Metamaterial-Enhanced MIMO Antenna for Multi-Operator ORAN Indoor Base Stations in 5G Sub-6 GHz Band

Abstract

This paper presents a novel, four-port, rectangular microstrip, inset-feed multiple-input and multiple-output (MIMO) antenna array, enhanced with metamaterials for improved gain and isolation, specifically designed for multi-operator 5G open radio access network (ORAN)-based indoor software-defined radio (SDR) applications. ORAN is an open-source interoperable framework for radio access networks (RANs), while SDR refers to a radio communication system where functions are implemented via software on a programmable platform. A 3 × 3 metamaterial (MTM) superstrate is placed above the MIMO antenna array to improve gain and reduce the mutual coupling of MIMO. The proposed MIMO antenna operates over a 300 MHz bandwidth (3.5–3.8 GHz), enabling shared infrastructure for multiple operators. The antenna’s dimensions are 75 × 75 × 18.2 mm³. The antenna possesses a reduced mutual coupling less than −30 dB and a 3.5 dB enhancement in gain with the help of a novel 3 × 3 MTM superstrate 15 mm above the radiating MIMO elements. A performance evaluation based on simulated results and lab measurements demonstrates the promising value of key MIMO metrics such as a low envelope correlation coefficient (ECC) < 0.002, diversity gain (DG) ~10 dB, total active reflection coefficient (TARC) < −10 dB, and channel capacity loss (CCL) < 0.2 bits/sec/Hz. Real-world testing of the proposed antenna for ORAN-based sub-6 GHz indoor wireless systems demonstrates a downlink throughput of approximately 200 Mbps, uplink throughput of 80 Mbps, and transmission delays below 80 ms. Additionally, a walk test in an indoor environment with a corresponding floor plan and reference signal received power (RSRP) measurements indicates that most of the coverage area achieves RSRP values exceeding −75 dBm, confirming its suitability for indoor applications.

1. Introduction

The radio access network (RAN) contributes significant capital expenditure (CAPEX) in any cellular network, and for 5G RAN, the cost increases from 45% to 65% [1,2]. Historically, this part of the network was vendor-dependent [3,4,5]. For any mobile network operator (MNO), the core of the network can be virtualized entirely, but as far as the RAN is concerned, it evolved through multiple stages: from distributed RAN, where the base band unit (BBU) and the radio resource unit (RRU) are colocated, to a centralized RAN, where the BBU and RRU are decoupled, and to a virtualized RAN, where the BBU or its functional parts are virtualized. These virtualized parts are deployed on a commercial off-the-shelf (COTS) server in the form of a software package [6]. And most recently, the ORAN has emerged, which not only increases virtualization but also breaks the vendor dependency by introducing open interfaces. Multiple split options are available for various applications, which increase the ORAN flexibility and adaptability [7]. The development of a promising MIMO antenna for the ORAN has become a top priority among researchers and the industrial community as it is a crucial part of the RF components needed for the 3GPP split option of a RAN [8]. MIMO technology is indeed widely used in various wireless communication systems, including wireless local area networks, as well as in public and private cellular networks. According to 3GPP Release 18, in Technical Specification 38.104, a single 5G MNO requires a 100 MHz band for uplink or downlink communication. From the perspective of a neutral host organization, 300 MHz is required to support three different MNOs [9]. Further shared infrastructure also provides flexibility in resource management.

Designing MIMO antennas involves addressing port isolation and gain enhancement while meeting requirements for bandwidth, cost, and size requirements. Researchers have explored various methods to reduce mutual coupling, including defected ground structures (DGSs) [10], which enhance isolation, and parasitic elements between radiating elements that act as absorbers [11]. Additionally, neutralization lines [12] and electromagnetic band gap (EBG) structures [13] have been used to suppress surface wave propagation and minimize interelement coupling in MIMO systems. These techniques have significantly improved the key performance parameters of MIMO antenna arrays.

Recent studies have further advanced MIMO antenna design by integrating innovative structural and spatial techniques. For instance, ref. [14] proposed an eight-port MIMO antenna comprising two equal modules. Each module consists of four coaxial probe feed patches with a shared radiator. Elements are separated by an additional central patch with ground shortening pins and four I-shaped slots. The arrangement of elements increases the operational bandwidth by exciting multiple characteristic modes of CMA, but port isolation and gain remain less than 12.5 dB and 2.5 dB, respectively. In another approach, [15] utilizes a frequency selective surface (FSS) positioned above four radiating elements to achieve a peak gain of 4.8 dB. Although the designed scheme achieves a wider bandwidth from 3.8 to 6 GHz, the isolation between elements remains at 20 dB even with the use of a low dielectric loss substrate. Similarly, the eight-port MIMO antenna presented in [16] utilizes E-shaped and I-shaped slots to achieve a broader bandwidth from 3.3 to 6 GHz. However, the designed antenna achieves 20 dB isolation even after introducing a single-ended spanner-shaped slot between radiating elements. The twin-element MIMO antenna proposed in [17] consists of a T-shaped hemispherical feeding patch, left-arm and right-arm radiating patches, and a conjoined triangular ground plane, which achieves a wider band (3.4–6 GHz) and isotropic gain up to 4.5 dBi. But the mutual coupling reached −15 dB. A multi-band non-planar antenna for 5G applications covering the sub-6 GHz (2–7 GHz) band, the Ku-band (13–17.5 GHz), and the mm-wave band (25–39 GHz) is suggested in [18], which demonstrates a wide frequency coverage across multiple 5G bands. The proposed U-shaped element configuration features two arms with three slots and one rectangular-shaped parasitic element in the patch, developed with a partial ground and two parasitic elements, achieving a maximum realized gain of 4.3 dB for the sub-6 GHz band. However, this design suffers from low mutual coupling of up to 21 dB and utilizes a 3D non-planar complex structure, a challenge to incorporate with 5G communication devices. A detailed comparison of these studies with the proposed MIMO antenna is presented in Section 3.

Metamaterial structures offer unprecedented benefits for applications in optics, acoustics, and telecommunications [19]. These metamaterials can alter the behavior of electromagnetic waves primarily through their structural design rather than through the inherent properties of the materials themselves. Structures possessing negative permittivity and negative permeability exhibit extraordinary electromagnetic behavior, enabling advanced wave manipulation through engineered material properties that deviate from the natural responses of conventional materials. The most famous metamaterials are split-ring resonators (SRRs); coplanar SRRs along radiating elements increase decoupling and improve port isolation. Metamaterials above radiating elements showing lens-type properties increase gain and improve port isolation among adjacent radiating elements [20,21].

In [22], a unit-cell-based methodology is employed to simplify the design of a four-port cavity-backed MIMO antenna for 5G indoor base stations, achieving wideband operation (1.68–4.15 GHz) with port isolation exceeding 16 dB. However, the design’s complexity, reliance on an x-shaped isolating block, and lack of real-world performance metrics limit its practical deployment. This underscores the need for compact, high-isolation antennas with validated field performance for multi-operator 5G ORAN applications. A low-profile omnidirectional antenna with Vivaldi slits, as presented in [23], offers a broad bandwidth of 58.8% (2.8–5.13 GHz) and a gain of 0.6–3.7 dBi for 5G indoor base stations. Its single-port configuration, however, lacks MIMO functionality, and the absence of real-world performance data restricts its suitability for multi-operator ORAN deployments. This highlights the need for compact MIMO designs with improved gain and validated field performance. A 16-port MIMO antenna array, detailed in [24], achieves a wide bandwidth 3.3–6.0 GHz and a gain of 5–6 dBi for indoor 5G base stations, with isolation exceeding 15 dB. Its moderate isolation, alongside the absence of field testing, limits its practicality for MIMO deployments where throughput and capacity are primary concerns. A comparison of sub-6 GHz indoor base station antennas with the proposed one is presented in

Table 1. Comparison of Sub-6 GHz Indoor Base Station Antennas.

Parameter	Proposed Design	[22]	[23]	[24]
Frequency Band (GHz)	3.5–3.8	1.68–4.15	2.8–5.13	3.3–6.0
Gain (dBi)	4.5–5	Not specified	0.6–3.7	5–6
Isolation (dB)	<–30	16	Not specified	>15
Size (mm3)	75 × 75 × 18.2	130 × 130 × 30	86 × 86 × 1	140 × 140 × 1.5
MIMO Ports	4	4	1	16
Substrate	FR-4	Not specified	F4BME	FR-4
Real-World Validation	200 Mbps throughput	None	None	None
Complexity	Low	High	Moderate	High

.

This investigation elucidates several groundbreaking advantages derived from the proposed antenna design:

• A novel microstrip patch antenna (MPA) was developed, incorporating meticulously optimized slits and slots. These features facilitate impedance matching and significantly enhance the operational bandwidth to 300 MHz, spanning the frequency range of 3.5–3.8 GHz. This bandwidth is crucial for supporting three distinct operators that utilize shared infrastructure. The unit cell was subsequently reconfigured into a four-port, orthogonally arranged multiple-input multiple-output (MIMO) structure.
• A double-negative metamaterial (MTM) unit cell was engineered, leveraging well-established constitutive effective parameters. This unit cell was optimized for spacing and transformed into a 3 × 3 array. By positioning the MTM structure as a superstrate above the MIMO antenna, the broadband gain was nearly doubled, achieving a stable range of 4.5–5 dB across the entire target frequency band, compared to the initial 1.2–3 dB. Furthermore, this configuration substantially reduced mutual coupling to below −30 dB for adjacent elements.
• The proposed MIMO antenna underwent rigorous real-world evaluation using an open-source 5G core and centralized unit/distributed unit (CU/DU) platform. Additionally, a walk test was conducted to benchmark its performance against commercial off-the-shelf (COTS) designs. These tests confirmed the antenna’s suitability for deployment in 5G indoor base stations.

The paper continues with the following sections. Section 2 details the design and results of the designed unit cell of the antenna, MTM, and array design. Section 3 provides an analysis of the proposed antenna, accompanied by an SDR of ORAN as a proof of concept for indoor wireless communication and a walk test for suitability of deployment. In the last section, the paper concludes with the Section 4.

2. Antenna Design

Designing an antenna for indoor communication is founded on three steps: the first step involves the design of a rectangular microstrip feed unit element, the second step involves the design of the unit element of the MTM and the whole 3 × 3 MTM superstrate, and the last step gives the design of a four-port MIMO array with a 3 × 3 MTM superstrate above a radiating array. Figure 1 illustrates the systematic design workflow for the four-port MIMO antenna array, detailing the iterative optimization process from the initial patch configuration to the final metamaterial-enhanced implementation.

The four-port MIMO configuration, combined with a 3 × 3 MTM superstrate, was chosen to balance relatively high gain (4.5–5 dBi) and low mutual coupling (<−30 dB), aligning with the requirements of ORAN for multi-operator indoor 5G base stations.

• A. Unit Antenna Element Design

The unit element is designed based on the transmission line model implemented on a commercially available 1.6 mm FR-4 substrate with a relative permittivity of 4.3 and a loss tangent of 0.02. The designed antenna incorporates a full ground plane, chosen to ensure compatibility with the ORU where it is placed on top of other electronic components. A partial ground plane could introduce arbitrary effects on the radiation pattern and other performance parameters due to interference from underlying electronics. Additionally, for ceiling or wall-mounted applications, the full ground plane minimizes backward radiation. The patch antenna was selected for this work due to its ease of integration with radio units and its capability to be camouflaged within the housing of radio units, which is particularly advantageous for discreet 5G base station deployments. The use of FR-4 as the substrate, despite its higher loss tangent (0.02) that may affect efficiency, is justified by its cost-effectiveness. An online analysis reveals that premium substrates, such as Rogers RO3003 or RO5880, are 20–30 times more expensive. For 5G small cell networks requiring hundreds to thousands of antennas, this results in a potential 2000–3000% increase in infrastructure costs at scale, making FR-4 a practical choice for mass deployment. The antenna dimensions obtained from the transmission line model were optimized to achieve optimal impedance matching at the target frequency of 3.65 GHz. The inset-feed mechanism is chosen due to its superior impedance matching capabilities and reduced feed radiation. Three rectangular slits of different lengths are carved on one side of the patch. A good impedance matching is achieved with another rectangular slit carved on the other side of the patch. These modifications adjust the effective electrical length and surface current distribution, reducing the quality factor ($Q)$ and broadening the antenna’s frequency response. The theoretical framework and calculations for incorporating slits and slots in the base design are based on the cavity model and transmission line theory.

The resonant frequency of a rectangular patch without slits is given by Equation (1).

$$f_r={\displaystyle \frac{c}{2L_{eff}\sqrt{{\epsilon }_{eff}}}}$$

(1)

where:

c = 3 × 10⁸ m/s (speed of light);

L_eff = L_p + 2ΔL (effective length, including fringing);

ΔL = $0.412h\left({\displaystyle \frac{\left({\epsilon }_{eff}+0.3\right)\left(\frac{W_p}{h}+0.264\right)}{\left({\epsilon }_{eff}-0.258\right)\left(\frac{W_p}{h}+0.8\right)}}\right)$ (Hammerstad and Jensen fringing field correction formula).

The slits and slot disrupt the TM10 mode current, causing it to meander, which increases the path length and excites higher-order modes. Edge slits add asymmetry, potentially tilting the pattern or introducing cross-polarization. The center slot splits the current, creating two parallel paths that enhance coupling and broaden the resonant response. The current flows perpendicular to the width of slits and length of the slot, therefore the effective length increase is approximately the cumulative current detour. The effective contribution of the slits and slots to the electrical length is a fraction of their physical length, typically adjusted empirically to match observed frequency shifts and bandwidth broadening [25]. For slits and slots, the additional length is approximated as

$$\Delta L_s\approx 0.15\left(S_{1x}+S_{2x}+S_{3x}+S_{4x}+S_5\right)$$

(2)

where:

S_1x = S_2x = S_3x = S_4x = Slit width;

S₅ = Slot length.

Therefore, L_eff for Equation (1) is given by Equation (3).

$$L_{eff}=L_p+2\Delta L+\Delta L_s$$

(3)

The bandwidth is inversely proportional to the quality factor [26]:

$$B.W={\displaystyle \frac{\Delta f}{f_r}} \approx {\displaystyle \frac{1}{Q}}$$

(4)

Slits and slots influence $Q$ by affecting the individual quality factors [27].

The addition of slits and slots alters the radiation pattern and increases the effective radiating area and introduces higher-order modes which reduce the radiation quality factor by enhancing radiation losses. Slits and slots modify the current distribution, potentially increasing dielectric losses in localized areas, which lowers the dielectric quality factor. Introduction of slits and slots excites surface waves, increasing surface wave losses and lowering the surface wave quality factor. The total $Q$ factor is:

$${\displaystyle \frac{1}{Q}}={\displaystyle \frac{1}{Q_r}}+{\displaystyle \frac{1}{Q_d}}+{\displaystyle \frac{1}{Q_c}}+{\displaystyle \frac{1}{Q_{sw}}}$$

(5)

where $Q_r, Q_d,Q_c, and Q_{sw}$ are radiation, dielectric, conductor, and surface wave quality factors, respectively.

Overall dimensions of the unit element are 37.5 × 37.5 × 1.6 mm³. Figure 2 clarifies the design stages, final design, and the corresponding S-parameters of the unit element.

Table 2. Parameter values of the single-element antenna.

Parameter	Value (mm)	Parameter	Value (mm)	Parameter	Value (mm)
Wg	37.5	Fw	2.98	S4	5
Lg	37.5	IL	3	S5	9.6
Wp	23.6	S1	11.95	X	1
Lp	17.95	S2	7.5	Y	1.5
FL	11.5	S3	2.5

provides detailed design parameters and their corresponding values in millimeters for the proposed antenna unit element. The proposed design achieves the required impedance bandwidth of more than 300 MHz (3.5–3.8 GHz).

• B. Metamaterial Unit Cell and Superstrate Design

To enhance MIMO antenna performance parameters a superstrate consisting of a 3 × 3 grid of identical structures is designed on a 1.6 mm thick FR-4 substrate as shown in Figure 3b. A single element of the grid is depicted in Figure 3a, which is based on a metallic rectangular patch of dimensions 13 × 13 mm² with an octagram star-shaped slot at its center. The octagram star-shaped slot was selected after iterative simulations comparing rectangular, circular, and star-shaped slots. The design parameters of the MTM unit cell and 3 × 3 MTM are listed in

Table 3. Parameter values of the metamaterial unit cell and 3 × 3 superstrate.

Parameter	Value (mm)	Parameter	Value (mm)	Parameter	Value (mm)
Wu	22	a	3.5	dy	9
Lu	22	b	2.25	W	75
WMTM	13	c	4.95	L	75
LMTM	13	dx	9

. The octagram design maximized resonance paths, resulting in a 3.5 dB gain enhancement. Starting from a conventional rectangular slot in a rectangular patch, the design was enhanced by superimposing a 45° rotated version of the slot, creating the octagram configuration. The star-shaped perforations were chosen over a simple slot loop as they create multiple resonant paths due to their multiple angular branches. These paths redirect the electromagnetic fields from the underlying MIMO radiators, resulting in enhanced gain and improved radiation characteristics.

The metamaterial unit cell was placed in a rectangular waveguide environment, with Port 1 positioned at a λ/4 distance from the structure facing the metallic elements, and Port 2 positioned at a λ/4 distance on the opposite side, facing the substrate, as shown in the simulation setup of Figure 4a. The port separation distance was calculated based on the center frequency of 3.65 GHz, resulting in λ/4 ≈ 20.5 mm. This λ/4 distance ensures optimal impedance matching and minimizes reflections at the port interfaces, providing accurate S-parameter measurements. The simulation employs periodic boundary conditions to replicate the behavior of an infinite array. Electric boundary conditions (Et = 0) were applied along the x-axis, creating perfect electric conductor (PEC) walls, while magnetic boundary conditions (H_t = 0) were applied along the y-axis, creating perfect magnetic conductor (PMC) walls. Open boundary conditions with additional space were implemented at Z_max and Z_min to allow electromagnetic wave propagation in the z-direction. These boundary conditions simulate the periodic nature of the metamaterial structure while enabling proper wave propagation analysis.

The effective constitutive parameters (permittivity ε, permeability µ, and refractive index n) were extracted from the S-parameters (S₁₁ and S₂₁) using the robust method proposed by [28]. This method improves upon the Nicolson–Ross–Weir approach by optimizing the effective slab boundaries. The effective slab thickness d was derived as 1.6 mm, corresponding to the FR-4 substrate thickness. The Nicolson–Ross–Weir relative permittivity and permeability are calculated using Equations (6) and (7).

$${\epsilon }_r={\displaystyle \frac{{\lambda }_0^2}{{\mu }_r}}\left({\displaystyle \frac{1-\Gamma }{1+\Gamma }}\right)\left({\displaystyle \frac{1}{{\lambda }_c^2}}-{\left({\displaystyle \frac{\ln \left(T\right)}{2\pi d}}\right)}^2\right)$$

(6)

$${\mu }_r={\displaystyle \frac{1+\Gamma }{1-\Gamma }}\sqrt{{\displaystyle \frac{{\lambda }_0^2}{{\epsilon }_r}}\left({\displaystyle \frac{1}{{\lambda }_c^2}}-{\left({\displaystyle \frac{ln\left(T\right)}{2\pi d}}\right)}^2\right)}$$

(7)

where:

$\Gamma ={\displaystyle \frac{S_{11}+S_{22}-S_{12}-S_{21}}{2}}$ is the reflection coefficient;

$T={\displaystyle \frac{S_{12}+S_{21}}{2}}$ is the transmission coefficient;

${\lambda }_0={\displaystyle \frac{c}{f}}$ free space wavelength;

${\lambda }_c={\displaystyle \frac{2}{\sqrt{{\left(\frac{m}{a}\right)}^2+{\left(\frac{n}{b}\right)}^2}}}$ is the cut-off wavelength.

Where a and b are the width and height of the waveguide. For TE10 dominant mode λ_c = 2a.

These were refined using [28] impedance z and refractive index n calculations given by Equations (8) and (9).

$$z=\pm \sqrt{{\displaystyle \frac{{\left(1+S_{11}\right)}^2-S_{21}^2}{{\left(1-S_{11}\right)}^2-S_{21}^2}}}$$

(8)

$$n={\displaystyle \frac{1}{k_{0}d}}\left\{\left[Im\left(ln\left(e^{in{k}_{0}d}\right)\right)+2m\pi \right]-iRe\left(ln\left(e^{in{k}_{0}d}\right)\right)\right\}$$

(9)

where:

$e^{in{k}_{0}d}=X\pm i\sqrt{1-X^2}$ and $X={\displaystyle \frac{1}{2S_{21}}}\left(1-S_{11}^2+S_{21}^2\right)$ with the sign of z selected to ensure $z^{\prime }\ge 0$ and adjusted via $e^{in{k}_{0}d}={\displaystyle \frac{S_{21}}{1-S_{11}\frac{z-1}{z+1}}} \mathrm{for} n^{″}\ge 0$. The real part $n^{\prime }$ branch was determined iteratively using Equation (9) ensuring continuity and satisfying $\left|n^{\prime }{z}^{″}\right|\le n^{″}{z}^{\prime }$. The final mu and epsilon were obtained as µ= nz and ε= n/z, with results validated across the frequency range. Figure 4b illustrates the magnitude of the S-parameters on a linear scale, along the left y-axis, and the phase characteristics of the proposed unit cell as obtained through waveguide analysis along the right y-axis. The reflection coefficients |S₁₁| and |S₂₂| demonstrate values below 0.25 approaching zero across the operational bandwidth, while the transmission coefficients |S₁₂| and |S₂₁| exhibit near-unity magnitude throughout the frequency range. Figure 4c presents the extracted constitutive parameters, revealing negative values for both relative permittivity and permeability across the band of interest.

The designed 3 × 3 superstrate sheet will be positioned above the radiating antennas at a distance of 0.182λ₀ (15 mm), where λ₀ represents the wavelength calculated at 3.65 GHz. The interelement spacing of 0.11λ₀ (9 mm) was selected to achieve optimal mutual coupling reduction while maintaining constructive interference between adjacent unit elements, resulting in enhanced gain and port isolation for the MIMO antenna system.

• C. Antenna Array with MTM Superstrate Design

Figure 5a shows a four-port MIMO antenna with a 3 × 3 metamaterial superstrate designed to enhance key performance indicators. The superstrate (Figure 3b, subsection B) is positioned 0.18λ₀ above the array (λ₀ calculated at 3.65 GHz). Initially, a λ₀/4 distance (approximately 20.5 mm at 3.65 GHz) was selected as the starting point, as this quarter-wavelength point is known to provide optimal impedance matching where reflections from the superstrate and antenna can constructively interfere, minimizing return loss. Following this initial choice, iterative parametric distances were investigated during the design phase to refine the height. The chosen distance of 0.18λ₀ was found to provide a balance between enhanced gain and improved port isolation, while maintaining acceptable impedance matching (S₁₁<−10 dB) across the operating band. The antenna is constructed on FR-4 with a thickness of 1.6 mm, a dielectric constant of 4.3, and a loss tangent of 0.02. The height of the antenna is 1.6 mm, but the overall antenna dimensions with the MTM superstrate are 0.913λ₀ × 0.913λ₀ × 0.22λ₀. The rectangular radiating elements designed in the first subsection of this section, as shown in Figure 2, are placed orthogonal to each other. The distance between each element is set to 0.2λ₀. The fabricated assembly consists of the MTM positioned above the MIMO antenna, separated by M3 nylon spacers at each corner (see Figure 5b.

Table 4. Parameter values of a four-port MIMO with a 3 × 3 superstrate.

Parameter	Value (mm)	Parameter	Value (mm)	Parameter	Value (mm)
H	15	D	18	dx	9
W	75	h	1.6	dy	9
L	75

provides optimized values for the positioning of each radiating element, the distance between MTM unit elements, and the gap between the MIMO antenna and the MTM superstrate.

Alternative methods for improving isolation in MIMO antennas, such as the use of a circular resonator (a parasitic element) in the same plane as [29], have been explored. This approach, effective at higher frequencies, enhances isolation without increasing thickness but may compromise other key MIMO parameters, such as gain or impedance matching. The proposed superstrate, positioned at 0.18λ_0, increases the thickness but improves gain and isolation while maintaining impedance matching across the band, creating a trade-off.

• Port Isolation and Reflection Coefficient Analysis:

The scattering parameters of the antenna, both with and without MTM, are analyzed in Figure 6. The reflection coefficients of all ports with and without MTM remain less than −10 dB for almost the whole band of concern (i.e., 3.5–3.8 GHz). In comparison, using MTM |S11|, the resonating frequency shifts from 3.6 GHz to dual frequencies, 3.6 GHz and 3.75 GHz. Specifically, the maximum port isolation scattering parameters of adjacent ports in the case of without MTM are above −30 dB. However, using the MTM superstrate, the port isolation is reduced compared to the case without MTM. Despite a minor shift in the reflection coefficient dip caused by the MTM, the structure improves the proposed antenna’s mutual coupling capability.

• MIMO Antenna Gain and Efficiency Analysis:

As shown in Figure 7, incorporating the MTM superstrate substantially enhances the MIMO antenna’s performance, elevating its gain from the original 1.2–3 dB range to a stable 4.5–5 dB throughout the entire 3.5–3.8 GHz frequency band. A superstrate with near-perfect transmission (0 dB) and minimal reflection (−35 dB) in the concerned band acts as an efficient impedance matching layer between the MIMO antenna and free space, allowing maximum power transfer with minimal losses. Additionally, it creates a resonant cavity effect that helps shape the radiation pattern and increase directivity, ultimately enhancing the overall gain of the MIMO antenna.

• Radiation Pattern Analysis:

A comparison of the antenna’s effectiveness with and without metamaterial integration is presented in Figure 8, which depicts the simulated radiation patterns for both (a) E-field and (b) H-field distributions. The E-field pattern demonstrates that the metamaterial implementation maintains the radiation characteristics while enhancing the pattern stability around the broadside direction. The H-field plots reveal improved cross-polarization suppression with the inclusion of metamaterials, achieving levels between −30 and −60 dB. This enhancement in polarization purity and pattern symmetry across both fields validates the effectiveness of the metamaterial structure. Figure 9 compares 3D radiation patterns at 3.6 GHz for the MIMO antenna with (a) and without (b) a metamaterial (MTM) layer. With the MTM, the patterns are more focused, showing higher intensity and directionality, which increases gain and reduces coupling between ports, improving isolation. This highlights the MTM’s role in enhancing directivity and port isolation.

• MIMO Diversity Parameters:

Four key diversity parameters must be considered for proper MIMO antenna operation in a wireless communication system: envelope correlation coefficient, total active reflection coefficient, diversity gain, and channel capacity loss. For antenna elements in a MIMO system, ECC measures the correlation between radiation patterns. The ECC ranges from 0 (complete isolation) to 1 (complete correlation). Generally, $\rho$ < 0.5 is considered acceptable for the appropriate operation of MIMO. The proposed MIMO antenna exhibits a very low value of ECC for any adjacent radiating element about less than 0.002 as shown on the left y-axis of Figure 10a. ECC is calculated from the formula given in Equation (10). It is noted that ECC remains lower when the MTM superstrate is used for frequencies ranging from 3.6 to 3.8 GHz.

$$\rho ={\displaystyle \frac{{\left|S_{ii}\ast S_{ij}+S_{ji}\ast S_{jj}\right|}^2}{\left(1-{\left|S_{ii}\right|}^2-{\left|S_{ij}\right|}^2\right)\left(1-{\left|S_{ji}\right|}^2-{\left|S_{jj}\right|}^2\right)}}$$

(10)

Multiple antennas increase the signal-to-interference ratio in a MIMO system; the diversity gain quantifies this increase. The DG also measures how much transmission power can be reduced without a performance loss. The more the DG, the better the performance of a MIMO antenna array, and its values range from 0 to 10. Calculated using the formula in Equation (11), the DG of the proposed MIMO antenna is plotted on the left y-axis of Figure 10b.

$$DG=10\sqrt{1-|\rho {|}^2}$$

(11)

In a multi-port MIMO antenna system, the overall input impedance matching and mutual coupling effects are characterized by a single parameter: TARC. For a MIMO antenna to function correctly, TARC values must remain less than −10 dB for the band of concern. Figure 10a (right y-axis) gives a comparison of TARC values in both cases. TARC of a MIMO antenna is calculated from S-parameters using the formula in Equation (12).

$$\mathrm{TARC} =-10\log \sqrt{{\displaystyle \frac{{\left(S_{mm}+S_{mn}\right)}^2+{\left(S_{nm}+S_{nn}\right)}^2}{2}}}$$

(12)

To estimate the expected loss in data throughput of a MIMO antenna communication system, CCL is calculated from scattering parameters using mathematical Equations (13) to (15). Lower the loss in throughput to improve the overall performance of the MIMO antenna. While the right y-axis of Figure 10b displays CCL curves with and without the use of the MTM superstrate, its value remains less than 0.2 bits/sec/Hz for the band of concern.

$$CCL=-lo{g}_2\left[det\left({\psi }^R\right)\right]$$

(13)

$${\psi }^R=\left[\begin{array}{llll}{\phi }_{11}\hfill & {\phi }_{12}\hfill & {\phi }_{13}\hfill & {\phi }_{14}\hfill \\ {\phi }_{21}\hfill & {\phi }_{22}\hfill & {\phi }_{23}\hfill & {\phi }_{24}\hfill \\ {\phi }_{31}\hfill & {\phi }_{32}\hfill & {\phi }_{33}\hfill & {\phi }_{34}\hfill \\ {\phi }_{41}\hfill & {\phi }_{42}\hfill & {\phi }_{43}\hfill & {\phi }_{44}\hfill \end{array}\right]$$

(14)

$${\phi }_{mn}=1-\left|{\sum }_{n=1}^4{S}_{mn}^{*}{S}_{nm}\right|$$

(15)

3. Performance Evaluation with ORAN Testbed

To evaluate the performance of the proposed antenna in a real-world scenario, a proof-of-concept testbed of a private 5G network based on ORAN split option 7.2 is implemented. Different components of the 5G network with the proposed antenna are illustrated in Figure 11. ACES in-house builds the radio frequency front end. For O-RU services, we utilized a commercially available, off-the-shelf SDR Ettus USRP B200. The srsRAN open-source software radio suite is installed on Ubuntu 22.04 for O-CU/O-DU services [30].

The metamaterial superstrate significantly enhanced both downlink and uplink throughput stability, maintaining consistent downlink throughput of approximately 200 Mbps and uplink throughput of nearly 70 Mbps throughout the 24 h period. In contrast, the antenna without MTM exhibited considerable throughput fluctuations, with dramatic drops of up to 30 Mbps in downlink and 25 Mbps in the uplink during peak usage hours (8:00–20:00), demonstrating the MTM design’s superior resilience to network congestion as depicted in Figure 12a. From Figure 12b, the implementation of the metamaterial resulted in a consistent 25–30 ms increase in transmission latency across all times, attributable to the additional signal processing complexity within the MTM structure. However, this design choice significantly reduced jitter by approximately 10 ms compared to the conventional antenna, demonstrating that the metamaterial superstrate effectively stabilizes signal propagation at the expense of slightly higher transmission delay.

A walk test was conducted in a designated indoor scenario to evaluate the performance of reference signal received power (RSRP) using INOS Digis Wireless RF processing tools, with the testing setup positioned in front of the exit door as indicated in the building’s floor plan in Figure 13. The results demonstrated robust signal coverage, with an excellent RSRP of more than −75 dBm observed across over 75% of the coverage area, ensuring strong signal quality for the majority of the space. Additionally, a good RSRP level exceeding −85 dBm was recorded in approximately 24% of the area, further highlighting the effectiveness of the setup. The signal strength distribution, ranging from −120 dBm to −75 dBm, was mapped across the floor plan, providing a comprehensive overview of the wireless performance in this indoor environment.

Table 5. Performance Comparison of Proposed Antenna with Recently Reported Antennas.

Ref	Array	Fc (GHz)	Antenna Size	Material	η %	PG dBi	PI dB	Technique	ECC	RWP
[12]	4 × 2	4.2	150 × 75 × 1.6	FR4	90	4	21	Neutralization Line	0.017	No
[14]	4 × 2	4.7	60 × 30 × 2	FR4	48	2.5	12.5	Ground Shortening Pins	0.14	No
[15]	2 × 2	4.5	50 × 50 × 12.5	Rogers 5880	80	4.8	20	Frequency Selective Surface	<0.1	No
[16]	2 × 8	4.65	160 × 68 × 0.8	FR4	68	N/G	15	Decouple with Spanner-Shaped Slots	<0.04	No
[17]	1 × 2	4.5	30.98 × 4.2	FR4	90	4.585	15	T-shaped Hemispherical Patch	< 0.001	No
[18]	3 × 2	4.5	70 × 70 × 70	Rogers 5880	80	4.3	21	Slots and Parasitic Components	<0.004	No
Pro	2 × 2	3.65	75 × 75 × 18.2	FR-4	50	4.8	30	Superstrate @ 0.22λ0	0.002	Yes

F_c = Center Frequency, η = Radiation Efficiency, PG = Peak Gain, PI = Port Isolation, ECC = Envelope Correlation Coefficient, RWP = Real-World Proof of concept. (FR4, Dk = 4.3, tanδ = 0.02), (Rogers 5880, Dk = 2.2, tanδ = 0.0009), Pro = Proposed.

presents a comparative analysis of the proposed MIMO antenna against previously reported state-of-the-art MIMO antennas operating in the sub-6 GHz band, evaluating parameters such as overall antenna profile, peak gain, port isolation, substrate material, techniques employed to enhance isolation, envelope correlation coefficients, and real-world proof-of-concept validation.

4. Conclusions

The proposed quad-port MIMO antenna, enhanced by a 3×3 MTM superstrate, achieves a 300 MHz bandwidth (3.5–3.8 GHz), high gain (4.5–5 dBi), and excellent isolation (<−30 dB), meeting ORAN’s multi-operator requirements. Real-world testing confirms stable 200 Mbps downlink and 70 Mbps uplink throughputs, with acceptable latency increases (25–30 ms) offset by reduced jitter, making it ideal for indoor 5G deployments. It is specifically designed for a shared infrastructure of ORAN-compliant software-defined radio implementations targeting indoor sub-6 GHz 5G deployments. The configuration aims for a suitable bandwidth for multi-operator, high-gain, and improved port isolation. The basic design features an inset-feed rectangular patch with four additional slits, a centered slot to enhance bandwidth by more than 300 MHz (3.5–3.8 GHz), and a 3 × 3 metamaterial superstrate layer at a height of ~0.002λ0. The presented design, with the addition of the MTM superstrate, achieves improved port isolation from greater than −30 dB to less than −30 dB for any adjacent ports, and a gain enhancement from 1.2 dB to 4.5 dB. Furthermore, during real-time proof-of-concept testing, throughputs increase by almost 20 Mbps and maintain consistency even during peak working hours. The presented design’s performance characteristics were validated through both numerical simulations and practical measurements. Additionally, thorough ORAN-compliant SDR testing confirms its suitability for indoor 5G wireless networks.