3, 4; j = 1, 2, 3, 4) denotes the single-ended S-parameters. As can be seen from the above formulas, the matching of the differential ports (Sdd11 and Sdd22) are related to the matching (S11 and S22) and isolation (S12 and S21) of the single-ended ports.

Port Distribution S parameter Diagrammatic Sketch
Two singleended ports Port matching: S11, S22 Port isolation: S21, S12 Two differential ports Port matching: Port isolation:

Strips-Fed Dual-Polarized Cylindrical DRA Using Single-Ended and Differential Ports
In this section, a strips-fed cylindrical DRA is studied to compare the performance of the dualpolarized antennas using the single-ended and differential ports. Figure 1 shows the structure of the DRAs, where Figure 1a uses two single-ended ports and Figure 1b employs two differential ports. The cylindrical DRA (εr = 6.85) has a radius of R = 12 mm and a height of H = 10 mm. For each port, a strip having a width of Ws = 1 mm and a length of Ls = 9 mm is adhered to the sidewall of the cylindrical DRA, which is used to excite its fundamental HEM111 mode. Figure 2 shows the port matching for both cases. Referring to the figure, the 10dB-impedance bandwidth of the single-ended (|S11| ≤ −10dB) and differential (|Sdd11| ≤ −10dB) versions are obtained as ~16% (3.98-4.65 GHz) and ~17% (3.77-4.49 GHz), respectively. Figure 3 shows the port isolation for both cases. It can be seen from the figure that the single-ended port isolation is only higher than 16 dB (3.98-4.65 GHz), while the port isolation of the differential version is higher than 45 dB . This shows the isolation of the differential version is better than that of the single-ended. The reason is given here. For the differential version (Figure 1b), ports 3 and 4 are symmetrical about the plane of ports 1 and 2. Under this condition, S13 = S14 and S23 = S24 are obtained, resulting in Sdd12 = 0. In other words, the isolation of the differential version is infinite in theory. Port where Sij (i = 1, 2, 3, 4; j = 1, 2, 3, 4) denotes the single-ended S-parameters. As can be seen from the above formulas, the matching of the differential ports (Sdd11 and Sdd22) are related to the matching (S11 and S22) and isolation (S12 and S21) of the single-ended ports.

Port Distribution S parameter Diagrammatic Sketch
Two singleended ports Port matching: S11, S22 Port isolation: S21, S12 Two differential ports Port matching: Port isolation:

Strips-Fed Dual-Polarized Cylindrical DRA Using Single-Ended and Differential Ports
In this section, a strips-fed cylindrical DRA is studied to compare the performance of the dualpolarized antennas using the single-ended and differential ports. Figure 1 shows the structure of the DRAs, where Figure 1a uses two single-ended ports and Figure 1b employs two differential ports. The cylindrical DRA (εr = 6.85) has a radius of R = 12 mm and a height of H = 10 mm. For each port, a strip having a width of Ws = 1 mm and a length of Ls = 9 mm is adhered to the sidewall of the cylindrical DRA, which is used to excite its fundamental HEM111 mode. Figure 2 shows the port matching for both cases. Referring to the figure, the 10dB-impedance bandwidth of the single-ended (|S11| ≤ −10dB) and differential (|Sdd11| ≤ −10dB) versions are obtained as ~16% (3.98-4.65 GHz) and ~17% (3.77-4.49 GHz), respectively. Figure 3 shows the port isolation for both cases. It can be seen from the figure that the single-ended port isolation is only higher than 16 dB (3.98-4.65 GHz), while the port isolation of the differential version is higher than 45 dB . This shows the isolation of the differential version is better than that of the single-ended. The reason is given here. For the differential version (Figure 1b), ports 3 and 4 are symmetrical about the plane of ports 1 and 2. Under this condition, S13 = S14 and S23 = S24 are obtained, resulting in Sdd12 = 0. In other words, the isolation of the differential version is infinite in theory. Port

Antenna
Port isolation:

Strips-Fed Dual-Polarized Cylindrical DRA Using Single-Ended and Differential Ports
In this section, a strips-fed cylindrical DRA is studied to compare the performance of the dual-polarized antennas using the single-ended and differential ports. Figure 1 shows the structure of the DRAs, where Figure 1a uses two single-ended ports and Figure 1b employs two differential ports. The cylindrical DRA (εr = 6.85) has a radius of R = 12 mm and a height of H = 10 mm. For each port, a strip having a width of W s = 1 mm and a length of L s = 9 mm is adhered to the sidewall of the cylindrical DRA, which is used to excite its fundamental HEM 111 mode. Figure 2 shows the port matching for both cases. Referring to the figure, the 10dB-impedance bandwidth of the single-ended (|S 11 | ≤ −10dB) and differential (|S dd11 | ≤ −10dB) versions are obtained as~16% (3.98-4.65 GHz) and 17% (3.77-4.49 GHz), respectively. Figure 3 shows the port isolation for both cases. It can be seen from the figure that the single-ended port isolation is only higher than 16 dB (3.98-4.65 GHz), while the port isolation of the differential version is higher than 45 dB (3.77-4.49 GHz). This shows the isolation of the differential version is better than that of the single-ended. The reason is given here. For the differential version (Figure 1b), ports 3 and 4 are symmetrical about the plane of ports 1 and 2. Under this condition, S 13 = S 14 and S 23 = S 24 are obtained, resulting in S dd12 = 0. In other words, the isolation of the differential version is infinite in theory.   Figure 1. The |S11| is for DRA 1 (single-ended port) and |Sdd11| is for DRA 2 (differential port).   Figure 1. The |S11| is for DRA 1 (single-ended port) and |Sdd11| is for DRA 2 (differential port). The |S 11 | is for DRA 1 (single-ended port) and |S dd11 | is for DRA 2 (differential port).

LS
Sensors 2020, 20, x 5 of 17 Figure 3. Port isolation of the strips-fed dual-polarized cylindrical DRAs shown in Figure 1. The |S12| is for DRA 1(single-ended port) and |Sdd12| is for DRA 2 (differential port). Figure 4 shows the configuration of the proposed broadband differential-fed dual-polarized  Figure 1. The |S 12 | is for DRA 1(single-ended port) and |S dd12 | is for DRA 2 (differential port).  Figure 4 shows the configuration of the proposed broadband differential-fed dual-polarized hollow cylindrical DRA locating on a ground plane of 12 × 12 cm 2 . The hollow cylindrical DRA made of frosted K9 glass (ε r = 6.85) has a radius of R = 11 mm and a height of H = 25 mm (H/R = 2.27). It should be mentioned that the value of H/R cannot be too small, otherwise the two higher-order modes (HEM 113 and HEM 115 modes) would be far away from the fundamental mode (HEM 111 mode) [41] and they will not be able to merge. In our experience, H/R should be larger than 2. Four approximate rectangular air holes are dug at the bottom of the DRA. Each air hole has a length of a = 6 mm, a width of b = 7 mm, and a height of d = 9 mm. Four metal strips were employed to excite the hollow cylindrical DRA, which has a length of Ls = 9 mm and a width of Ws = 1 mm. These strips are connected with four single-ended ports (1, 2, 3, 4), which form two differential ports (1+, 1−) and (2+, 2−), respectively.  Figure 1. The |S12| is for DRA 1(single-ended port) and |Sdd12| is for DRA 2 (differential port). Figure 4 shows the configuration of the proposed broadband differential-fed dual-polarized hollow cylindrical DRA locating on a ground plane of 12 × 12 cm 2 . The hollow cylindrical DRA made of frosted K9 glass (εr = 6.85) has a radius of R = 11 mm and a height of H = 25 mm (H/R = 2.27). It should be mentioned that the value of H/R cannot be too small, otherwise the two higher-order modes (HEM113 and HEM115 modes) would be far away from the fundamental mode (HEM111 mode) [41] and they will not be able to merge. In our experience, H/R should be larger than 2. Four approximate rectangular air holes are dug at the bottom of the DRA. Each air hole has a length of a = 6 mm, a width of b = 7 mm, and a height of d = 9 mm. Four metal strips were employed to excite the hollow cylindrical DRA, which has a length of Ls = 9 mm and a width of Ws = 1 mm. These strips are connected with four single-ended ports (1, 2, 3, 4), which form two differential ports (1+, 1−) and (2+, 2−), respectively.

The Effect of the Air Region
The effect of the air region on the antenna performance is discussed here. Figure 5 shows the solid cylindrical DRA (Figure 5a), as well as three kinds of hollow DRAs (Figure 5b-d) with different air region distributions. The excitation method and ground plane used in Figure 5 is the same as that in Figure 4. For brevity, the ground plane is not shown here. The corresponding simulated |S dd11 | at differential-port 1 is shown in Figure 6. It should be mentioned that the definition of the differential S parameters for a differential-fed dual-polarized antenna has been given in Table 1. With reference to the result of the |S dd11 | for DRA I, the HEM 111 and HEM 113 mode of the cylindrical DRA are excited at 3.31 and 4.78 GHz, respectively. However, its matching around 4.0 GHz is not very good, which makes it impossible to achieve a wideband antenna. A similar result of |S dd11 | is observed for DRA II having a rectangular hollow region at its center. To further enhance the bandwidth, DRA III and DRA IV with four approximate rectangular air holes are designed, in which the air holes are connected in DRA III and disconnected in DRA IV. With reference to the results of |S dd11 | for DRA III and DRA IV, the HEM 111 , Sensors 2020, 20, 6448 6 of 16 HEM 113 and HEM 115 modes of the DRA are successfully excited to form a broadband DRA. This is reasonable because the insertion of the air region inside the DRA reduces its effective dielectric constant, and therefore enhances overall bandwidth. Furthermore, as can be seen from Figure 7, the antenna gain of the DRA III ranging from 6.5 to 7.0 GHz has a sharp decline; this is because the hollow region at the center destroys the first standing wave of the HEM 115 mode more seriously. The above discussion implies that the total volume of the hollow region should be large enough so that the differential impedance bandwidth can be obviously increased. However, the hollow region should be avoided in the middle of the DRA, so as to minimize the influence on the antenna gain of the higher-order HEM 115 mode. In addition, it should be mentioned that good isolation |S dd12 | between the two differential-ports is obtained for the four cases, with their results shown in Figure 8.
to the result of the |Sdd11| for DRA I, the HEM111 and HEM113 mode of the cylindrical DRA are excited at 3.31 and 4.78 GHz, respectively. However, its matching around 4.0 GHz is not very good, which makes it impossible to achieve a wideband antenna. A similar result of |Sdd11| is observed for DRA II having a rectangular hollow region at its center. To further enhance the bandwidth, DRA III and DRA IV with four approximate rectangular air holes are designed, in which the air holes are connected in DRA III and disconnected in DRA IV. With reference to the results of |Sdd11| for DRA III and DRA IV, the HEM111, HEM113 and HEM115 modes of the DRA are successfully excited to form a broadband DRA. This is reasonable because the insertion of the air region inside the DRA reduces its effective dielectric constant, and therefore enhances overall bandwidth. Furthermore, as can be seen from Figure 7, the antenna gain of the DRA III ranging from 6.5 to 7.0 GHz has a sharp decline; this is because the hollow region at the center destroys the first standing wave of the HEM115 mode more seriously. The above discussion implies that the total volume of the hollow region should be large enough so that the differential impedance bandwidth can be obviously increased. However, the hollow region should be avoided in the middle of the DRA, so as to minimize the influence on the antenna gain of the higher-order HEM115 mode. In addition, it should be mentioned that good isolation |Sdd12| between the two differential-ports is obtained for the four cases, with their results shown in Figure 8.

Resonant Modes
The resonant modes of the proposed hollow cylindrical DRA (DRA IV) are verified by the corresponding near field in this section. The H-fields (x-z plane) of the proposed DRA at the resonance frequencies are given in Figure 9. The analysis is similar with that in [42]. With reference to Figure 9a, the typical H-field of the HEM 111 y mode was found at 3.56 GHz, which has one energy concentration zone inside the DRA. Referring to Figure 9b,c, two and three energy concentration zones along z-direction were observed, respectively. This denotes the resonant modes at 4.88 GHz and 6.08 GHz are caused by the HEM 113 y and HEM 115 y modes, respectively.
resonance frequencies are given in Figure 9. The analysis is similar with that in [42]. With reference to Figure 9a, the typical H-field of the HEM111 y mode was found at 3.56 GHz, which has one energy concentration zone inside the DRA. Referring to Figure 9b,c, two and three energy concentration zones along z-direction were observed, respectively. This denotes the resonant modes at 4.88 GHz and 6.08 GHz are caused by the HEM113 y and HEM115 y modes, respectively.

Parametric Study
In this part, parametric study is carried out to characterize the proposed DRA. Figure 10 shows the simulated |Sdd11| versus frequency with the length of the strips Ls = 7, 8, and 9 mm. Referring to the figure, increasing Ls would improve the matching of the DRA, and a good result is obtained as Ls = 9 mm. Its inset gives the simulated |Sdd11| versus frequency with the width of the strips Ws = 0.5, 1, and 1.5 mm. As can be seen from the figure, altering Ws has little effect on the matching of the DRA. Next, the sizes of the hollow region on the effect of the DRA performance is investigated. Figure 11 shows the simulated |Sdd11| versus frequency with the length of the hollow region a = 4, 5, and 6 mm. Its inset shows the simulated |Sdd11| versus frequency with the width of the hollow region b = 5, 6, and 7 mm. With reference to the figure, increasing a and b would shift the third resonance modes obviously upward, and thus enhance the differential impedance bandwidth of the DRA. Figure 12 presents the simulated |Sdd11| of the proposed DRA versus frequency for different d = 8, 9, and 10 mm. Referring to the figure, changing d has a small effect on the differential impedance bandwidth of the DRA. However, a larger d would have a negative effect on the antenna gain around 6.5 GHz. Based on the above analysis, a design guideline is also concluded as follows:

Parametric Study
In this part, parametric study is carried out to characterize the proposed DRA. Figure 10 shows the simulated |S dd11 | versus frequency with the length of the strips L s = 7, 8, and 9 mm. Referring to the figure, increasing L s would improve the matching of the DRA, and a good result is obtained as L s = 9 mm. Its inset gives the simulated |S dd11 | versus frequency with the width of the strips W s = 0.5, 1, and 1.5 mm. As can be seen from the figure, altering W s has little effect on the matching of the DRA. Next, the sizes of the hollow region on the effect of the DRA performance is investigated. Figure 11 shows the simulated |S dd11 | versus frequency with the length of the hollow region a = 4, 5, and 6 mm. Its inset shows the simulated |S dd11 | versus frequency with the width of the hollow region b = 5, 6, and 7 mm. With reference to the figure, increasing a and b would shift the third resonance modes obviously upward, and thus enhance the differential impedance bandwidth of the DRA. Figure 12 presents the simulated |S dd11 | of the proposed DRA versus frequency for different d = 8, 9, and 10 mm. Referring to the figure, changing d has a small effect on the differential impedance bandwidth of the DRA. However, a larger d would have a negative effect on the antenna gain around 6.5 GHz. Based on the above analysis, a design guideline is also concluded as follows: (1) Determining the dimension of the cylindrical DRA: R~0.47 λ d and H~1.07 λ d , in which λ d is the wavelength in the dielectric corresponding to the center frequency; (2) Introducing four rectangular hollow regions (a~0.26 λ d , b~0.3 λ d and d~0.39 λ d ) into the DRA, and then adjusting L s to obtain a good matching; (3) Slightly adjusting a and b to obtain an optimal differential impedance bandwidth of the DRA; (4) Slightly adjusting d to optimize the antenna gain of the high frequency band.
(1) Determining the dimension of the cylindrical DRA: R ~ 0.47 λd and H ~ 1.07 λd, in which λd is the wavelength in the dielectric corresponding to the center frequency; (2) Introducing four rectangular hollow regions (a ~ 0.26 λd, b ~ 0.3 λd and d ~ 0.39 λd) into the DRA, and then adjusting Ls to obtain a good matching; (3) Slightly adjusting a and b to obtain an optimal differential impedance bandwidth of the DRA; (4) Slightly adjusting d to optimize the antenna gain of the high frequency band.   (1) Determining the dimension of the cylindrical DRA: R ~ 0.47 λd and H ~ 1.07 λd, in which λd is the wavelength in the dielectric corresponding to the center frequency; (2) Introducing four rectangular hollow regions (a ~ 0.26 λd, b ~ 0.3 λd and d ~ 0.39 λd) into the DRA, and then adjusting Ls to obtain a good matching; (3) Slightly adjusting a and b to obtain an optimal differential impedance bandwidth of the DRA; (4) Slightly adjusting d to optimize the antenna gain of the high frequency band.

Results
In this part, the measured results of the proposed broadband differential-fed dual-polarized hollow cylindrical DRA are reported. The DRA was fabricated using frosted K9 glass, with its

Results
In this part, the measured results of the proposed broadband differential-fed dual-polarized hollow cylindrical DRA are reported. The DRA was fabricated using frosted K9 glass, with its prototype shown in Figure 13. Figure 14 shows the measured and simulated differential S-parameters of the proposed DRA at two differential-ports. With reference to the figure, the three resonance frequencies for the differential-port 1 are measured as 3.58 GHz, 4.88 GHz and 6.26 GHz, respectively. This agrees with the simulated resonance frequencies of 3.56 GHz (0.55% error), 4.88 GHz (0.00% error), and 6.08 GHz (2.87% error). A similar result was obtained for the differential-port 2. The measured differential impedance bandwidth (|S dd11 |&|S dd22 | ≤ −10 dB) at the differential-port 1 and 2 are obtained as 68.4% (3.23-6.59 GHz) and 69.4% (3.22-6.64 GHz), respectively. The above results are summarized in Table 2 for the ease of reference. Figure 15 shows the measured and simulated isolation between two differential-ports. As can be seen from the figure, the isolation between differential-port 1 and 2 (|S dd12 |) is less than 46 dB across the whole operating band.

Results
In this part, the measured results of the proposed broadband differential-fed dual-polarized hollow cylindrical DRA are reported. The DRA was fabricated using frosted K9 glass, with its prototype shown in Figure 13. Figure 14 shows the measured and simulated differential S-parameters of the proposed DRA at two differential-ports. With reference to the figure, the three resonance frequencies for the differential-port 1 are measured as 3.58 GHz, 4.88 GHz and 6.26 GHz, respectively. This agrees with the simulated resonance frequencies of 3.56 GHz (0.55% error), 4.88 GHz (0.00% error), and 6.08 GHz (2.87% error). A similar result was obtained for the differential-port 2. The measured differential impedance bandwidth (|Sdd11|&|Sdd22| ≤ −10 dB) at the differential-port 1 and 2 are obtained as 68.4% (3.23-6.59 GHz) and 69.4% (3.22-6.64 GHz), respectively. The above results are summarized in Table 2 for the ease of reference. Figure 15 shows the measured and simulated isolation between two differential-ports. As can be seen from the figure, the isolation between differential-port 1 and 2 (|Sdd12|) is less than 46 dB across the whole operating band.      The simulated and measured radiation patterns of the proposed DRA at differential-port 1 for the three DRA modes are shown in Figure 16. In the measurement, a broadband external 180° hybrid coupler (2-18 GHz) was applied to provide the differential signals. Referring to the figure, symmetrical broadside radiation patterns were presented for the three resonant modes. In the boresight direction, the co-polarized field is stronger than its cross-polarized counterpart by more than 20 dB for the E-and H-planes. It is also observed that the level of the cross-polarization field of all angles is smaller than −20 dB, which is due to the use of differential feeding. The simulated and measured radiation patterns of the proposed DRA at differential-port 1 for the three DRA modes are shown in Figure 16. In the measurement, a broadband external 180 • hybrid coupler (2-18 GHz) was applied to provide the differential signals. Referring to the figure, symmetrical broadside radiation patterns were presented for the three resonant modes. In the boresight direction, the co-polarized field is stronger than its cross-polarized counterpart by more than 20 dB for the E-and H-planes. It is also observed that the level of the cross-polarization field of all angles is smaller than −20 dB, which is due to the use of differential feeding. Figure 17 presents the simulated and measured antenna gains at the differential-port 1. Referring to the figure, the three measured peak gains are obtained as 3.54 dBi (@3.1 GHz), 6.06 dBi (@5.0 GHz) and 9.67 dBi (@6.8 GHz), which are due to the HEM 111 , HEM 113 , and HEM 115 modes, respectively. It is found that the peak gain of the three DRA modes increase in turn, showing that the air holes have a relatively small effect on the antenna gain of the higher-order modes. The measured antenna efficiency of the proposed DRA at the differential-port 1 is also given in Figure 18. With reference to the figure, its range is from 0.74 to 0.95 in the usable frequency band (3.23-6.59 GHz). The radiation patterns, antenna gain and efficiency at the differential-port 2 are very similar with that of the differential-port 1. For brevity, the results are not given here. Figure 19 shows the simulated and measured envelope correlations (ECs) of the proposed DRA. The formula provided in [43] is used to calculate this parameter. With reference to the figure, both the simulated and measured ECs are smaller than 0.005 across the whole frequency band, satisfying the requirement of the MIMO system (EC ≤ 0.5).  Figure 17 presents the simulated and measured antenna gains at the differential-port 1. Referring to the figure, the three measured peak gains are obtained as 3.54 dBi (@3.1 GHz), 6.06 dBi (@5.0 GHz) and 9.67 dBi (@6.8 GHz), which are due to the HEM111, HEM113, and HEM115 modes, respectively. It is found that the peak gain of the three DRA modes increase in turn, showing that the air holes have a relatively small effect on the antenna gain of the higher-order modes. The measured antenna efficiency of the proposed DRA at the differential-port 1 is also given in Figure 18. With reference to the figure, its range is from 0.74 to 0.95 in the usable frequency band (3.23-6.59 GHz). The radiation  Figure 17 presents the simulated and measured antenna gains at the differential-port 1. Referring to the figure, the three measured peak gains are obtained as 3.54 dBi (@3.1 GHz), 6.06 dBi (@5.0 GHz) and 9.67 dBi (@6.8 GHz), which are due to the HEM111, HEM113, and HEM115 modes, respectively. It is found that the peak gain of the three DRA modes increase in turn, showing that the air holes have a relatively small effect on the antenna gain of the higher-order modes. The measured antenna efficiency of the proposed DRA at the differential-port 1 is also given in Figure 18. With reference to the figure, its range is from 0.74 to 0.95 in the usable frequency band (3.23-6.59 GHz). The radiation  Figure 17 presents the simulated and measured antenna gains at the differential-port 1. Referring to the figure, the three measured peak gains are obtained as 3.54 dBi (@3.1 GHz), 6.06 dBi (@5.0 GHz) and 9.67 dBi (@6.8 GHz), which are due to the HEM111, HEM113, and HEM115 modes, respectively. It is found that the peak gain of the three DRA modes increase in turn, showing that the air holes have a relatively small effect on the antenna gain of the higher-order modes. The measured antenna efficiency of the proposed DRA at the differential-port 1 is also given in Figure 18. With reference to the figure, its range is from 0.74 to 0.95 in the usable frequency band (3.23-6.59 GHz). The radiation patterns, antenna gain and efficiency at the differential-port 2 are very similar with that of the differential-port 1. For brevity, the results are not given here. Figure 19 shows the simulated and measured envelope correlations (ECs) of the proposed DRA. The formula provided in [43] is used to calculate this parameter. With reference to the figure, both the simulated and measured ECs are smaller than 0.005 across the whole frequency band, satisfying the requirement of the MIMO system (EC ≤ 0.5).   Finally, the performance of the proposed differential-fed dual-polarized DRA is assessed. Table  3 summarizes the different dual-polarized DRAs. As can be seen from the table, using the differential feeding ( [27] and proposed) can obtain a better isolation than the single-ended feeding. In addition, compared with other works [4][5][6][7][8]27], our DRA has a broader impedance bandwidth and higher antenna peak gain, with an excellent isolation and medium cross-polarization level obtained.  Finally, the performance of the proposed differential-fed dual-polarized DRA is assessed. Table  3 summarizes the different dual-polarized DRAs. As can be seen from the table, using the differential feeding ( [27] and proposed) can obtain a better isolation than the single-ended feeding. In addition, compared with other works [4][5][6][7][8]27], our DRA has a broader impedance bandwidth and higher Finally, the performance of the proposed differential-fed dual-polarized DRA is assessed. Table 3 summarizes the different dual-polarized DRAs. As can be seen from the table, using the differential feeding ( [27] and proposed) can obtain a better isolation than the single-ended feeding. In addition, compared with other works [4][5][6][7][8]27], our DRA has a broader impedance bandwidth and higher antenna peak gain, with an excellent isolation and medium cross-polarization level obtained. Additionally, its consistence of the radiation patterns observed at two ports is good. However, the cost is that a DRA of a relatively large size is required.

Conclusions
A broadband differential-fed dual-polarized hollow cylindrical DRA has been investigated. The HEM 111 , HEM 113 , and HEM 115 modes of the hollow cylindrical DRA have been used for the design. The effect of the air region distribution on the performance of the hollow DRA has been investigated. It has been found that the hollow DRA with four disconnected air holes can have a broad differential impedance bandwidth and good antenna gain. It shows that the proposed DRA has a measured differential impedance bandwidth of~68% for each differential-port. Furthermore, a good differential-port isolation higher than~46 dB has been obtained. The proposed DRA covers the frequency bands including the 5G-n77, 5G-n79, WLAN-5.2 GHz, and WLAN-5.8 GHz. Additionally, the proposed design concept is potentially suitable for designing the on-chip DRA at high frequency.
Author Contributions: Conceptualization, X.F. and Y.S.; Supervision, X.F.; software, K.S.; writing-original draft preparation, X.F.; writing-review and editing, X.F. All authors have read and agreed to the published version of the manuscript.