With the advancement of commercial and consumer wireless communication devices, the electromagnetic spectrum at ultrahigh frequency keeps getting congested and restricted [1]. This draws attention to high frequency bands from 2 GHz to 10 GHz and at the same time demands ultra-wideband (UWB) characteristics. UWB technology applies to numerous areas like satellite communication, radars, bio detection, pulse communication as well as high pulsed sources [2–4]. As frequency increases the antenna size reduces but to incorporate low frequencies in the UWB antenna, simultaneously maintaining the modern compact size, the integration requires miniaturization of the design [4]. The Vivaldi antenna adapts to all these challenges. It is linearly polarized, has an ultra-wide bandwidth (BW) with a stable radiation pattern and gives incremental gain to a broader extent over the wide frequency range [4].

It behaves as a travelling wave antenna at higher frequencies while acts as a resonant antenna at lower frequencies [1, 3]. However, it comes at the expense of a complex fan shaped feeding structure. This structure is a transition from microstrip to a slot line which works as a feeding balun, this limits its BW especially the lower cutoff frequency and gives distorted gain, radiation pattern, and higher sidelobes as well grating lobes at higher frequencies [3, 4].

The Vivaldi antenna, first introduced by Gibson in 1979 as the Aerial Vivaldi, is commonly referred to as the tapered slot antenna (TSA) or coplanar Vivaldi antenna [5]. This coplanar end-fire radiator provides wide impedance bandwidth but suffers from limited gain and radiation pattern instability at higher operating frequencies. To overcome these limitations, Gazit proposed the antipodal Vivaldi antenna (AVA) in 1988 [6]. In this configuration, the exponential flares are printed on opposite sides of the substrate in a symmetric fashion, with one arm functioning as the radiator while the other provides the ground reference. Compared to the coplanar TSA, the AVA exhibits improved impedance matching over a wide frequency range, stable end-fire radiation, and higher gain with enhanced directivity [6].

Despite these advantages, further research has focused on extending the intrinsic performance of the AVA. Reported approaches include the use of anisotropic metamaterials and low-permittivity/loss-tangent substrates to improve realized gain, bandwidth, and return loss [7, 8]. Other studies have investigated modifications to the exponential taper, including leaf-shaped, binomial, and elliptical profiles, or the introduction of multiple tapered slots, thereby enabling multiband operation and gain enhancement [9–12].

Another notable direction has been the integration of parasitic patches which act as director structures beyond the radiating aperture of the Vivaldi antenna. These director-loaded designs, inspired by Yagi–Uda configurations, employ dipole-like or strip directors aligned in the end-fire direction to reinforce radiation and significantly enhance gain, often achieving 10–12 dBi. However, such external directors are highly sensitive to geometric spacing, leading to increased design complexity and fabrication tolerance issues [3, 13]. More importantly, the additional parasitic elements narrow the impedance bandwidth, elevate sidelobe levels, and enlarge the overall antenna profile, thereby compromising one of the fundamental advantages of the Vivaldi class: UWB performance with compact physical size [13].

In contrast to external directors, recent studies have demonstrated that parasitic conductive patches embedded within the aperture of the AVA can effectively enhance gain and directivity without substantially degrading bandwidth. These patches, typically fabricated from metals such as copper, couple electromagnetically to the tapered flares and extend the effective aperture through localized current induction. Various geometries, including circular, diamond, trapezoidal, elliptical, and staircase shapes, have been proposed [13–19]. While each geometry exhibits distinct frequency-dependent responses, the overall mechanism is consistent: electromagnetic (EM) coupling between the patches and the main radiator increases directivity in the end-fire axis. A typical trade-off is that certain patch geometries, such as staircase structures, enhance gain substantially but reduce bandwidth and induce multiband behavior [18], its mitigation is discussed in Section V.

Building upon this body of research, the present work introduces a modified AVA with optimized director-like parasitic patches integrated inside the aperture, thereby combining the gain-enhancing properties of Yagi-inspired designs with the bandwidth-preserving characteristics of embedded patch loading. Unlike conventional director-loaded Vivaldi antennas that extend the aperture and restrict usable frequency range, the proposed architecture maintains wideband impedance matching while achieving high gain and directivity enhancement in the end-fire direction. This yields a compact, high-performance solution suitable for applications demanding ultra-wideband radiation stability and elevated gain, including ground-penetrating radar (GPR), microwave imaging, and ultra-wideband wireless communication systems.

Fig. 1 shows the radiator flare and the ground flare of the fabricated prototype of the AVA with bar patches. 0.3 mm thick copper is used as the conductor for the AVA flares as well as for the bar patches. The substrate is FR4 with relative permittivity (ɛ_r ) of 4.5, loss tangent (tanα) of 0.025 and thickness is 1.5 mm. The schematic of the AVA is shown in Fig. 2. This configuration of AVA is also called as “Exponentially tapered slot antenna” (ETSA) [11].

The outer flare height (H_fo ) should be at least λh4 and the inner flare length (L_fi ) should be at least λh3 where λ_h is the minimum frequency wavelength. L_fo is the length of outer flare and H_fi represent the inner flare height. The equation for the flare design given in [14] is

where x is the horizontal component of the flare and the constants C₁ and C₂ are given as

Here y₂, y₁, x₂, and x₁ are the start and end points of the exponential profile and k is the expansion factor, the rate at which the taper profile grows. Whereas f_i is the feedline length and f_w is the feedline width. Fig. 3 shows the parameters for the bar patches. After numerous simulations and design iterations it was found that the length of the bar patches can be approximated by Eq. (4). The objective of these simulations was to find the optimum length of the bar patches and these simulations are discussed in detail in Section VI.

Here L is the initial length of the first patch element and x ranges from 1–4. Lp_x is the length of the bar patch, and the width of the patch is represented by Wp_x which is constant. This antenna uses parallel plate transmission line but since its impedance is greater than 50 Ω, it is transitioned to a microstrip feed with 50 Ω impedance.

The AVA behaves as a travelling wave antenna at higher frequencies as mentioned in [3], so the radiation comes from coupling of electric field (E-field) between the two flares. When the patches are present in the vicinity of the two flares, this high frequency E-field couples to the patches and produces strong radiation in the end fire direction. The dimension of these patches relates directly to the frequency. The higher the frequency, the smaller the length of the patch. The bar patches and the antenna flares are conductors with the FR4 as a dielectric between them. This is also referred to as capacitive EM coupling.

Fig. 4(a) shows E-field distribution on the surface of AVA with units of volts per meter (V/m) at 8 GHz without bar patches. Weak EM coupling is observed along the antenna axis in the end fire direction almost equal to 300 V/m. The E-field in this case is relatively widely distributed and results in a wider half-power beamwidth (HPBW) and so eventually less directivity is obtained. Fig. 4(b) shows E-field distribution at 8 GHz with bar patches and here on the contrary, strong EM coupling is observed with the bar patches along the antenna axis in the range of 800 V/m. Here most of the bar patch length is being utilized and contributes to the EM coupling and so the E-field is focused and will result in a narrower HPBW, hence more directivity is obtained. Fig. 5 represents E-field comparison in the end fire direction. The results also prove a direct relation between the gain and E-field as the highest E-field value corresponds to high gain. This plot is obtained using far-field approximation at 1 cm over the entire BW. From this, it is proved that the bar patches significantly increase the E-field strength by EM coupling.

Fig. 6 visualizes where and how intensely current flows on the device’s surface at 8 GHz. This simulation result substantiates the E-field distribution results discussed earlier in Fig. 4. So, strong coupling leads to higher current distribution in the bar patches which leads to a higher gain. This proves the application of EM-coupling to enhance antenna radiation.

The simulation was carried out in CST Microwave Studio. The time domain solver was used with the hexahedral mesh and 20 cells per wavelength near to model. The fraction of maximum cell near to model was 100 and the cells far from model were only 2. The boundary setup was open and added space perfectly matched layer (PML) with frequency from 2 GHz to 10 GHz. To produce high quality results, adaptive mesh refinement was used with a maximum of 4 passes and as a result the total mesh cell count was progressively over 80 million.

The multiband problem arises when the patches are too close to the inner flares. Refer to Fig. 2 when D is 24 mm and Fig. 7 when D becomes 1 mm. To resolve this multiband issue and make the antenna gain higher without compromising the BW, the position of the bar patches relative to the inner flare makes the most difference. The term multiband refers to an antenna that is capable of operating effectively on two or more distinct frequency bands. Each of these bands corresponds to different radio frequency ranges, allowing the antenna to transmit and receive signals at multiple frequencies rather than being limited to a single frequency range. In UWB application the multiband feature is considered undesirable as it reduces the absolute BW of the antenna. The same Multiband problem is present in [18] where the staircase patches are used to increase gain of the AVA.

Fig. 8 shows us that when the bar patches are in very close proximity to the inner flares, the multiband problem arises, and the absolute BW is reduced. This problem was reported in [18] when staircase patches were introduced to improve the radiation characteristics of the AVA. So, by increasing the value of D the multiband issue can be easily resolved, but this is not the only problem, because the gain of the AVA is also dependent on the parameter D. Fig. 9 represents how the gain in the end fire direction that is theta 0° and phi 0° is affected at higher frequencies by different values of D. By keeping all the parameters same as shown in Table 1 and manipulating only D the results in Fig. 6 are obtained. At D = 24 mm the gain curve shows the highest value at higher frequencies.

The bar patch lengths contribute differently to the gain of AVA at variant frequencies. To find the best case for fabrication of the design, a parametric study of different bar lengths was done with D = 24 mm. Table 2 gives the dimensions of the bar patches. These values are obtained from Eq. (4) by substituting the values of L. Each value of L derives the four different lengths of the bar patches. Fig. 10 shows the simulation results of the effect of different bar patch lengths on the gain of AVA in the end fire direction. From these curves, it is deduced that the curves align and show no difference at the starting and mid frequencies but after 8 GHz the gain plummets to a meager value of 4 dB for the two larger bar lengths, that is for L = 10 mm and 9 mm. The other two curves, for L = 8 mm and for L = 7 mm, give similar results with a difference of ±0.2 dB. Keeping manufacturing tolerances in mind L = 8 mm is the optimum option for the bar patch lengths.

Fig. 11 shows the effect of bar patch lengths on the reflection coefficient of AVA. For all the bar lengths the reflection coefficient remains below −10 dB from 2–10 GHz, and it can be deduced that it has no effect on the absolute BW of the AVA.

Fig. 12 shows the fabricated antenna under test (AUT), in over-the-air (OTA) antenna chamber. A 50-Ω PCB mount SMA connector is used to feed the antenna. The AUT is mounted on the turning arm in an azimuthal plane.

Fig. 13 illustrates a comparison between simulated and measured reflection coefficient. The reflection coefficient value is below −10 dB from 2 GHz to 10 GHz and beyond but since the antenna is designed and optimized for up to 10 GHz, the results up to only 10 GHz are presented. Both the curves of S₁₁ are of AVA with bar patches. The results confirm that the −10 dB BW of the AVA was not compromised with the introduction of bar patches. The absolute BW of AVA is 8 GHz with the fractional bandwidth (FBW) of over 130%. Fig. 14 shows the realized gain vs frequency plot of simulation of both cases, with and without the bar patches as well as the measured gain of AVA with bar patches. It is evident from the measurement that the bar patches significantly increase the gain. The measured gain is from 3.8 dB to a maximum of 7.8 dB, having gain greater than 5 dB through frequency range of 3–10 GHz. The small deviation between the slopes of measured curve and simulated curve is possibly due to the alignment error between the AUT and the reference antenna in the OTA antenna chamber.

Table 3 shows the comparison between different antenna’s gain measurements in the North Atlantic Treaty Organization (NATO) E-band that is from 2–3 GHz and the 2–10 GHz BW of their respective frequency range. The antenna size is given in terms of the wavelength in the dielectric medium to account for the effect of the permittivity of substrate. The wavelength in the medium is calculated as

where c is the speed of light in meter per second, f is the lowest frequency of the absolute bandwidth that is 2 GHz. From the results in Table. 3, it is evident that the proposed design with bar patches has the highest gain in the NATO E-band with the most compact size in terms of λ_m. This is an improvement to the best of author’s knowledge over the AVA designs present in the published literature as the other antenna designs considerably underperform in NATO E-band with the gain hardly close to 3 dB meanwhile the proposed AVA has E-band gain ranging from 3.8–5 dB. Some AVA designs have a max gain 2–3 dB higher than the proposed AVA, for instance [3] and [20] but their size in terms of wavelength is also greater, and the gain is directly related to the size of antenna [3].

Hence for a compact size, the proposed AVA has the highest gain in the NATO E-band and maintains a high gain in the range of 5–7.8 dB through the remaining BW. Using low permittivity substrates can further improve the characteristics of the AVA as it is directly related to the frequency losses [21]. Moreover, FR4’s optimum frequency range is from a few megahertz to 4 GHz and at higher frequencies the FR4 substrate incurs high losses [22]. Besides all this FR4 remains an adequate choice for research purposes with its ease of availability and cost effectiveness.

Fig. 15 shows the polar radiation patterns of E-plane at nine different frequencies of the measured AVA superimposed on the simulated AVA pattern. The measured pattern is represented with solid blue line while the simulated pattern is represented with dashed red color. The plot angles are represented here on a 360° scale with an angle theta ranging from 0° to 359°. The inner circles on the plot are on a scale of dB ranging from -30 dB to 10 dB. From these polar plots, it can be easily acknowledged that the measured patterns align perfectly with the simulated patterns and only slight deviations are observed at higher frequencies. These deviations are due to the fabrication tolerances, as higher frequencies become more prone to even the minimal manufacturing tolerance values.

Fig. 16 shows the total efficiency of the proposed AVA, and it ranges from 55% to a maximum of about 87%.

The novel bar-shaped multiple patches are introduced in the vicinity of the AVA flares without compromising on the UWB characteristics of the AVA. This approach significantly improves the gain at the most often ignored NATO radio frequencies E-band without doubling the antenna size and maintains a high gain greater than 5 dB through 85% of the BW. It also increases directivity and stabilizes the radiation pattern at high frequencies with main lobe or boresight shifting only a couple of degrees from the end-fire axis of the antenna.