Wideband Eight-Antenna Array Designs for 5G Smartphone Applications

Abstract

This paper proposes a broadband eight-antenna array design suitable for Fifth Generation New Radio (5G NR) smartphone applications. To cover the 5G NR bands n77/n78/n79 (3300–5000 MHz) and 5G NR-U n46 band (5150–5925 MHz), the single antenna array unit applied is a modified loop antenna element (MLAE) that can generate three different loop modes. To yield good multi-input multi-output (MIMO) performances, the designed MLAE is further arranged as an eight-antenna array, and the experimental results show that the overlapping 6 dB bandwidth can cover the bands-of-interest (3300–5925 MHz) with good isolation and total efficiency of >10 dB and 51–84%, respectively. Finally, good MIMO performances, such as an envelope correlation coefficient (ECC) of lower than 0.1 and desirable channel capacity (CC) of 37–40 bps/Hz, were calculated across the bands-of-interest.

1. Introduction

In today’s modern society, the construction of Fifth Generation (5G) Mobile Communication is gradually being completed. Due to the increasing demands for data throughput and bandwidth utilization from technologies such as the Internet of Things (IoT), electric vehicles, and smart-home appliances, the bandwidth provided by Fourth Generation Mobile (4G) Communication is inadequate to fulfill the demands of the market. To enable higher data transmission within the 5G bands, MIMO technology has become crucial in smartphone and base station technology. Additionally, MIMO technology can effectively enhance spectrum utilization and channel capacity (CC) [1,2]. In 2018, the 3rd Generation Partnership Project (3GPP) officially defined the specifications for 5G NR Release 15 [3]. According to these specifications, 5G NR in the sub-6 GHz range includes three operating bands: Band n77 (3.3–4.2 GHz), Band n78 (3.3–3.8 GHz), and Band n79 (4.4–5.0 GHz). Consequently, countries around the world have selected their respective 5G bands from these frequencies.

With the increasing market and technological demands, not only the aforementioned licensed bands, but also the application of an unlicensed spectrum has gained significant attention. The main relevant frequency band for this is the Unlicensed National Information Infrastructure (U-NII) band, and corresponding to this band, LTE Band 46 has been proposed. In 2020, 3GPP officially introduced Release 16 [4], which highlights the 5G New Radio Unlicensed (5G NR-U) band. This unlicensed band is due to the development of the unlicensed spectrum LTE-U (LTE in Unlicensed Spectrum) and Licensed-Assisted Access (LAA) from 3GPP Release 13/14. As a result, NR-U equipment will operate within 6 GHz band for the first time. In the United States, the NR-U is divided into four sub-bands, namely, U-NII-5 to U-NII-8 (5925–7125 MHz), and is classified as the new version of Wi-Fi, widely known as Wi-Fi 6E. Finally, 3GPP’s Release 16 officially added the LTE Band 46 (5.15–5.925 GHz) to the 5G NR FR 1, designated as 5G NR-U n46; the 5.925–7.125 GHz band is designated as 5G NR-U n96.

In the early designs of 5G mobile phone antennas, to induce a resonance mode at 3.5 GHz, a common approach was to etch an open-slot antenna with a length of λ/4 in the ground plane [5,6,7] or to generate a quarter-wavelength mode using a monopole antenna [8]. However, present smartphone antennas are often collocated along the phone’s frame, leading some researchers to design inverted-F antennas (IFAs) [9] and loop antennas [10] on the frame to excite a 3.5 GHz mode. However, narrowband antennas for smartphone applications are no longer sufficient to meet the market demands. Consequently, many papers have reported combined IFAs and loop antennas to form hybrid antenna designs [11,12,13], which can produce dual-band or wideband modes. In the limited space of a smartphone, accommodating multiple antennas presents a significant challenge for maintaining isolation between antennas.

Various methods to improve isolation have been proposed in the open literature; for instance, decoupling structures [14,15,16,17,18,19,20,21] can effectively reduce the coupling energy between antennas, and orthogonal polarization [22,23,24,25,26,27] uses polarization diversity to enhance antenna isolation. Besides the above techniques, the self-isolated designs [28,29,30,31,32,33] can also improve isolation efficiency by reducing the impact of ground plane currents. For example, [31] capture currents on the ground plane of the antenna (wave trap) to minimize the influence on adjacent antennas, while [32,33] found that symmetric structures effectively reduce the ground plane current, thereby lowering interference between antennas. Notably, the work in [34] points out that the antenna feed position affects isolation performance, suggesting that placing antennas at current null points can improve isolation. Additionally, [35] mentions that adding capacitors or inductors between adjacent antennas can effectively eliminate the coupling.

As stated in [36], the construction of 5G mobile communication infrastructure for the sub-6 GHz band has been gradually completed in many developed countries. To achieve higher data throughput in the 5G operating frequency bands below the sub-6 GHz band, one of the leading 5G technologies is MIMO technology, which plays a pivotal role in both smartphone and base station technology. As aforementioned, many single-band or dual-band operation MIMO antenna arrays have been investigated for 5G smartphones; however, these antenna designs can only cover part of the 5G bands (3300–5000 MHz). Therefore, it is vital to implement a wideband operation 5G MIMO antenna array design that can cover the 5G bands of interest [37,38,39,40,41,42,43,44,45,46]. In [38,39,40], MIMO antenna array designs for metallic frames are adopted. In [38,39], an open slot is cut into a metallic frame with an extended slot loaded into the ground plane, and an additional tuning stub is implemented to achieve a good 6 dB overlapped impedance bandwidth across the bands-of-interest. In comparison, [40] requires two slots (inverted T-shaped and inverted C-shaped) to be loaded into the metallic frame, and these slots are fed by a meticulously designed L-shaped feed-line.

Besides applying antenna array structures with metallic frames, other designs have been explored. In [41], the individual array element consists of an inverted-F antenna with shorting branches and parasitic elements. Impedance matching for the mid-band at 5.2 GHz can be fine-tuned using an L-shaped parasitic element and a slot. In [42], an inverted-F antenna (IFA) unit structure is implemented for 4 × 4 MIMO operation, and it is coupled-fed by a meandering line coupled feeding structure. Notably, a 4 × 4 MIMO operation IFA pair is also implemented in [43], featuring a two-antenna conjoined type with two identical IFA structures (back-to-back) linked to the same grounding structure. However, [43] only applied the IFA pair for 4 × 4 MIMO operation, and such a design requires the excitation of three different modes (loop, IFA, and slot) to achieve wideband operation, resulting in diverse radiation patterns across the bands-of-interest. Compared to the IFA pair design in [43,44], a different integrated slot antenna pair for 8 × 8 MIMO operation is implemented. This design employs a dual-mode decoupling technique specifically for closely spaced open-slot antennas, but a horizontal connecting line must be used to link the two open-slot antennas to achieve desirable isolation of >10.8 dB.

In this paper, a modified loop antenna element (MLAE) is proposed as the antenna unit for an 8 × 8 MIMO antenna array. This MLAE design is derived from a modified loop antenna type fed by a T-shaped coupled feeding structure [46]. However, the loop antenna in [45] could only excite a fundamental loop mode, resulting in a narrow operating bandwidth of 3400–3600 MHz. To achieve a better impedance bandwidth, the design in [46] was further modified into a wideband loop antenna. This wideband design was achieved by disassembling the loop antenna into four sections and applying a coupling method, where two sections located on one side of the substrate are coupled to the two collocated on the other side. This reconfigures the loop antenna into a segmented coupled loop antenna type. By applying this method, the coupling structure can excite an additional 0.75-wavelength loop mode. Combining this mode with the 0.5-wavelength fundamental loop mode and the 1.0-wavelength higher resonant loop mode yields a wideband operation covering the 5G bands-of-interest (3300–5925 MHz). Nonetheless, to avoid the complexity of the segmented coupled loop antenna, which requires printing antenna structures on both sides of the substrate, we propose a novel MLAE, and its corresponding contributions are as follows:

• It offers a streamlined manufacturing process and cost savings by printing the loop antenna structure solely on one side of the substrate. In comparison, the loop antenna described in [46] necessitates printing on both sides.
• It can produce three distinct loop modes (at 3.6, 4.0, and 5.1 GHz), resulting in a broader operating band width compared to [46].
• The CC demonstrated is an impressive 37–40 bps/Hz, just marginally below the theoretical upper limit of 46 bps/Hz.

Both the design methodology of the MLAE and the MIMO performance with an array of eight MLAEs are confirmed through simulation and experimental measurement.

2. Antenna Geometry and Design Evolution

Figure 1 depicts the MLAE antenna design, integrated on the inner side of the smartphone’s longer frame. The antenna, measuring 21 mm × 6 mm, is fed by a 50-ohm microstrip line at point F, with its loop structure grounded via shorting point S on the system ground plane beneath the substrate. This MLAE features a main structure comprising two P-type radiating elements arranged back-to-back, connected by a bridge and adorned with two protruding stubs, each measuring 1 mm (L₁) × 0.8 mm, as shown in Figure 1. Notably, each P-type radiating element is loaded with a U-shaped slot (with parameter L₂ = 1.5 mm), which excites the three loop modes (at 3.6, 4.0, and 5.1 GHz) to cover the bands-of-interest (3300–5925 MHz).

Figure 2 shows the proposed eight-antenna array, with four MLAEs arranged along each long side frame of the smartphone casing, maintaining a gap distance of 16.5 mm (d1 = d2 = d3) between any two MLAEs. The side frame measures 150 mm × 7 mm in area, while the system ground covers an area of 148 mm × 73 mm on a substrate that is 150 mm × 75.6 mm. The smartphone casing, including the frame and ground, is made of FR4 substrate of thickness 0.8 mm.

Figure 3 shows the design evolution of the proposed MLAE. It began with a simple conventional loop antenna structure (similar to [10]), denoted as Ref. 1, which can excite two distinct modes at 3.54 and 5.34 GHz. To enhance the operating bandwidth, Ref. 2 was constructed by reconfiguring the loop structure into two P-shaped elements (back-to-back) linked by a narrow bridge. In this configuration, besides the previous modes from Ref. 1 (shifted to 3.59 and 5.62 GHz), an additional middle mode at 4.19 GHz is also excited. However, as the lower frequency of Ref. 2 is at 3370 MHz, it is insufficient to cover the bands-of-interest (3300–5925 MHz). To further improve the matching of Ref. 2 and satisfy the bands-of-interest, the connected bridge of Ref. 2 was enlarged from 1.4 mm to 1.8 mm, denoted as Ref. 3. Although Ref. 3 excited three resonant modes at 3.7, 4.1, and 4.86 GHz, the upper frequency could not cover the high band of n46 (5925 MHz). Therefore, a rectangular slot forming the P-shaped element was reconfigured into a U-shaped type (with parameter L₂), and two stubs (with parameter L₁) were protruded from the connected bridge. This modification shifted the higher resonant mode from 4.86 to 5.10 GHz, while the lower and middle modes were slightly deviated to 3.6 and 4.0 GHz, respectively. Here, the proposed MLAE exhibits a desirable 6 dB bandwidth of 3.15–6.75 GHz, and the impedance matching is better as compared to the conventional loop antenna (Ref. 1).

Figure 4a,b illustrate the effects of fine-tuning the two parameters, L₁ and L₂. By linearly increasing the length of parameter L₁ by 0.5 mm (0.5 to 1.5 mm), the three modes are also shifted linearly towards the higher spectrum, demonstrating that adjusting the length of L₁ affects all resonant modes. In comparison, by linearly expanding the parameter L₂ by 0.5 mm (1 to 2 mm), the middle resonant mode shifts to the lower spectrum and combines with the lower resonant mode, forming a dual resonant effect. Since fine-tuning L₁ and L₂ does not affect the overall impedance bandwidths of the proposed MLAE, we selected L₁ = 1 mm and L₂ = 1.5 mm as the optimum values.

3. Findings and Analysis

The front and back views of the fabricated eight-antenna array prototype are depicted in Figure 5. Figure 6a,b display the simulated reflection coefficient (return loss) and transmission coefficient (isolation) of the eight-antenna array. For brevity, we only investigate Ant.1 to Ant.4. Figure 6a shows that the stacked 6 dB bandwidth (S11, S22, S33, and S44) adequately covers the bands-of-interest (3300–5925 MHz). Regarding the transmission coefficients (S21, S32, and S43), good isolation levels better than 10 dB were achieved.

Figure 7a presents the measured reflection coefficients (S11, S22, S33, and S44) of the proposed eight-antenna array. It demonstrates that the array covers the bands-of-interest with a 74.1% (3100–6750 MHz) stacked 6 dB bandwidth. Figure 7b shows that the measured transmission coefficients (S21, S32, and S43) closely match the simulated results in Figure 6b. Notably, the simulated and measured data of the proposed array are well validated, with slight differences likely due to fabrication tolerances.

To further understand the excitation of the three loop modes in the proposed MLAE, Figure 8 illustrates the current distributions at 3.6, 4.0, and 5.1 GHz. The current path at 3.6 GHz represents a half-wavelength (0.5λ loop mode) distribution that flows from the feeding point F to the shorting point S, passing through a null point at the right upper corner of the P-shaped element. At 4.0 GHz, the current flows from the feeding point F to the P-shaped element (located on the left side of the MLAE), reaching a null point before returning to the shorting point S. This forms another half-wavelength (0.5λ loop mode) distribution. Lastly, at 5.1 GHz, the current path follows a conventional one-wavelength (1λ loop mode) distribution, extending from point F to point S via the connected bridge, with null points located at both ends of the P-shaped elements.

Figure 9 is a photo of the proposed eight-antenna array placed in an anechoic chamber for measurement. The simulated and measured total efficiencies of the proposed array (Ant.1–Ant.4) are shown in Figure 10a,b, respectively. Ant.2 and Ant.3, positioned at the center of the array, exhibit slightly lower efficiencies due to the influence of adjacent antennas, compared to the outer elements (Ant.1–Ant.4). Despite this, the total efficiencies of all elements remain robust across the bands-of-interest, with simulated values ranging from 62% to 85% and measured values from 51% to 84%. These performance levels meet the typical standards (>40%) required for practical smartphone applications.

Figure 11a–c display the normalized radiation patterns of the proposed array (Ant.1–Ant.4) at 3.6, 4.0, and 5.1 GHz. The radiation patterns of Ant.1 and Ant.4 are very similar, as are those of Ant.2 and Ant.3. Despite slight variations across the three resonant modes, the main beam directions of the patterns are consistently oriented towards the −X and +Z directions. This orientation facilitates excellent envelope correlation coefficient (ECC) values, enhancing the CC and overall performance of the MIMO system when employing this eight-antenna array.

To further validate the findings presented in Figure 11, the antenna field pattern is plotted in a planar expansion diagram using HFSS, where theta angles (0 to 180 degrees along the y-axis) and phi angles (0 to 360 degrees along the x-axis) are represented, as depicted in Figure 12. Across the three frequencies, Antennas 1 to 4 exhibit a consistent dominant component (gain phi/gain theta) as depicted in the diagram. Specifically, the planar expansions at 3.5 GHz and 4 GHz appear quite similar for Antennas 1 to 4. However, at 5.1 GHz, Antennas 1 to 4 show a different dominant component, indicating a shift to a conventional loop mode, distinct from the modes observed at 3.5 GHz and 4 GHz (refer to Figure 8).

The investigation of the envelope correlation coefficient (ECC) is crucial for assessing the performance of MIMO antennas in diverse applications. Generally, a lower ECC indicates a higher CC, and for mobile phone MIMO antennas, an ECC below 0.3 is considered the design standard. Assuming an isotropic distribution of incident signals, the ECC can be calculated from far-field patterns using the following formula stated in [1]:

$${\displaystyle \frac{{\iint }_{4\pi }^{ }{E}_1\left(\theta ,\varphi \right)·E_2^{\ast }\left(\theta ,\varphi \right)d\mathsf{\Omega }}{\sqrt{{\iint }_{4\pi }^{ }{E}_1\left(\theta ,\varphi \right)·E_1^{\ast }\left(\theta ,\varphi \right)d\mathsf{\Omega }{\iint }_{4\pi }^{ }{E}_2\left(\theta ,\varphi \right)·E_2^{\ast }\left(\theta ,\varphi \right)d\mathsf{\Omega }}}}$$

(1)

In Equation (1), E₁ and E₂ represent the far-field radiation patterns produced by exciting ports 1 and 2, respectively. By applying this formula, the correlation between these two radiation patterns can be determined. The calculated ECC for the array, based on both simulation and measurement, is illustrated in Figure 13a,b. The simulation results display ECC values below 0.05, while the measurement outcomes show values under 0.1, significantly lower than the standard threshold of 0.3. These results confirm a low correlation between adjacent elements, indicating that the proposed array is highly effective in multipath fading environments.

Figure 14 shows the calculated CC of the proposed array in an 8 × 8 MIMO system. The results indicate a desirable CC of 37–40 bps/Hz, slightly lower than the calculated upper limit of 46 bps/Hz.

4. Hand Effect on the Proposed Eight-Antenna Array

This section will analyze the outcome of user hand interaction on the Antenna A array. The simulation is divided into single-hand mode (SHM) for data mode, and double-hand mode (DHM) for read mode. To make the simulation results more applicable to real-world scenarios, the relative permittivity and electric conductivity of the hand are referenced from Speag’s product SHO-V3RW-C/LW-C. The permittivity and conductivity values corresponding to different frequencies (between 3 and 6 GHz) are shown in [47].

4.1. Single-Hand Mode (SHM)

The relative positions of the user’s hand in SHM and the smartphone drawn in the simulation software 2021 HFSS (High-Frequency Structure Simulator) are shown in Figure 15. The reflection coefficient outcome of SHM is illustrated in Figure 16a. It can be observed that the antenna bands tend to shift towards lower frequencies. However, most antennas still cover the bands-of-interest, with only a few antennas not fully covering the target bands. Figure 16b shows the total efficiency of the proposed antenna affected by the single-hand mode. Among them, Ant.1 and Ant.5, which are farther from the hand, are less influenced by hand interference (efficiency > 60%). Other antennas such as Ant.2, Ant.3, Ant.4, Ant.6, and Ant.8 exhibit efficiencies above 30%, while Ant.6 shows an efficiency between 20% and 25%. In this mode, Ant.6 consistently shows poorer antenna efficiency. The reason for this is that Ant.6 is located at the position of the thumb crotch, as shown in Figure 14, resulting in more contact with hand tissues and reducing total efficiency to as low as 20%.

4.2. Double-Hand Mode (DHM)

The relative positions of the user’s hand in DHM and the smartphone drawn in the simulation software HFSS (High-Frequency Structure Simulator) are shown in Figure 17. Subsequently, Figure 18a presents the reflection coefficient of the proposed antenna influenced by the DHM, indicating a tendency for antenna modes to shift towards lower frequencies. However, overall, all eight antennas meet the bands-of-interest (3300–5925 MHz). Figure 18b illustrates the total efficiency affected by the double-hand mode, indicating that most antennas are impacted, with their efficiency dropping to over 40%. The cause of this situation may be interference from the positions where the hands are placed, affecting most of the antennas.

5. Battery Component Simulation

The smartphone battery is one of the largest components in the device. Therefore, an actual lithium battery from the Xiaomi 8 Pro (model BM3F) with dimensions of 76.9 mm × 64.2 mm × 3.7 mm is applied and integrated into the antenna array simulation. The battery cell is made of aluminum alloy and covered with a 0.3 mm thick plastic film. Figure 19 shows the corresponding position of the battery on the ground plane of the smartphone. Figure 20a displays the simulated reflection coefficient of the proposed antenna affected by the battery. The battery’s position, being near the feed locations of the four middle antennas (Ant.2, Ant.3, Ant.6, and Ant.7), causes changes in the operating frequency bands of these antennas, especially at the 4 GHz resonance frequency. Nevertheless, all eight antennas meet the bandwidth requirements of 3300–5925 MHz. Figure 20b shows the simulated total efficiency of the proposed antenna affected by the battery. The four middle antennas are significantly affected, with their efficiency slightly reduced to 60% (from 62%) in the low band, due to the influence of the feeding lines by the battery, while the efficiency of the other antennas remains nearly unchanged.

6. Performances Comparison

Table 1. Comparison with other reported sub-6 GHz 8-antenna array designs.

Ref.	Ant. Size (mm2)	Frame Profile (mm)	BW (GHz)	Iso. (dB)	ECC	Peak Channel Capacity (bps/Hz)
[39] ^	2 × 7frame 9 × 2ground slot	7	3.2–5.93 10-dB	>12	<0.1	43.93 (8 × 8)
[40] ^	17 × 7frame	7	3.3–6 6-dB	>18	<0.05	NA
[41]	13.9 × 6	7	3.1–6.0 10-dB	>10	<0.1	39 (8 × 8)
[42]	16 × 5	7	3.3–6 6-dB	>12	0	NA
[43]	30 × 6.2 *	7	3.3–7.5 6-dB	>10	<0.06	NA
[44]	28 × 7 &	5	3.3–5 6-dB	>10	<0.3	NA
[46]	21.5 × 6	7	3.3–5 6-dB	>14.5	<0.1	38 (8 × 8)
PA	21 × 6	7	3.1–6 6-dB	>10	<0.1	40 (8 × 8)

PA: propose antenna, NA: not applicable, * IFA pair, ^ metallic frame, and ^& integrated slot antenna pair with connected line of width 1.8 mm.

shows the performance comparison between the proposed antenna and other wideband designs reported in the open literature. By observing Table 1, Refs. [39,40] are for metallic frame applications. Even though the works in [41,42] show much smaller sizes than the proposed one, they require very complicated structures, and the feeding technique applied is coupled-fed type. Regarding the proposed eight-antenna array that utilizes a modified loop antenna design, when compared to other loop antenna structures such as [46], it demonstrates significantly improved bandwidth performance without requiring the loop antenna to be printed on both sides of the substrate. This characteristic facilitates a simplified manufacturing process and reduced production costs for the antenna.

7. Indoor Throughput Measurement

To evaluate the proposed antenna’s throughput in a real propagation environment, an indoor measurement was conducted by BWant Co., Ltd. in Taipei, Taiwan. Due to equipment limitations, only 4 × 4 channel measurements were performed instead of 8 × 8. As shown in Figure 21a, the experiment setup involved four Vivaldi antennas with an operating frequency band of 0.5–6 GHz as transmitting antennas, mounted 0.9 m above the ground. The receiving antennas (the proposed antenna) were positioned 2.5 m away from the transmitter, also at a height of 0.9 m. Figure 21b illustrates the arrangement of the transmitting antennas, with a pair of Vivaldi antennas placed vertically and the other pair positioned orthogonally (horizontally) for this measurement. At different frequencies of 3.6 GHz, 4 GHz, and 5.1 GHz, the measured throughput for the 4 × 4 channel measurement was consistently 2023 Mbps. Although the ideal 4 × 4 channel throughput is calculated to be 4430 Mbps, considering the indoor multipath fading effects, the measured throughput of 2023 Mbps (or 2.023 Gbps) is considered a very good result.

8. Conclusions

An eight-antenna array design operating across the 5G NR-bands (3300–5925 MHz) for smartphone applications has been successfully investigated. The measured 6 dB stacked bandwidth and isolation were 3.15–6.17 GHz and >10 dB, respectively. The measured total efficiency of the eight-antenna array was 51–84%. Regarding MIMO performance, the ECC remained under 0.1 across the bands-of-interest, and the maximum CC for an 8 × 8 MIMO system reached >40 bits per second per Hertz (bps/Hz). Finally, an indoor throughput measurement was conducted, and the 4 × 4 channel throughput was consistently measured at 2023 Mbps across the bands of interest.