The application of reconfigurable antenna arrays (RAAs) in millimeter wave-based communication systems, particularly in the Ka-band, has recently gained significant attention, driven by the increasing demand for high data rates. However, cross-polarization in these antennas often leads to signal degradation and unwanted interference from different polarization modes, impacting the overall performance of communication systems. Therefore, minimizing cross-polarization in dual-linearly-polarized antenna arrays is critical to ensure efficient and reliable millimeter-wave (mmWave) communication.

To efficiently switch between linear polarization modes, various antenna architectures have been investigated over the years, including those based on dual-feed ports [1, 2], micro-electromechanical systems (MEMS), radio frequency (RF) chips, and PIN diodes. However, each of these approaches has its own advantages and disadvantages. For instance, dual-feed configurations are frequently employed in lower frequency ranges, such as the S-band, because of their durability and appropriateness for larger elements within an antenna array [3]. However, it is difficult for them to adapt to the reduced sizes necessary for Ka-band applications with two traditional ports, thus limiting their use in linear antenna arrays [4]. Alternatively, to deal with a planar array, a modern RF chip attached in a package structure along with a complicated multilayer design has been proposed, but this method introduced its own set of complexities [5, 6]. Meanwhile, PIN diodes are widely used to generate linearly polarized radiation across a broad frequency range, starting from low RF frequencies to approximately the X-band [7–9]. Their popularity stems from their compact size, affordability, ease of integration, control of printed circuit boards (PCBs), and high isolation in the OFF state.

Achieving a higher or lower cross-polarization discrimination (XPD) relies on having sufficient isolation in the OFF state of the PIN diode. However, at higher frequencies, particularly in the Ka-band, a single PIN diode may not be able to block RF currents effectively. In such cases, instead of focusing on achieving linear polarization switching, the PIN diode is often used to control the current flow on the antenna patch and to induce phase shifting [10, 11]. Although other switching devices, such as the SPST diode, offer excellent isolation in the OFF state, implementing a dual-linearly-polarized reconfigurable system remains a significant challenge. One of the main difficulties is the small antenna size, which limits the incorporation of an additional DC bias line into the antenna configuration to control the diode.

Drawing on these observations, we propose a cost-effective approach for constructing a compact dual-linearly-polarized antenna array with high XPD in the Ka-band using PIN diodes. To achieve this, we first implemented the image configuration technique to suppress cross-polarization in the dual-linearly-polarized array using two PIN diodes. Notably, the use of a dual-feed structure when implementing this method has been extensively studied in various papers [12–14]. By positioning the PIN diodes at specific positions corresponding to the 0° or 180° port phases, we configured the proposed single-port antenna elements into various 2×2 subarray arrangements to examine cross-polarization degradation. Subsequently, we evaluated the performance of a 4×4 antenna array by conducting simulations and experimentation using an external beamformer, thereby validating the reduction in cross-polarization while ensuring that the effect remains consistent during beam steering.

Fig. 1 depicts the geometrical structure of the proposed antenna element, with its dimensions measuring 6 mm × 6 mm × 1.076 mm. The antenna comprises four metallic layers on two identical Rogers RT/duroid 5880 substrates (with ε_r = 2.2 and tanδ = 0.0009) that are bonded together by an adhesive layer of Copper Clad Laminate DS-7402 (BS #1067, with ε_r = 3.7 and tanδ = 0.02). The patch antenna is located on the top metal layer (L1) and fed by a metalized via (diameter, 500 μm) positioned at the center. For polarization switching, two PIN diodes (MACOM MADP-000907-14020W Flip Chip) are embedded into the patch antenna. They are controlled by a bias line printed on the second metal layer (L2) of the first substrate.

A bias network consisting of microstrip radial stubs is connected to the patch by metalized vias to ensure high RF signal isolation. Ground connection is established using a short-circuit stub (L2), connected to the ground plane using metalized vias (from L1 to L4). Notably, in addition to its role as the patch antenna ground (L3), the second substrate also serves as the ground for the SMPM connector (L4). Notably, the PIN diode model employed in this study, adopted from previous on-PCB measurements [15] and the MACOM PIN diode datasheet, consists two series circuit (one with R_ON = 4.2 Ω and L_ON = 0.05 nH, and the other with L_OFF = 0.05 nH and C_OFF = 42 fF) in the forward (I_bias = 10 mA) and reverse (V_bias = −1.2 V) states, respectively. All relevant geometrical dimensions pertaining to the proposed antenna element are summarized in Table 1.

The two PIN diodes were configured in opposite states for each polarization mode through a shared bias line, as shown in Fig. 1. When PIN diode D1 is in the ON state, D2 is forced to be in the OFF state, the O-slot is short-circuited at one of its ends, and the patch antenna radiates horizontally polarized outgoing waves. Conversely, for the vertically polarized radiated wave, PIN diode D2 is switched ON. Furthermore, Fig. 2 clearly depicts the current distribution on the edge and main bars, producing corresponding polarizations. The E-field of the two operating modes exhibits stronger radiation on the sides without the activated PIN diodes. Fig. 3 presents the simulated S₁₁ and the co- and cross-polarization performance of the antenna based on two outgoing wave modes at 29 GHz. The proposed antenna achieved a 5 dB realized gain, an XPD of less than 18 dB, and a bandwidth of 1.2 GHz (−10 dB S₁₁) in both polarization modes. Nonetheless, a low XPD could pose many challenges for array implementation at other beam steering angles, particularly in scenarios where achieving high isolation is crucial to mitigating interference from the remaining polarizations in 5G communication systems.

Fig. 7 depicts the antenna array under test in an anechoic chamber. For this analysis, we employed a beamformer, with its insertion loss within the operating frequency range effectively corrected to approximately 0 dB. The measured phase and magnitude of the 16 output ports exhibited variances of 35° and 2.5 dB, respectively, relative to each other. Furthermore, the antenna array was equipped with two main bias lines (DC1 and DC2) to control the PIN diode states for each individual element within the array. It was verified that the proposed antenna is capable of working in two linear polarizations, while also offering beam scanning capability.

Due to the limited sample size of the 4×4 antenna array, we focused on measuring the high directive beam from −30° to +30°, as shown in Fig. 8. The scanning ability of the array was restricted by fluctuating sidelobes and gain reduction. Nonetheless, good agreement was observed between the measured and simulated patterns for co-polarization. For the linear H- and V-polarizations, the peak gain reached 15.2 dB, and the XPD was about 25 dB. Although the observed XPD exceeded 20 dB, the cross-polarization deteriorated by roughly 10 dB compared to the simulated pattern. Notably, this degradation occurred within the operating frequency range, as shown in Fig. 9.

Fig. 6 confirms that cross-polarization suppression is highly sensitive to unidentical elements. This implies that although random magnitude and phase errors were accounted for in the simulation, the main factors contributing to performance degradation are manufacturing inaccuracies and imperfections, which occur particularly during the manual soldering of connectors and small PIN diodes (0.3 mm × 0.7 mm in size). At mmWave frequencies, these sensitivities become more noticeable, significantly affecting the effectiveness of the image configuration technique. As shown in Fig. 8, the scan loss at 30° was found to be approximately 3.5 dB in the orthogonal plane of active polarization and about 2.7 dB in the active polarization plane. Notably, this phenomenon, which resulted from the cancellation of mirroring elements, was predicted in the simulation. The SLL also experienced an imbalance when the beam steered to the orthogonally activated polarization plane, declining to −7 dB from −8.5 dB at ±30° in the activated polarization plane. The XPD was estimated to be more than 20 dB in the operating frequency range, along with a radiated efficiency of 72%, which is a good result compared to the other works reported in Table 2 [16–20].

In this study, we examined a dual-linearly-polarized RAA in the Ka-band by employing the cross-polarization suppression technique. While the individual antenna elements exhibited a low XPD, the overall cross-polarization of the antenna array was effectively suppressed by mirroring the antenna elements with respect to the H- and V-planes. Furthermore, a 4×4 prototype was fabricated and subjected to experimentation in an anechoic chamber using a beamformer. The compact design of the antenna array and its efficacy in cross-polarization suppression underscore its potential for integration with contemporary Ka-band phase shifters. One such example is the use of a transmitarray with 1-bit phase quantization (0°/180°) that employs an O-slot structure with PIN diodes or an RF chip featuring a single port for each element, as opposed to a conventional dual-port setup.