Design of Compact Antenna Array for MIMO Implementation Using Characteristic Mode Analysis for 5G NR and Wi-Fi 6 Applications

In this paper, we present a compact antenna for 5G new radio (NR) and Wi-Fi 6 applications. Using characteristic mode analysis (CMA), two orthogonal modes have been separated and then a gap-coupled shaped feed-line has been used to excite the radiator. An open stub is added to achieve the wide operating bandwidth of the antenna. The size of a single antenna element and the 1 <inline-formula> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula> 2 array-unit (AU) are 20 <inline-formula> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula> 21.5 mm2 and 40 <inline-formula> <tex-math notation="LaTeX">$\times $ </tex-math></inline-formula> 21.5 mm2, respectively. The operating and fractional bandwidths (FBW) of the AU are 5.18–7.71GHz and 39.25%, respectively. The proposed AU is used to develop a number of radiating element-independent multiple-input-multiple-output (MIMO) antennas comprising of 2AUs, 4AUs, and 8AUs with identical electrical characteristics and good isolation (<inline-formula> <tex-math notation="LaTeX">${I}$ </tex-math></inline-formula>) ≥ 18.8 dB. All the AUs have realized gain ~ 3.0 dBi, and the total antenna efficiency ~ 70% making it suitable for the single-user MIMO (SU-MIMO) and multiple users MIMO (MU-MIMO) needed for Wi-Fi 6 applications. To show the application of these antennas, the deployment scenario of a 16AUs MIMO antenna on a large size ground plane which mimics the smart TV panel and router chassis, etc., is also presented. These AUs may find potential applications in radiator-number independent MIMO systems.


I. INTRODUCTION
T WO DECADES back, an alliance was formed by wireless technology companies to distribute the data wirelessly to the Internet, which later became the wireless fidelity (Wi-Fi) technique [1]. With each passing day, it has matured itself and has impacted all spheres of human lives, whether it is the home, the public or the office places [2]. To meet the public demand for uninterrupted Wi-Fi connectivity, different wireless standards and techniques have been developed. The success of these technologies is largely dependent on the antennas used [3]. Since antennas are essential for wireless communication systems, to comply with Wi-Fi standards; various antennas have been developed [4], [5], [6], [7]. The wide coverage area is the main focus of the Wi-Fi and 5G new radios (5G NR) and using various techniques, single port antennas have been developed for low-data rate applications [8], [9], [10], [11], [12].
The low-data rate limitation can be overcome by using the multiple-input-multiple-output (MIMO) antennas [13]. In the lower frequency bands, a number of MIMO antennas have been reported where they serve as the backbone of the 4 th generation (G) and 5G communication systems [14], [15]. Recently, similar to the single port, using metasurface patch arrays, a three-port Wi-Fi antenna with an overall size of 0.48λ 0 × 1.34λ 0 × 0.05λ 0 operating in the 4.68-5.75 GHz band has also been reported where λ 0 is the free-space wavelength corresponding to the center frequency [16]. This type of structure insights into the use of patch arrays for Wi-Fi MIMO antenna design.
Nowadays, apart from the electronic gadgets used in the public, offices, and dwelling places, over-the-top (OTT) platforms like Netflix, Hotstar, YouTube, etc., are becoming indispensable to human life, and these OTT platforms are being embodied in the smart televisions (TVs). However, to facilitate these services and to connect to multiple devices simultaneously, high data rate Wi-Fi connectivity is essential [2]. To meet these emerging demands, the data transmission rate is to be increased to 5 Gbps which is possible by using the Wi-Fi 6 technology [17], [18], [19], [20]. A part of the C-band of the spectrum extending from 5.925-7.125 GHz has been allocated for this where it has been further sub-divided into four bands from 5.925-6.425 GHz, 6.425-6.525 GHz, 6.525-6.875 GHz, and 6.875-7.125 GHz. The success of the technology is largely dependent on effective MIMO antenna implementations [21], [22], [23]. In most Wi-Fi applications, it has been focused to develop the single-user MIMO (SU-MIMO) but the multi-user MIMO (MU-MIMO) technique can harness the significant technological benefit by connecting the host router/user equipment's (UEs) to the multiple devices at the same time [24], [25]. In both these MIMO techniques, to meet the high data rate requirements, the number of antenna elements must be increased [26]. To comply with the need for multiple antennas in the MIMO implementations, a 12-port MIMO antenna with separated ground planes and varying electrical characteristics is reported in [27]. A voluminous 8-port spatially decoupled dielectric resonator antenna (DRA) to operate in 5. 45-6.15 GHz covering a part of the Wi-Fi 6 spectrum is reported in [28]. For the X-band applications, a 2 × 2 circular patch MIMO antenna designed on a 1.34λ 0 × 1.34λ 0 sized ground plane with a 37.62% FBW is reported in [29]. Here, it has also been focused to improve the isolation using metamaterial-inspired slot-via walls and the edge-to-edge separation among antenna elements. In another case, a 3 × 3 antenna array in the mm-wave range has been developed [30]. In this 3-D cavity-backed antenna array, also the isolation has been improved by the spatial separation and the cavity walls. Further, a comprehensive study of different isolation improvement techniques for the MIMO antenna arrays is reported in [31].
Wi-Fi antenna for the antenna-on-display made of 13-unit cells using the leaky wave concept is reported in [32]. However, it also covers a part of the Wi-Fi 6 spectrum extending up to 6.5 GHz only with a maximum gain of 1.8 dBi at 6 GHz. Recently, a two-port Wi-Fi 6 antenna designed using the half-open cavity with a horizontally polarized radiation pattern and an FBW = 37.7% has been reported in [33]. However, due to the use of two ports only, it may find limited applications in the MU-MIMO implementations.
At the other side, with the increase in the number of ports, the antenna characteristics change, and the design complexity increases.
The multiport MIMO antennas designed using characteristic mode analysis (CMA) have also been proposed [34], [35]. In these antennas, the number of radiators is fixed; by increasing it, the radiation characteristic changes. From the literature, it is inferred that most of the Wi-Fi 6 antennas cover a part of the allocated spectrum [28], [32], [33]. Further, depending on the ground plane size, the multiport MIMO antennas have a fixed number of radiating elements [34], [35]. Since, in these topologies, the antennas share the common ground plane or radiators, the electrical properties largely depend on the number of elements used [22], [34], [35]. Seldom, the number of radiating element-independent MIMO antenna has been investigated [36]. Of late, it has been observed that in place of the single radiating element, their array can also be used for the MIMOs implementation [29], [30], [31] which have certain benefits of increased gain, wide bandwidth, etc. [10], [11], [12], [16]. However, for the successful implementation of large-size MIMO antennas for the Wi-Fi 6 applications, stable electrical characteristics are important [9]. Thus, it is necessary to design compact antennas to support SU-MIMO and MU-MIMO for Wi-Fi 6 standards with flexible numbers of elements and stable responses. To meet these requirements, we propose the MIMO antennas with the novel contributions outlined in Table 1.
The paper is organized as follows. Section II demonstrates a method to separate two orthogonal modes using CMA. In Section III, using the separated mode structure, a gapcoupled monopole antenna design and its feeding mechanism is presented. In Section IV, an improved-performance compact array-unit (AU) comprising of 1 × 2 elements is designed. In Section V, the use of the AU in MIMO implementation is demonstrated. In Section VI, the large-size MIMO arrays with a flexible number of AUs for consistent performances are investigated. In Section VII, an application scenario using 16AUs in MIMO configuration is demonstrated. In Section VIII, to qualify for the MIMO implementation, the envelope correlation coefficients (ECC) and the mean effective gain (MEG) of two large-size antenna arrays made of 8AUs in collinear and orthogonal arrangements are calculated. Section IX presents a discussion on the design of the number of AUs independent MIMO antennas. In Section X, the proof of concept by prototyping a 4AUs MIMO antenna is presented. In Section XI, the state-of-the-art comparison is made. Finally, Section XII concludes the work.

II. MODE SEPARATION USING CMA A. CHARACTERISTIC MODE ANALYSIS (CMA)
The CMA has been widely used in antenna placement, isolation (I) improvement, and bandwidth enhancement in handheld devices. However, it can also be used in compact antenna designs [37], [38].
For an arbitrarily shaped perfectly electric conductor (PEC) body, the characteristic modes can be calculated from the Eigenvalue equation (1) where R, X, λ n , and J n are the real part of the generalized impedance matrix Z, the imaginary part of the Z, Eigenvalue for the characteristic current, and n th mode characteristic current, respectively [39], [40].
Matrix Z can be obtained by applying the method of moment (MoM) to the electric field integral equation and for the normalized J n and J m where they are the modal currents, the orthogonal relations of (2)-(3) are satisfied where δ mn is the Kronecker delta function [39], [40].
The orthogonal relationships of the electric (E n ) and magnetic (H n ) fields can be obtained from the Poynting theorem given by (4) [39].
From (4), it is evident that the Poynting vector which is the function of the modal currents is dependent on the [Z] matrix. The [Z] matrix is structure-dependent and thus by manipulating the shape of the metallic body, the [Z] matrix can be changed to separate the orthogonal current components. Thus, the resonating behavior and the frequency of different modes can be changed which can be presented by the change in the modal significance (MS) which is given by (5) [35], [41].
Therefore, by the shape modification of a PEC sheet, two orthogonal modes can also be separated to enhance the operating bandwidth of a radiating element.

B. ORTHOGONAL MODE SEPARATION OF A PATCH
A 9 mm × 9 mm PEC patch inclined at an angle θ = 45 • with respect to (w.r.t.) the x-axis obtained using CST Microwave Studio is shown in Figure 1(a). Its first modal resonance frequency occurs when the dimensions (L 1 , L 2 ) are close to the half-wavelength [42]. Thus for the L 1 = L 2 = 9 mm, the first modal resonance frequency can be obtained near 16.6 GHz. Further, for the square patch, due to the dependency of the modes on the shape of the structure as discussed previously, the first two orthogonal modal resonance frequencies must coincide with each other. To understand it, the MS of the first five modes is shown in Figure 1(b). From the curves in the figure, it is observed that the first MS = 1 depicting the resonance conditions appear at 15.5 GHz which is close to the predicted value of 16.6 GHz [42]. The orthogonal second mode also follows the Mode#1 curve. In the simulated 5-30 GHz band, the fourth mode appears at 17.5 GHz, and the remaining two modes (#3 & #5) have MS < 0.9. Thus, only these three modes satisfy the criterion MS ≥ 0.9 for the antenna design with the antenna reflection coefficient at the i th port (S ii ) ≤ −10 dB. However, the orthogonal modes, Mode#1 and Mode #2, which superimpose each other, have MS ≥ 0.9 over the wide frequency band; and they can be exploited to design a compact wideband antenna.
The modal surface current density (J s ) of orthogonal Mode#1 and Mode#2 at 15.5 GHz are shown in Figures 1(c-d). The maximum J s in these modes are along the two orthogonal edges of the patch and thus by modifying the edges; they can be separated to resonate at two distinct frequencies.
To separate these two orthogonal modes, the square patch is truncated from one corner as shown in Figure 2(a) where, W 1 and W 2 are the widths of two sides and thus, the metal of the dimension (L 1 -W 2 ) × (L 2 -W 1 ) is removed. Its effect on MS is shown in Figure 2 With the decrease in W 1 and W 2 (by removing metal), the first modal resonance frequency is shifted to the lower frequency. However, movement in the second resonance position is respectively low and it is remaining around the previous value, i.e., 15.5 GHz. To understand the cause of this downward shifting trend of the first modal resonance frequency, the J s on the body for the different values of W 1 and W 2 are drawn in Figure 3. In Figure 3(a), when W 1 = W 2 = 8 mm, the J s for two orthogonal modes are similar. The current nulls appear at the corners of the patch and they have MS ≥ 0.9 at 15.5 GHz. When W 1 and W 2 are reduced to 3 mm, the J s pattern changes. At the corner joining L 1 and L 2 , J s is continuous in Mode#1 which causes the reduction in the modal resonance frequency to 10 GHz. However, in Mode#2, at this corner, it has a disjoint J s and thus there is no appreciable change in its modal resonance frequency. For the third case when W 1 = W 2 = 1 mm, in Mode#1, J s is continuous at the corner causing an increase in the resonating length to downshift the modal resonance frequency to 8.6 GHz which is governed by L 1 + L 2. However, in Mode #2, still, there is the disjoint J s which makes the structure to resonate at a higher frequency and thus two orthogonal modes are separated. In this case, Mode#2 still appears near 15.5 GHz, and in conclusion, by changing the dimensions, the mode reinforcement and mode separation phenomenons are noticed. The mode reinforcement causes the structure to resonate at much lower than the natural resonance frequency.
Therefore, two orthogonal modes can be separated and down-shifted to design compact antennas. Since, in the present case, for W 1 = W 2 = 1 mm, the modal resonance frequency of the first mode has been shifted to 8.6 GHz, hence, by suitably exciting, a compact antenna for the intended applications can be developed.

C. CMA OF FEED NETWORK AND GROUND PLANE
1) Feed network: To design the compact radiator and to make it resonate at a lower than the modal resonance frequency, it is necessary to suitably feed the structure shown in Figure 2(a) while preserving the Js pattern of Mode#1. Thus, as shown in Figure 4(a), the radiator is fed from the side using the feeding network comprising microstrip lines of lengths LF1-LF3. The effect of the feeding network on the Js is shown in Figure 4(b). It is noticed that the continuous Js pattern of Mode#1 on the radiating element is retained. As shown in Figure 4(c), it resonates at 8.6 GHz. Further, due to the feeding-network loading, Mode#2 frequency has shifted downward to 11.6 GHz. It is also noticed that only the first two modes with MS ≥ 0.9 are available in the simulated frequency band which is separated by 2.6 GHz. Thus, these modes can be used to design a wideband or multiband antenna.
2) Effect of the ground plane on the modal resonance: In compact antennas, the smaller ground plane also influences the radiation characteristics and the operating bandwidth [37]. Thus, to understand its effect, a metallic ground plane of size S W × G L = 20 × 10 mm 2 has been placed at h = 0.8 mm below the radiating patch. The structure with the ground plane and the MS are shown in Figures 5(a-c), respectively. Due to the ground plane, two modal resonance frequencies have been further lowered which can be exploited to obtain a single larger operating bandwidth by placing the suitable excitation [43].

III. GAP COUPLED MONOPOLE ANTENNA DESIGN
Using the design parameter annotations outlined in Figure 4    gap g 1 is shown in Figure 6. The optimized design parameters for a single probe-fed antenna are given in Table 2.
The effect of g 1 on the impedance matching is shown in Figure 6(c). For g 1 = 0 mm, the antenna resonates at 6.25 GHz with a single resonating mode because it is tightly coupled to the feed line. When g 1 = 0.2 mm, it has a wider ≤ −10 dB impedance-matched bandwidth in the  5.53-6.32 GHz band. With a further increase in the g 1 , due to reduced coupling, the impedance matching is reduced. Next, for the case g 1 = 0.2 mm, two distant resonating modes R 1 , and R 2 can also be noticed at 5.46 GHz and 6.17 GHz, and their vicinity to each other is helpful in increasing the operating bandwidth.

A. BANDWIDTH ENHANCEMENT
The operating bandwidth of the structure can be increased by lowering the quality factor. Thus, a narrow open stub (L 3 × W 3 ) as shown in Figure 7(a), is added to L 1 . Next, its position (P), w.r.t., L 2 is varied to obtain the best impedance matching as shown in Figure 7(b). It is noticed that for P = 0 mm, it has poor impedance matching. With the increase in it, the impedance bandwidth improves. For P = 1.5 mm, the impedance matching of the antenna is significantly improved covering 4.96-7.59 GHz. In this case, three distinct resonance frequencies (R 1 , R 2 , and R 3 ) at 5.14 GHz, 6.29 GHz, and 7.45 GHz can also be noticed. With the further increase in P to 1.75 mm, the lower and upper resonating modes shift out of the intended operating band. Thus, by varying P, the operating frequency band can be controlled.
To understand the genesis of these three resonating modes, the J s distribution at 5.2 GHz, 6.0 GHz, and 7.0 GHz are shown in Figure 8 (a-c), respectively. At 5.2 GHz, the antenna operates in the first mode and the total resonating length (high current density) induces it [44].  Except for the phase reversal, the J s distribution is similar to Mode#1 Figure 4(b). Further, due to the ground plane and the dielectric loading, the resonating bands have shifted downward. As shown in Figure 8(b), at 6.0 GHz, another frequency, although, the J s is in the same direction, the radiating arm length has shortened to increase the operating frequency. This reduction in resonating length also affects the realized gain and the directivity at this frequency. At 7.0 GHz, the second mode is dominant for which the current path is shown in Figure 8(c). Since, in this case, the current is mainly concentrated around the joint of L 1 − L 3 segments, the current path is further reduced to increase the operating frequency and reduce the realized gain. In all three cases, the open stub L 3 is playing an important role in governing the resonating behavior and thus the antenna characteristics.

B. RADIATION CHARACTERISTICS
The 3-D radiation pattern of the antenna at three frequencies 5.2 GHz, 6.0 GHz, and 7.0 GHz are shown in Figures 9(a-c), respectively. It is noticed that the antenna has wide beamwidth coverage to make it suitable for Wi-Fi applications. The simulated directivity (D S ), the simulated IEEE gain (G AS ), and the simulated realized gain (G S ) of the antenna are shown in Figure 10(a). The antenna efficiencies are shown in Figure 10(b). The realized gain of the antenna varies in the range of 3.36 -0.54 dBi in the 5.2-7.2 GHz band. The radiation and the total simulated antenna efficiencies (η S ) are close to 80% and ≥ 67% in the intended frequency band, respectively.

IV. PERFORMANCE IMPROVEMENT OF COMPACT ANTENNA
The low realized gain at the high frequency is one of the constraints of this antenna which can be solved by designing a 1 × 2 antenna array. However, the paucity of space  aggravates the design challenge in implementing the corporate feed for this type of antenna. Here, when placed side by side, the available area between two antennas is 18.4 mm × 10 mm = 0.31λ L × 0.17λ L where λ L is the wavelength corresponding to the intended lowest operating frequency of 5.2 GHz. Thus to implement a 1 × 2 antenna array, a corporate feed with a newly designed power divider is proposed.

A. 1 × 2 COMPACT ANTENNA IMPLEMENTATION
The single unit of the compact antenna has been replicated to design a 1 × 2 array as shown in Figure 11. In Figure 11(a), two probe-fed antennas with individual excitation are shown. However, in Figure 11(b), the corporate feed has been placed and thus the antenna array has been excited using the edge launched 50 sub-miniature adapter (SMA) connector. To incorporate the feed, the power divider of the dimensions PL 1 , PW 1 , PL 2 , and PW 2 are used where these are the length of the main feed line, the width of the main feed line, the length of the branch line, and the width of the branch line, respectively. The physical dimensions of these parameters are given in Table 3 and their electrical dimensions are 0.06λ L , 0.03λ L , 0.14λ L , and 0.01λ L, respectively. Thus, it is a compact power divider design. The comparison of the responses of Figures 11(a-b) is shown in Figure 11(c). When two antennas are placed side by side and individually excited, in comparison to the stand-alone antenna of Figure 7, the impedance matching reduces. Similarly, when a 1 × 2 array antenna is excited, the impedance matching  is poor which can be improved by controlling the open stub position P and modifying the power divider.

B. FEED NETWORK FOR COMPACT ANTENNA ARRAY
To improve the impedance matching, a circular patch of radius R has been placed on the arm PL 2 as shown in Figure 12(a). The position of the patch (PL 3 ), w.r.t. the junction and R have been parametrically optimized to obtain the best matching. After several parametric studies, it has been observed that the circular patch must be placed away from the junction to achieve wider impedance matching which is placed at 5.7 mm = 0.10λ L away from it. The R is another important parameter to control the impedance matching level. With the increase in the R, the impedance matching improves, and the best matching is obtained when R ≈ 2 mm. With the further increase to R = 2.5 mm, the impedance matching deteriorates. For the sake of brevity, only for R = 2 mm and 2.5 mm, the responses have been shown in Figure 12(b).
The S 21 = S 31 are also close to −3 dB at the resonance frequency, and it is above −6 dB in the entire simulated band. The I between waveguide Ports #2&3 is better than 5 dB over the band and thus by placing the circular patch on the power divider line; the impedance matching can be improved.
Finally, after incorporating this power divider, various parameters of the 1 × 2 antenna array named array-unit (AU) are optimized using the genetic algorithm to achieve a wideband response and improved gain. The antenna structure and its S 11 response are shown in Figures 12(c-d), respectively. The −10 dB impedance bandwidth extends from 5.19-7.75 GHz. The finalized design parameters are given in Table 3.
Since the circular patch helps in improving the impedance matching; the surface current density on the antenna is shown in Figure 13 where it inhibits the current flow by creating nulls and making the lines to act as the impedance transformer.
The directivity, IEEE gain, and realized gain are shown in Figure 14(a). The directivity and the IEEE gain of the antenna are better than 4.5 dBi and 3.5 dBi over the 5.2-7.2 GHz, respectively. Their pattern over the frequencies is almost flat which shows the advantage in comparison to the single-element responses of Figure 10(a). The calculated array factor is 2.71 dB and therefore in comparison to the 0.54 dBi gain of the single element, the gain of the array can be increased up to 3.25 dBi at 7.2 GHz. Thus, as shown in Figure 14(a), the minimum simulated realized gain of the AU at 7.2 GHz is 3.2 dBi. The efficiencies are shown in Figure 14(b). The total efficiency has also improved in

FIGURE 15. The radiation pattern of an AU at (a) 5.2 GHz, (b) 6 GHz, and (c) 7 GHz.
comparison to Figure 10(b) which is above 70% as shown in Figure 14(b).
The 3-D radiation pattern of the AU at three frequencies 5.2 GHz, 6 GHz, and 7 GHz are shown in Figures 15(a-c), respectively. These patterns exhibit nearly-omnidirectional characteristics in the yz-plane and uni-directional in the xzplane, respectively.

V. MIMO IMPLEMENTATION OF AU
The AU has the benefits of increased gain, antenna efficiency, and control over the beam direction by introducing progressive phase shifts in the array. Thus, in place of the single antenna elements, the AUs made of the compact antenna elements can find suitability in the SU-MIMOs and MU-MIMOs implementation for better Wi-Fi 6 connectivity.

A. 2AUs MIMO ANTENNA
1) 2AUs MIMO Antenna Design: To design the 2AUs (1×4 radiating element array) MIMO antennas, two AUs have been placed side by side as shown in Figure 16(a). In this case, the inter-element spacing, the edge-to-edge distance between radiating elements, and the port-to-port separation are 20 mm = 0.346λ L , 6.6 mm = 0.114λ L , and 40 mm = 0.692λ L at 5.2 GHz, respectively. The S-parameters of the antennas are shown in Figure 16(b). In this figure, S ii and S ji represent the reflection and transmission coefficients between two ports. For i, j = 1&2, the −10 dB impedance bandwidth and FBW are from 5.19-7.68 GHz and 38.69%, respectively. Further, it is comparable to the bandwidth of the single AU extending from 5.19-7.57 GHz as shown in Figure 12(d) and the isolation is better than 22 dB. The antenna gain and  efficiencies are shown in Figure 17. From Figure 17(a), it is observed that the realized gain of the two antennas is not identical. However, the total antenna efficiency as shown in Figure 17(b) is similar to Figure 14(b) and it is better than 72%.

B. GAIN BALANCING OF 2AUs MIMO ANTENNA
From Figure 17(a), it is observed that two antennas have dissimilar realized gains which may affect the performance of the Wi-Fi systems [9]. To understand the cause of the dissimilarity in the realized gain of the similar AUs, the J s on the antenna has been studied in Figure 18.
In Figure 18(a), AUs have the same orientation along the x-axis. For this case, J s on the antenna including the ground plane at 6 GHz with one port excited and the other matched terminated, is shown. The comparison of the ground plane J s reveals that when A#1 is excited, it has the J s minima on the ground plane at two places (a) at the left side of A#1, and (b) at the extreme right side of A#2. However, when A#2 is excited, the same is not reciprocated. There is no current minimum at the right side of antenna A#2 and the left side spread is less in comparison to the previous case. Further, the J s confinement near A#2 region increases the gain of the antenna.
This current imbalance can be minimized by flipping the AU-2 about the y-axis as shown in Figure 18(b). Under this arrangement, it can be observed that when A#1 is excited, the current minimum is at the extreme right. Similarly, when A#2 is excited, the J s minimum is at the extreme left of the ground plane which is balanced.
The complete 2AUs (1×4 radiating element array) MIMO implementation is shown in Figure 19(a). The S-parameter responses, the realized gain, and the efficiencies are shown in Figures 19(b-d). From Figure 19(b), it is observed that the −10 dB impedance matched bandwidth is similar the Figure 16(b) (without flipping AU-2). The port-to-port isolation is reduced by 3.6 dB in comparison to Figure 16 (b) but it is still better than 18.8 dB over the operating frequency band.
The main advantage of flipping the AU-2 can be noticed by comparing the realized gain pattern of Figure 17(a) and Figure 19(c). Although the two antennas are similar, they have dissimilar realized gain patterns in Figure 17(a). This problem is solved when the AU (AU-2 in this case) is flipped. Thus the gain is balanced as shown in Figure 19 (c). For one AU flipped, the peak realized gain is 3.4 dBi at 5.6 GHz. The total efficiency as shown in Figure 19(d) is better than 70% in the Wi-Fi 6 band. Thus, here, a method to balance the gain of the AUs for efficient MIMO implementation is also presented.
The 3-D radiation pattern of the antenna at 5.2 GHz, 6.0 GHz, and 7.0 GHz are shown in Figures 20(a-c), respectively. From these results, again it can be confirmed that the antennas have a nearly omnidirectional pattern in the yzplane. Secondly, it is unidirectional in the xz-plane which is suitable for the low envelope correlation coefficient (ECC) based MIMO antenna design.

VI. LARGE-SIZE MIMO IMPLEMENTATION
The proposed AU can be easily arranged to operate in different MIMO implementations with a flexible number of array units (AUs) to meet the emerging Wi-Fi Router, Smart TV, and device-to-device connectivity requirements. The electrical performances of the antenna in different MIMO implementations are demonstrated in this section.

A. 4AUs MIMO ANTENNAS
Using the flipped array of Figure 20(a), a 4AUs (1×8 radiating element array) MIMO antenna is developed in   Figures 21(b-d), respectively. The −10 dB impedance bandwidth extends from 5.18-7.72 GHz with a 39.37% FBW and the isolation is better than 18.8 dB.
The realized gain of the 4AUs antenna is shown in Figure 21(c). It is noticed that A#1 and A#4, which are the outer elements, have similar realized gain patterns but due to their position, with the increase in operating frequency above 5.5 GHz, it decreases. However, it is better than 3.0 dBi in the entire Wi-Fi 6 band. For the inner AUs A#2 and A#3, with the increase in frequency the realized gain increases. Thus, the antenna gain of A#1-A#4 and A#2-A#3 are in pairs which may be useful for MU-MIMO applications. Further, in the intended frequency band, the maximum deviation in the realized gain between antenna pairs is ≤ 0.7 dB. From Figure 21(d), it is noticed that the antennas have total efficiency of ≥70% over the operating band.

B. 8 AUs MIMO ANTENNAS
An 8AUs (1×16 radiating element array) MIMO antenna is shown in Figure 22(a). Its responses are shown in Figures 22(b-d), respectively. Similar to the 2AUs and 4AUs MIMO antennas, it has a −10 dB operating bandwidth from  Figure 22(c). The realized gain of the outer elements A#1-A#8 are following the gain pattern of A#1-A#4 of the 4 AUs MIMO of Figure 21(c). Further, the antenna pairs A#2&A#7, A#3&A#6, and A#4&A#5 have a similar gain pattern. The antenna efficiencies as shown in Figure 22(d), are identical and ≥ 70%.

C. ORTHOGONAL 8AUs MIMO ANTENNAS
Using two 4AUs, an 8AUs orthogonal MIMO antenna is designed and shown in Figure 23(a).
The S-parameters, the realized gain, and the efficiency are shown in Figures 23(b-d), respectively. The S-parameters are identical to earlier cases. However, the realized gain values of the antennas have changed but these are still in pair and ≥2.94dBi. Further, as shown in Figure 23(d), the total antenna efficiencies are ≥70%.

VII. AN APPLICATION SCENARIO OF MIMO ANTENNAS
To understand the application scenarios of the MIMO implementation, the 16 AUs are placed around a ground plane size of 160 mm × 160 mm. The structure and its S-parameter are shown in Figures 24(a-b), respectively.
These AUs also show similar characteristics as analyzed in the preceding sections. Thus, it is suitable for the applications like Smart TVs and Wi-Fi routers, etc. Since AUs have identical behavior, varying numbers of AUs can be mounted at the sides of the larger smart TV panels also to meet the different panel size industrial requirements.

VIII. MIMO CHARACTERISTICS ANALYSIS
To examine the MIMO performance of the antennas, apart from the port isolation, the ECC and the mean effective gain (MEG) of all the antennas have been investigated using the methods described in [14], [15], [22]. In all the previous cases, ECC is lesser than 0.022 but for the sake of brevity, only the result of the antennas reported in Figures 22(a) and 23(a) are shown in Figures 25(a-b), respectively.
The MEG is another important MIMO characteristic for indoor and urban environment applications [22]. Thus, it has been calculated at 5.2 GHz, 6 GHz, and 7 GHz for the cross-polar discrimination (XPD) = 0 dB and 6 dB, which are noted in Table 4. The MEG of antennas is within the acceptable range −12 dB ≤ MEG ≤ −3 dB.   Fig. 22(a) and Fig. 23(a).

IX. DISCUSSION ON THE NUMBER OF AUs OF MIMO
The objective of the manuscript is to develop the number of AU-independent MIMO antennas that can be used in the varying panel size of Smart TV and Wi-Fi routers and other similar applications. Thus, it is necessary that the electrical characteristics of the AUs should be independent of the MIMO size. Thus, the comparison of the electrical properties of the different topologies in the Wi-Fi 6 band is presented in Table 5 where the antenna characteristics are similar. From Table 5, it is observed the different topologies of the AUs have similar operating and the fractional bandwidth. The total antenna and radiation efficiencies are also similar. The realized gain is close to 3.0 dBi and the isolation in all the cases is better than 18.8 dB. Similarly, the ECC is also very low in all the MIMO implementations.
The performance of MIMOs is not significantly dependent on the number of AUs used and thus they can be treated as the number of AU-independent MIMO arrays. This can find various applications such as enhancing the channel capacity and data rate in the SU-MIMO and the MU-MIMO. Although not shown here, apart from this, due to the array factor of the AUs, by introducing progressive phase shift, these topologies can also be made to act as a 1-D beam steering phased array antenna with increased gain to direct the beam to a specific direction which is useful for the MU-MIMO environment.

X. PROOF OF THE CONCEPT
To validate the concept, a 4AUs (1×8 radiating element array) as shown in Figure 21 Figures 26(a-b), respectively.

A. MEASURED SCATTERING PARAMETERS
The simulated and measured S-parameters are shown in Figures 27(a-b), respectively. Compared to the simulated −10 dB impedance from 5.18-7.72 GHz, the measured bandwidth for A#1, A#2, A#3, and A#4 are from 5.19-7.78 GHz, 5.19-7.9 GHz, 5.19-7.9 GHz, and 5.28-7.69 GHz, respectively. Further to re-assure the accuracy of the concept, this antenna has also been re-simulated using Ansys HFSS, and the S ii results, as shown in Figure 27(a), are also in good agreement. The measured transmission parameters S 21 , S 32 , and S 43 are shown in Figure 27(b). The simulated and measured isolation is better than 18.8 dB and 20 dB making the antenna suitable for the MIMO implementation.

B. MEASURED RADIATION PATTERN
The measured 2-D radiations pattern at 5.2 GHz and 7.0 GHz are shown in Figure 28 and Figure 29, respectively. The simulated and measured gain-patterns are nearly omnidirectional in the = 90 • plane at both frequencies. The maximum realized gain variation among these antennas at these frequencies is 1.16 dB and 0.8 dB, respectively. It ensures the wide-beam coverage area for the communication systems.  The gain pattern in the θ = 90 • is also demonstrated to show the angular separation of the radiating beam of two consecutive antennas which is suitable for the MIMO implementation. For example, at 5.2 GHz, the radiated beam from A#1 and A#2 are squinted to 60 • , and 120 • , respectively. A similar phenomenon can also be noticed between A#3 and A#4.

C. MEASURED REALIZED GAIN
The realized gain of the antenna has been measured using two identical antennas placed at 1.7m from each other in θ = 0 • direction. The measured realized gain (G M ) of the antennas over the frequencies is shown in Figures 30(a-b), respectively. The gain patterns over the frequencies are similar. However, some disagreements among them have been observed due to the main beam directions which are not exactly aligned with θ = 0 • direction at each frequency. Further, due to the losses in two antennas, the gain has dropped and it has a minimum of 2.45 dBi realized gain in Wi-Fi 6 bands. The difference among the measured realized gain values of the antennas is lesser than 0.80 dB and 1.16 dB in the Wi-Fi 6 band (5.9-7.125 GHz) and the entire common operating band of the MIMO (5.28-7.69 GHz), respectively. The variation can be further reduced by improving the measurement setup.

D. DIRECTIVITY AND EFFICIENCY FROM THE MEASURED RESULTS
In this section, the computed directivity from measured results (D M ) and antenna efficiency (η M ) have been coarsely verified using (6)-(7) [45].
To verify the directivity, the half-power-beamwidth (HPBW) from the measured 2-D patterns in the principle planes ( = 0 • & = 90 • planes) at 5.2 GHz and 7.0 GHz have been extracted and included in (6). The computed directivity from measured results (D M ) along with the simulated Directivity (D S ) is given in Table 6. Next, using (7), the antenna total efficiency is also calculated and compared to the simulated results as shown in Table 6. The results are in reasonable agreement due to the use of the approximate formula (6).

E. MIMO PERFORMANCE
Finally, to characterize the MIMO performance, the MEG for XPD = 6 dB has been computed from the measured gain in θ = 90 • plane using (8) and it is placed in Table 7 [22].
, φ dφ (8) Further |MEG i -MEG j | ≤ 3dB is satisfied by each antenna. The ECC is another MIMO important parameter. However,  from Table 5 and Figure 25, it is evident that the ECC is significantly low, and thus, for the sake of brevity, it has not been included.

XI. STATE OF THE ART COMPARISON
The state-of-the-art comparison to some recently published Wi-Fi and sub-6 GHz 5G antennas is made in Table 8. In terms of size, the single-element antenna and AU are comparable to others except [32]. Further, among them, [6], [9], [10], [11] are standalone antennas that may find limited applications. The antenna proposed in [32] is compact but its maximum gain is only 1.8 dBi. Further, the comparison of the gain variation shows that in our antennas, even with the increase in the number of AUs, the gain is stable over the operating frequency band and the maximum 0.7 dBi gain variation ( ) in simulation is recorded irrespective of the number of AUs in 1×n array where n is the number of elements. Apart from this, the majority of Wi-Fi 6 antennas such as [11], [28], [32] only partially cover the Wi-Fi 6 band. Except for the proposed design, only [33] is capable of covering all four sub-bands of the spectrum but it has only two MIMO elements which may restrict its application in the MU-MIMO system.
In addition to Table 8, the design method, decoupling technique, isolation, the distance between elements, and design complexity are other evaluation parameters. The antennas reported in [6], [28] use dielectric resonators which are 3-D structure and thus voluminous and complex. The antenna in [9] is reconfigurable structure which uses p-i-n diodes and thus it needs biasing arrangement. Similarly, antennas in [10], [11], [16] are 3-D metasurface antennas in which feeding complexity is increased. The antenna proposed in [32] is planar in nature but needs photolithography technique which increases the prototyping challenges. In [33], the antenna has been realized using open cavity and thus it is also complex in nature. The antenna proposed in [34] uses slot resonators. Against these antennas, the proposed structure in this manuscript is a low-cost, low-complexity; planar antenna developed using conventional printed circuit board (PCB) technique. Apart from this, the performance of various antennas on the basis of isolation can also be made. The MIMO antennas reported in [16], [28], [33], [34] have 25 dB, 20 dB, 22.67dB and 15 dB isolation, respectively, against the reported isolation of 18.8 dB in this paper. In the reported papers, the isolation has been enhanced by orthogonal [16], [28], and back-to-back [33] antenna placement. Since, antennas in [6], [9], [10], [11], [32] are single element, the isolation is not available. The distance between radiating elements is also comparable to others. In [10], [11], [16], [28] the center-to-center distance is 0.51λ L , 0.65λ L , 0.546λ L , 0.726λ L , respectively. Against these values, the port-to-port separation and edge-to-edge separation of the AUs presented in this manuscript are 0.692λ L , and 0.114λ L , respectively.
Nevertheless, in addition to comparable design parameters, as shown in Table 8, the main advantage of the manuscript can be found in terms of flexible MIMO implementation of AUs.

XII. CONCLUSION AND FUTURE SCOPE
In this paper, a compact wideband antenna with 39.25% FBW has been designed. It has been rigorously analyzed and a method to separate two orthogonal modes using CMA has been proposed. To make the compact antenna suitable for the intended applications in sub-6 GHz 5G and Wi-Fi 6, gain balancing, and a new impedance matching technique to design the array using circular patches are proposed. Next, compact antenna arrays with stable gain and the nearlyomnidirectional radiation characteristics with the realized gain ∼3dBi have been developed. The measured realized gain variation in these antennas is ≤1.16 dB. Further, due to the simple design, by augmenting the AUs, 2 AUs (1 × 4 array), 4 AUs (1 × 8 array), 8 AUs (1 × 16 array), and 16 AUs MIMO antenna implementations with stable radiation characteristics have been demonstrated. The isolation among AUs is better than 18.8 dB. Apart from this, in MIMO implementation of AUs, the ECC is ≤ 0.022. The MEG of the AUs are ∼ −5dB and the MEG ratio between two antennas is ≤ 3 dB making it suitable for the MIMO implementation. The electrical performances of these antennas are consistent and thus the numbers of radiating-element independent MIMO antennas have been developed. Due to the flexible design, the AUs are suitable for size-independent platforms, and therefore they may find various applications in the 5G NR and Wi-Fi 6 bands.