A New Arraying Technique for Band-Switchable and Polarization-Reconfigurable Antenna Arrays With Wide Bandwidth

Modern antenna arrays desire antenna apertures which can support a wide frequency range and reconfigurable polarization states. It is challenging to realize such antenna apertures with high isolation among different antenna elements when they are closely packed, particularly for wide-band (>2:1 frequency ratio) operations. In this paper, we demonstrated switchable-band antenna arrays which can be reconfigured from S to C band with an instantaneous bandwidth close to 2:1. The major novelty in this antenna array is that a new arraying technique is adopted to reduce the number of feeding ports by 50%. As a result, the isolation among antenna elements can be significantly improved, thereby increasing the overall radiation efficiency. It should be noted that vertical, horizontal and circular polarizations are supported by this antenna array when all ports are provided with appropriate phases. Herein two arrays are demonstrated. Array I is based on a fractal-shaped slot-ring structure. Each antenna element is fed by a microstrip along the diagonal direction. The common antenna aperture can be reconfigured to operate at 2.5-4.0 GHz or 4.2-7.2 GH, using PIN diode switches. Array II is extended from Array I but with an Artificial Magnetic Conductor (AMC) ground to enhance the instantaneous bandwidth to 2:1. The length of the microstrip feeding lines are reduced to enable modular design. In addition, PIN diode switches are embedded in the feeding lines to further improve the isolation and radiation efficiency. The return losses, port isolations, radiation patterns, gains, radiation efficiencies and IIP3 are simulated and measured.

Typically, a dual-polarized antenna requires two feeding ports which excite the orthogonal polarization states, respectively [17], [18], [19], [20], [21]. For a large antenna array, the large number of ports leads to energy loss due to the limited isolation between the feeding lines within the same antenna element or between neighboring elements. It is observed that there is a general tradeoff between the isolation level and instantaneous fractional bandwidth (FBW) when a dual-polarized antenna element is concerned [17], [18], [19], [20], [21].
For a narrow-band tunable slot-ring antenna [22], the isolation between the vertical (V) and horizontal (H) feeding ports degrades from 34 to 16 dB when the frequency is tuned from 0.93 (0.5% FBW) to 1.6 GHz (5% FBW). By incorporating fractal shapes into the square slot-ring, FBWs up to 66.7% (2:1 frequency ratio) were demonstrated [23], [24]. This wide FBW led to up to 5/4 dB isolation between the two orthogonal ports at S/C band, respectively [24]. The radiation efficiencies are reduced from 80%(S band)/90%(C band) to approximately 50% at both bands, due to this limited isolation.
A new arraying technique is developed here by feeding each antenna element along the diagonal direction. In this configuration, only one feed is required for each antenna element. The different polarization states can be achieved by feeding the antenna ports with appropriate phases which will be articulated in Section II. There are two main advantages: (1) the number of antenna ports is reduced by 50%; (2) the isolation issue between the two orthogonal polarization states inside the same antenna element is completely mitigated.
This arraying concept is similar to [25] but with distinctions whereas a circularly polarized (CP) antenna array be formed using linearly polarized (LP) antenna elements through sequential rotation. It is noted that neither the operation frequency nor the polarization state can be changed in [25]. In another development [26], antenna elements are linearly polarized along ±45 • directions. However, the polarization state of the entire array is also along ±45 • directions. The array in [26] can reconfigure the radiation pattern from broadside to monopulse with the assistance from a complex switching network.
In this paper, we will present two arrays which can realize vertical (V), horizontal (H), and circular polarization states at the array level while each element is polarized along the diagonal direction. It should be noted that both Arrays I&II are band-switchable antenna arrays with wide instantaneous FBW. The works presented here are part of Huang's doctoral dissertation [27].
In Array I, a ground plane using copper is placed λ 0 /4 (@4.7 GHz) below the radiating aperture to realize single-sided radiation. However, due to the wide bandwidth (a total of 4:1 frequency ratio) of this array, the realized gain cannot stay flat across the entire frequency range. Therefore, Artificial Magnetic Conductor (AMC), a form of Electromagnetic bandgap (EBG) structures [28], [29], [30], [31], [32], is used in Array II to replace the copper ground plane in Array I. In earlier works [33], [34], [35], bandwidths up to 10:1 were demonstrated for antennas using AMC. On the other hand, AMC can be used to support dual-band antenna designs [36], [37], [38], [39].
There are two extra efforts in Array II designs. First, the microstrip feeding lines are shrunk to the perimeter of the slot ring, which is necessary to extend the array to a larger size (modular design). Second, to further improve the isolation between the S and C band operation states, a few PIN diode switches are embedded in the feeding lines at both bands, which lead to higher radiation efficiencies and realized gains.
The design of Array I using the proposed arraying technique is presented in Section II. The fabrication and measurement results of Array I are discussed in Section III. The design of Array II with AMC is explained in Section IV while its fabrication and experimental results are shown in Section V. The comparison of both arrays with the State of the Art is summarized in Section VI.

II. ARRAY I DESIGN A. ARRAY I RADIATION APERTURE DESCRIPTION
As shown in Fig. 1, Array I is composed of several layers. The top layer is a substrate with the fractal-shaped slot rings on the bottom side and microstrip feeding lines on the top. At λ g /4 (@5 GHz) below the top layer is a copper plate. Multiple micro-coaxial lines are vertically connected to the microstrip lines and pass through small holes drilled inside the copper plate ( Fig. 1(d)). The center pins of the microcoaxial lines pass through the substrate and are soldered to the microstrip lines while the outer conductors of the micro-coaxial lines are soldered to the slot rings.
Array I is composed of 2×2 S-band antenna elements as shown in Fig. 1(a). Each S-band element is fed by two perpendicular microstrip lines connected at the center of the antenna. As shown in Fig. 1(b), 8 PIN diode switches are embedded inside the radiation aperture. When all of them are turned off, the thin meandering lines connecting the switches are not shorted, therefore the slots in the entire Sband element can form multiple resonances at S band. When all the PIN diode switches are turned on, the S-band element is divided into 4 smaller slot-rings. One smaller slot-ring is illustrated in Fig. 1(c) and can create multiple resonances at C band. The C-band element is fed by a microstrip line along the diagonal direction. It should be noted that this switchableband design can maintain half-length array element spacing across a 4:1 frequency bandwidth [24].
All PIN diode switches share the same biasing voltages since they are turned on and off at the same time. As shown in Fig. 1(c), the metal area connected to the positive (negative) terminal of the diodes is labeled as Area I (II). Area II is connected to the outer conductors of the micro-coaxial lines therefore can be considered as ground. The DC voltage of Area I is provided by an enameled wire which is soldered at the conner of the array elements. Since the outer conductors of the micro-coaxial lines for the C-band elements are also connected to Area I, a circular ring is cut around the C-band element feeding line in order to provide DC voltage isolation. The gap size in the circular ring and location of the bias wire are verified in ANSYS High Frequency Structure Simulator (HFSS) to present negligible effects on the antenna performance.

B. ARRAYING TECHNIQUE
A square slot-ring antenna can be excited by microstrip lines at different locations as shown in Fig. 2(a). When the microstrip lines are placed at the center of the vertical slots, H polarizations can be excited. Antennas I and II represent two different scenarios considering the location of the coaxial line. Both Antennas III and IV can provide LP along the diagonal direction. All four feeding methods provide similar frequency bandwidths as shown in Fig. 2(b). Fig. 3 illustrates the operation principle of this new arraying technique. In Fig. 3(a), when a 1 -d 1 are provided with the phases shown in the figure, the horizontal components of the radiated fields cancel while the vertical components   add up. H polarization can be realized similarly as shown in Fig. 3(b). This arraying technique was first presented in [40] by the same authors. However, the design in [40] requires a physical rotation of the antenna aperture when the operation frequency band is switched from S(C) to C(S) band. In the two designs here, no physical movement of the  antenna aperture is needed when either operation frequency or polarization state is reconfigured.

C. ARRAY I DESIGN PROCEDURE
The fractal-shaped slot-ring antennas in [23], [24] can achieve up to 66.7% FBW at both S and C bands. However, the shape of the S-band element resembles that of the C-band element but with a 45 • rotation. This leads to the aforementioned issue in [40]. Therefore, new fractal shape designs are necessary to accommodate the new arraying technique.
The array element design starts from the S-band slot ring. A first-order fractal shape is incorporated into a square slot-ring as shown in Fig. 4. The fractal design introduces a second resonance in addition to the fundamental mode at approximately 2.4 GHz. This fundamental resonance decreases in frequency when the slot size L f 1 ×W f 1 increases [41], [42]. The second resonance becomes stronger around 3.3 GHz when L f is swept from 2 to 5 mm for W f of 11 mm as shown in Fig. 4(c). As suggested by the electric field distribution in Figs. 4(a) and 4(b), the polarization of this antenna element is along 45 • direction similar to the Antenna III ( Fig. 2(a)). Fig. 4(d) studies the effect of W f 1 when L f 1 is fixed at 5 mm. It is observed that W f 1 exhibits a much less impact on S 11 of the antenna. W f 1 of 11 mm and L f 1 of 5 mm yield the best antenna matching for now.
The S-band antenna impedance matching can be further improved by using dual-branch stepped impedance microstrip lines as shown in Fig. 5. To minimize the design variables, the length of each branch is set to 28 mm while W m2 is fixed at 1.2 mm. In Fig. 5(a), it is observed that the larger impedance contrast renders better impedance matching. W m3 of 0.1 mm yields the best result. However, in the middle of the targeted operational frequency band (2-4 GHz), the S 11 is above −10 dB in the middle of the band. Finally, W m1 is studied in Fig. 5(b) and it was found that W m1 of 16 mm can provide <−10 dB S 11 within the entire S band.
Next, the C-band elements are added to the optimized S-band element as shown in Fig. 6(c). The C-band element is also a first-order fractal-shaped slot-ring but with slightly different arrangement. In Fig. 6(c), t of 1 mm can equally excite two resonance modes at C band. The E fields shown in Figs. 6(a) and 6(b) resembles that of Antenna IV (Fig. 2a), showing a LP along -45 • direction.
In Array I element design, we used two different feeding line along the diagonal directions to avoid the overlap between them. The C-band feeding line is also a steppedimpedance microstrip line. Figs. 6(d) and 6(e) study the effect of W mc1 and W mc2 , respectively. The combination of W mc1 of 0.6 mm and W mc2 of 1.2 mm can provide S 33 <−10 dB from 4.5 to 7.2 GHz. This is the best result we can achieve in Array I though the target is 4-8 GHz.
The design procedure is illustrated in Fig. 7. It is noted that the optimization of C-band elements slightly perturbs the matching of the S-band element. A second iteration might be necessary to obtain good impedance matching in both bands.

A. FABRICATION
The antenna substrate is 31-mil-thick Rogers RT/Duroid 5880 (ε r = 2.2 and tanδ = 0.0009). The fractal-shaped slot-ring antenna array and its feeding lines are etched on both sides of the substrate as shown in Figs. 8(a) and 8(b). The diameters of the inner and outer conductors of the micro-coaxial line are 0.58 and 1.68 mm, respectively. 32 PIN diode switches (Skyworks DSM-8100) are mounted using silver paste (SPI 05002-AB from Structure Probe, Inc.). The equivalent circuit and power consumption of these PIN diode switches were investigated by the same authors [43].
According to Fig. 1, the holes in the ground plane should be barely larger than the out conductor of the micro-coaxial line to avoid backside radiation. However, this makes it difficult to replace a failing PIN diode switch since the ground plane cannot be removed without de-soldering all coaxial lines. A dual-layer ground plane design is illustrated in Figs. 8(c) and 8(d). The holes in Ground Plane I are large enough to allow the connectors to pass through. Meanwhile the holes in Ground Plane II are still very small. However, additional slots are carved in Ground Plane II to expose all the switches. When a repair is deemed necessary, Ground Plane I can be removed, and the de-soldering of PIN diode switches can be done though the slots.
As shown in Fig. 8(e), the micro-coaxial lines pass through the holes in Ground Plane II and then are soldered to the antenna substrate. A zoom-in view of this connection for a C-band element is shown in Fig. 8(f). Next, Ground Plane I is mounted. 3-D printed plastic support structures are used to assembler the antenna array as shown in Figs. 8(g)-(i). It should be noted that the length of the micro-coaxial cables can be significantly reduced (all the way to the ground plane) without unfavorably affecting the antenna performance.

B. RETURN LOSSES AND MUTUAL COUPLING
The return losses of all four S-band ports are presented in Fig. 9(a). The actual operational bandwidth (S11<−10 dB) is 2.5-4.0 GHz. The measured responses of the four S-band ports are consistent with each other and agree with the simulations. In Fig. 9(b), the return losses of all four C-band ports (within the top left S-band element) are presented. The actual operational bandwidth is 4.2-7.2 GHz.
Since this array can provide beam-steering capability, it is important to check the mutual coupling among different ports. First, the active S parameters at S(C) band are shown to be close to S parameters in Fig. 9(c) ( Fig. 9(d)). This implies that the mutual coupling does not significantly impact the impedance matching of each antenna element inside the entire array. In addition, the mutual coupling for several different cases is illustrated in Fig. 9(e)-(f) and Fig. 10, respectively. Fig. 9(e) (Fig. 9(f)) presents the coupling from an S-band port (all C-band ports) to the four C-band ports (the S-band port) within the same aperture at S-band (C-band) operation state. The measured coupling level is less than −12 dB at both bands of operation. The mutual coupling between any two of the four S-band ports (C-band ports in the same S-band aperture) is shown in Fig. 10(a) (Fig. 10(b)-(c)) and is lower than −25 dB (−15 dB) in most of the frequency range. Regarding the mutual coupling between the C-band ports inside different S-band elements, only a 4 , b 4 , c 4 , and d 4 are studied and presented in Fig. 10d(d) since they are close to each other. This mutual coupling level is below −16 dB. For the brevity of presentation, the mutual coupling between C-band ports in other cases are not shown here and also found to be much smaller in HFSS simulations. It is noted that the mutual coupling shown in Fig. 10(a)-(d) does not impact radiation efficiency since they represent in-band coupling.
The coupling between ports of different bands does contribute to losses. In Fig. 9(e)-(f), such coupling in the same S-band element is shown. This coupling between ports of different bands within different S-band elements is also simulated and measured in Fig. 10(e)-(f). As expected, these coupling levels are very low (<−20 dB) in both operation bands.

C. RADIATION PATTERN
The radiation patterns of S and C band arrays for the vertical polarization are simulated and measured. In a digital array, the amplitude and phases at each port can be arbitrary. However, to demonstrate the radiation patterns herein, customized feeding networks are used to provide the 180-degree phase at some ports as illustrated in Fig. 11.
The co-polarization (Co-pol.) and cross-polarization (X-pol.) patterns in both E and H planes at four different frequencies in two bands are illustrated in Fig. 12. The measured patterns agree well with HFSS simulations. At the C-band state, the measured X-pol. levels are 23 dB below the Co-pol. levels at the boresight direction. At S-band, the simulated X-pol. level is less than −40 dB. However, the measured X-pol. level is <−10 dB. This might be due to the small ground plane size. Any misalignment between the ground plane and slot-ring aperture can cause increase X-pol levels, particularly at S band. In HFSS simulations, this misalignment does not exist.

D. REALIZED GAIN
The realized gain is measured using an Agilent network analyzer (N5230A) and the anechoic chamber at UCF. The reference plane is calibrated to the micro-coaxial connectors.

E. IIP 3
The 3rd-order intercept point at input (IIP 3 ) is characterized in measurement for several difference situations. The setup is illustrated in Fig. 14(a).
When the array operates at C band, all switches are forward biased. The IIP 3 at C band is measured to be 47.9 dBm at 6 GHz as shown in Fig. 14(b). While in the S band operation sate, all switches are reverse biased. It is expected that the IIP 3 increases against reverse bias voltage. As shown in Fig. 14(c)-(e), at f 0 = 2.5 GHz, IIP 3 is measured to be 45.2 dBm (10V), 52.6 dBm (25V) and 58.6 dBm (50V), respectively.

F. BEAM STEERING CAPABILITY
The beamsteering at both bands (3 and 5 GHz) is demonstrated. All ports are excited to realize vertical polarization. Fig. 15(a) (Fig. 15(b)) shows the beamscanning to 10 • in E plane (15 • in H plane) at 3 GHz. While Fig. 15(c) (Fig. 15(d)) shows the beamscanning to 15 • in E plane (20 • in H plane) at 5 GHz. It should be noted that several delays lines were built and used to emulate the phases as specified in Fig. 15(e) to facilitate the beamsteering pattern measurement.
In a digital array, multiple signals can be fed into one antenna port to enable independent multiple beams scanning
The simulated active S parameters for the 2×2 S-band/4×4 C-band array with different scan angles are presented in Fig. 17. The active S parameters for the S-band port is below −10 dB when the beam scans in the E plane. However, in the H plane, the active S 11 rises to −6 dB when beam scans to 35 • . On the other hand, the active S parameters for the C-band ports remain below −10 dB at different scan angles in both E and H planes. The active S parameters at some frequency points rise to −5 dB.

G. CIRCULAR POLARIZATION CAPABILITY
This array can easily support CP for a digital array. The phases at each antenna port for right-handed CP (RHCP) operations are illustrated in Fig. 18(a). When the phases are sequentially rotated in the other direction, lefthand CP (LHCP) can be achieved. The simulated axial ratios (ARs) are below 0.2 dB at S band and 0.8 dB at C band. as shown in Fig. 18(b)-(c). The radiation patterns at

FIGURE 16. The 2×2 S-band array (a) polarizes along X-axis and scans along Y-axis and (b) polarizes along Y-axis and scans along X-axis at 3GHz. (c) Independent dual-beam
scanning at 3 GHz.

A. MOTIVATION OF IMPLEMENTING AMC
In Array I, a solid ground plane is used below the radiation aperture. The electrical distance between the radiation aperture and ground plane is λ 0 /4 at 4.7 GHz. At the edges of the entire frequency range of 2-8 GHz, this λ 0 /4 condition can be hardly satisfied, leading to reduced operational bandwidth.
In Array II, an AMC layer is added in between the radiation aperture and ground plane. This AMC can provide the equivalent λ 0 /4 condition at 2.8 GHz and 6.6 GHz, respectively. As a result, the bandwidth, gain and overall efficiency of the antenna array can be significantly improved.

B. FRACTAL-SHAPED SLOT-RING ANTENNA APERTURE
The fractal shape design at both S and C bands is very similar to Array I. However, there are a few critical improvements in Array II. First, The C-band ports are moved from the outside corners to the inside corners as shown in Fig. 19(a). The benefits of this change are twofold: (1) the size of the S-band antenna element is much smaller; (2) the unfavorable effects on the realized gain of the Array I S-band element are minimized since the two of the C-band ports are at the maximum electric field of the S-band antenna element ( Fig. 2(a) Antenna III). Second, the S-band feeding lines are bent at the edge of the S-band element so that λ/2 spacing can be met for a modular array. Third, it was found that the four solid metal strips surrounding the S-band port in Array I deteriorates the gain of the C-band elements. Since these metal strips also act as the ground plane of the S-band microstrip feeding lines, A meander design is implemented in Array II to avoid the aforementioned issue while at the same time it still provides a ground for the S-band feeding lines. Finally, since the four C-band ports are much closer in Array II, four L-shaped floating metal strips (marked in yellow in Fig. 19) are used to improve the isolation among the four ports.
As shown in Fig. 19(a), eight PIN diode switches are embedded in the radiation aperture allowing the switching of operational frequency bands similar to Array I. In array II, as shown in Fig. 19(b), additional switches are embedded in both S and C band feeding lines to further improve the isolation between the two frequency bands. In Section V, it will be shown that the isolation can be improved by an additional 20(10) dB at S(C) band. To provide the DC voltage bias for all these switches, metals in the radiation aperture are segmented and shown in different colors shown in Fig. 19. The contour and feeding lines of the S-band element (both in purple color) are connected to V 1 through the bias wire at the right edge. A 10-k resistor is used to connect the contour and feeding lines together. Approximately half of the contour of the C-band elements and their feeding lines (both in orange color) are connected to ground through the outer conductor of the coax lines at Ports 2-5. Again, these two pieces of metals are connected through 10-k resistors. V 2 and V 3 are provided through the inner conductor of Port 1 and Ports 2-5, respectively.

C. AMC LAYER DESIGN
Periodic square loops [44] is adopted here to achieve a bandstop resonance around 10 GHz as shown in Fig. 20(a). This structure is equivalent to a transmission line loaded with a series LC resonator. The AMC layer is designed on a 31-mil-thick Rogers RT/Duroid 6002 substrate (ε r = 2.9, tanδ = 0.0012) and the dimensions of the AMC structure are given in Fig. 20.
When the AMC layer is incorporated into Array II as shown in Fig 20(b). the entire structure can be modeled as the equivalent circuit in Fig. 20(b). The simulated S 11 phase of a unit cell is equal to 0 • at both 2.8 and 6.6 GHz.

B. MEASUREMENT RESULTS
In S-band operation state, the simulated and measured S 11 satisfies the -10dB criteria in most of the frequency range of 2-4 GHz (Fig. 22(a)). The measured isolation between the S band port and the four C-band ports is better than 30 dB at most of the frequency range ( Fig. 22(b)), compared to 10 dB in Array I. In C-band operation state, the simulated and measured reflection coefficients are below −10 dB in most of the frequency range of 4-8 GHz (Fig. 22(c)). The simulated active S parameters for C-band ports (Fig. 22(d)) are very close to the reflection coefficients. The measured isolation between the C-band ports and S-band port is better than 20 dB at most of the frequency range of 4-8 GHz (Fig. 22(e)), compared to 10 dB in Array I. It is noted that when Array II operates in S(C) band, the PIN diodes in the C(S) band feeding lines are turned off. The improvement in  the isolation shown here is critical since both cases represent losses in the radiation aperture.
The simulated realized gain is 5-8 dBi at S band as shown in Fig. 22(f) while the measured realized gain closely matches the simulation except for a dip around 3.5 GHz, which is due to the widened radiation pattern around 3.5 GHz instead of loss. The overall efficiency at S band is above 68% in most of the frequency range, except for 3.1-3.2 GHz (still above 50%). The simulated realized gain is 5-12 dBi at C band in Fig. 22(g). The measured gain agrees well with the simulation. The simulated overall efficiency at C band is from 67 to 83%.
The radiation patterns are measured at 2.2 and 3.8 GHz at S band and 4.8 and 7.0 GHz at C band and shown in Fig. 23. The overall agreement between simulation and measurement is apparent. The measured X-pol. levels are much better at S band and similar at C band compared to Array I. It is noted here the ground plane in Array II is significantly larger than the S band antenna element.

VI. COMPARISON WITH STATE OF THE ART
Arrays I and II are compared with other wideband fixed frequency or reconfigurable antenna arrays in Table 1. The antenna arrays in [3], [4] exhibit one port at each antenna element and therefore can achieve only one polarization state. The antenna arrays in [5], [8], [16], [21], [22], [24] have two ports per element, one for each polarization state. The isolation between the two ports is compared in Table 1. It should be noted that the radiation pattern and frequency bandwidth and form factor for these arrays are very different, therefore the isolation performance needs to be carefully considered in this table. The array in [26] can reconfigure its radiation patterns and polarization in a specific way but is not suitable to be extend to a large array, potentially.
The array in [24] achieves a 4:1 frequency ration using a band-switchable approach. This arrangement can maintain λ/2 spacing at two frequency bands without the need of overpopulated T/R modules. The port isolation, however, is its major weakness. The Array I (II) presented herein fundamentally solves this issue while almost maintain (maintain) the wide bandwidth performance. The radiation efficiency and realized gain are significantly improved compared to [24].

VII. CONCLUSION
A novel arraying technique has been demonstrated for polarization-reconfigurable array applications. Each array element has one feeding port only. All three polarization states can be dynamically reconfigured by providing proper phases at each antenna port.
This new arraying technique is demonstrated inside two reconfigurable wideband (4:1 frequency ratio) arrays since wide-band multifunctional antenna arrays are the future. The band-switchable approach alleviates the need to of ultrawideband T/R modules and over-population by providing near λ/2 spacing at each band. Therefore, high-performance single-band T/R modules can be used in Arrays I and II.
Array I is simpler in structure with one ground plane but suffers from the reduced efficiency and gain, particularly at the far ends of the operation frequency range. The performance of Array II is improved by incorporating an AMC layer. In addition, Array II is optimized in size for the design of modular arrays and enhanced in isolation performance.