Let’s first consider the magneto-electric dipole (MED) antenna design focusing on the antenna element of an array radiating at broadside only, i.e., a fixed-beam array. This is also the starting point for the scanning phased array antenna element design. The considered MED antenna is based on the design approach outlined in Li and Luk (2015). The proposed MED antenna element is shown in Figure 1 and consists of four metallic square patches that function as the electric dipoles and the space between the metallic vias that function as the equivalent magnetic dipole. The excitation of the antenna is achieved by an L-probe feed. The design dimension parameters are shown in Figure 1 and their numerical values are tabulated in Table 1. The width and the length are denoted by W _MED, and L _MED, respectively. The substrate thickness is denoted by H _sub. The width and the length of the dipole patches are given by W ₁, L ₁, respectively. L ₂ denotes the length of the L-probe foot, while g ₁ and g ₂ denote the separation distance between the dipole patches as illustrated in Figure 1. The diameter of vias is d _vias and the minimum pad size is denoted by g _vias.

The unit cell size of the considered MED antenna element is W _MED = λ _cf, where λ _cf is the wavelength of the center operating frequency considered here, i.e., f _cf ≈ 28 GHz. Since the unit cell dimension is large, it thus can only be used for fixed-beam applications where broadside radiation is of interest. Otherwise, grating lobes will appear while scanning.

We now investigate the impact of the substrate dielectric constant ɛ _r on the |S ₁₁| ≤ −10 dB bandwidth of the MED antenna element for fixed-beam applications. The considered substrates in this study are the Rogers RO 5880, RO 3003, RO 4350B, and the Panasonic Megtron-6, which electrical and dimensional characteristics are summarized in Table 2. More specifically, we consider 2.2 ≤ ɛ _r ≤ 3.6, where 2.2 and 3.6 are the dielectric constants of the Rogers 5880 and Rogers 4350 B, respectively. It is worthwhile to note that the dielectric constants of the most commonly used substrates for mmWave applications with low-loss performance typically fall within this range. The computations are done by first optimizing the MED dimensions for the Rogers RO 5880 substrate, which values are shown in Table 1. These values were then kept constant, while ɛ _r was increased.

Figure 2 shows a comparison of the computed reflection coefficient |S ₁₁| for the MED antenna element as a function of different dielectric constant values ɛ _r . As can be seen from Figure 2, increasing the ɛ _r results in the shift of |S ₁₁| toward lower frequencies. This is expected, as the basic theory of microstrip antennas predicts that when the ɛ _r of the substrates increases, the effective wavelength of the antenna will decrease accordingly, e.g., see Kildal (2015). Another important observation is that increasing ɛ _r can be exploited to reduce the size of the antenna unit cell with is in line with the study presented in Schaubert et al. (1989). In addition, it is worthwhile to note that increasing or decreasing ɛ _r by 10%, or about 0.2 in absolute value, will shift the entire operating frequency by approximately 3%, down or up, respectively. Finally, the fractional bandwidth also varies from 43% to 45% as ɛ _r changes within the considered interval.

Our next step is to use the MED antenna element described above to design a phased array antenna. However, as stated above, the MED antenna dimensions are too large for this purpose. Therefore, the dimensions of the proposed MED are firstly altered by choosing the dimensions satisfying the non-radiating grating lobes condition. Hence, the unit cell dimension of the MED antenna must satisfy the well-known relationship for the inter-element spacing Kildal (2015) W MED ≤ λ h 1 + sin θ scanning , (1) where the λ _h is the wavelength of the highest operating frequency of interest, i.e., f _h ≈ 30 GHz for the targeted design, and θ _scanning is the scanning angle. Assuming that the non-grating lobes condition for scanning shall be satisfied for θ _scanning ≤ 60° we use 1 to arrive at the unit cell dimension of W _MED = L _MED ≈ 0.52λ _h for all the cases that will be discussed from this point on. In choosing this value we also considered that sufficient room will be left for optimizing the MED unit cell dimension. In addition, we fix the thickness of the MED to H _sub = 1.52 mm in order to give a sufficient fair comparison for all different substrates given in Table 2. Excluding the aforementioned parameters, all the other parameters, i.e., W ₁, L ₁, L ₂, g ₁ and g ₂ (see Figure 1) have been optimized to ensure the scanning impedance bandwidth, also known as active impedance. Table 3 tabulated the optimized parameters of the MED antenna for phased array based on different dielectric substrates. The optimization is done to maintain the bandwidth at the broadside to cover the 24–30 GHz band. These values are sufficiently good to provide the adequate scanning performance of the phased array. Nonetheless, optimization of the parameters did also take into account the fabrication tolerances/limitations such as the minimum diameter of vias within d _vias ≥ 0.3 mm and a pad size within g _pad ≥ 0.1 mm.

The computed scanning impedances for both E- and H-planes scanning of the proposed unit cell MED antenna are depicted in Figure 3, for Roger 5880, Roger 3003, Roger 4350 B, and Panasonic Megtron-6. The proposed MED phased array antenna is intended to operate from 24–30 GHz with acceptable degradation of the scanning impedance from −10 dB to −7 dB, which is equivalent to an additional 0.5 dB gain drop due to the mismatching caused by beam scanning Hansen (2009). For instance, with scanned impedance of −10 dB, the power loss due to the mismatch is approximately 0.5 dB and is increased to 1 dB when the scanned impedance has deteriorated to −7 dB. The power loss due to the beam scanning mismatching can be computed using mismatch loss, L _mismatch = 1 − |Γ|², where Γ is the reflection coefficient Pozar (2011). There are several observations regarding the impact of the dielectric substrate on the scanning impedance as follows next. The scanning performance of the MED based on RO 5880 and RO 3003 is generally better than for RO 4350B and Megtron-6. For instance, considering the operating frequency of 24–30 GHz, in the E-plane scanning, the MED based on RO 5880 and RO 3003 can scan up to 60° at S _11,scanned ≤ −7 dB, while the MED based on RO 4350B and Megtron-6 can only scan up to 50°. Similar performance is observed for H-plane scanning where MED based on RO 5880 and RO 3003 can scan up to 50°, whereas the MED based on RO 4350B and Megtron-6 can only scan up to 40°. In addition, it is also worthwhile to note that as the dielectric constant of the substrate increases, the lower-end operating frequency (24 GHz) provides better active matching while scanning. This is mainly due to the shifting in the operating frequency which is in line with the earlier observation of the fixed beam MED antenna. For example, considering the −7 dB scanned impedance at scanning angle θ = 40°, the lowest operating frequency is approximately 23.8 GHz for MED based on RO 5880, whereas this lowest frequency for MED based on RO 3003, RO 4350 B and Megtron-6 is approximately 22 GHz. Nevertheless, this shifting is not as significant as it is in broadside, as the mismatching due to the mutual coupling and surface waves will also impact the overall performance of the phased array when the beam is scanned away from the broadside Hansen (2009).

As well known, the performance of the center elements can be approximated fairly well by an infinite array model. However, in realistic array antennas with a finite number of antenna elements, the edge element patterns, and active reflection coefficients will suffer from inaccuracies due to the asymmetric environment. Therefore, to further investigate the beam scanning performance of the proposed MED phased array antenna based on various dielectric materials, an 8 × 8 finite array antenna, as seen in Figure 4, is simulated in CST Microwave Studio with an open boundary condition. Since the proposed MED antenna was evaluated using the unit cell boundary condition in the previous investigation, for finite array antenna, to minimize the undesirable edge effects, an additional row/column of dummy elements is added to each side of the 8 × 8 array layout.

Figure 5 shows the computed radiation pattern of the proposed 8 × 8 MED phased array antennas at 24 GHz (lower end of the targeted operating frequency) and 30 GHz (higher end of the targeted operating frequency) based on various types of dielectric substrates as explained in the figure caption. Results are provided for scanning in the E- and the H-plane. There are several relevant observations that can be drawn from the computed radiation patterns. First, taking into account a 3 dB loss in gain when the radiation beam is scanned away from the broadside direction, the considered MED phased array exhibits better scanning performance at the E-plane than at the H-plane for all substrates. It is worthwhile noting that this effect is more significant at the 30 GHz than at 24 GHz. This is explained by the fact that the wavelength at 30 GHz is closer to the maximum allowed inter-element distance avoiding the appearance of grating lobes given by 1. This results in a more significant mismatching of the phased array antenna impedance. This finding is consistent with the conclusion based on the computed scanning impedance of the MED unit cell phased array shown in Figure 3. Indeed, a better scanning impedance matching is achieved for the E-plane scanning as compared to the H-plane scanning of the proposed MED phased array antenna. Second, the scanning performance of the MED phased array antennas based on the RO 5880 and the RO 3003 are comparable. The MED phased array antenna based on these substrates could typically deliver full scanning, i.e., up to ±60°, for the E-plane and up to approximately ±50° for the H-plane. It is important to note that, for the H-plane scanning when the beam is scanned up to ±60°, although the gain drop is larger than 3 dB in both cases, the MED phased array antenna based on RO 3003 exhibits a more substantial drop in gain compared to the one based on RO 5880. This is due to the more significant mismatching in the scanning impedance of the RO 3003 MED phased array antenna, as demonstrated in the unit cell evaluation in Figure 3. On the other hand, the scanning performances of the MED phased array based on RO 4350 B and Megtron-6 are comparable. Scanning up to ±50° and ±40° are possible for the E- and H-planes, respectively, with a gain drop of less than 3 dB. This observation of the scanning performance is consistent with the assessment of the scanning impedance of the MED phased array antenna unit cell evaluated with the infinite boundary condition.

In addition, another key parameter of the radiation performance of the phased array antenna is the side lobe levels (SLLs). To attempt a fair comparison for all the considered substrates, we fixed the larger scanning angle to 50° and 40° for E- and H-planes, respectively. It is worth noting that, in the broadside direction, the SLLs of the proposed MED antenna for all considered substrates are as low as −13 dB. Nevertheless, when the beam is scanned away from the broadside, an intriguing observation can be made. Firstly, at the lower-end frequency (24 GHz), the SLLs for the MED based on RO 4350B and Megtron-6 are approximately equal to −11.3dB, which are generally better than the MED based on RO 5880 and RO 3003, with SLLs approximately −11 dB. This effect can be explained by the better matching at the lower-end frequency due to the shifting of operating frequency as the result of the increase of dielectric constant as aforementioned. On the other hand, at the higher-end operating frequency (30 GHz), the SLLs for the MED based on RO 5880 and RO 3003 are generally better than the MED based on RO 4350B and Megtron-6. The SLLs for E-plane scanning are approximately −11.0 dB, −10.9 dB, −10.5 dB and −10.6 dB for RO 5880, RO 3003, RO 4350B, and Megtron-6, respectively. Moreover, the SLLs for H-plane scanning are approximately −12.0dB, −11.8 dB, −10.6 dB, and −11.1 dB for RO 5880, RO 3003, RO 4350B, and Megtron-6, respectively. This result can be explained by the surface waves in the dielectric materials. The cutoff frequency of the excitation of the surface wave shifts toward lower frequencies as the dielectric permittivity increases Horsfield and Ball (2000). Therefore, for a fixed dimension of the unit cell MED antenna, the surface wave propagating in a dielectric substrate with a higher dielectric permittivity value will appear at a smaller scanning angle as compared to the MED antenna based on lower dielectric permittivity, reducing the beam scanning capability of the MED phased array antenna.

Figure 6 shows the computed gain and the maximum cross-polarization level X − Pol based on different dielectric substrates for broadside and scanning angle θ = 60° degrees at E- and H-planes, respectively. As can be seen from Fig.- 6, the array antenna gains of the proposed MED phased array based on the various dielectric substrates are comparable to each other. The minor difference observed is primarily caused by the different dielectric losses of the materials. Nonetheless, when the antenna beam is scanned away from the broadside to θ = 60° in both the E- and H-planes, the computed gain performance varies significantly at the higher end of the operating frequencies depending on the dielectric substrate. This fluctuation in the computed gain at the higher operating frequencies of the proposed MED phased array is mainly caused by mismatching in the scanning impedances, as discussed above. Furthermore, the relative cross-polarization level is sufficiently low in the broadside direction and E-plane scanning for all substrates. However, the level of cross-polarization for H-planes at θ = 60° is high. This is primarily due to the fact that the proposed MED does not perform properly at θ = 60° for the H-plane, as revealed in the scanning impedance study presented above.

The performance evaluation of the MED phased array antenna based on the considered dielectric substrates discussed earlier is summarized in Table 4. In addition to the scanning performance, the manufacturability of the proposed MED antenna for the mmWave band is a significant factor. As previously stated, the AiP is preferable for minimizing transmission line losses because it allows both active and passive RF components to be incorporated into a single package with minimal transmission line lengths. Although designing the AiP is not the primary objective of this research, the manufacturability of the proposed MED antenna as an AiP integration solution should be investigated in the future. Despite RO 5880 providing the most promising overall performance, in theory, this substrate type is generally unsuitable for AiP development. This is mostly due to the fact that the RO 5880 is composed of glass microfibre reinforced PTFE composites ro5 (2022). The fundamental issue with these substrates is that they are very flexible and can be easily bent when multilayer stack-up designs such as AiP are considered, causing them to break effortlessly ro5 (2022); Chou et al. (2021). Although this sort of substrate has been extensively proposed in academic research for mmWave applications due to its exceptional performance, it is too expensive for large-scale fabrication, making it unfavorable to industry applications Chou et al. (2021). On the other hand, dielectric substrates with strong mechanical properties, such as the RO 3003, RO 4350, and Panasonic Megtron-6 are being examined. RO 3003 is a low-loss low-dielectric constant substrate and, as demonstrated here, a scanning performance close to that of RO 5880; however, it is typically a more expensive one. Apart from that, the RO 4350 and Panasonic Megtron-6 substrates, which can be manufactured at a lower cost, are considered the most favorable solutions by the majority of the industry Chou et al. (2021). Nonetheless, future research should address improving the scanning performance of the MED phased array developed on these substrates.

The Computer Simulation Technology (CST) Microwave Studio was utilized for all computations of the proposed MED phased array antenna. There were two distinct boundary conditions that were applied. Firstly, the infinite unit cell with periodic boundary conditions is utilized to simulate the unit cell MED phased array. The scanning impedance at various scanning angles for the E- and H-planes was obtained from frequency domain simulations in CST. To analyze the scanning performance of the finite 8 × 8 MED phased array antenna, the open boundary condition of the CST is utilized with the time domain simulation. Moreover, to speed up the simulation, the acceleration token in the CST was employed for finite array simulation. It is worth noting that this acceleration token is only applicable for time domain simulation.

All the simulations were completed using the radio system group server at the University of Twente. The server is installed with the NVIDIA-Tesla-P100-16 GB which is one of the essential components to activate the acceleration token.