Beam-Switching Antenna Array Using Rotman Lens for Railway Communication at 28 GHz

Article information

J. Electromagn. Eng. Sci. 2026;26(3):271-279

Publication date (electronic) : 2026 May 31

doi : https://doi.org/10.26866/jees.2026.3.r.363

Hyeongjun Cho¹, Jaehyun Kim¹, Ashwini K. Arya², Chang-Woo Kim^1,, Sanghoek Kim^1,

¹Department of Electronics and Information Convergence Engineering and the Institute for Wearable Convergence Electronics, Kyung Hee University, Yongin, Korea

²School of Engineering, Jawaharlal Nehru University, New Delhi, India

^*Corresponding Author: Sanghoek Kim (e-mail: sanghoek@khu.ac.kr), Chang-Woo Kim (e-mail: cwkim@khu.ac.kr)

Hyeongjun Cho and Jaehyun Kim contributed equally to this work.

Received 2025 September 5; Revised 2025 October 23; Accepted 2025 November 21.

Abstract

I. Introduction

Wireless communication plays a vital role in transportation systems [1–4]. Specifically, in railway transportation systems, antennas are typically mounted on the rooftop of the vehicle [5–7]. The communication schemes supported by railway antennas include long-term evolution (LTE), 5G FR1 (below 6 GHz), and 5G FR2 (near 28 GHz) [8–10]. In the upper 5G band (FR2), high gain is essential to compensate for the severe path loss in millimeter waves and to elongate the communication distance, which can be achieved through appropriate array configuration [11].

A high-gain antenna oriented in the static beamforming direction can be easily designed and fabricated using a fixed feeding network. Such an antenna is well suited for straight track intervals (Point A to Point B), as illustrated in Fig. 1.

Fig. 1

Typical scenario depicting a wireless railway link communicating with base station antennas positioned alongside the track.

However, in reality, transportation tracks include curved sections, such as the interval between Points B and C in Fig. 1, where the beamforming direction must vary over time. In practical railway environments, track curvature can result in turning angles of up to −20° to 20° [12, 13], depending on line geometry and train speed. Accordingly, the antenna system must be designed to enable beam steering from −30° to 30°, which would be sufficient to compensate for the track curvature. Moreover, to build a compact antenna with a controllable beam pattern, beam steering needs to be implemented electronically. In electronic steering for controllable beamforming, the phase and amplitude of the array elements must be adequately adjusted to achieve constructive interference among the radiated waves, thereby concentrating the total transmitted signal toward a targeted beam direction. In a multiple antenna system, the wavefront direction is controlled by phase shifters. By changing the phase of individual signals in an antenna array, a beam can be steered toward a particular angle, causing the plane wave to propagate in the desired direction [11, 14–17]. However, while phase shifters are often implemented using microwave diodes, field-effect transistors, or ferrite toroids, it is tricky to build a high-resolution phase shifter when the wavelength is short. Moreover, this would also increase the price of the component, especially in the case of millimeter waves.

In this context, the Rotman lens can be employed as an alternative beamforming network to replace phase shifters [18]. This lens-based beamformer has emerged as an attractive solution due to its low cost, reliability, simple design, and wide beam-scanning capabilities. A conceptual configuration of the proposed beam-switching system is illustrated in Fig. 2. In this system, the RF signal from the source is routed through a switching network to one of the input ports (P1–P4) of the Rotman lens, which serves as a feeding network that provides the required phase distribution to the antenna array. Each input port corresponds to a different beam direction, which is determined by the geometrical design of the Rotman lens.

Fig. 2

Schematic of the proposed beam-switching system.

This system avoids the complexity arising from the use of multiple phase shifters for wide-angle beam steering in an electronic scanning array Since the Rotman lens is a time-delay device, it produces beam steering that is relatively insensitive to the operating frequency. As a result, it is more capable of supporting wide-band operations compared to other beamforming networks [18–21].

In the upper 5G band (28 GHz), high directivity is essential for long-distance communication. If the beam direction is static, a finite curvature of the track (from Point B to Point C) may interrupt communication. Therefore, to ensure seamless wireless communication at low system cost, we present a beam-steerable antenna with a Rotman lens.

A notable drawback of the Rotman lens is that the number of available beam directions is limited by the discrete input ports, unlike systems using phase shifters that allow continuous beam steering. Nonetheless, we claim that the Rotman lens is particularly suitable for railway applications because the relative position of the transceivers along the tracks is rather restricted, meaning that there is no need for continuous steering of the beamforming direction over a wide range. Moreover, since the relative position of the transceivers is restricted in the vertical direction, a one-dimensional array equipped with a Rotman lens is sufficient to cover the beamforming range.

Therefore, this work proposes a feeding network that uses the Rotman lens for phase distribution, thereby forming multiple beams in railway applications. The array antenna fed by the Rotman lens attained a gain higher than 10 dBi. Simultaneously, the size of each antenna element was minimized, allowing the entire antenna to fit inside a shark-fin cover.

The choice of antenna element can make the antenna polarization either linear or circular. In the case of wireless communication through linear polarization (LP), an obstacle (e.g., a tunnel wall) located between the transmitter and the receiver may change the antenna polarization from horizontal to vertical, and vice versa [22]. Therefore, significant polarization loss may occur in LP due to non-line-of-sight (NLOS) coupling between transceivers. In the case of circular polarization (CP), such an obstacle does not change the polarization direction from left-handed to righthanded, and vice versa. Therefore, communication performance in CP remains robust, even under NLOS conditions. In this study, we investigated both linearly and circularly polarized antenna arrays as antenna elements for 5G railway communications.

Overall, we demonstrate that the proposed antenna outperforms previously reported antennas in terms of size, directivity, and bandwidth, in addition to the cost savings realized by replacing phase shifters with a Rotman lens. These features make the proposed antenna a compelling choice for railway communications at 28 GHz.

II. Proposed Antenna Design Studies

We designed one antenna element for LP and another for CP. For both designs, full-wave electromagnetic simulations were conducted using CST Microwave Studio [23]. Subsequently, the feeding lens was designed for a carrier frequency of 28 GHz, and then integrated with the antenna elements to form the complete antenna with beamforming operations. The feeding lens produces four distinct beams, enabling uninterrupted communication between the antenna on the train’s roof and the base stations located along the track.

1. Antenna Design

As shown in Fig. 3(a), a rectangular patch antenna with an inset feed was designed as the linearly polarized antenna element, with the operating frequency being 28 GHz. The antenna size was 2.5 mm × 2.4 mm, and it was fabricated on an FR4 substrate (dielectric constant 4.4 with thickness 0.6 mm) of size 5.5 mm × 5.5 mm. A single 1-mm inset feed line (feed line width = 0.25 mm) was connected to the antenna to yield good input impedance matching.

Fig. 3

Design geometries of the proposed antenna element for (a) linear polarization and (b) circular polarization (unit: mm).

The circularly polarized antenna was designed in two steps— one pertaining to the single antenna element and the other concerning the dual antenna element. Notably, CP is obtained when two orthogonal modes of equal amplitude are excited with a 90° phase difference. In this study, a diagonal slot was added to the rectangular patch to enhance CP performance, thus forming the single antenna element, which is depicted in Fig. 3(b) as the area within the rectangular dashed lines.

To construct the dual antenna element, we added one more patch to create a series-fed configuration, as shown in Fig. 3(b). The center-to-center distance between the two patches was adjusted to 7 mm so as to equalize the excitation phases and increase the gain in the broadside direction.

The dual element enhanced both CP performance and gain. To further examine it, the axial ratio, which defines CP performance, was simulated for the single and dual elements with respect to the broadside angles at 28 GHz. Fig. 4 shows that the axial ratio of the dual element remains below 3 dB for a wider range of broadside angles than that of the single element. Fig. 5 illustrates the axial ratio of the single and dual elements with respect to frequency. At 28 GHz, the axial ratio of the dual element is about 1 dB less than that of the single element, while the axial ratio bandwidth of the dual element reached 500 MHz. This confirms that the dual element outperforms the single element in terms of both spatial angle and absolute value at the carrier frequency.

Fig. 4

Simulated axial ratio in the observation plane at 28 GHz.

Fig. 5

Axial ratio with respect to frequency for the circularly polarized single element, the circularly polarized dual element, and the circularly polarized antenna array. For the circularly polarized antenna array, all elements were fed the same amplitude and phase.

To further increase the gain and enable steerable beamforming, a five-element linear polarized antenna array (LPA) with an element separation of 5 mm was employed. The same procedure was applied to build a circularly polarized antenna array (CPA), where the separation between the adjacent elements was maintained at 5.5 mm. The antenna arrays for LP and CP are depicted in Fig. 6(a) and 6(b), respectively. The total size of the array is 14 mm × 26.5 mm. In Fig. 6, the y-direction of the coordinate space is set along the direction of the arrangement of the antenna elements. This implies that the radiation pattern in the yz-plane, which is our plane of interest, can be enhanced by the array pattern.

Fig. 6

Proposed antenna design geometries for (a) the linearly polarized array and (b) the circularly polarized array (unit: mm).

The simulated S-parameters, including mutual coupling between adjacent elements (indexed by i and j) in the required frequency range (from 24 to 32 GHz), are presented in Fig. 7(a). Here, index i in j = i ± 1 refers to the port number in the 5- port antenna array configuration. Notably, since all the antenna elements were the same in appearance, the results for only one port (i = 3) in the antenna array configuration are presented. A bandwidth of around 2 GHz is observed in both LPA and CPA cases (|S₁₁ < −10 dB|), while the mutual coupling between adjacent elements is below −20 dB. Furthermore, the axial ratio for CP radiation in the xz- and yz-radiation planes is presented in Fig. 7(b), showing that it is below 3 dB for both planes in the broadside direction (θ = 0°). When an input signal was fed through Port 1, the maximum array factor was directed toward +30°. Likewise, when the input signal was fed through Ports 2, 3, and 4, the maximum array factor was produced at +15°, −15°, and −30°, respectively. This feature enabled beam steering in a desired direction based on input port selection (Fig. 8).

Fig. 7

Antenna array performance: (a) S-parameters (j = i ± 1) and (b) axial ratio of the circularly polarized antenna array.

Fig. 8

The Rotman lens and its array factor: (a) the Rotman lens for beam steering with four input ports, five output ports, and five dummy ports and (b) the simulated array factor.

III. Results and Discussion

In this section, the proposed beamforming concept is discussed, based on the measurement results obtained from the fabricated antenna array with the feeding network. Since the antenna array and feeding lens were designed using the same substrate, proper connections could be established without the need for separate coaxial connectors. The antenna arrays (LPA and CPA) and the feeding lens were fabricated on FR4 substrate with the aim of investigating the beam scanning angles of various input ports. The fabricated prototypes for the LPA and CPA are presented in Fig. 9(a) and 9(b). The complete antenna configuration features four input ports for the four respective beams, along with five dummy ports. Each input port provides appropriate phase distributions for the antenna array to achieve the desired beam direction.

Fig. 9

Lens antenna fabrication and measurement results: optical images of (a) the linearly polarized antenna array and (b) the circularly polarized antenna array integrated with the Rotman lens; (c) S-parameters; and (d) the antenna sample in an anechoic chamber for radiation pattern measurement.

Measurements were performed with the input signal applied to one port, while the remaining ports were loaded with 50- ohm impedance. The antenna’s S-parameters were measured using a vector network analyzer (VNA). All S-parameter curves remained below −10 dB for the entire range of frequencies owing to the wideband characteristics of the Rotman lens. The S-parameters of P1 for both arrays, i.e., LPA and CPA, are depicted in Fig. 9(c), showing negligible reflection across the bandwidth ranging from 27 GHz to 29 GHz, which is sufficient to support the 5G millimeter-wave band in South Korea, Japan, and the USA [25, 26].

The radiation patterns were measured in an anechoic chamber (Fig. 9(d)) at the operating frequency of 28 GHz. The proposed antenna (both LPA and CPA), considered the receiving antenna, was rotated from −90° to +90° in the yz-plane. The beam observations are presented in Fig. 10(a) and 10(b). The dashed curves represent the simulation results, and the solid ones denote the measurement results. The results for both LPA and CPA show good agreement. Furthermore, the beam directions are observed at certain angles, such as +30°, +15°, −15° and −30° for excitation ports P1, P2, P3, and P4, respectively.

Fig. 10

Simulated and measured radiation beam patterns of (a) the linearly polarized array and (b) the circularly polarized array.

Furthermore, the axial ratio for CP was measured in the desired observation plane (yz-plane). Since the axial ratio is considered the primary metric for evaluating CP performance, the axial ratios for the four input ports were measured. The results in Fig. 11 show that the axial ratio is below 3 dB in the +33° direction for P1. Furthermore, when the signals are fed through P2, the axial ratio is below 3 dB in the 14° direction. Similarly, when P3 and P4 are the input ports, the axial ratios are below 3 dB in the −14° and −33° directions, respectively. Overall, the axial ratio is below 3 dB at certain points near the required beam scan angles. The other antenna performance parameters, such as directivity, peak realized gain, efficiency, and sidelobe level (SLL), for each beam are presented in Table 1.

Fig. 11

Simulated and measured axial ratios of the circularly polarized array.

Table 1

Antenna performance parameters

Lastly, Table 2 presents a performance comparison of the proposed antenna with other antennas reported in the literature for operation at 28 GHz [27–31]. The antenna gain achieved in [32] is 7.3 dBi, with an axial ratio of 2 dB and a patch size of 3.3 mm × 3.3 mm. When the feed line is included, the total size of the antenna reaches approximately 10 mm × 19 mm. Another work on the 5G antenna presented an antenna size of 26.4 mm × 20.4 mm, which managed a gain of 7.03 dBi at 28 GHz [33]. In these works, the large size of the antennas contributed to a high gain. In contrast, the miniaturized antenna with a radius of 0.19λ in [27], where λ (≈11 mm at 28 GHz) denotes the free-space wavelength, exhibited an almost omnidirectional radiation pattern. Notably, this antenna size could be minimized by adopting a substrate made of high dielectric material. Another small antenna with a size of 0.5λ × 0.5λ × 0.08λ was reported in [28]. In this case, shorting vias and proper matching with the balun were required at the operating frequency. Furthermore, a patch antenna array of size 39.3 mm × 30.65 mm attained a bandwidth of 308 MHz and a high gain of 17 dBi, but with a static beamforming direction [29].

Table 2

Comparison with other antennas at 28 GHz

In contrast, the proposed antenna offers a good compromise between size, gain, and controllability. The compact size of the arrays enables the antenna, along with the feeding network, to fit inside a shark-fin cover. Moreover, steerable beamforming with a gain higher than 10 dBi makes the antenna suitable for railway communication, even with a possible curve on the track.

IV. Conclusion

In this paper, a novel method that allows beamforming at the 28-GHz band without the involvement of phase shifters is presented for application in railway communication. The Rotman lens is employed to introduce the necessary phase shifts for beamforming, thereby simplifying the feeding network and reducing implementation costs. However, since the lens generates discrete phase distributions defined by its geometry, the beam can be steered only in a finite number of directions. This inherent trade-off between simplicity, cost, and steering flexibility is acceptable in railway communication systems, where only limited beam coverage is required and low-cost implementation is highly desirable. The proposed concept is examined using both LP and CP antenna arrays. In the case of CP, the axial ratio is measured to be below 3 dB for all ports. Compact antennas of size 5.5 mm × 5.5 mm for LP and 14.0 mm × 4.5 mm for CP are employed for the antenna arrays, achieving a maximum directivity of 10.8 dBi and 11.2 dBi, respectively. These sizes are small enough for the antennas to comfortably fit inside a shark-fin cover during vehicular operations [9]. Overall, high gain, compact size, steerable beamforming, and reduced system cost by eliminating expensive phase shifters make the proposed antenna attractive for railway applications in the upper 5G band.

Article information Continued