A Broadband Ultra High Frequency (UHF) Fat-Dipole Antenna for Digital TV Applications

Featured Application

This antenna can be used in the receiving front-end Digital TV indoor applications currently under test, where the transmission and reception benefits of MIMO operation are investigated.

Abstract

A broadband and compact planar UHF antenna is presented for use in MIMO Digital TV indoor applications. The choice of the element and its physical implementation is oriented towards a low-profile, low-cost, and single-layer deployment. Since it is intended to be used in a 2 × 1 MIMO array, two antennas were constructed and tested to minimize their coupling to provide larger MIMO gains. Regarding the excitation, two different structures were evaluated to observe their impact on the final impedance profile. The option of an infinite balun provided a 62% fractional bandwidth (394 MHz to 747 MHz), covering the desired band and with good isolation levels when analyzed in the array. Results regarding MIMO parameters are provided herein, such as Envelope Correlation Coefficient and Total Active Reflection Coefficient, proving the potential of the proposed antenna array to be used in MIMO Digital TV applications in covering the UHF band.

1. Introduction

Digital TV (DTV) 3.0 guidelines in Brazil [1] propose the use of indoor multiple-input multiple-output (MIMO) antennas for reception and transmission, aiming to increase data throughput while using the same channel. To cover the entire allocated band of 230 MHz in the UHF range, centered at a frequency of 585 MHz, a broadband response is required (39% fractional bandwidth). The size consideration is significant: a classical half-wavelength dipole at the central frequency of this band measures approximately 25 cm in length. This dimension is already substantial for a single element and becomes even more pronounced when considering a two-element MIMO array. In addition to performance and size requirements, the construction of these antennas must be as simple as possible to ensure that they are low-cost enough for mass production. All these factors must be taken into account during the design process.

UHF antennas typically occupy large areas due to their resonant wavelengths. One approach to achieve operation with electrically shorter lengths utilizes the space-filling properties of fractal structures [2]. The antenna in this study is an improved version of a former found in [3], where a fat-dipole planar antenna is introduced with optimized dimensions for reduced areas, but with not much emphasis on the low-coupling requirement necessary for MIMO. The previous design, in turn, was based on a fractal approach for broad bandwidth response, combining the ideas from [4,5] varying dimensions along the dipole length and with fat strips operating as planar dipoles. Antennas intended for MIMO applications, in particular, require high electromagnetic isolation to generate uncorrelated signals. Several studies have explored methods to decrease coupling without increasing inter-element spacing, such as T-shaped slots etched on the ground plane layer of an inverted F antenna (IFA) array [6]. In terms of similar antennas close to this frequency range, a bow-tie element excited by a microstrip balun achieved a band from 0.85 GHz to 3.25 GHz; however, the implemented solution is not fully planar [7]. When the structure is allowed to be 3D, this extra degree of freedom enables a larger potential for broad bandwidths, such as using two metallic plates excited by a discone-like monopole excitator, resulting in a band between 30 MHz and 3000 MHz [8]. The use of active devices associated to broadband non-Foster networks that implement negative slopes with respect to the frequency resulted in a planar inverted F antenna (PIFA) element covering the 100 MHz to 500 MHz range [9]. Another approach [10] achieved the low profile characteristic at the VHF range by loading a half-loop antenna with ferrite bars (frequency range 30 MHz to 300 MHz). Considering situations where the antenna is embedded, a U-shaped monopole DTV element was placed inside a laptop with reduced dimensions, taking advantage of the the mutual interaction with the internal chassis structure and resulting in an operational range from 470 MHz to 780 MHz [11]. With the aim of DTV applications as well, a quasi-Yagi antenna using two planar layers was implemented (470 MHz to 806 MHz, considering the VSWR below 2 criterion), whereby its final dimensions were 24 cm by 20 cm [12]. For the particular case of embedded Digital TV antennas (DVB-H) with a band ranging from 470 MHz to 750 MHz in mobile terminals, ref. [13] describes the use of non-resonant capacitive coupling element (CCE)-type antennas, where the radiation efficiency is sacrificed in order to achieve elements that fit into small volumes. For the same DVB-H application, another solution intended to be used in the receiver had a planar folded radiating patch with a folded parasitic arm, covering the 470 MHz to 806 MHz range (VSWR < 3 or insertion loss < 6 dB) [14].

This article reports on the design process of a planar antenna for DTV applications, intended for deployment in a 2 × 1 MIMO array. Among the various DTV standards, the Advanced Television Systems Committee (ATSC 3.0) proposes MIMO as a way to enhance robustness by transmitting two independent data signals over a single RF channel, thereby enabling spatial multiplexing [15]. The additional capacity provided by MIMO facilitates DTV features that allow for the transmission of multiple channels within the same bandwidth or a return path for user interaction. Figure 1 illustrates the antenna with a generic Digital TV receiver block diagram. The relevant design constraints are presented, along with implementation and 3D electromagnetic simulation details, which are compared with measurements. Specific MIMO parameters are discussed and applied to the antenna design.

2. Antenna Design

The basic requirements for this design are as follows:

• Compact dimensions suitable for indoor applications.
• Low-cost implementation.
• Operating UHF frequency range of 470–700 MHz (current allocated DTV bandwidth).
• A 75 Ω impedance matching, which is the standard for TV systems.
• Single-layer planar structure.
• Minimal coupling in 2-element array configuration.

The antenna is printed on a low-cost FR–4 substrate, with a 1.6 mm thickness, ${ϵ}_r=$ 4.3, and loss tangent of 0.025. The basic idea starts from a fat-dipole, which has a broader bandwidth since they show a lower quality factor (Q). Given their larger surface area, their current is less concentrated in comparison to thin elements, resulting in a lower inductance per unit length. Figure 2 shows an antenna with the same length but different widths, using a 1.6 mm thick FR–4 dielectric slab. Both antennas are referenced to a 75 Ω impedance.

Table 1. S11 comparison of impedance behavior of dipole versions.

	Band [MHz]	Fractional Bandwidth [%]
Thin	451–522	14.6
Fat	480–658	31.3

contains the simulated results for both cases, and the bandwidth is defined as the range where the S11 parameter is less than or equal to −10 dB. It can be seen that the fat element generates a bandwidth response twice as large.

The design started from a fat uniform dipole, which was progressively cut in order to further improve its frequency response. The CST Microwave Studio version 2024 optimization yielded the widths and lengths that achieve low reflection losses while maintaining compact dimensions. In addition to the cuts, two parasitic patches were added to enhance design flexibility, similar to the approach reported in [16], where two inverted F-antennas achieved operation in both the LTE 700 MHz band and the 2500–2700 MHz range using parasitic elements.

The evolution of antenna design, starting from a fat uniform dipole to the final version, is depicted in Figure 3, and the results in terms of bandwidth are presented in

Table 2. S11 obtained at −10 dB for the proposed antennas.

Design	Band [MHz]	Fractional Bandwidth [%]
A	462–583	23.0
B	456–600	26.8
C	458–602	27.2
D	468–645	31.7
E	476–659	32.1
F	483–704	37.1

. The final optimized dimensions and geometry are shown in Figure 4. From the evolution results, it can be seen that the cuts progressively increase the bandwidth, at the expense of shifting upwards its minimum frequency, resulting in a larger element. Therefore, there is a design trade-off between size and bandwidth, which needs to be taken into account for keeping the element with reasonable size.

The antenna can be represented as an RLC circuit, in a first-order approximation. Figure 5 shows the passive network and the respective comparison with the results from the simulated antenna. Their impedances are depicted in both S11 and Smith Chart. Using the Laplace variable s, the input impedance of the passive network $Z_{in}s$ can be expressed as:

$$Z_{in}\left(s\right)={\displaystyle \frac{R_1{L}_{1}s}{R_1{C}_1{L}_1{s}^2+L_{1}s+R_1}}+L_{2}s+{\displaystyle \frac{1}{C_{2}s}}$$

(1)

A particular concern regarding this dipole-like antenna is the balun. Simulations were conducted using a so-called discrete port, whose terminals provide an ideal balanced excitation between its two extreme points. However, this representation is idealized and may not accurately reflect real-world implementation. To investigate the actual excitation, two solutions were employed: one with an F-type connector directly attached to the antenna, with the shield and inner conductor soldered to each of the dipole elements, and another using an infinite balun, where the coaxial shield is soldered to one of the elements and the inner conductor contacts the other dipole element. Due to the antenna’s two central slots, the infinite balun was positioned off-center, offset from the geometric midpoint. The presented simulations used the off-centered discrete port to closely match the prototype configuration.

Field and surface current plots provide valuable insights into how each antenna component affects overall performance. These visualizations reveal the spatial distribution of electromagnetic effects, complementing the global performance data obtained from insertion loss curves. Figure 6 displays the computed surface currents at the lowest, middle and highest frequencies. The off-centered feed position creates an asymmetric current pattern, with the largest intensity concentrated in the central area. This observation validates the strategic placement of slots and parasitic patches in this high-sensitivity region. At 710 MHz (upper frequency), the current distribution exhibits limited propagation towards the dipole extremities due to the shorter wavelength.

The farfield pattern for three different frequencies of a single element is shown in Figure 7. The plots exhibit characteristics consistent with those expected from a standard dipole, despite the presence of cuts and parasitic elements. The omni-directional nature of the radiation pattern across the XY plane is particularly relevant for broadcast applications, where signals can arrive from multiple directions and the indoor placement of antennas is often random, depending on each user’s specific situation. The observed asymmetry is caused by the off-axis placement of the feed, due to the presence of slots.

For MIMO applications, where both vertical and horizontal deployments are utilized for antenna arrays, it is essential to achieve good isolation between co-polarized and cross-polarized components to prevent signal absorption from orthogonal polarizations. In indoor applications, where the wireless channel rarely contains a line-of-sight (LOS) path, received signals arrive from multiple directions. This necessitates an isolation analysis, defined as the difference in dB between the gains for the desired co-pol polarization and the orthogonal cross-pol, for each point across the entire 3D space. Figure 8 presents a 2D projection of the 3D isolation patterns at two frequencies, obtained through full-wave simulations. The isolation performance degrades at the upper frequency limit, particularly along the z-axis (θ = 0° and 180°).

3. Results

The antenna element was initially tested to assess the influence of the balanced-to-unbalanced conversion. An F-type connector was directly soldered to each of the dipole arms, and the results were later compared to those obtained using an infinite balun. All measurements were conducted in a laboratory environment without shielding or absorbing walls, as an anechoic chamber was not available. Nevertheless, this setup aligns with the intended condition for indoor applications. It is important to note that the measurement results may vary depending on the antenna’s position and placement.

In addition to laboratory evaluations, virtual simulations allow for antenna verification prior to prototyping, thereby providing confidence during the design phase. To validate this specific case, two different 3D numerical methods were employed using a commercial full-wave electromagnetic field solver. Ideally, these two numerical methods and their respective meshes should yield similar results when analyzing the same structure, demonstrating the consistency of the simulation. The numerical methods used are as follows:

• CST Microwave Studio transient solver, based on the Finite Integration Technique (FIT) [17].
• CST Microwave Studio frequency domain solver, based on Finite Element Method (FEM) [18].

Figure 9 illustrates the simulation methods alongside the results obtained with the direct connection excitation. It is evident that the results from the two simulation methods vary; however, within the frequency band of interest, the FIT method yielded the best results. Both simulations used a discrete port, which represents an ideal balun with signals 180° out of phase between its terminals. The difference in simulation results can be attributed to the use of different meshes (tetrahedral for the FEM and hexahedral for the FIT), as well as variations in internal convergence criteria. In the plots, a rectangle delineates the bandwidth of interest.

Figure 10 presents a comparison between the two excitation approaches (balun and direct connection), along with the simulation performed using the transient solver of CST Microwave Studio. The balun is constructed by soldering the coaxial inner conductor and the shield to each of the dipoles. The length of the shield section that is soldered is proportional to the wavelength; in this case, approximately 4 cm was used. A realistic 3D simulation of an infinite balun is not feasible because accurately representing the coaxial braided shield requires very fine modeling of the copper mesh, taking into account the holes and gaps in the braided area [19]. Although a section of a coaxial cable was simulated in 3D for electromagnetic compatibility (EMC) purposes, it was performed with a limited length [20]. This geometrical constraint, resulting from the imperfect shielding and the overall length of the cable, leads to a very large aspect ratio with respect to the antenna dimensions. Therefore, the virtual models and the prototype are not rigorously identical due to the presence of the balun. It is evident that the ideal excitation implemented in the virtual simulation is of paramount importance. Unless a real broadband balun is utilized, additional geometric optimizations are relatively insignificant, as the sensitivity of the prototype connection to the feed cable has a much greater influence on the results compared to minor structural changes. During measurements, it was also observed that common-mode currents flowed in the vector network analyzer cable when touched during direct excitation, generating visible changes in the S11 curve being measured. This indicates a poor conversion between the unbalanced currents from the network analyzer and the balanced termination presented by the antenna. The first observed out-of-band resonance in the measurements with the balun occurred at approximately 420 MHz, and it was likely caused by the balun itself; therefore, it has a non-radiating character.

Table 3. Bandwidth comparison.

	Band [MHz]	Fractional Bandwidth [%]
Simulation	483–712	38
Direct connection	513–562	9
Infinite balun	394–747	62

summarizes the results in terms of fractional bandwidth (considering the −10 dB threshold). It is clear that the direct connection, despite its simplicity and mechanical ruggedness, imposes a significant constraint on the coupling of currents to and from the antenna dipole, resulting in a much narrower bandwidth and unstable operation, given the observed common-mode currents flowing in the cable.

The antenna element gain (in dB) at the broadside direction was simulated and measured using a 75 Ω input impedance network analyzer in a laboratory environment, as illustrated in Figure 11. Two identical antennas were positioned at a known distance to compute the individual gain from the $S_{12}$ parameter using the Friis free-space propagation formula. Three-dimensional electromagnetic simulations predicted a uniform gain across the frequency range, as the virtual model employed Perfect Matched Layers (PMLs) to absorb reflections at the simulation boundaries. In contrast, experimental measurements revealed significant gain oscillations, primarily attributed to the cluttered indoor environment that introduced numerous reflections and multipath effects. A notable gain reduction was observed from 625 MHz to the upper limit of the measured frequency band. This phenomenon can be attributed to two primary factors: (1) increased reflection and refraction effects as the wavelength becomes shorter, and (2) potential energy absorption by the implemented balun. Although the balun showed no apparent mismatches in the S11 reflection coefficient (Figure 10), it may not have been effectively radiating the entire input energy.To mitigate these issues, implementing a 1:1 75 Ω balun could potentially improve the antenna’s performance, albeit at increased cost and system complexity.

Since the antennas were intended for deployment in a MIMO operation, a 2-element array was investigated. Minimizing the mutual coupling between the elements was essential to reduce the correlation and thereby enhance diversity gain. The first test examined the influence of the relative positions of the antennas, considering both collinear and orthogonal deployments. Both elements were positioned 51 cm apart, and as shown in Figure 12, the orthogonal arrangement was favored due to a lower level of coupling, except for a small band near 600 MHz and at the lower end of the band. The distances are defined as depicted in Figure 12, and this choice is justified when compared to the conventional center-to-center length, as the primary interest here is the overall array dimension. In other words, the parameters $d_o$ and $d_c$ (representing orthogonal and collinear distances, respectively) were always set to be equal, to ensure that the MIMO antennas are of similar size. It is important to note, however, that to maintain this condition, the orthogonal antennas were electrically closer than their collinear counterparts.

Focusing solely on the orthogonal case, Figure 13 presents the coupling results for different $d_o$ distances. The minimum distance (32 cm) is observed when the antenna substrates are touching each other, and in this situation, an isolation level larger than 20 dB was measured across the entire DTV band.

The Envelope Correlation Coefficient ($ECC$) is a key parameter used to describe the isolation between different communication channels [21]. It can be expressed using either the far-field antenna patterns or the scattering parameters, as shown in Equation (2):

$$ECC={\displaystyle \frac{|s_{11}^{*}{s}_{12}+s_{21}^{*}{s}_{22}|}{(1-|s_{11}{|}^2-|s_{21}{|}^2\left)\right(1-|s_{22}{|}^2-|s_{12}{|}^2)}}$$

(2)

Diversity gain ($DG$), Equation (3), can be extracted directly from the $ECC$, as [22]

$$DG=10\sqrt{1-EC{C}^2}$$

(3)

Another important parameter that describes the concept of return loss applied to MIMO antennas is the Total Active Reflection Coefficient ($TARC$). This coefficient takes into account the mutual influence of the other array antennas on the individual antenna input. For a 2-element array, it is expressed according to Equation (4):

$$TARC=\sqrt{{\displaystyle \frac{|s_{11}+s_{12}{|}^2+{|s_{21}+s_{22}|}^2}{2}}}$$

(4)

Computed $ECC$ and $TARC$ results for the case with orthogonal antennas spaced 32 cm apart are shown in Figure 14. $ECC$ values below 0.1 are considered adequate for digital transmission [21] and, therefore, suitable for use in DTV. The $TARC$ results remain below 10 dB across the DTV frequency range, which are slightly close to the lower band limit of 470 MHz.

4. Application in a MIMO DTV Channel

Figure 15 shows the designed antenna in a 2 × 2 MIMO channel, where both the transmission and reception operate with two channels. $h_{ij}$ stands for the channel impulse response for each case and $x_i$ and $y_{i}i$ denote the input and output sequences, respectively.

The received signals can be written as

$$y_1=h_{11}{x}_1+h_{12}{x}_2+n_1$$

(5)

$$y_2=h_{21}{x}_1+h_{22}{x}_2+n_2$$

(6)

where $n_i$ represents the noise. It can be observed that there are four distinct paths available, each represented by a complex variable. MIMO DTV 3.0 applications consider different information circulating in each channel, meaning that $x_1$ and $x_2$ are not identical. Furthermore, knowledge of the channel impulse responses $h_{ij}$ enables more efficient data throughput; these channel impulse responses (CIRs) are computed using pre-defined sequences transmitted in pilot tones during transmission and are subsequently recovered at the receiver end.

The channel capacity C can be increased if different signals are transmitted by each transmitter antenna. That is the case for DTV 3.0, which employs MIMO-OFDM (Orthogonal Frequency Division Multiplexing). In ATSC 3.0, a 2 × 2 deployment, orthogonal (vertical and horizontal) antennas are utilized at both receiver and transmitter ends, allowing for the transmission of two distinct data streams using both polarizations [23]. Similarly, Advanced ISBD-T recommends a MIMO 2 × 2 configuration as a means to achieve channel capacity in scenarios where $C/N⩽0$ dB, where C and N represent carrier and noise energy, respectively [24].

5. Conclusions

A low-cost Digital TV planar antenna is proposed for use in a MIMO configuration at the receiving site. Its design is based on a fat-dipole approach, incorporating cuts, slots, and two parasitic patches inserted to optimize its performance while achieving a compact size and broad bandwidth. Two feeding methods are analyzed: direct connection and an infinite balun. The latter method demonstrated better results, highlighting the significant impact of proper connection on the final outcomes. The full-wave simulation of the antenna is discussed using two distinct numerical methods and meshes, along with limitations of the feeding method in comparison to real-world implementation. The MIMO array exhibited good performance in terms of physical dimensions, bandwidth, and isolation, making it attractive for indoor applications. MIMO parameters TARC and ECC indicate that this antenna can potentially be deployed in a 2 × 1 array format.

Table 4. Similar planar broadband antennas. ${\lambda }_o$ refers to the center frequency. In case of a MIMO array, dimensions are relative to a single element. NI = not informed; $G_{max}$/$G_{min}$ = maximum and minimum gains within the band, respectively.

Ref.	Band [MHz]	Frac. Bandwidth [%]	Dim. [${\mathit{\lambda }}_{\mathit{o}}$]	MIMO	Radiator Type	Techn.	${\mathit{G}}_{\mathit{max}}$/${\mathit{G}}_{\mathit{min}}$ [dB]
[6]	2450–2480 and 5150–5825	3.2 and 12.3	0.71 × 1.06	Yes	IFA	2-layer	0/N.I.
[9]	100–500	133	0.065 × 0.154	No	PIFA	2-layer + active circuit	20/−40
[10]	3–300	196	0.02 × 0.30	No	Monopole	2-layer, ferrite-loaded	−9/−22
[11]	470–780	49.6	0.01 × 0.02	No	U-shaped monopole	3D	1.8/0.8
[12]	470–806	52.7	0.51 × 0.42	No	Yagi	2-layer	4.6/3.5
[21]	3000–12,400	122	0.46 × 2.90	Yes	Elliptical monopole	1-layer	5/0
[25]	1750–3880	75	0.43 × 0.70	No	Monopole	2-layer	3.6/2
[26]	2000–5600	95	0.69 × 0.50	Yes	Electric dipole + magnetic slot	1-layer	5.7/2.8
[27]	2320–2950	23.9	0.31 × 0.23	Yes	Dipole + parasitic	1-layer	5.5/3
[28]	3000–6750	77	0.32 × 0.16	Yes	Slot	2-layer	4.9/4
[29]	20,000–120,000	142	4.0 × 3.43	No	4× microstrip patches	1-layer	15.11/7.88
[30]	2300–4400	62.7	0.45 × 0.45	Yes	Dipole over V-shaped ground	2-layer + vias	2.8/0.9
[31]	1470–2650	57.3	1.22 × 1.22	No	Crossed dipoles	2-layer	8.7/7.8
this study	393–747	62	0.48 × 0.13	Yes	Fat dipole	1-layer	3.9/−6.9

presents a comparison with other planar antennas, demonstrating its good performance. Additionally, its construction requires only a single layer, making it easier and more cost-effective to manufacture.