**A Wide Frequency Scanning Printed Bruce Array Antenna with Bowtie and Semi-Circular Elements** †

**Zeeshan Ahmed 1,2,\* , Patrick McEvoy <sup>1</sup> and Max J. Ammann 1,2**


Received: 26 October 2020; Accepted: 25 November 2020; Published: 27 November 2020 -

**Abstract:** A printed edge-fed counterpart of the wire Bruce array antenna, for frequency scanning applications, is presented in this paper. The unit-cell of the proposed antenna consists of bowtie and semi-circular elements to achieve wide bandwidth from below 22 GHz to above 38 GHz with open-stopband suppression. The open-stopband suppression enables a wide seamless scanning range from backward, through broadside, to forward endfire. A sidelobe threshold level of −10 dB is maintained to evaluate efficient scanning performance of the antenna. The antenna peak realized gain is 15.30 dBi, and, due to its compact size, has the ability to scan from −64◦ to 76◦ .

**Keywords:** meander line antenna; periodic structure; millimeter-wave antenna; frequency scanning antenna; leaky-wave antenna

#### **1. Introduction**

In 1931, Edmond Bruce patented the idea of the Bruce array antenna (BAA) in which a long wire antenna was bent in equal and periodic meandered intervals. The antenna was designed for amateur radio applications in which bi-directional broadside radiation and high gain are required. Figure 1 shows a typical BAA fed from the center of the structure using a twin-line feed mechanism. In the figure, the lengths and directions of the arrows are representations of the magnitudes and flow of the current, respectively. The horizontal and vertical segments of the meander line were both kept equal in size, i.e., approximately λ/4 for ham radio applications, except for the last two inward bent segments, which are half the length of the other segments. The currents in the horizontal segments, represented by light grey colored arrows in Figure 1, flow in opposite directions so as to add together destructively, thus cancelling out radiation in ideal circumstances. These horizontal elements are, therefore, considered interconnecting segments. The currents in the vertical segments flow in the same direction, adding constructively in phase to give broadside radiation, which is why these segments are termed radiating elements. The half-length segments, which are bent inward on either ends of the structure, have little to no magnitude of current; therefore, they maintain reasonably low cross-polarized radiation [1]. As the number of radiating elements are added to the structure, the half power beamwidth (HPBW) becomes sharp with the increase in peak realized gain and the radiation pattern in the broadside becomes so compressed and narrow that it can be classified as a highly directive fan-beam radiation pattern. The BAA offers reduced complexity, substantially greater bandwidth than other wire antennas (such as the bobtail curtain and half-square antennas), and, for a relatively low

height requirement, it can achieve the maximum possible gain in a given area [2]. Suited to a particular installation, the wire BAA can also be fed at points other than the center, the lengths can be varied to tune the resonant frequency, and while it usually does not require a ground system, an extensive ground system can be deployed under the BAA to mitigate the losses if there is adequate space [2].

**Figure 1.** A typical twin-line fed 8-element wire Bruce array antenna (BAA).

The BAA has been around for over a century, but, in spite of its simplicity, modern researchers have overlooked its development and utilization in modern day antenna applications. There are only a handful of ideas proposed to make use of the structure, some of which include Nakano et al.'s concatenation of the Bruce and Franklin antennas' performance at 12.50 GHz [3], Chen's twin-line fed slot-type BAA planar equivalent [4], a tri-band mm-wave printed counterpart of BAA [5], and an edge-fed printed BAA [6].

In 1940,W.W. Hansen patented the first waveguide-basedleaky-wave antenna (LWA) [7]. Several other researchers later elaborated the concept in their research [8–11], but it was A. A. Oliner who streamlined the working mechanism in 1984 [12]. In IEEE standard 145-2013, an LWA is defined as "An antenna that couples power in small increments per unit length, either continuously or discretely, from a travelling-wave structure to free space" [13]. LWAs are generally divided into two categories, namely uniform and periodic LWAs [14], the latter of which are widely used in mm-wave frequency regions as well as other scanning applications because of their ability to scan a wider area than the uniform LWAs [15]. Planar periodic LWAs are low-profile, relatively easier to fabricate, and can scan in the backward or the forward endfire direction with a fan-beam radiation pattern with frequency tuning. Several types of LWAs based on a range of technologies have been proposed in the scientific literature, including periodically meandered rampart array [16], sharpening the bends [17], squarely modulated reactance surface (SquMRS) [18], composite right/left-handed structures (CRLH) [19,20], slot or coplanar lines [21], substrate integrated waveguide (SIW) structures [22–27], Goubau line structures [24], spoof plasmon transmission line (SSP-TL) structures [25], and periodically loaded microstrip structures [26].

In the case of periodic LWAs, a steep gain-drop is usually observed around the broadside when scanning from the backward to forward endfire, because of which the antenna suffers from pattern degradation. This is because of the presence of the so-called open-stopband (OSB) at which the LWA, which usually supports a traveling-wave, exhibits standing-wave characteristics with equal excitation of the unit-cells. At OSB frequency, the incident power from the unit-cell that is supposed to radiate outwards instead reflects into the source due to the coupling of a contra-directional pair of space harmonics (Floquet modes) [28]. There are numerous periodic LWAs that have either overcome or suppressed this problem. Balanced transmission lines are used in Metamaterial LWAs to enable seamless scanning through the broadside [29,30]. Other than that, SIW structures use shorting vias [31], unequal width in transversal elements of meander lines [16], and non-identical elements in their unit-cells [32]. A lattice-network based TL model [33] has also been reported to suppress the OSB.

This paper presents a modification of planar, edge-fed, periodic array using meandering concept of wire BAA geometry and the suppression of the OSB around the broadside by replacing horizontal and vertical segments with semi-circular and novel bowtie elements, respectively. The unit-cell is the repetitive part of the structure designed at the broadside frequency. The optimizations and simulations were performed using CST Microwave Studio, in which the dielectric and metallization losses were considered. Finally, the prototype antenna was fabricated and measured responses were compared against the simulated results, from which a satisfactory agreement was attained.

#### **2. Unit-Cell and Antenna Geometry**

The configuration of the unit-cell of modified printed BAA is shown in Figure 2. The vertical and horizontal segments of the meander-line BAA antenna, shown in Figure 1, were replaced with bowtie and semi-circular elements, respectively. Either ends of the bowtie had the same width as the semi-circular segment, i.e., *wc*. Compared to the BAA, the meandered segments' lengths are approximately λ/4, but in the mm-wave region, this corresponds to a very small size which gives rise to coupling between the vertical segments; therefore, these lengths are varied. However, the length and diameter of the unit-cell, *l<sup>v</sup>* and *lc,* respectively, are kept the same, as shown in Table 1. The periodic unit-cell's dominant mode does not radiate on its own because of its slow-wave characteristics; the free space wavenumber, *k0*, is less than the phase constant (β > *k*0). Floquet's theorem states that as the unit-cells are combined in series, the periodicity introduces an infinite number of Floquet modes in a leaky-wave structure. These Floquet modes are represented by phase constant β*n* [14].

$$
\beta\_n p\_{\rm unit} = \beta\_0 p\_{\rm unit} + 2\pi n; \; n = 0, \pm 1, \pm 2, \; \pm 3 \tag{1}
$$

where *punit* is the period of the unit-cell defined by 4 × (*l<sup>c</sup>* − *wc*/2), *n* is the *n*th number space harmonic, and β*<sup>0</sup>* is the phased constant of the dominant mode of the now modulated and uniform waveguide. From Equation (1) the β*<sup>0</sup>* is slow-wave, but the structure is designed in a way that the other modes are fast. In order to scan a single beam in a directive manner, the first space harmonic, i.e., *n* = −1, is substituted into Equation (1) and is written as

$$
\beta\_{-1} p\_{\text{unit}} = \beta\_0 p\_{\text{unit}} - 2\pi \tag{2}
$$

**Figure 2.** Geometry of the modified BAA unit-cell with vertical bowtie and horizontal semi-circular segments printed on a 0.254 mm thick grounded Arlon DiClad 880.

**Table 1.** Table of parameters of unit-cell.


The scanning direction of the periodic LWA can be expressed using [14]

$$
\sin \theta\_m = \frac{\beta\_{-1}}{k\_0} \tag{3}
$$

where θ*<sup>m</sup>* is the maximum beam angle deviation from the broadside and *k<sup>0</sup>* is the free space wave number. Subsequently, the beamwidth is given by

$$
\Delta\theta \approx \frac{1}{\left(\frac{L}{\lambda\_0}\right)\cos\theta\_m} \tag{4}
$$

where *L* is the overall length of the leaky-wave antenna structure.

The geometry of the periodic modified BAA antenna, with vertical bowtie and horizontal semi-circular segments, is presented in Figure 3a. As is the case with linear arrays, the addition of unit-cells in series increases the gain and decreases the beamwidth along the length of the antenna, but a large number of unit-cells prevents the increase in gain due to the lower power delivered to the last unit-cells. Thirteen unit-cell elements, presented in Figure 2, were connected in series, and the configuration presented in Table 1 was used for simulation and prototyping of the structure. The periodic modulation of the geometry assisted with radiation along the length of the antenna. The structure was fed using transmission line of length 6.18 mm and width 0.76 mm; the last unit-cell element was terminated using a similar transmission line and another 50 Ω port that acted like a resistor to avoid reflections. Arlon DiClad 880 substrate, with a thickness of 0.254 mm, ε<sup>r</sup> = 2.2 and tan δ = 0.0009, was used to fabricate the prototype presented in Figure 3b; the measurements were performed using a Rhode and Schwarz Vector Network Analyzer (ZVA40). The overall dimensions of the antenna were 83.60 <sup>×</sup> 18.0 <sup>×</sup> 0.254 mm<sup>3</sup> .

**Figure 3.** Top view of the proposed periodic 13 unit-cell antenna fabricated on 0.254 mm thick grounded Arlon DiClad 880 substrate. (**a**) Geometry. (**b**) Prototype.

#### **3. Parametric Analysis**

Figure 4 shows the effects on |S11| for the structure with 13 unit-cell elements presented in Figure 3 by simultaneously varying *l<sup>v</sup>* and *l<sup>c</sup>* with *w<sup>v</sup>* = 0.76 mm, without the vertical bowtie element. The multiplying factor of the wire BAAs, λ/4, was increased to avoid unwanted resonances in the

mm-wave region, due to close separation distance at λ/4 between vertical elements; it varied from 1.69 × λ/4 to 1.89 × λ/4 (3.30 mm to 3.70 mm). With this configuration, mismatching was observed around the broadside frequencies, for which |S11| > −10 dB indicated the presence of OSB. Additionally, an increase in *l<sup>v</sup>* and *l<sup>c</sup>* by 0.1 mm tuned down the OSB mismatched region by approximately 1 GHz. The |S11| above and below the OSB, between 20 GHz and 40 GHz, remained less than −10 dB.

**Figure 4.** Effect of varying length, *lv*, and diameter, *lc*, simultaneously for a 13 unit-cell element periodic structure with *w<sup>v</sup>* = 0.76 mm unit-cell segments on |S11|.

The effect of parameters associated with the vertical bowtie segments, *w<sup>v</sup>* and *l<sup>t</sup>* , of the 13 unit-cell structure, on |S11|, is shown in Figure 5. The other parameters, presented in Table 1, remain unchanged. A noticeable improvement in impedance matching, around the OSB region, was observed as the width, *wv*, was increased from 0.50 mm to 0.60 mm, with *l<sup>t</sup>* fixed at 1.68 mm, in which |S11| improved from −7.24 dB to −3.54 dB at 28.0 GHz. In the other case, where *w<sup>v</sup>* was kept constant at 0.65 mm, the variation in *lt* from 1.45 mm to 1.65 mm improved the impedance matching significantly without any upward or downward tuning of frequency.

**Figure 5.** Effect of varying vertical bowtie segment parameters, *w<sup>v</sup>* and *l<sup>t</sup>* , of 13 unit-cell element periodic structure independently on |S11| while the other parameters remain the same as Table 1.

Figure 6 shows the effect of independently varying *l<sup>v</sup>* and *l<sup>c</sup>* with the fixed *w<sup>v</sup>* = 0.65 mm bowtie vertical unit-cell segment on |S11|. The frequency was tuned down by approximately 1.15 GHz and the impedance matching deteriorated when *lv* was increased by 0.20 mm, between 3.40 mm and 3.60 mm, and *l<sup>c</sup>* was kept constant at 3.50 mm. A downward shift in frequency of 0.70 GHz was observed when *l<sup>v</sup>* was fixed at 3.50 mm and *l<sup>c</sup>* was increased from 3.40 mm to 3.60 mm, with a minute effect on impedance matching.

**Figure 6.** Effect of varying length, *lv*, and diameter, *lc*, independently of 13 unit-cell element periodic structure with fixed *w<sup>v</sup>* = 0.65 mm bowtie segments on |S11|.

Figure 7 shows the realized gain of the 13 unit-cell antenna structure without the vertical bowtie element at between 22 GHz and 38 GHz with *l<sup>v</sup>* = *l<sup>c</sup>* = 3.50 mm and *w<sup>v</sup>* = 0.76 mm. The gain rose gradually between 23.0 GHz and 26.0 GHz and, while still under 12.0 dBi at 26.0 GHz, a sharp gain drop was observed around 28.0 GHz. The realized gain around 28.0 GHz was considerably less than the realized gain in the rest of the forward endfire region. This is consistent with the OSB region, identified in Figure 4, where |S11| > −10 dB.

**Figure 7.** Realized gain of 1-D periodic BAA-LWA with *l<sup>v</sup>* = *l<sup>c</sup>* = 3.50 mm and *w<sup>v</sup>* = 0.76 mm unit-cell showing gain degradation around 28.0 GHz.

#### **4. Results and Discussion**

Figure 8 shows the simulated and measured response of the |S11| of the 13-element structure with the bowtie vertical segment unit-cells that were shown in Figure 2 and the parameters presented in Table 1. The mismatched OSB frequency range, for when *l<sup>v</sup>* and *l<sup>c</sup>* is 3.50 mm (~1.79 × λ/4) and *w<sup>v</sup>* = 0.76 mm as presented in Figure 4, improved without any frequency tuning, resulting in the mitigation of the OSB. The |S11|, at 28.0 GHz, improved to −12.01 dB with the vertical bowtie element, compared to −7.73 dB without the bowtie element. With the bowtie element, the |S11|was ≤ −10 dB from below 22.0 GHz and above 40.0 GHz with a fractional bandwidth of more than 67%.

**Figure 8.** Simulated and measured |S11| of 13 unit-cell element periodic structure with bowtie vertical unit-cell segments.

Figure 9 shows the realized gain plot of the antenna structure with the modified bowtie unit-cell shown in Figure 2. The sharp decline in realized gain around 28.0 GHz region, shown in Figure 7, considerably improved with this arrangement using the parameters presented in Table 1. The gain profile gradually increased around 23.0 GHz, and onwards, with peak realized gain of 15.30 dBi at 35.0 GHz.

**Figure 9.** Simulated and measured realized gain comparison of a 13 unit-cell element periodic structure with *w<sup>v</sup>* = 0.76 mm and a modified bowtie vertical unit-cell segments.

The 3D radiation patterns for the simulation of the 13-element structure, with bowtie and semi-circular unit-cell, are presented for backward endfire, broadside and forward endfire regions in Figure 10a–c, respectively. The patterns, at 24.0 GHz, 28.0 GHz and 35.0 GHz, showed a fan-beam scanning with an increase in frequency and radiation angles of −42◦ , 0◦ and 56◦ , respectively. The scanning range of the proposed antenna is presented in Figure 11. Figure 11a shows the scanning range from backward endfire approaching towards the broadside. The antenna scanned from −64◦ at 22.87 GHz. Figure 11b shows the scanning range from the broadside to the forward endfire. The antenna scanned seamlessly through the broadside, due to the mitigation of OSB, until 76◦ , i.e., 37.0 GHz.

**Figure 10.** Three-dimensional radiation pattern visualizing scanning at (**a**) 24.0 GHz, (**b**) 28.0 GHz, and (**c**) 35.0 GHz.

Figure 12 shows the main beam direction and sidelobe level (SLL), in the yz-plane, of the proposed antenna. From Figure 8, it can be seen that the antenna has wide bandwidth below 22.87 GHz and above 37.0 GHz, but these frequencies are not considered as part of the scanning range in Figure 11 because an SLL threshold of −10 dB is maintained to efficiently define the scanning region which. As the mainlobe of the scanning range approaches forward endfire after 76◦ , the rise in the antenna's backlobe and the increase in SLL makes it unsuitable to scan in a single direction as efficient as the rest of the considered bandwidth.

**Figure 11.** Scanning range of the proposed 1-D periodic modified BAA-LWA with bowtie and semi-circular unit-cell (**a**) backward quadrant and **(b**) forward quadrant.

**Figure 12.** Sidelobe level and main beam direction in the yz-plane of the proposed LWA.

Figure 13 shows the radiation efficiency of the proposed antenna array. The antenna had more than 60% radiation efficiency throughout the scanning range, and, between 25.0 GHz and 37.0 GHz, the efficiency was more than 80%. The HPBW in both xz and yz-planes, across the entire scanning range, is also shown in the figure. As the beam approaches broadside from backward endfire, the HPBW in xz-plane increased and stabilized before dropping again as it approached forward endfire shown in Figure 10c. From the yz-plane, it can be seen that the antenna had a narrow radiation beam throughout the scanning range which can be classified as fan-beam radiation pattern.

**Figure 13.** Radiation efficiency and half power beamwidth (HPBW) plots of the proposed LWA.

The performance of the proposed antenna's scanning characteristics was compared with other antennas in the scientific literature and is presented in Table 2. The proposed antenna had wider bandwidth and scanning range along with the peak realized gain than [34–37] and in [35], only the SLL at the broadside is mentioned. There is no mention of SLL threshold for efficient scanning in any articles except [6,27,38,39]. The presented antenna had better scanning range than the first continuous scanning range of [6]. In [39], although the realized gain and bandwidth are better, it has overall dimensions of <sup>133</sup> <sup>×</sup> <sup>93</sup> <sup>×</sup> 21 mm<sup>3</sup> , with a narrower scanning range; it is also difficult to fabricate geometry compared to the presented structure. If not for the SLL ≤ −10 dB threshold maintained throughout this work, the proposed antenna may have had a wider scanning range due to the available bandwidth below 22.87 GHz in the backward endfire and above 37.0 GHz in the forward endfire. The proposed antenna can thus be used for 28 GHz 5G and Ka-Band millimeter-wave imaging applications [40].


**Table 2.** Table of performance comparison between proposed LWA and LWAs in literature.

CRE—complimentary radiation elements; TL—transmission line; HW-MLWA—half width microstrip LWA; WG—waveguide; CTS—continuous transverse stub; N/A—not available; \* OSB from 5.30 GHz to 6.20 GHz.

#### **5. Conclusions**

A wide backward to forward endfire scanning leaky-wave antenna is proposed in this paper, as well as a discussion of the results. The initial concept to design the unit-cell of the antenna is taken from meandered wire Bruce array antenna and transformed to printed geometry. The horizontal and vertical segments of the meandered unit-cell were replaced with semi-circular and bowtie segments, respectively, of which the latter assists in the mitigation of the open-stopband at broadside. The length and diameter of both vertical and horizontal segments, respectively, are kept equal at 3.50 mm. The proposed antenna has a wide operational bandwidth from below 22.0 GHz to above 38.0 GHz; however, an SLL threshold of −10 dB was enforced to define an efficient scanning range between 22.87 GHz and 37.0 GHz. The 13 unit-cell periodic antenna has a compact size, offers a scanning range between −64◦ to 76◦ , and has peak gain of 15.30 dBi.

**Author Contributions:** Conceptualization, Z.A.; methodology, Z.A.; software, Z.A.; validation, Z.A.; formal analysis, Z.A.; investigation, Z.A.; resources, Z.A., P.M., and M.J.A.; data curation, Z.A.; writing—original draft preparation, Z.A.; writing—review and editing, Z.A., P.M., and M.J.A.; visualization, Z.A.; supervision, P.M. and M.J.A.; project administration, Z.A. and M.J.A.; funding acquisition, M.J.A. All authors have read and agreed to the published version of the manuscript.

**Funding:** This publication has emanated from research conducted with the financial support of Science Foundation Ireland (SFI) and is co-funded under the European Regional Development Fund under Grant Number 13/RC/2077.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Compact N-Band Tree-Shaped Multiplexer-Based Antenna Structures for 5G**/**IoT Mobile Devices**

**Amélia Ramos 1,2,\*, Tiago Varum 1,2 and João N. Matos 1,2**


Received: 9 October 2020; Accepted: 7 November 2020; Published: 8 November 2020

**Abstract:** This paper presents a simple, compact and low-cost design method that allows one to obtain low-profile multi-band antennas for the overcrowded future generation networks, which are widely versatile and very heterogeneous in the K/Ka bands. The proposed antennas comprise *n* radiating monopoles, one for each of the desired operating frequencies, along with a frequency selective feeding network fed at a single point. This concept enables a single antenna to be shared with different radio-frequency (RF) frontends, potentially saving space. Typically, with *n-band* structures the biggest challenge is to make them highly efficient and here this is assured by multiplexing the frequency, and thus isolating each of the monopoles, allowing the design of scalable structures which fit the 5G applications. Based on the vision proposed here, a dual-band and a tri-band structures were built and characterized by their main parameters. Both prototypes achieved peak efficiencies around 80%, with adequate bandwidths and gains, as well as great compactness.

**Keywords:** frequency multiplexed; IoT; millimeter-waves; multi-band; n-band antenna; antenna as a sensor

#### **1. Introduction**

It is clear that the global tendency of exchanged data traffic is growing exponentially and, with it, the number of interconnected devices, pushing existing systems to their limits. In addition to the increasing traffic on each network, the variety of the services supported is increasing too, which naturally gives rise to great concerns about energy consumption. In this sense, 5G wireless systems must fulfil three main requirements: (i) have high throughput; (ii) serve many users simultaneously; and (iii) have less energy consumption [1], the latter being probably the biggest driving force of 5G. Associated with this, the high user mobility will force new antenna designs and new concepts to be implemented [2].

In addition, migration to mmWaves becomes mandatory, because despite implying higher propagation issues, it is probably the most effective way to achieve the necessary bandwidth [3]. In these frequencies, new challenges arise, given the reduced dimensions, however, there is an inherent opportunity to produce compact solutions and an increasing demand for robust multipurpose dual-band or multi-band antenna systems for 5G applications. Today, mobile devices comprise a wide range of applications and features, most of them involving several communication frontends, and also multiple antennas, requiring space, which can create problems such as the coupling between them.

A possible solution to combat the lack of available space resulting from this growing incorporation of communication systems, is to develop a single antenna, operating in each frequency band, and shared by the various communication systems. Figure 1 clarifies the scenario referring to a mobile device where a single antenna structure interacts with multiple radio-frequency frontends, which can be one of the main characteristic nowadays, since even a common smartphone gathers multiple antennas, enabling the connection with the most varied services.

**Figure 1.** Single N-Band Antenna Shared by Multiple Frontends.

There are several ways to design an antenna that meets these requirements. One is through the bandwidth, i.e. producing an ultra-wideband antenna, however, it this (obtaining a band that includes all the necessary frequencies) is an almost impossible condition, especially given the disparity of frequency bands used. In addition, isolation between bands and frontends would not be guaranteed, decreasing the efficiency.

Another possibility is to develop antennas resonating in different bands, which are the commonly named multi-band antennas. This option consists of a process with a high degree of randomness, since in most cases this is done at the expense of deformations of the radiating structure, presenting little freedom to the control of the antenna's properties. There is also the group of reconfigurable antennas (in this case, in terms of frequency), although these antennas are much more complex because they are active and require additional control to select the operating frequency. In addition, these antennas use switches in the form of discrete elements, which consume energy. Due to all this, an alternative design is proposed.

This paper is divided into 6 sections, starting with this introduction. Section 2 presents a careful revision of the related works, focusing on multi-band antennas and in frequency-selective feeding structures. Then, in Section 3 the design principle is presented, clarifying the concept proposed. Section 4 contains all methods and presents the structures developed, each in a different sub-section. The results, both measured and simulated can be found in Section 5, as well as the prototypes built. Finally, major conclusions can be found in Section 6.

#### **2. Related Work**

In this way, several authors have tested the feasibility of reducing the number of antennas in each equipment and thereby saving space. In [2] a structure based on a dual-band slot is proposed and is designed to operate at 28/38 GHz. Along with the operating frequencies, this antenna exhibits adequate bandwidth values, as well as adequate gains for incorporation in a 5G mobile device. Despite the interesting results, this work lacks measures and the authors end up defending their proposal by comparing CST with HFSS.

In [4] another dual-band antenna in a massive MIMO arrangement is described. This structure is based on a series-fed antenna array and one can denote high gains (of more than 12 dB at each band) and good radiation patterns, however, once again, only simulation values are shown.

On the contrary, in [5] the measurement results support the work on a substrate integrated waveguide (SIW) antenna array for the Ka-band. There, authors implemented a linear array to improve the overall antenna's performance and they were able to achieve satisfactory bandwidths (lower than 5%) and proper gain measurements (around 5 dBi). The authors claim that by using the presented structure, multi-band antennas can be designed, but no implementation is shown to prove this statement.

Additionally, a dual-band dual-circularly-polarized antenna operating in the Ka-band is presented in [6]. The work is supported by simulated and measured results, however, a mediocre correspondence can be denoted between simulations and measurements. The gain, both in the uplink and downlink bands, is improved when the radiating element is implemented into an array, however measurements regarding axial ratio are poor, the complexity is huge and both the antenna's scalability and the robustness are questionable.

When it comes to tri-band antennas, neither the microstrip patch antenna presented in [7] nor the antenna array with defected ground seen in ref. [8] show any measures confirming their respective simulations. Both structures have resonances in the mmWaves region, along with suitable bandwidths and gains. In [7] the efficiency values shown are quite high, whereas in [8] these values are not mentioned.

Additionally, a scalable structure can be found in [9]. It proposes a compact dual-band and small slot antenna without compromising its performance. Measurements show good correspondence with simulations, as well as good gain, considering the typical values for slot antennas. Apart from these good results, one could argue that the major contribution of the work is its scalability, since authors defend that by adding more slots to the structure the desired multi-band behavior can be achieved. Nevertheless, the structure proposed operates at 2.4/5.2 GHz, very low frequencies for the 5G context.

Recently, an aperture-sharing integration methodology implementing a 3.5/28 GHz antenna with mmWave beam steering capability was proposed [10]. The main concept is to share the aperture of a linear 28 GHz array, comprised by four separately fed dipoles, with a 3.5 GHz dipole antenna. Favorable results were obtained regarding a stable mmWave beam at different scanning angles, meanwhile with broad impedance bandwidths in both operating bands (over 20%). Nevertheless, only one of the resonating frequencies suits the mmWave spectrum region and the overall size of the structure is close to the size of a single 3.5 GHz dipole antenna, which, in many of the future applications may be inappropriate.

In [11] an interesting approach is conducted on using the half-mode substrate integrated waveguide (HMSIW) technique to design low-profile cavity-backed multi-band antennas. Authors designed single, dual and triple band structures to validate the multi band responses which indicated favorable results on the radiation patterns' stability and the front-to-back ratio. Moreover, the radiation efficiency being higher than 80% at the operating frequencies is a quite encouraging result to test this concept of introducing U-shaped strips outside the aperture of an HMSIW cavity in the mmWaves region of the electromagnetic spectrum, since the tests made were at the C-band.

In [12], an architecture whose main objective is to solve the bandwidth limitations of phased arrays was proposed. The suggested design includes five printed quasi-Yagi antennas, which should be placed in the upper edge of a mobile device. Their placement, and the orientation of the active element and the directors are crucial to solve such bandwidth limitations. With the suggested configuration and by not using phase shifters and simply switching the feeding to one of the quasi-Yagi elements it is possible to scan the desired areas. However, this switchable antenna system results in a physically larger setup, since only one antenna is used at a time, which when compared to the phased array is a disadvantage, as in these structures, antenna area can be saved, as the whole aperture is exploited.

Another structure which explores alternative designs is presented in [13]. Considering the advantages of omnidirectional radiation patterns in communicating regardless of direction, a modified fork-shaped microstrip monopole antenna with a probe feed line shows wideband and multi-band characteristics. Here, the impedance bandwidth is improved by designing a dual-triangle portion of the ground plane, yet the resonant frequencies are quite low, suitable for example for the GSM band.

Regarding the emerging MIMO systems, in [14] a MIMO antenna system for multi-band 5G (mmWave) and wideband 4G application is shown. This structure works at triple bands (28, 37 and 39 GHz) for 5G and the wideband (1.8–2.6 GHz) for 4G. Each one of the MIMO elements consists of a slot in the ground plane and two microstrip feeding ports in the top layer (the isolation is also enhanced by using a low pass filter). Indeed, this design can work as a tapered slot antenna for 5G, covering 27.5–40 GHz or as an open-ended slot antenna for 4G covering 1.8–2.6 GHz. However, as noted, in the 5G band one denotes wideband operation, instead of multiple resonances at the frequencies of interest.

One of the major challenges of a multi-band antenna structures is its efficiency, a specification that gains further importance when operating in the mmWave region. The major contribution of this work is the microstrip feeding arrangement, which allows to section the antenna into *n* frequency-selective parts. In the literature, other techniques can be found to improve the feeding network's efficiency.

In [15], an antenna array consisting of five radiating elements is designed and measured, operating at the 2–4 GHz range, achieving a beamwidth around 24 degrees. The frequency selective feeding network delivers the signal to the selected elements at 2*f* <sup>0</sup> and gradually switches the signal between elements as the frequency decreases to *f* <sup>0</sup>. The major advantage of this strategy is an almost constant beamwidth over a broad frequency range.

A similar concept is presented in [16] where a six-element antenna array operating in 1.75–3.5 GHz frequency range is seen. In this report, the feeding network uses a directional filter where the adjustment of coupled-line sections allows for the flexible selection of transmission coefficients. This structure permits a constant beamwidth in an octave frequency range the signal is redirected from the center elements to the outer elements. In the end, this feeding network's achievements are reached at the expense of a single directional filter and equal split power dividers.

More recently in [17] an interesting approach on multi-band filtering slot antennas is proposed. There, authors realize a duplexing and filtering antenna by integrating a multi-band antenna and the multimode resonator. The work is validated since three antennas were designed, fabricated and measured. Although the correspondence between simulations and measurements is quite satisfactory, neither the antenna's scalability is a priority nor the operating frequency is suitable for the future generation of mobile communications. Above all, the concept proposed in [17] requires the usage of two feeding ports, which differs from the goal of this work which lies on having a single antenna, fed in a single point, capable of interacting with multiple RF frontends, isolating each frequency.

All of these suggested frequency selective networks are associated with higher complexity in the structure's design and fabrication. They demand for additional components and the antennas were designed for lower frequencies than what is expected within the 5G context. The work presented here defends an innovative concept where neither the manufacturing is compromised nor the feeding network imposes the usage of any additional components.

In this sense a new concept to design a structure for a multi-band antenna operating in the K/Ka bands is proposed. The design idea lies on sectioning a n-band antenna into *n* parts, having each section properly isolated (in frequency) in order to maximize the efficiency, and that is achieved by applying the adequate impedance matching as it will be explained later. This frequency multiplexing concept represents the main difference between the state-of-the-art studied and the structures proposed in this report. With this alternative it is possible to several antennas in a single structure with a single feeding point. Two prototypes were built and tested representing the cases of a dual and a tri-band antenna, ensuring the scalability to other resonances.

#### **3. Design Principle**

Devices which operate in multiple bands are highly interesting, given the wide and varied protocols that they are equipped to operate. To exemplify, the architecture of a current smartphone includes capabilities to operate with different communication technologies such as Wi-Fi, Bluetooth, GSM, 3G/4G (LTE), GPS, NFC, among others. In addition to multiple radios, these devices require several antennas, increasing interferences and couplings, and it is an ever more difficult task to

accommodate them due to the reduced space available. It is in this context that a new concept for designing a multi-band antenna, is inserted.

Figure 2 presents a schematic example of a multi-band antenna for modern terminals. As noted, the antenna structure in the example operates in the downlink and uplink bands of 5G, SATCOM and LTE. In the illustration, instead of five antennas, with this concept it is possible to design a single antenna operating in all desired bands, with only one feeding point.

**Figure 2.** Schematic of the Main Design Concept.

Being an antenna shared by different radios, it is important to assure isolation between all bands, that is, that each RF frontend receives/transmits information using only its respective resonating element. The project begins with the design of all *n* resonant antenna elements, one for each operating frequency. After having the *n* radiating elements, the feeding structure ensures the high global efficiency of the antenna, forcing theoretically infinite impedances in the operating bands of neighboring elements, clearly conducting the signal towards/from the respective resonant element.

As an example, Figure 2 can be analyzed in depth, where in the middle schematic the user wants to receive data in the SATCOM downlink band, and thus the path is clear for the respective frequency band B1, and for all the other operating bands the impedance seen will be infinite. This concept is scalable for the *n* different bands of the antenna. In short, this concept allows to design a *n-band* structure as efficient as the summation of several resonant antennas in each of the *n* bands.

#### **4. Methods and Structures**

The great advantage of having the prototype composed of *n* sections is to guarantee that for each operation frequency, the input impedance is only imposed by the respective resonant element. This leads to a very efficient feeding network, in line with the challenging requirements of 5G, massive IoT and the future satellite communications.

Nowadays, wireless communications systems are mainly equipped using printed antennas [18,19] mostly due to their low cost and ease of fabrication. Another important aspect is the ability to produce various structures, with unlimited design shapes and typologies. Bearing in mind the need to ensure communication regardless of direction, monopole and dipole-based structures present interesting advantages.

In order to verify the practicability of this concept, it was decided to demonstrate it, designing two prototypes, a dual band and a tri-band antenna. Starting with the dual-band model, it is composed by two printed monopoles, each resonant at a different frequency, 28 GHz and 38 GHz respectively. Additionally, confirming the scalability of the concept, a tri-band antenna was developed, and for this, a radiator at 20 GHz was added. Both schemes are presented in Figure 3, clarifying the arrangement and the impedances expected in different parts of the structure.

The key aspect of these antennas is the feeding structure. The sectioning of this antenna into different parts lies on the existence of the connection points presented in Figure 3, thus, to scale this concept to a *n-band* antenna, *n-1* connection points are required. These connection points represent the division of the feeding network. In the example, for a dual-band antenna, there is one single connection point, however, for the tri-band structure, two connection points are required, and so on.

 **Figure 3.** Schemes Proposed for the (**a**) Dual-Band and (**b**) Tri-Band antenna.

The resulting parts of the overall antenna have an adequate matching at the resonant frequency of the respective radiating element, while simultaneously showing a close to infinite impedance at the frequency of the opposite radiating element. Therefore, at each resonant frequency, a parallel association of infinite impedances with a characteristic impedance of interest is observable.

Clarifying with Figure 3a, at the connection point and looking at the 28 GHz element: at the resonant frequency, 28 GHz, an adequate matching is verified ensuring that it works properly, and, at the same time, the feeding network warrants that at 38 GHz a very high impedance appears. These principles are applied symmetrically to the opposite side of the structure.

*ε* δ Both antennas proposed are designed over a single layer of the dielectric substrate Rogers RO4350B, which main properties include a dielectric constant ε<sup>r</sup> = 3.48, thickness *h* = 0.254 mm and dissipation factor of tan(δ) = 0.0037 @ 10 GHz. All simulations have been carried out using the electromagnetic simulator software which is the Computer Simulation Technology Microwave Studio Suite (CST-MWS).

#### *4.1. Dual-Band Antenna Design*

Ω Figure 4 identifies the main design parameters of the dual-band structure. Two monopoles with dimensions *LMp28* × *WMp28* and *LMp38* × *WMp38* were designed. A 50 Ω impedance microstrip line, along with a quarter-wavelength transformer lead the input signal to the connection point referred above. Additionally, in order to maximize the structure's compactness, meander lines were used for the greater stubs.

From the connection point up to the monopoles, there is a group of microstrip lines which guarantees the adequate matching (at the desired frequency) and the close to infinite impedance at the adjacent operating band.

The acceptable matching is achieved by the colored identified lines highlighted in Figure 4a. These lines' dimensions and placement were implemented bearing the main concepts of microwave propagation. Starting with the open-ended stub named *Stub 1*, which has a 90◦ electrical length of 90◦ at 38 GHz, it turns infinite impedance into a short circuit. Naturally, *Line 2*, having the same electrical length at 38 GHz, converts the above-mentioned short circuit once again into a very high impedance at 38 GHz.

Through this pair of lines, the first requirement at the connection point is guaranteed: an (close to) infinite impedance at 38 GHz. However, these lines surely also influence the impedance in this part of the antenna, at 28 GHz. Therefore, the third line, *Stub 3*, was placed to compensate for this undesired influence, ensuring the impedance matching at 28 GHz.

Symmetrically, on the other side of the connection point, there is an open-ended stub, immediately followed by another stub, both with an electrical length of 90◦ at 28 GHz, enabling the open circuit in the connection point at 28 GHz. Subsequently, a third stub once again compensates the influence in the antenna matching observed at 38 GHz.

**Figure 4.** Dual-Band Antenna: (**a**) Design Parameters, (**b**) Layers Identification and (**c**) Final Structure.

After bearing all theoretical propagation considerations, the dual-band antenna optimized design parameters are presented in Table 1 and the feeding network line width is constant throughout the antenna (0.2 mm), with the exception of the quarter-wavelength transformer (*Lti* × *Wti)* and the 50 Ω input line (*Lin* × *Win*), connected to the single feeding point.

**Table 1.** Dual-Band Antenna's Dimensions.


#### *4.2. Tri-Band Antenna Design*

Ω

Using the same method as before, and considering both the impedance requirements showed in Figure 3 and the design principle of Section 3, the tri-band antenna was designed over the same dielectric substrate used for the dual-band structure and its main design parameters are presented in Figure 5. λ λ λ

Ω

**Figure 5.** Tri-Band Antenna: (**a**) Design Parameters, (**b**) Layers Identification and (**c**) Final Structure.

The three resonant elements are easily denoted in Figure 5 through their associated design parameters *LMpx* × *WMpx*. The main design parameters are listed in Table 2. All microstrip lines designed have 0.2 mm width, apart from the 50 Ω input line (*Lin* × *Win*) and the quarter-wavelength transformer (*Lti* × *Wti*), as in the dual-band example. In this case, the final version of the monopoles has lengths equivalent to 0.4λd, 0.45λ<sup>d</sup> and 0.37λd, regarding the 20, the 28 and the 38 GHz monopoles, respectively.


#### **5. Results and Measurements**

The previously described prototypes were simulated and built and they are shown in Figure 6. It is important to mention that regarding the tri-band structure, Figure 6b presents a connectorized version of the antenna, where the length of the 50 Ω input feed line is excessive and just needed to place the connector. The antennas were measured, and the main results obtained are presented throughout this section. Ω Ω

**Figure 6.** Prototypes Built: (**a**) Dual-Band and (**b**) Tri-Band Antennas.

Figure 7 exhibits the simulated and measured results of the reflection coefficient for the dual-band structure.

**Figure 7.** Simulated and Measured Reflection Coefficient of the Dual-Band Antenna.

For this prototype, a good impedance matching was achieved, with both curves presenting a similar behavior despite some frequency shifts of 680 MHz and 410 MHz from the desired 28 and 38 GHz, respectively. These small discrepancies seen might be justified either by slight construction imperfections (which is something natural given the antennas' small size). Nevertheless, the measured values of S<sup>11</sup> at the operation frequencies are −13.64 dB and −16.94 dB, respectively, making of these results quite satisfactory.

Regarding the bandwidth, and considering the S<sup>11</sup> < 10 dB criteria, the prototype achieved an operating band of 2.55 GHz [27.54–30.09 GHz], representing 8.8% bandwidth around 28 GHz. On the other hand, around 38 GHz, a 1.98 GHz [37.2–39.18 GHz] band was reached, meaning 5.2%. Table 3 summarizes the measured and simulated results of the dual-band radiating element.


**Table 3.** Dual-Band S<sup>11</sup> Parameter Characteristics.

The second prototype built proves the concept here proposed in a tri-band antenna. Figure 8 identifies a reasonable impedance matching between simulated and measured results in respect to the reflection coefficient, since the minimum value for S<sup>11</sup> occurs at the frequencies of interest, despite the slight frequency shifts around 20 GHz and 38 GHz.

**Figure 8.** Simulated and Measured Reflection Coefficient of the Tri-Band Antenna.

These minor mismatches can be once again justified by the same reasons presented for the previous prototype: construction imperfections, or dielectric permittivity variations. Regarding the first operation band (at 20 GHz), the antenna shows a measured bandwidth of 0.62 GHz, equivalent to 3%. At the other operation frequencies, 0.4 GHz and 2.16 GHz are seen at 28 and 38 GHz respectively. Table 4 summarizes the measured and simulated results of the tri-band radiating structure.



Through the reflection coefficient of the latest prototype analyzed it is possible to confirm the adequacy of the prototypes built and more importantly the principle of design proposed seems to guarantee the impedance matching.

Other aspect that sustains the appropriate performance of these antennas in 5G and the massive IoT scenario is the information about the antennas' efficiency, which is presented in Figures 9 and 10, regarding the dual and the tri-band radiating structures, respectively.

**Figure 9.** Efficiency Variation Over Frequency of the Dual-Band Antenna.

**Figure 10.** Efficiency Variation Over Frequency of the Tri-Band Antenna.

Figure 9 presents the simulated efficiency of the dual-band antenna over the frequency. Naturally two local maximums are highlighted. These maximums are according to expected since they appear adjacent to the frequencies of operation. Their magnitude is higher than 85% in both operating frequencies, an important value given the structure's compactness and size.

Similarly, for the tri-band antenna, there are clearly three local maximums in Figure 10, which are once again, near the frequencies of interest, in this case 20, 28 and 38 GHz, confirming the proper functioning of the antenna.

Besides the efficiency peaks, the surface current distribution also confirms the proper functioning of the feeding network, verifying once again the frequency multiplexed structure achieved. Figure 11 starts by presenting the surface current distribution at the operating frequencies of the dual-band antenna, 28 GHz and 38 GHz.

**Figure 11.** Surface Current Distribution at (**a**) 28 GHz and (**b**) 38 GHz.

It can be easily observed that, for both frequencies, only the respective resonant monopole is being fed. The opposite monopole has no current flowing at all.

Figure 12 exhibits the respective results for the tri-band structure, denoting the surface current distribution for the three operating frequencies.

**Figure 12.** Surface Current Distribution at (**a**) 20 GHz, (**b**) 28 GHz and (**c**) 38 GHz.

Once again at each one of the resonant frequencies, only the respective monopole has any current through it, all the others remain not fed. With this, it is possible to state that the frequency multiplexing concept proposed allowed to isolate, in frequency, each radiating element, maintaining a single input feed and without requiring any additional filters.

Figures 13 and 14 show the gain variation of both prototypes over frequency, highlighting the respective values at the operating frequencies. It is possible to verify that, in the tri-band case, the gain values vary between 2.5 dBi and 4.5 dBi, in the frequencies of interest and they are mainly dependent on the antenna structure.

**Figure 13.** Gain Variation Over Frequency of the Dual-Band Antenna.

**Figure 14.** Gain Variation Over Frequency of the Tri-Band Antenna.

Regarding the gain of the antennas, in Figure 15 it is possible to observe the simulation and measured results of the two main radiation planes at both frequencies of interest, 28/38 GHz, for the dual band antenna.

**Figure 15.** Simulated and Measured Normalized Radiation Pattern of the Dual-Band Antenna at 28 GHz and 38 GHz.

As it is natural in very compact structures, particularly in multi-band antennas, the radiation patterns obtained seem a bit distorted from what was expected to be seen in monopoles. Apart from that, the dual-band antenna presents a maximum simulated gain of 3.3 dBi at 28 GHz, while at 38 GHz that value is 3.4 dBi.

When it comes to the tri-band antenna, the same two radiation planes are shown, but now for the three frequencies of interest. These results are presented in Figure 16 and a similar distortion in the radiation pattern shapes is observed. This radiating structure achieved a maximum simulated gain of 3.1, 2.5 and 4.5 dBi at 20, 28 and 38 GHz.

**Figure 16.** Simulated and Measured Normalized Radiation Pattern of the Tri-Band Antenna at 20/28/38 GHz.

The main achievements of these structures are summarized and compared with the current state-of-the-art of dual and tri-band antennas in Table 5, analyzing the publication year (PY), the operating frequencies, bandwidth, size, scalability and the isolation in frequency factor.


**Table 5.** State-of-the-Art of Multi-Band Antennas: Main Results.

Knowing that one of the main concerns of this letter is the microstrip frequency-selective feeding network able to maximize the efficiency as well as easing he interaction of a single antenna with multiple RF Frontends, Table 6 compiles some of the state-of-the-art works related to the frequency multiplexing argued, showing how it was obtained and the main characteristics of the antenna structure.



**Table 6.** *Cont.*

#### **6. Conclusions**

In this paper, a new multi-band antenna design concept is proposed, explained and proved through some practical prototypes. The main basis is the use of a feeding structure which ensure the impedance matching at a given operating frequency, isolating the structure at the remaining frequencies. When compared with an eventual alternative of a radiator with dual resonance, the option here presented, besides maximizing the efficiency, it avoids the usage of any bulky filters at the input.

This strategy was implemented in a dual-band and a tri-band prototypes, both suitable to integrate in IoT sensors. Individually these antennas present good impedance bandwidths for all frequencies of interest, since the smallest measured value is 400 MHz.

Regarding the antennas' gain, and considering that these structures are based in monopoles, these values are also satisfactory, especially when allied to the efficiency values witnessed. More importantly, the proposed concept of producing a *n-band* antenna has proven not only to work, but it also assures higher efficiency values than what is commonly seen in the state-of-the-art for multi-band antennas.

**Author Contributions:** All the authors have contributed to this paper. A.R. has designed the proposed antennas, performed the simulations and the measurements, and wrote the paper. T.V. and J.N.M. have supervised the entire research, the approach used, the results analysis, and discussion. Additionally, both have strongly contributed to the writing of the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work is funded by FCT/MCTES through national funds and when applicable co-funded EU funds under the project UIDB/50008/2020-UIDP/50008/2020This work was partially supported by FCT/MCTES through national funds and when applicable cofunded EU funds under the project UIDB/50008/2020-UIDP/50008/2020 and the European Regional Development Fund through the Competitiveness and Internationalization Operational Program, Regional Operational Program of Lisbon, Regional Operational Program of the Algarve, in component FEDER, and the Foundation for Science and Technology, Project RETIOT, POCI-01-0145-FEDER-016432.

**Conflicts of Interest:** Authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Letter* **A Single-Fed Multiband Antenna for WLAN and 5G Applications**

**Zakir Khan <sup>1</sup> , Muhammad Hunain Memon <sup>1</sup> , Saeed Ur Rahman <sup>2</sup> , Muhammad Sajjad <sup>2</sup> , Fujiang Lin <sup>1</sup> and Liguo Sun 1,\***


Received: 22 September 2020; Accepted: 20 October 2020; Published: 6 November 2020

**Abstract:** In this paper, a slotted conical patch connected to a small triangular patch multiband antenna for both microwave and millimeter-wave applications is presented. The designed antenna has three characteristics. The proposed antenna is a multiband, having a compact size of 0.35λ<sup>0</sup> × 0.35λ<sup>0</sup> × 0.004λ<sup>0</sup> at its lowest operational frequency, i.e., 2.4 GHz, and more importantly, it can cover both the microwave and millimeter-wave bands with a single feeding. According to the −10 dB matching bandwidth, experimental results show that the antenna operates at (2.450–2.495) GHz, (5.0–6.3) GHz, and (23–28) GHz. The reduced size, simple design, and multiband large bandwidth are some of the advantages over the reported designs in the latest literature. Both simulated and experimental results show a good agreement, and the proposed antenna can be used for wireless local area network (WLAN) applications and fifth-generation (5G) wireless communication devices.

**Keywords:** multibandoperation; slotted antenna; microwave; millimeter-wave band; WLAN; 5G

#### **1. Introduction**

A wireless local area network (WLAN) is a widely used network for short-range wireless communication applications. According to 802.11b/g and 802.11a standards, the bands used for WLAN are (2.400–2.484) GHz, (5.15–5.35) GHz, and (5.725–5.825) GHz [1]. Moreover, the need for high quality videos and other high data rate applications require wider bandwidth. Therefore, to fulfill the requirement of wider bandwidth, 28/38 GHz frequencies are seen as the most promising choices for fifth-generation (5G) technology [2].

Numerous researches have been exerted over the last few years pertaining to the evolution of existing standards of the wireless communication system to future 5G wireless communication standards, which is likely going to be implemented in the 2020s [3,4]. For that reason, more and more requirements have been made in the design of an antenna, in terms of size, multiband operation, and radiation pattern [5]. Many researchers are focusing on the advancement of an antenna system to operate in both current and future standards of wireless communication systems. Therefore, the acceptable way to be considered is the designing of a multiband antenna that can be integrated as a single element in many standards [6], such as WLAN, global positioning system (GPS), and other wireless communication applications.

For designing multiband antennas, different techniques were used previously to achieve multiband operating frequency standards [7–17]. The following study, of different multiband antennas covering the microwave band for WLAN applications, has been conducted in [1,7–14]. A defected ground structure (DGS) monopole antenna operating at triple frequencies for WLAN applications is presented in [6]. The radiating patch and ground of the antenna were etched on both sides of a printed-circuit board (PCB). The ground plane was modified by two equal-shaped slots on the right and left sides. Similarly, a multiband characteristic of the antenna in [9] was generated by a rectangular slot on the upper side of the antenna substrate loaded with differently shaped stubs on each side of the slot. In [10], a slotted monopole antenna, having a C-shaped patch introduced by a G-shaped parasitic strip and a partial ground plane, is used to obtain a larger bandwidth of 3.5 GHz at (3.92–7.52 GHz). Two elements of a multiple-input–multiple-output (MIMO) antenna etched with a different slot is reported in [13]. Similarly, a triple-band antenna for 2.4, 5.2, and 5.8 GHz applications in [1] and a dual-band antenna operating at 2.4 GHz and 5.2 GHz in [14] are presented. Meandering slots etched in the patch and a slotted ground DGS is used, respectively, to obtain the triple and dual-band characteristics. In [15], a 28-GHz mm-wave antenna of size 30 mm × 20 mm for 5G is reported, which is the combination of a waveguide aperture and several microstrip patches. Further, the study of antennas covering both microwave and mm-wave bands simultaneously were performed in [16,17]. In [16], a multi-layer antenna system having a dual-element MIMO on the top layer operating in the microwave band, and an antenna array at the bottom layer for the 5G band, is presented. A multiband antenna operating in both microwave and mm-wave is introduced in [17], which consists of a monopole antenna operating at 2.4/5.5 GHz and a rectangular patch covering the mm-wave 5G band. The comparison of the proposed work with the available designs of [7–17] in terms of bandwidth, multiband operation, substrate availability, design complexity, the number of layers, and feeding used is shown in Table 1. It was observed from the comparison table that the available designs have large size, complex geometry, multi-ports, and they can only cover the microwave band, or only mm-wave band, but cannot cover both bands with one feeding. Thus, the challenging part of this work is to design an antenna that can cover both the microwave band and mm-wave band with a single feeding and a compact size.


**Table 1.** Comparison of different multiband designs.

To solve these problems (of large size and complex geometry), a compact, multiband antenna covering the microwave band and the mm-wave band is proposed. Rogers RT/Duroid 5880, a widely available and inexpensive substrate was used to design the proposed antenna. We modified the radiating patch by truncating the corners of the two rectangles to form a cone and a triangle. We etched

different slots in the conical patch, which increased the electrical length of the antenna and made it more compact. Additionally, corners of the ground plane were truncated and cut by different slots to form a DGS, unlike the conventional solid ground plane. The mentioned design techniques were applied to make the antenna resonate at about (2.4, 5.2, 5.8) GHz and (28 GHz). To endorse the concept and validate the simulated results, a prototype is fabricated and results are measured. The simulated and measured results suggest that the designed antenna is the best candidate for various wireless communication applications in terms of multiband operations, compactness, large bandwidth, ease of design, and low cost. In Section 3, an explanation of a parametric study has been discussed to properly select optimized dimensions of the proposed design and achieve good results of multiband. In Section 4, different types of results were discussed followed by the comparison and conclusion of the paper in Sections 5 and 6.

#### **2. Geometry of the Antenna**

The detailed geometry of the proposed antenna will be discussed in this section. The three-dimensional (3D) electromagnetic wave solver, computer simulation technology (CST) microwave studio [18] was used for numerically investigating and optimizing the configuration of the proposed designed antenna. The front and back view of the antenna is depicted in Figure 1e. From figure, the blue color is the copper used at the top and bottom layer and the brown color is the substrate. Rogers RT5880 (ε*r*= 2.2, tan δ = 0.0009) is used as a substrate to design the antenna. The overall dimension is 30 <sup>×</sup> <sup>30</sup> <sup>×</sup> 0.508 mm<sup>3</sup> . The final geometry consists of a slotted conical patch connected to a small triangle by narrow lines. The conical patch covers the WLAN band and the triangular patch covers the 5G band. Two meandering slots and a triangular slot were etched on a conical patch. Feeding is given through a 50 Ω microstrip line. A defected ground plane etched with six slots and truncated corners are used at the bottom layer of the antenna.

The final geometry of the designed antenna was obtained by different modifications in two rectangular patches antenna (Antenna1) shown in Figure 1a. The Antenna1, patch dimensions such as width and length were obtained by Equations (1) and (2) [19]. From Figure 2a,b the Antenna1 is resonating at (24.32–24.54) GHz and (25.5–25.7) GHz. It also gives resonance at (5.3–6.0) GHz but that's not below −10 dB. We used corners truncation and meandering slits for compactness and multiband operation [6,9]. These two techniques lower the frequency of operation of the antenna by increasing the electrical length that results in the compactness of the antenna [6,9,20,21]. The ground plane used at the bottom layer is defected with different meandering slots and truncated corners to get higher bandwidth in both operating bands of the antenna. More details of truncating the corners of the square patches to form a conical and triangular patch are also discussed in Section 3.1 of this article. The next stage (Antenna2) in Figure 1b had its corners truncated from rectangular patches and the ground plane to form a cone-like and a small triangle. From Figure 2a, the resonating frequency of Antenna2 is lowered to (21.80–22.42) GHz and (26.2–27.6) GHz. Moreover, resonating at 4.7 and 7.0 GHz, but the resonance was not below −10 dB. In the next stage, Antenna3, a slot of optimized value in the conical patch and two slots in the truncated ground were etched as shown in Figure 1c. From Figure 2a, the higher resonating frequency turns into a broad band, i.e., (22–26) GHz. While at the lower band, the antenna resonates at 3.9 GHz, which is below −10 dB. It also resonates at 2.60 and 5.05 GHz, but that is above −10 dB. Moving to the next stage, Antenna4, one more slot in the patch and two more slots in the ground plane were etched, shown in Figure 1d. The antenna resonates at 2.3 GHz and (5.0–6.3) GHz, but there is a mismatch at 5.2 GHz. Thus, further improvement is needed. The final stage is to etch a triangular slot in the patch along with two more slots in the ground plane, shown in Figure 1e. From Figure 2a,b it can be seen that the antenna is exactly resonating at (2.46–2.49) GHz, (5–6.3) GHz, and (23–28) GHz.

*Sensors* **2020**, *20*, 6332

*ε δ* 

Ω

‐

**Figure 1.** Evolution of proposed antenna front and back view (**a**)Antenna1, (**b**) Antenna2, (**c**) Antenna3, (**d**) Antenna4, and (**e**) proposed antenna.

$$\mathcal{W} = \frac{(2\mathcal{N} + 1)}{\sqrt{(\varepsilon\_r + 0.5)}} \times \frac{\lambda o}{2} \tag{1}$$

$$L = \frac{(2N+1)}{\sqrt{\varepsilon\_{eff}}} \times \frac{\lambda o}{2} - 2\Delta L \tag{2}$$

−

−

−

ሺ ሻ ඥሺℇ ሻ

> ൈ

ሺ ሻ ඥℇ

ൈ 

െ Δ

‐

ൌ

ൌ

−

Δ

Ɛ λ

E<sup>r</sup> = dielectric constant, λ*<sup>o</sup>* = free space wavelength. ℇ ൌ 2 2

Δ ℎ

ൌ ቈ

ℇ

Ɛ Ɛ െ

ℇ

 ௪ ௪ 

ඥሺ ℎ ሻ

൩

**Figure 2.** Simulated S<sup>11</sup> (**a**) antenna 1 to 5, (**b**) primary and final design.

∆*L* is the effective length and can be found by the Equation (3)

$$\frac{\Delta L}{h} = 0.412 \left[ \frac{\varepsilon\_{eff} + 0.3}{\varepsilon\_{eff} - 0.258} \right] \left[ \frac{\frac{w}{h} + 0.264}{\frac{w}{h} + 0.813} \right] \tag{3}$$

where E*e*ff = effective relative permittivity of the substrate

$$
\varepsilon\_{eff} = \frac{\varepsilon\_r + 1}{2} + \frac{\varepsilon\_r + 1}{2} \sqrt{(1 + 12h/2w)}\tag{4}
$$

#### **3. Parametric Study**

To understand the impact of various parameters on different results and to achieve the best optimized dimensions of the final design, a parametric analysis has been done on different parameters of the antenna. All other parameters were kept at their final value during the parametric study.

#### *3.1. E*ff*ect of Truncating Corners of the Patch*

The corners of the rectangular patches shown in Figure 3 were truncated at three different values to form a conical and triangular shape patch. A visible effect at both operating frequency bands i.e., microwave and mm-wave band, was observed. At first value, the antenna is only resonating at 2.4 GHz in the microwave band while at (21 GHz–22.5 GHz) in the mm-wave band. At the second value, the antenna resonating frequencies are 2.4 GHz, (4 GHz–5 GHz), and (22 GHz–23.5 GHz). At the third value, the antenna gives resonance at 2.4 GHz, 5.8 GHz, and (21 GHz–24.8 GHz). The effect of different values along with the optimized value results are shown in Figure 3. ‐ ‐ ‐ ‐

**Figure 3.** Effect of truncating the corners of patches.

#### *3.2. E*ff*ect of Truncating Corners of the Ground*

‐ ‐ The DGS also has a very high impact on both microwave and mm-wave bands of operation of the antenna. Two types of techniques were used in this paper for defected ground. The first method was a truncation of the ground at three different values at all corners. At first value, the antenna resonated at 2.3 GHz, whereas the other resonating frequency shifted to 6.5 GHz. Moreover, there was a mismatch at (22.5–24.3) GHz. At the second value, the resonance frequency shifted to 2.4 GHz and 6.4 GHz, and a mismatch at (25.5–26.5) GHz. Finally, when the value increased from its optimized value, the resonance frequency moved to 4 GHz and 5.8 GHz. Whereas, at mm-wave, there was mismatching at (25.7–27.1) GHz. All of the results (of truncating the corners of the ground at different values along with its optimized values) are shown in Figure 4. ‐ ‐

**Figure 4.** Effect of truncating corners of the ground.

#### *3.3. E*ff*ect of Di*ff*erent Values of Slots in the Ground*

The second technique used for DGS was to etch different values of meandering slots at the bottom layer. Widths of the ground slots were varied at different values to analyze its performance at all the operating frequencies. Each slot width decreased to 0.5 mm from its optimized value (1 mm) and it was observed (from the result shown in Figure 5) that the antenna was not resonating at 2.4, 5.2, and 5.8 GHz. Again, when width of the slots increased from 0.5 to 0.8 mm, the resonance was above −10 dB. Finally, when the width of each slot increased to 1.2 mm, the antenna resonated only at 5.8 GHz, with the maximum resonance of −12 dB. There is no clear effect on the mm-wave operation band apart from a little mismatch at 0.5 mm on (21–25) GHz. The effect of variation of slots in the ground layer is shown in Figure 5. − − ‐ −

**Figure 5.** Effect of different values of slots in the ground.

#### *3.4. E*ff*ect of Distance between Two Patches*

‐ The antenna performance was analyzed by different values of spacing between a triangle and a conical patch. As illustrated in Figure 6, by increasing the distance between the patches, the resonance at the mm-wave band deteriorated with every variation, and had almost no effect on the lower frequencies. Initially, the distance between two patches was kept at 0.7 mm and a mismatch was observed at (25.5–27.8) GHz. When the distance further increased to 1.1 mm, the resonating frequency emanated to (22–24.2) GHz. Finally, when the distance was kept at 1.5 mm, then the resonance came further down to (21–23.5) GHz. ‐

**Figure 6.** Effect of distance between patches.

#### *3.5. E*ff*ect of Di*ff*erent Values of Slots in the Patch*

The conical patch slots width were varied at three different values and the results were analyzed. In the first step, the widths of the slots were kept 0.5 mm, and it was noted that the antenna was resonating at 5.2 GHz and 5.8 GHz, and the resonance frequency of 2.4 GHz shifted to 2.6 GHz. When widths of the slots increased to 1.3 mm, it was observed that the antenna was only resonating at 5.8 GHz, whereas there was no resonance at 2.4 GHz and 5.2 GHz. Finally, when the width of the slots further increased to 1.7 mm, again, the antenna only resonated at 5.8 GHz, and there was no resonance at 2.4 GHz and 5.2 GHz. From Figure 7, there is no effect of patch slot variation at the mm-wave operating frequency. ‐

**Figure 7.** Effect of different values of slots in the patch.

#### **4. Simulated and Measured Results Discussion**

The prototype antenna, shown in Figure 8a, is fabricated and measured to confirm simulated results. Different results of the proposed antenna, such as reflection coefficient, radiation pattern, current density, and antenna gain will be discussed in the subsections below.

‐ **Figure 8.** (**a**) Simulated and measured S-parameter, (**b**) measurement setup 2–9 GHz, and (**c**) measurement setup 21 to 29 GHz.

#### *4.1. Reflection Coe*ffi*cient*

The proposed antenna simulated along with measured S<sup>11</sup> results are depicted in Figure 8a. The S<sup>11</sup> result from 2 to 7 GHz was measured by the subminiature version A (SMA)-1 connector (D550B51H01-118) and the S<sup>11</sup> result from 23GHz to 28 GHz was measured by SMA-2 2.92 (D360B50H01-118). The S<sup>11</sup> value is below −10 dB at all frequencies of operation. The antenna operates at different frequency bands, i.e., in microwave band at (2.45–2.495) GHz, (5.0–6.3) GHz and in mm-wave band at (23–28) GHz. The demonstrated measurement setup for S<sup>11</sup> is shown in Figure 8b,c, measuring the microwave and mm-wave band respectively. In Figure 8a–c, a very good agreement can be seen between the results of simulated and measurement. However, the slight dissimilarity between the two results, especially at the mm-wave band of operation, can be noticed, and it could be possible because of the practical factors, which include SMA connector loss and hand soldering of the SMA-D360B50H01-118 to the antenna.

#### *4.2. Current Density*

‐

‐

‐ ‐

‐ ‐

‐

To understand further explanation of the multiband operation, the surface current distribution of the designed antenna was analyzed at 2.4, 5.2, 5.8, and 28 GHz. As shown in Figure 9a–c, the maximum current density is along the different parts of the conical shape patch at 2.4 GHz, 5.2 GHz, and 5.8 GHz. Whereas in Figure 9d, it can be realized that the maximum current strength of the antenna is mainly associated with the smaller triangle at 27.5 GHz.

**Figure 9.** Surface current density (**a**) 2.4 GHz, (**b**) 5.2 GHz, (**c**) 5.8 GHz, and (**d**) 27.5 GHz.

‐

#### *4.3. Radiation Pattern*

The far-field radiation pattern and gain were measured at 2.4 GHz, 5.2 GHz, 5.8 GHz, and 26.5 GHz. Both the E-plane and H-plane radiation patterns were given in Figure 10a–d. The radiation pattern in E-plane at 2.4 GHz, 5.2 GHz, and 5.8 GHz is nearly a dumbbell shape, whereas it is nearly omnidirectional in H-plane, which makes it suitable for multiple wireless systems. Similarly, the E-plane and H-plane radiation pattern at 26.5 GHz nearly has a directional pattern, as shown in Figure 10c. The measured radiation pattern results have a good agreement with the simulated results. However, again, a minor discrepancy can be noticed, and it could be possible as a result of the hand soldering of the 2.92 mm SMA D360B50H01-118 connector and measurement errors.

‐ ‐ **Figure 10.** Radiation pattern in E-plane and H-plane at (**a**) 2.4 GHz, (**b**) 5.2 GHz, (**c**) 5.8 GHz, and (**d**) 26.5 GHz.

‐

#### *4.4. Antenna Gain*

The antenna gain was calculated using an anechoic chamber at different frequencies of operation in both microwave and the mm-wave bands are depicted in Figure 11. A horn antenna was used as a reference antenna and the measurement setup in the anechoic chamber can be seen in Figure 12. From Figure 11, it can be seen that the antenna gives a maximum gain of 3.55 dB at 5.2 GHz, 4.72 dB at 5.8 GHz, and 5.85 dB at 26.5 GHz, respectively.

**Figure 11.** Proposed antenna gain over frequency.

**Figure 12.** Measurement setup for radiation pattern.

#### **5. Comparison**

‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ The comparison of the proposed design with other state-of-the-art designs is presented in Table 1. It can be seen in Table 1 that the antenna designs in the literature can only operate in microwave band for WLAN applications or in mm-wave band for 5G applications while the proposed designed antenna in this paper can operate in both microwave and mm-wave bands. The proposed antenna can be useful for two different communication technologies. Moreover, the proposed design can cover a large bandwidth as compared to the available designs. Further, the proposed antenna designed in this paper

gives better performance in terms of multiband operation, light weight, low profile, low fabrication cost, simple geometry, and compactness.

#### **6. Conclusions**

In this paper, a DGS slotted double patch antenna, having a compact size, single feeding, simple design, and multiband characteristics, was designed and measured. The designed multiband antenna consists of (2.4, 5.2, and 5.8) GHz slotted conical patch antenna and 28 GHz triangular patch antenna. Good results were achieved in both the microwave band and mm-wave band. High-quality results compared to the latest literature were obtained by optimizing different parameters of the antenna. Measured results confirmed the proposed antenna is a suitable candidate for WLAN (5.0–6.3) GHz and 5G (23–28) GHz applications.

**Author Contributions:** Conceptualization, Z.K.; methodology, Z.K.; software, Z.K.; supervision, F.L., L.S.; validation, S.U.R., M.S.; Resources, S.U.R.; formal analysis, Z.K.; writing—original draft, Z.K.; writing—review and editing, Z.K., M.H.M., L.S., F.L., S.U.R.; Funding acquisition, Z.K. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Key R&D Program of China under Grant 2019YFB2204601.

**Acknowledgments:** The authors would like to thank the Information Science Laboratory Center, University of Science and Technology of China (USTC), for hardware and software services. The authors would also like to thanks Jalil Ul Rehman Kazim for providing technical help.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Letter* **Integration of Microstrip Slot Array Antenna with Dye-Sensitized Solar Cells**

#### **Bowen Bai \* , Zheng Zhang, Xiaoping Li, Chao Sun and Yanming Liu**

The Key Laboratory of Information and Structure Efficiency in Extreme Environment, The Ministry of Education of China, and The School of Aerospace Science and Technology, Xidian University, Xi'an 710071, China; zzhang\_1992@stu.xidian.edu.cn (Z.Z.); xpli@xidian.edu.cn (X.L.); sunc@xidian.edu.cn (C.S.); ymliu@xidian.edu.cn (Y.L.)

**\*** Correspondence: bwbai@xidian.edu.cn; Tel.: +86-13572025103

Received: 15 September 2020; Accepted: 31 October 2020; Published: 2 November 2020

**Abstract:** This paper describes the integration of microstrip slot array antennas with dye-sensitized solar cells that can power array antennas at 5.8 GHz, ensuring normal communications. To appraise the antennas, a 2 × 2 circularly polarized microstrip slot array antenna integrated with dye-sensitized solar cells using a stacked design method was analyzed, fabricated and measured. The size of the entire array is 140 mm × 140 mm, where the size of each solar cell is 35 mm × 35 mm. The results show that the effect of the antenna has a slight influence on the output performance of the solar cells, and the interference of the output current of the solar cells to the antenna feeding system is negligible. The gain of the array antenna increases by 0.12 dB and the axial ratio is reduced to 1.50 dB after the integration of dye-sensitized solar cells. The integration saves a lot of space, and has the ability of self-sustaining power generation, thus providing reliable and long-term communication for various communication systems.

**Keywords:** dye-sensitized solar cells; integration; antenna array; solar antenna

#### **1. Introduction**

Recently, solar energy has received more and more attention as a clean renewable energy source, and the solar antenna (solant) has drawn a large amount of concern because it can not only transmit and receive electromagnetic waves, but also generate electricity [1]. The research on the fusion technology of solar cells and antennas can be traced back to 1995. Tanaka et al. [2] took the lead in designing a fusion device of solar cells and patch antenna, and the device was successfully applied to a microsatellite. Compared with the simple juxtaposition of the antenna and solar cell, the integration of the antenna and the solar cell has certain advantages in volume, weight, appearance and electrical performance. Both amorphous (a-Si) and crystalline (c-Si)-type silicon solar cells with integrated antennas have been reported [3–8]. A single crystal silicon solant is proposed [9]; several solar cells are placed 5 mm above the microstrip slot antenna, indicating a poor combination between the solar cells and the antenna. S. V. Shynu et al. [10] integrated a double-slot antenna with an amorphous silicon solar cell by covering a dual-band WLAN. K Yang et al. [11] replaced the copper grounding plane of the vivaldi antenna with an amorphous silicon solar cell; while the solant achieves a high degree of integration, there is a certain interference between the solar cell and the antenna. M. Danesh et al. [12] used a monopole antenna in combination with an amorphous silicon solar cell, and placed only the solar cell in the radiating portion of the monopole antenna, resulting in low space utilization. Therefore, it is still a challenge to integrate antennas with the solar cell to the greatest extent and eliminate the interference between antennas and solar cells.

Dye-sensitized solar cells have been widely studied and applied [13–15] due to their lower processing cost compared with crystalline silicon cells. With the development of research, the conversion efficiency of dye-sensitized solar cells has been improved [16], and can be printed on flexible conductive plastic layers for enhanced integration [17,18]. However, there is little literature on the integration of antennas with dye-sensitised solar cells. The first dye-sensitised solar cell antenna in a proof-of-concept dipole configuration was studied in [19]; series-connected dye-sensitised cells could produce 1.49 V and 15.5 mA, which meets design requirements. However, the interference caused by the integration of antennas with dye-sensitised solar cells still needs to be analyzed. A compatible integration of a circularly polarized omnidirectional metasurface antenna with solar cells has been reported in [20]. While the antenna gain of type IV is 4.1 dBi, it can be predicted that the output power generated by the solar cell array is not high due to the fact that the solar cells are not connected to each other, which limits its practical application.

In this paper, an integration of a 2 × 2 circularly polarized microstrip slot array antenna with dye-sensitized solar cells is designed. A novel stack design method makes the solar cell and array antenna well integrated and the simulation and measurement results show that the gain of the array antenna increased by 0.12 dB, reaching 6.60 dBi, and the axial ratio was reduced to 1.50 dB after the integration of dye-sensitized solar cells. The solar cells and the microstrip slot array antenna are perfectly integrated. The integration saves a lot of space, especially when the proposed antenna is used in satellite communication. Compared with the existing circularly polarized microstrip slot array antenna, the proposed antenna adds the output of dye-sensitized solar cells into the voltage regulation circuit to form a stable power supply for the radio frequency system, which ensures the operation of the microstrip slot array antenna. In other words, the antenna has the ability of self-sustaining power generation capabilities, so as to provide reliable and long-term communication for the communication system when the power is not easy to obtain.

#### **2. Integrated Design Array Antenna and Solar Cells**

Due to the influence of climate, environment and other factors, the linear polarization wave easily causes polarization deflection loss. These factors have little effect on the polarization deflection of circularly polarized waves, and, concerning circularly polarized antennas, polarization mismatch does not easily occur. In order to suit practical applications, a circularly polarized microstrip slot antenna is used in this paper. The geometry of a 2 × 2 circularly polarized microstrip slot array antenna integrated with dye-sensitized solar cells is illustrated in Figure 1a. The size of the entire array is 140 mm × 140 mm, where the size of each solar cell is 35 mm × 35 mm. The circularly polarized waves are excited by arranging the four slots crosswise and designing the microstrip feeder network reasonably.

The dielectric substrate of the microstrip slot antenna uses a Rogers R4350b plane, a dielectric constant of 3.48, a dielectric loss angle of 0.0037 and a thickness of 0.5 mm. The equal division Wilkinson power divider is used in the feed network to obtain the excitation signal with equal amplitude and phase difference of 90◦ , as shown in Figure 1b. The isolation resistance is 100 Ω, and the impedance of the feed port is set to 50 Ω. Through optimization analysis, the width of microstrip lines is W1 = 1.15 mm (50-Ω) and W2 = 0.63 mm (70.71-Ω). The slot size is 14.5 mm × 1.8 mm, and the eccentric distance of the feed point is 3 mm. The microstrip slot array antenna gain and the available solar cell area on the antenna surface need to be taken into consideration, concerning a slot spacing of 44 mm.

Dye-sensitized solar cells are a new type of photovoltaic technology developed by simulating the principle of plants in nature using solar energy for photosynthesis. DSSCs are based on dye sensitizers and nano-TiO2, which can make the photoelectric conversion efficiency reach a better level. At the same time, dye-sensitized solar cells (DSSCs) are rich in raw materials, non-polluting and low in cost; the manufacturing cost is only one fifth to one tenth that of silicon solar cells [21]. Therefore, dye-sensitized solar cells were chosen for this paper. Dye-sensitized solar cells adopt a stacked structure similar to the microstrip slot antenna, so the metal ground plane of the microstrip slot antenna can be used as the substrate of the solar cell.

**Figure 1.** (**a**) The geometry of a 2 × 2 circularly polarized microstrip slot array antenna integrated with dye-sensitized solar cells; (**b**) microstrip feeder network.

The integrated design does not change the intrinsic structure of the antenna, and will minimize the impact of the antenna. Due to the conductive material in the dye-sensitized solar cell, it will block the radiation of electromagnetic waves, so it is necessary to retain the slot of the antenna. The dye-sensitized solar cell and the microstrip slot antenna share a metal ground plane, and the cell structure is as described in [22]. The Components of the integration of a circularly polarized microstrip slot antenna with a dye-sensitized solar cell are shown in Figure 2. The dielectric constant and conductivity of the dye-sensitized solar cell materials at room temperature are shown in Table 1. The dye-sensitized layer is an electrolyte containing I − and I −3 , mixed with sensitizers, potassium chloride, etc.

**Figure 2.** Component of the integration of a circularly polarized microstrip slot antenna with a dye-sensitized solar cell.

**Table 1.** Solar cell material properties.


According to Figure 2 and Table 1, for the antenna structure shown in Figure 1, the following two spectral domain integral equations can be obtained by using the boundary condition that the total electric field tangential direction of the antenna surface is 0:

$$\iint \left[\tilde{\mathbf{G}}\_{\mathbf{x}\mathbf{x}} \tilde{\mathbf{J}}\_{\mathbf{x}} + \tilde{\mathbf{G}}\_{\mathbf{y}\mathbf{y}} \tilde{\mathbf{J}}\_{\mathbf{y}}\right] \exp\{-\mathbf{j}(\mathbf{k}\_{\mathbf{x}}\mathbf{x} + \mathbf{k}\_{\mathbf{y}}\mathbf{y})\} \mathrm{d}\mathbf{k}\_{\mathbf{x}} \mathrm{d}\mathbf{k}\_{\mathbf{y}} = \iint \tilde{\mathbf{G}}\_{\mathbf{x}\mathbf{z}} \tilde{\mathbf{J}}\_{\mathbf{z}} \exp\{-\mathbf{j}(\mathbf{k}\_{\mathbf{x}}\mathbf{x} + \mathbf{k}\_{\mathbf{y}}\mathbf{y})\} \mathrm{d}\mathbf{k}\_{\mathbf{x}} \mathrm{d}\mathbf{k}\_{\mathbf{y}}\tag{1}$$

$$\iint \left[\tilde{\mathbf{G}}\_{\text{Y}\mathbf{X}} \tilde{\mathbf{J}}\_{\text{x}} + \tilde{\mathbf{G}}\_{\text{YY}} \tilde{\mathbf{J}}\_{\text{y}}\right] \exp\{-\mathbf{j}(\mathbf{k}\_{\text{x}}\mathbf{x} + \mathbf{k}\_{\text{y}}\mathbf{y})\} \mathrm{d}\mathbf{k}\_{\text{x}}\mathrm{d}\mathbf{k}\_{\text{y}} = \iint \tilde{\mathbf{G}}\_{\text{Y}\mathbf{Z}} \tilde{\mathbf{J}}\_{\text{z}} \exp\{-\mathbf{j}(\mathbf{k}\_{\text{x}}\mathbf{x} + \mathbf{k}\_{\text{y}}\mathbf{y})\} \mathrm{d}\mathbf{k}\_{\text{x}}\mathrm{d}\mathbf{k}\_{\text{y}}\tag{2}$$

where ∼ Gxx, ∼ Gxy, ∼ Gxz, ∼ Gyx, ∼ Gyz, ∼ Gyy are the components of Green's function in the electric field spectral domain, ∼ J<sup>x</sup> and ∼ J<sup>y</sup> are the x and y spectral components of the unknown current on the surface, and ∼ Jz are the spectral components of the feed current of the coaxial probe. The antenna is covered with multi-layer dielectric plates, where each layer of the dielectric plate is a lossy medium, their relative dielectric constant is εri, the thickness of each covering layer is h<sup>i</sup> and the permeability of each layer of the medium is µ<sup>0</sup> . According to reference [23], the analytical calculation formula of spectral domain Green's functions of the electric field on the antenna surface can be obtained as follows:

$$\tilde{\mathbf{G}}\_{\rm xxi}(\mathbf{k}\_{\rm x}, \mathbf{k}\_{\rm y}) = \mathbf{k}\_{\rm 0} \boldsymbol{\eta}\_{\rm 0} [\mathbf{B}^{\rm h} \mathbf{k}\_{\rm y}^2 / \mathbf{k}\_{\rm t}^2 + \mathbf{B}^{\rm e} \mathbf{k}\_{\rm x}^2 / \mathbf{k}\_{\rm t}^2] \tag{3}$$

$$\tilde{\mathbf{G}}\_{\rm xyi}(\mathbf{k}\_{\rm \lambda}, \mathbf{k}\_{\rm y}) = \mathbf{k}\_{\rm 0} \boldsymbol{\eta}\_{\rm 0} \mathbf{k}\_{\rm x} \mathbf{k}\_{\rm y} / \mathbf{k}\_{\rm t}^2 [\mathbf{B}^{\rm e} - \mathbf{B}^{\rm h}] \tag{4}$$

$$
\tilde{\mathbf{G}}\_{\rm{yzi}}(\mathbf{k}\_{\mathbf{x}}, \mathbf{k}\_{\mathbf{y}}) = \tilde{\mathbf{G}}\_{\rm{xyi}}(\mathbf{k}\_{\mathbf{x}}, \mathbf{k}\_{\mathbf{y}}) \tag{5}
$$

$$\tilde{\mathbf{G}}\_{\rm Yyi}(\mathbf{k}\_{\rm x}, \mathbf{k}\_{\rm y}) = \mathbf{k}\_{\rm 0} \boldsymbol{\eta}\_{\rm 0} [\mathbf{B}^{\rm h} \mathbf{k}\_{\rm x}^{2} / \mathbf{k}\_{\rm t}^{2} + \mathbf{B}^{\rm e} \mathbf{k}\_{\rm y}^{2} / \mathbf{k}\_{\rm t}^{2}] \tag{6}$$

$$
\tilde{\mathbf{G}}\_{\rm xzi}(\mathbf{k}\_{\mathbf{x}}, \mathbf{k}\_{\mathbf{y}}) = \mathbf{j} \mathbf{k}\_{0} \mathbf{\eta}\_{0} \mathbf{B}^{\mathbf{e}} \mathbf{k}\_{\mathbf{x}} / \gamma^{2} \tag{7}
$$

$$
\tilde{\mathbf{G}}\_{\rm yzi}(\mathbf{k}\_{\rm x}, \mathbf{k}\_{\rm y}) = (\mathbf{k}\_{\rm y}/\mathbf{k}\_{\rm x}) \tilde{\mathbf{G}}\_{\rm xzi}(\mathbf{k}\_{\rm x}, \mathbf{k}\_{\rm y}) \tag{8}
$$

where:

$$\mathbf{B}^{\mathbf{h}} = (\mathbf{A}\_{11}^{\mathbf{h}} + \mathbf{A}\_{21}^{\mathbf{h}}) / \mathbf{U}^{\mathbf{h}} \tag{9}$$

$$
\begin{bmatrix}
\left(\mathbf{A}\_{11}^{\mathbf{h}}\right) & \left(\mathbf{A}\_{12}^{\mathbf{h}}\right) \\
\left(\mathbf{A}\_{21}^{\mathbf{h}}\right) & \left(\mathbf{A}\_{22}^{\mathbf{h}}\right)
\end{bmatrix} = \prod\_{i=1}^{N} \frac{1}{1 + \mathbf{R}\_{i}^{\mathbf{h}}} \begin{bmatrix}
\mathbf{e}^{\mathbf{j} \mathbf{y}\_{i} \mathbf{h}\_{i}} & \mathbf{R}\_{i}^{\mathbf{h}} \mathbf{e}^{\mathbf{j} \mathbf{y}\_{i} \mathbf{h}\_{i}} \\
\mathbf{R}\_{i}^{\mathbf{h}} \mathbf{e}^{-\mathbf{j} \mathbf{y}\_{i} \mathbf{h}\_{i}} & \mathbf{e}^{-\mathbf{j} \mathbf{y}\_{i} \mathbf{h}\_{i}} \\
\end{bmatrix} \tag{10}
$$

$$\mathbf{R}\_{\mathbf{i}}^{\mathbf{h}} = (\boldsymbol{\gamma}\_{\mathbf{i}} - \boldsymbol{\gamma}\_{\mathbf{i}-1}) / (\boldsymbol{\gamma}\_{\mathbf{i}} + \boldsymbol{\gamma}\_{\mathbf{i}-1}) \tag{11}$$

$$\mathbf{R}\_{\mathbf{i}}^{\mathbf{e}} = (\varepsilon\_{\text{ri}}\boldsymbol{\gamma}\_{\text{i}+1} - \varepsilon\_{\text{ri}+1}\boldsymbol{\gamma}\_{\text{i}}) / (\varepsilon\_{\text{ri}}\boldsymbol{\gamma}\_{\text{i}+1} + \varepsilon\_{\text{ri}+1}\boldsymbol{\gamma}\_{\text{i}}) \tag{12}$$

$$\mathbf{U}^{\mathbf{h}} = \left\{ \mathbf{j}\gamma \cot \mathbf{y} (\gamma \mathbf{h}) \Big[ (\mathbf{A}\_{11}^{\mathbf{h}}) + (\mathbf{A}\_{21}^{\mathbf{h}}) \Big] - \gamma\_1 [(\mathbf{A}\_{11}^{\mathbf{h}}) - (\mathbf{A}\_{21}^{\mathbf{h}})] \Big\} \tag{13}$$

$$\mathbf{U}^{\mathbf{e}} = \left| \mathbf{j}(\mathbf{k}\_0^2 \varepsilon\_\mathbf{r}/\gamma) \cot \mathbf{y}(\gamma \mathbf{h}) \right[ (\mathbf{A}\_{11}^{\mathbf{e}}) + (\mathbf{A}\_{21}^{\mathbf{e}}) ] - (\mathbf{k}\_0^2 \varepsilon\_\mathbf{r}/\gamma) \left[ (\mathbf{A}\_{11}^{\mathbf{e}}) + (\mathbf{A}\_{21}^{\mathbf{e}}) \right] \tag{14}$$

$$
\gamma = \sqrt{\varepsilon\_\mathrm{r} \mathbf{k}\_0^2 - \mathbf{k}\_t^2} \tag{15}
$$

$$\gamma\_{\mathbf{i}} = \sqrt{\varepsilon\_{\mathbf{r}} \mathbf{k}\_0^2 - \mathbf{k}\_{\mathbf{t}}^2} \tag{16}$$

$$\mathbf{k}\_{\mathbf{t}} = \sqrt{\mathbf{k}\_{\mathbf{x}}^2 + \mathbf{k}\_{\mathbf{y}}^2} \tag{17}$$

Replace the superscript 'h' in Equations (9) and (10) with 'e' to obtain the expressions for each component of B e , A e <sup>11</sup>, A e <sup>12</sup>, A e <sup>21</sup> and A e <sup>22</sup>. k<sup>0</sup> is the free space wave number, and η<sup>0</sup> is the free space wave impedance. Equations (3)–(8) are the new analytical calculation formulas for the spectral domain Green's function of the microstrip slot antenna structure covered by the multilayer dielectric. After obtaining the calculation formula of Green's function in the spectral domain, the solution of integral Equations (1) and (2) can be discussed, as below. Assuming that the coaxial probe is fed at point (xp, yp) and there is a constant current I<sup>0</sup> on the probe, the formula for calculating the spectral domain of the current on the probe can be obtained as follows:

$$\tilde{\mathbf{J}}\_{\mathbf{z}} = \mathbf{I}\_{0} \exp[\mathbf{j}(\mathbf{k}\_{\mathbf{x}} \mathbf{x}\_{\mathbf{p}} + \mathbf{k}\_{\mathbf{y}} \mathbf{y}\_{\mathbf{p}})] \tag{18}$$

Let the unknown current on the antenna surface be:

$$
\tilde{\mathbf{J}}\_{\mathbf{s}}(\mathbf{x}, \mathbf{y}) = \mathbf{J}\_{\mathbf{x}}(\mathbf{x}, \mathbf{y})\tilde{\mathbf{x}} + \mathbf{J}\_{\mathbf{y}}(\mathbf{x}, \mathbf{y})\tilde{\mathbf{y}}\tag{19}
$$

Jx (x, y) and J<sup>y</sup> (x, y) are expanded by a set of basis functions, and then Fourier transform is used to obtain the spectral domain expression:

$$\tilde{\mathbf{J}}\_{\mathbf{x}}(\mathbf{k}\_{\mathbf{x}}, \mathbf{k}\_{\mathbf{y}}) = \sum\_{n=1}^{N\_{\mathbf{x}}} \mathbf{C}\_{\mathbf{x}n} \tilde{\mathbf{J}}\_{\mathbf{x}n}(\mathbf{k}\_{\mathbf{x}}, \mathbf{k}\_{\mathbf{y}}) \tag{20}$$

$$\tilde{\mathbf{J}}\_{\mathbf{y}}(\mathbf{k}\_{\mathbf{x}\prime}\mathbf{k}\_{\mathbf{y}}) = \sum\_{\mathbf{n}=1}^{N\_{\mathbf{y}}} \mathbf{C}\_{\mathbf{yn}} \tilde{\mathbf{J}}\_{\mathbf{yn}}(\mathbf{k}\_{\mathbf{x}\prime}\mathbf{k}\_{\mathbf{y}}) \tag{21}$$

∼ Jxn (kx, ky) and ∼ Jyn (kx, ky) in Equations (20) and (21) are the spectral domain expressions of the selected basis functions. By introducing Equations (18), (20) and (21) into Equations (1) and (2), the integral Equations (1) and (2) can be solved using the Galerkin method to obtain the current coefficients Cxn and Cyn. Then the relevant characteristic parameters of the antenna can be further calculated.

The electromagnetic simulation software ANSYS HFSS is used to obtain the relevant electromagnetic parameters of the antenna [24]. The center frequency of the antenna is set to 5.8 GHz and the polarization mode is right-handed circular polarization. The microstrip slot antenna and the microstrip slot antenna integrated with dye-sensitized solar cell are simulated separately, and the related reflection coefficient and radiation efficiency are shown in Figure 3. It can be seen that the solar cell has little influence on the impedance matching of the antenna, and the microstrip slot antenna integrated with dye-sensitized solar cell has an impedance bandwidth of 2.33 GHz, from 4.23 to 6.56 GHz. After the integration of the dye-sensitized solar cell, the radiation efficiency of the antenna decreases from 76.1% to 68.9% when the operation frequency is 5.8 GHz. The radiation efficiency of the antenna decreases slightly, which indicates that the dye-sensitized solar cell has little influence on the circular polarization radiation characteristics of the antenna.

The simulation results show that the presence of the solar cell has little effect on the performance of the antenna, but the solar cell generates a corresponding current when receiving visible light irradiation. Therefore, when the cell is in working condition, the influence on the antenna performance needs to be measured.

**Figure 3.** *Cont*.

**Figure 3.** (**a**) Reflection coefficient; (**b**) radiation efficiency of the microstrip slot with or without solar cell.

#### **3. Results and Discussion**

The photograph of the 2 × 2 circularly polarized microstrip slot array antenna integrated with the dye-sensitized solar cells is presented in Figure 4. It can be seen that the slot divides the entire antenna into nine parts, and each part is tightly connected to a dye-sensitized solar cell. Among them, three independent dye-sensitized solar cells are connected in series to improve the output voltage of the solar cells, and three series connected solar cells are connected in parallel to increase the output current of the solar cells. The dye-sensitized solar cell uses glass as the substrate, the cell and the antenna are grounded together by welding the negative electrode of the cell to the metal ground plant of the microstrip slot antenna.

**Figure 4.** Photograph of the 2 × 2 circularly polarized microstrip slot array antenna integrated with dye-sensitized solar cells.

The high-illuminance xenon lamp is used as the light source to illuminate the microstrip slot array antenna integrated with dye-sensitized solar cells at a suitable distance. Figure 5a shows the related experiments on the energy output characteristics of the microstrip slot array antenna integrated with the dye-sensitized solar cells. The open-circuit voltage and short circuit current of dye-sensitized solar cells are 1.94 V and 99 mA, respectively, whether the antenna works or not. The dye-sensitized solar cells are externally connected with a sliding rheostat whose resistance varies from 0 to 100 Ω. The relationship between the output voltage and the output power of the solar cell under the two conditions of antenna operation and non-operation is measured by changing the value of the sliding rheostat, as shown in Figure 5b. It can be seen that the curves coincide basically when the antenna works or does not work, which indicates that, whether it works or not, the antenna has little effect on the performance of dye-sensitized solar cells. The 2 × 2 circularly polarized microstrip slot array antenna integrated with dye-sensitized solar cells is measured in an anechoic chamber to further validate its design; the illumination intensity was maintained at 150 W/m<sup>2</sup> and the relevant measurement environment is shown in Figure 6.

**Figure 5.** (**a**) The related experiments on the energy output characteristics; (**b**) output power versus output voltage.

**Figure 6.** The experimental setup and the measurement environment.

Figure 7 shows the measured reflection coefficient and radiation pattern of the microstrip slot array antenna and the array antenna integrated with dye-sensitized solar cells. From the measurement results in Figure 7a, when the array antenna is integrated with dye-sensitized solar cells, the current generated when the solar cells work will have a certain influence on the reflection coefficient of the antenna, but the reflection coefficient performance of the array antenna integrated with dye-sensitized solar cells is still good on the whole. From the measurement results of Figure 7b, the gain of the microstrip slot array antenna is 6.48 dBi at 5.8 GHz, and the gain of the array antenna integrated with dye-sensitized solar cells is 0.12 dB higher than that of the microstrip slot array antenna, reaching 6.60 dBi. Generally, the existence of dye-sensitized solar cells has little effect on the gain performance of microstrip slot array antenna.

Figure 8 shows the measurement results of the influence of the axial ratio parameters of the microstrip slot array antenna integrated with dye-sensitized solar cells. It can be found from Figure 8a that the axial ratio of the microstrip slot array antenna is 1.65 dB at 5.8 GHz; the axis ratio of the array antenna integrated with dye-sensitized solar cells is 1.50 dB. The existence of dye-sensitized solar cells has little effect on the circular polarization radiation performance of the microstrip slot array antenna. The axial ratio radiation patterns in Figure 8b show that the working solar cells have almost no effect on the antenna axis ratio.

The results of normalized radiation patterns of the array antenna integrated with dye-sensitized solar cells are shown in Figure 9. It can be seen that the existence of dye-sensitized solar cells has little influence on the directivity of the microstrip slot array antenna. The above results show that, when the antenna is integrated with dye-sensitized solar cells, the measurement results in the figures will show slight changes, which are mainly caused by the interference between the solar cells and the mutual coupling between the wires. To sum up, the dye-sensitized solar cells and the microstrip slot array antenna are perfectly combined; the interference between the solar cells and the antenna is minimal. When solar cells with higher cost but better photoelectric conversion efficiency are used, the antenna size can be reduced to further improve the performance.

**Figure 7.** (**a**) Reflection coefficient; (**b**) gain of the microstrip slot antenna integrated with dye-sensitized solar cells varies with frequency.

**Figure 8.** *Cont*.

**Figure 8.** (**a**) The axial ratio varies with frequency; (**b**) axial ratio radiation pattern of the proposed antenna.

**Figure 9.** Normalized radiation patterns of the proposed antenna.

#### **4. Conclusions**

In this paper, a 2 × 2 circularly polarized microstrip slot array antenna integrated with dye-sensitized solar cells is designed. A novel stack design method makes the solar cells and the array antenna well integrated. The simulation and measurement results show that the gain of the array antenna increases by 0.12 dB and the axial ratio decreases to 1.50 dB after the integration of the dye-sensitized solar cells. Whether the antenna works or not has little influence on the performance of the dye-sensitized solar cells. The array antenna integrated with dye-sensitized solar cells has a similar radiation performance to the traditional microstrip slot array antenna, and can also provide electricity. Compared with the existing circularly polarized microstrip slot array antenna, the proposed antenna adds the output of dye-sensitized solar cells into the voltage regulation circuit to form a stable power supply to the radio frequency system, which ensures the operation of the microstrip slot array antenna. In other words, the antenna has the ability of self-sustaining power generation capabilities, so as to provide reliable and long-term communication for the communication system when the power is not easy to obtain.

**Author Contributions:** Conceptualization, B.B.; methodology, B.B., C.S. and Z.Z.; software, B.B., Z.Z.; validation, B.B., Z.Z. and C.S.; formal analysis, Z.Z. and C.S.; investigation, B.B., Z.Z. and X.L.; resources, B.B., X.L. and Y.L.; writing—original draft preparation, Z.Z.; writing—review and editing, B.B. and C.S.; visualization, Z.Z.; supervision, X.L., Y.L.; project administration, B.B., X.L.; funding acquisition, B.B. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported in part by the National Natural Science Foundation of China under Grant 61801343 Grant 61701381, Grant 61431010, and Grant 61627901, in part by the Natural Science Basic Research Plan in Shaanxi Province of China under Grant 2019JM-177, and in part by the Chinese Postdoctoral Science Foundation.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Letter* **Timestamp Estimation in P802.15.4z Amendment** †

#### **Ioan Domuta \* , Tudor Petru Palade , Emanuel Puschita and Andra Pastrav**

Communication Department, Technical University of Cluj-Napoca, 400027 Cluj-Napoca, Romania; tudor.palade@com.utcluj.ro (T.P.P.); Emanuel.Puschita@com.utcluj.ro (E.P.); Andra.PASTRAV@com.utcluj.ro (A.P.)

**\*** Correspondence: ioan.domuta@2dd.ro; Tel.: +40-264-440000

† This paper is an extended version of our paper: Domuta, I.; Palade, T.P.; Puschita, E.; Pastrav, A. Localization in 802. 15.4z Standard. In Proceedings of the 2020 International Workshop on Antenna Technology (iWAT), Bucharest, Romania, 25–28 February 2020; pp. 1–4, doi:10.1109/iWAT48004.2020.1570615511.

Received: 9 July 2020; Accepted: 20 September 2020; Published: 22 September 2020

**Abstract:** Due to the known issue that the ranging in the 802.15.4™-2015 standard is prone to external attacks, the enhanced impulse radio (EiR), a new amendment still under development, advances the secure ranging protocol by encryption of physical layer (PHY) timestamp sequence using the AES-128 encryption algorithm. This new amendment brings many changes and enhancements which affect the impulse-radio ultra-wideband (IR-UWB) ranging procedures. The timestamp detection is the base factor in the accuracy of range estimation and inherently in the localization precision. This paper analyses the key parts of PHY which have a great contribution in timestamp estimation precision, particularly: UWB pulse, channel sounding and timestamp estimation using ciphered sequence and frequency selective fading. Unlike EiR, where the UWB pulse is defined in the time domain, in this article, the UWB pulse is synthesized from the power spectral density mask, and it is shown that the use of the entire allocated spectrum results in a decrease in risetime, an increase in pulse amplitude, and an attenuation of lateral lobes. The paper proposes a random spreading of the scrambled timestamp sequence (STS), resulting in an improvement in timestamp estimation by the attenuation lateral lobes of the correlation. The timestamp estimation in the noisy channels with non-line-of-sight and multipath propagation is achieved by cross-correlation of the received STS with the locally generated replica of STS. The propagation in the UWB channel with frequency selective fading results in small errors in the timestamp detection.

**Keywords:** UWB; 802.15.4z; timestamp detection; ranging; multipath; frequency fading

#### **1. Introduction**

The wireless localization is a key part of many emerging technologies: internet of things (IoT), intelligent transportation systems (ITS), autonomous robots, or unmanned aerial vehicles. For many critical applications, localization accuracy is a basic requirement of localization systems. Due to its large bandwidth, the impulse radio ultra-wideband (IR-UWB) technology provides the best precision in range measurement by time-of-flight (ToF) estimation.

The basic feature in the accuracy of estimating the ToF is the shape of the pulse, more precisely the speed of increase of the pulse front. Nowadays the IEEE Task Group 4z, (TG4z), [1] is working on the enhanced impulse radio (EiR) project focused on localization safety improvement. This new amendment proposes a new UWB reference pulse and a time domain mask. In this paper, the UWB pulse is synthesized from power spectral density (psd) specified by the 802.15.4-2015 standard [2]. Two shapes of pulses were synthesized and compared, the first pulse being synthesized using only the central lobe of the psd mask and the second one being synthesized from the entire allocated spectrum.

The EiR amendment proposes that the timestamp estimation is validated by the cross-correlation of locally generated STS replica with received STS sequence. To avoid the interferences, the pulses are spread out on a symbol. This article proposes a supplementary spreading by a bit position modulation with a randomly generated sequence and shows, by simulated experiments, that lateral lobes of cross-correlation are mitigated by this modulation.

The behavior of the proposed methods is analyzed in a noisy radio channel with non line-of-sight (NLOS) and multipath propagation. The channel impulse response is estimated and subsequently used for the generation of a local replica of STS.

The main contributions of this article are that it:


This paper is organized as follows. Section 2 presents state of the art research in the field. In Section 3, the UWB pulses are synthesized. Section 4 presents the random spreading of STS sequence and timestamp estimation in noisy channels and NLOS propagation. All sections incorporate simulated experiments, and because the simulation results from a subsection are used in the subsequent ones, they will be presented along with the theoretical aspects in the corresponding subsection.

#### **2. Literature Review**

The UWB radio holds a large bandwidth, but the harmonized standards [3] impose upper limits for power spectral density (−41.3 dBm/MHz, which is under the noise floor), resulting in great difficulty in the extraction of signal from noise. The research in [4,5] shows the presence of intra-symbol interference (IASI), inter-symbol interference (ISI), and multipath interference (MUI). In order to minimize the interferences, the pulse should have a small duration of the leading lobe and a high attenuation of the side lobes. It must be noted that this small duration of the main lobe can lead to an excess in bandwidth. The pulse shape has to be a compromise between the regulatory compliance, the need of low voltage and low power supply, low duration for maximization of data rate, ranging accuracy, and minimalization of interferences. In most cases, the UWB pulse is synthesized from Gaussian impulse [6], its derivatives [7], or a linear combination of Gaussian pulses [8]. Keshavarz et al. [9] infer the weights of the derivatives in the impulse structure by particle swarm optimization (PSO) algorithm, and PSO algorithm is used for optimization of the architecture of a UWB transmitter [10]. A linear combination of Gaussian monocycles with weight optimization by semidefinite programming is used for pulse synthesis [11]. Baranauskas and Zelenin present a direct waveform synthesis of UWB pulse by high speed DAC [12].

The EiR amendment specifies two pulses [13] as boundaries for the UWB pulse and a time domain mask [14] as a constraint for pulse shape. The UWB pulse synthesized from the entire allocated spectrum falls into the time domain mask, has high energy, and can be used as a reference in UWB pulse design.

In the 802.15.4 standard [2], the UWB PHY is specified in detail and, the transceivers manufactured in this technology are widely used in localization. However, several researches show that the range measurement in the current technology is prone to external attacks. Francillon et al. [15] present a relay attack, Taponecco et al. [16] show a delay attack and Singh et al. [17] propose a modulation scheme that secures the distance measurement against relay attack. The EiR project [1] brings a lot of improvements, including UWB reference pulse shape, preamble symbols revision, addition of scrambled timestamp sequence for secure ranging, an increase in data rate and PHY payload length, and the modification and addition of a new MAC primitive for key management. An overview of the EiR standard is presented in the work of Sedlacek et al. [18].

This article is limited to estimating the timestamp, without going into detail regarding ranging or location methods. Alarifi et al. perform a deep analysis of ultrawide band indoor positioning Technologies [19]. Several works deal with the wireless localization in internet of things (IoT) [20–22] and many research depict the localization in vehicular technologies [23–25].

#### **3. UWB Pulse Synthesis**

The pulse shape plays an essential role in the ranging accuracy and in reaching the maximum distance, while maintaining regulatory compliance. The 802.15.4-2015 standard, hereinafter called 'old standard' has been defined as a root raised cosine reference pulse. As this pulse has a precursor, it can mask the attenuated first path signal. The TG4 proposes [13] that the transmitted pulse shape p(t) to be constrained by the time domain mask, specified by the standard. The EiR specifies that the pulse risetime, 10–90%, for 500 MHz channels has to be maximum 2 ns.

This paper proposes the synthesis of UWB pulse from the compliant power spectral density mask (psd) [2] by Kolmogorov factorization (detailed in Appendix A) [26], because this method provides a minimum phase pulse, as it is specified by the EiR. The old standard specifies the psd mask as it is depicted in Figure 1a, namely the trace 802.15.4a mask. The power is expressed in Watts, in order to get the pulse amplitude in volts (1 Ω load). For pulse synthesis, a new spectral mask is designed, trace 1.5 ns risetime impulse, using a raised cosine profile Ω =൞ ு |− | < (1 − ) ൬ ு + 1 ൰ ൬1 + 2 ( − − 1 + )൨൰ + (1 − ) ≤ |− | ≤ (1 − ) 0 ℎ

$$H = \begin{cases} \begin{pmatrix} p\text{sd}\_{H} & \left| f - f\_{\text{c}} \right| < (1 - \beta)F\_{\text{c}}\\ \frac{\left| p\text{sd}\_{H} + p\text{sd}\_{L} \right|}{1} \left( 1 + \cos\left[ \frac{\pi}{2\beta} (\frac{f - f\_{\text{c}}}{F\_{\text{c}}} - 1 + \beta) \right] \right) + p\text{sd}\_{L} & (1 - \beta)F\_{\text{c}} \le \left| f - f\_{\text{c}} \right| \le (1 - \beta)F\_{\text{c}}\\ 0 & \text{otherwise} \end{cases} \tag{1}$$

where central frequency *f<sup>c</sup>* = 0, cutoff frequency *F<sup>c</sup>* = 315 MHz, roll-off factor β = 0.25 and *psd<sup>H</sup> and psd<sup>L</sup>* are the high and low value of psd mask. The sampling frequency is *f<sup>s</sup>* = 10 GHz and the window length is *N* = 10<sup>4</sup> samples for 1 MHz frequency resolution. = 1 MHz ඥ [MHz] 0 dBm on 50 MHz

**Figure 1.** UWB impulse synthesis: (**a**) Spectral masks; (**b**) The pulses in time domain.

Different from standard [3], where the pulse energy is averaged on a 1 ms interval, in the paper, the mean power is computed on a 1 µs interval, considering that the pulse repetition frequency *PRF* = 1 MHz. Accordingly, the pulse amplitude is inferred based on this PRF. The EiR defines many mean PRFs and, in order to comply with the regulations, the determined amplitude has to be scaled


with p *PRF* [MHz]. Furthermore, the regulation imposes the pulse peak power to a value that shall not exceed 0 dBm on 50 MHz bandwidth, and the impulse has to respect this restriction too.

Figure 1b shows the synthesized pulse, trace synthesized 1.5 ns impulse, compared to EiR compliant pulses for the 499.2 MHz bandwidth (i.e., 1.7 ns risetime impulse and 1.2 ns risetime impulse traces). The synthesized pulse is situated between recommended pulse limits, so it respects the specifications. The traces 1.7 ns risetime and 1.2 ns risetime in Figure 1a show that the pulses suggested by the EiR do not fit exactly in the standard psd mask, the first exceeding the maximum psd and the second exceeding the bandwidth.

Recently, TG4z has defined a time domain mask for UWB impulse [12], as illustrated in Figure 2b. Based on this mask, it is appropriate to search for a new UWB pulse shape which falls in this mask for impulse energy maximization and risetime reduction. The Cramer–Rao lower bound in ToF estimation is inversely proportional to effective bandwidth [27], so it is convenient to use the lateral lobes of low power for pulse synthesis. In order to do this, a new psd mask is designed using a sum of raised cosine profiles, Figure 2, having the following parameters: (*Fc*<sup>1</sup> = 315 MHz, β<sup>1</sup> = 0.05); (*Fc*<sup>2</sup> = 400 MHz, β<sup>2</sup> = 0.05); (*Fc*<sup>3</sup> = 500 MHz, β<sup>3</sup> = 0.05). ൫ଵ = 315 MHz, <sup>ଵ</sup> = 0.05൯; ൫ଶ = 400 MHz, <sup>ଶ</sup> = 0.05൯; (ଷ = 500 MHz, <sup>ଷ</sup> = 0.05)

**Figure 2.** UWB impulse synthesis from complete spectral mask: (**a**) Spectral mask from multiple raised cosine profiles; (**b**) Time domain mask and synthesized pulses.

0.187 V/1 Ω/1 μs ଵ%→ଽ% = 1.2 ns 0.164 V/1 Ω/1 μs ଵ%→ଽ% = 1.5 ns Figure 2b shows that the pulse synthesized from complete mask has a smaller risetime and a stronger attenuation of lateral lobes compared to the pulse synthesized from partial psd mask. As such, the pulse synthesized from complete mask allows for more precise timestamp estimation and less interferences. The pulse synthesized from complete mask has the amplitude 0.187 V/1 Ω/1 µs, higher energy, low risetime *T*10%→90% = 1.2 ns and smaller lateral lobes than the pulse synthesized from partial psd mask which has an amplitude of 0.164 V/1 <sup>Ω</sup>/1 <sup>µ</sup><sup>s</sup> and a risetime of *<sup>T</sup>*10%→90% = 1.5 ns. The pulse risetime plays a key role in timestamp estimation precision. Therefore, it is preferable to choose the pulse synthesized from complete mask as UWB reference pulse.

The degree of spectrum usage is evaluated by normalized effective signal power (NESP) [6]

$$\text{NESP} = \frac{\int \left| \mathbf{S} \frac{f}{f} \right|^2 df}{\int \mathbf{M}(f) df} \tag{2}$$

 where *<sup>S</sup>*(*f*) 2 is spectral density of the impulse and *M*(*f*) is allocated spectral mask.

= ඨ

$$\int\_{228}^{f} \int\_{}^{f} $$

<sup>ଶ</sup>

<sup>ଶ</sup>

*Sensors* **2020**, *20*, 5422

Because the signal does not have a uniform distribution on the entire spectrum, the effective bandwidth β is defined

$$\beta = \sqrt{\frac{\int f^2 \left| \mathcal{S}(f) \right|^2 df}{\int \left| \mathcal{S}(f) \right|^2 df}} \tag{3}$$

Usually, the quality of the UWB pulse synthesis is defined in terms of occupied bandwidth and pulse duration. Table 1 compares the quality parameters of proposed pulse with P802.15.4z reference pulse.


**Table 1.** The synthesis quality parameters.

The minimum uncertainty, σ, for range estimation in the time of arrival method is quantified by Cramér–Rao lower bound (CRLB) [28] ,

$$
\sigma = \frac{c}{\beta \sqrt{8\pi^2 SNR}}\tag{4}
$$

where β is effective bandwidth, *c* is speed of the light, and *SNR* is signal to noise ratio. 

Relation (100) clearly shows that for minimizing the uncertainty the effective bandwidth has to be maximized, but within the limit of regulations.

#### **4. Timestamp Estimation in the 802.15.4z Standard**

#### *4.1. Timestamp Estimation by Random Spreading of STS*

= 8 chips

The new standard brings in a new physical protocol data unit (PPDU) structure by incorporating the STS for secure ranging. The STS is encrypted using the AES-128 algorithm, the time of arrival estimation is achieved on STS, and the range measurement is validated only if the received STS cross correlated with the locally generated reference exceeds the "match level" [29] threshold.

The default PHY frame format proposed by EiR [29] is depicted in Figure 3, where SHR is the synchronization header (preamble), STS is the scrambled timestamp sequence, and PHR is the PHY header.

**Figure 3.** PHY frame format in the 802.15.4z amendment.

 ∑ () ( − ) ఋಽିଵ ୀ () () = 2 To avoid the inter-pulse interferences, [30] stipulates that every component *B<sup>k</sup>* of Ipatov ternary symbol (ITS), or STS, is spread out on a symbol of length δ*<sup>L</sup>* by Pδ*L*−<sup>1</sup> *n*=0 δ(*n*) *p*(*t* − *nTch*); where δ(*n*) is Kronecker delta, *p*(*t*) is UWB pulse and *Tch* = 2 *ns* is chip duration. To mitigate the side lobes of correlation, this paper proposes a supplementary spreading by a randomly generated sequence *S<sup>k</sup>* . The sequence of length *N* is:

$$s(t) = \Sigma\_{k=0}^{N-1} B\_k \Sigma\_{n=0}^{\delta\_L - 1} (\delta(n) \cdot p(t - (k \cdot \delta\_L + S\_k \eta\_s + n)T\_{ch})).\tag{5}$$

<sup>ଷ</sup> ++1

= 128 = 1 − 2

≅ 10 ≅ 3.33

= 0

The STS sequence of length *N* = 128, *s*(*t*), is generated by taking *B<sup>k</sup>* = 1 − 2*A<sup>k</sup>* , A being the result of AES encryption. The peak pulse repetition frequency (PRF) is 499.2 MHz, the mean PRF is 62.4 MHz, resulting δ*<sup>L</sup>* = 8 chips. Figure 4a depicts the STS sequences: STS standard spread out is generated according to the standard specifications, *S<sup>k</sup>* = 0, STS reference signal is the signal needed to achieve correlation and for STS randomly spread out, the *S<sup>k</sup>* is generated based on a linear feedback shift register with the characteristic polynomial *x* <sup>3</sup> + *x* + 1. The cross-correlation of the reference signal and STSs are shown in Figure 4b. The ratio between the main lobe and maximum side lobe is η 10 for cross-correlation randomly spread out and η 3.33 for cross-correlation standard spread out, which proves the efficiency of random spreading.

**Figure 4.** Timestamp estimation by STS: (**a**) UWB pulse position in STS sequence; (**b**) cross-correlation of STS sequences with reference signal.

Figure 4a shows that the random spreading increases the risk of IASI. Therefore, it is necessary to perform the analysis of timestamp estimation for propagation in noisy channel and for multipath propagation.

#### (, ) *4.2. Sounding the Channel with Multipath Propagation*

 ௦ ଵ

(, ) = ೃೣ(ௗ,) = ଵ ்௫ோ௫ (/ ) షమ(ഉశభ) The channel sounding is the estimation of the channel impulse response (CIR) using preamble sequence, in order to remove the noise and design the channel equalizer.

ೣ() ଶ (ௗ/ௗబ) ோ௫(, ), ்௫() ்௫, ோ௫ The frequency-dependent path gain *G*(*d*, *f*) is modeled considering an isotropic radiation pattern, with the "antenna attenuation factor" [31] of <sup>1</sup> 2 :

$$\mathcal{G}(d,f) = \frac{P\_{\text{Rx}}(d,f)}{P\_{\text{Tx}}(f)} = \frac{1}{2} \left. \mathcal{G}\_0 \eta\_{\text{Tx}} \eta\_{\text{Rx}} \frac{(f/f\_c)^{-2(\kappa+1)}}{(d/d\_0)^n} \right. \tag{6}$$

=0 = 1.2 = 1.76 = 4.58 where *PRx*(*d*, *f*), *PTx*(*f*) are received and transmitted power, η*Tx*, η*Rx* are the transmission and reception antenna gains, *G*<sup>0</sup> is the path gain at reference distance *d*0, *d* is the distance between transmitter and receiver, *n* is the path gain exponent, *fc* is the carrier frequency, *f* is the frequency and κ is the frequency decaying factor.

ℎ () ℎ () = ∑ ∑ , (∅,)( − − ,) ୀ ୀ , ∅, The frequency decaying factor follows the Friis equation and, in Equation (6), it has the value κ = 0. The path loss varies from *n* = 1.2 in industrial LOS, *n* = 1.76 in outdoor LOS, to *n* = 4.58 in residential NLOS. It is noted that multipath propagation in the industrial environment leads to an increase in the path gain.

, Λ Using the Saleh-Valenzuela statistical model, the propagation paths are designed as the sum of clusters, every cluster having multiple rays. The impulse response *h <sup>m</sup>*(*t*) is

$$dh^m(t) = \sum\_{l=0}^L \Sigma\_{r=0}^K a\_{r,l} \exp\left(j\varpi\_{r,l}\right) \delta\left(t - T\_l - \tau\_{r,l}\right),\tag{7}$$

ቁቁ (− <sup>ఛ</sup>ೖ,

ఊభ )

ఊೞ

௨௦௧

{,} ∝ ቀ1 − ቀ− <sup>ఛ</sup>ೖ,

where *ar*,*<sup>l</sup>* is the tap weight of *<sup>r</sup>* ray in cluster *<sup>l</sup>*, ∅*r*,*<sup>l</sup>* is the ray phase, *T<sup>l</sup>* is the delay of *l* cluster and τ*r*,*<sup>l</sup>* is the delay of *r* ray relative to cluster *l* front. The intervals between the time arrivals of the clusters is modeled as Poisson process with the arrival rate Λ, and the ray delays inside the cluster are modeled as a mixture of two Poisson processes with arrival rates λ1, λ<sup>2</sup> and mixing weight β. The mean power of arriving clusters follows an exponential decay with time constant Γ, having a normal distribution around the mean value σ*cluster*, and the cluster shape also bears to an exponential decay with time constant γ.

In the above CIR model, the first path has the highest energy. In non-line-of-sight propagation, there are cases when the first path is strongly attenuated. For such situations, [31] proposes a new modeling for the first path.

$$E\{a\_{k,0}\} \propto \left(1 - \chi \exp\left(-\frac{\tau\_{k,l}}{\mathcal{I}\_{r\text{rise}}}\right)\right) \exp\left(-\frac{\tau\_{k,l}}{\mathcal{I}\_1}\right),\tag{8}$$

where χ describes the attenuation of the first path, γ*rise* determines how fast the power delay profile (PDP) increases, and γ<sup>1</sup> determines the profile decay. By joining Equations (7) and (8), the CIR, *h*(*t*) can be found.

The preamble sequence consists of a string of 32 or 64 Ipatov ternary symbols, every symbol having 91 elements with 81 non-zero elements [30]. An Ipatov symbol has "perfect" periodic autocorrelation, i.e., all side lobes of autocorrelation are zero, and using Wiener-Hopf equation, by cross-correlation of the received signal *y*(*t*) with the input Ipatov sequence *I*(*t*) results immediately the CIR, *h*(*t*)

$$\ast\_{\bullet} y(t) \star I(t) = h(t) \ast I(t) \star I(t) = h(t) \ast \mathbb{R}\_{I,I}(t), \tag{9}$$

where autocorrelation of ITS is: *RI*,*I*(0) = N;*RI*,*I*(*t*) = 0 for *t* , 0.

For outdoor, NLOS and multipath propagation environment, the channel PDP, *h*(*t*), is modeled based on Equations (7) and (8), with the parameters retrieved from [31], and synthesized in Table 2.


**Table 2.** The UWB channel parameters for outdoor NLOS propagation.

For channel sounding, the EiR standard specifies a mandatory preamble sequence of 32 or 64 ITSs, the Ipatov symbol having a number of 91 elements with 10 zero elements. In this paper, the CIR is estimated only using one ITS with a total of 57 elements including 8 zero elements [32]. The transmitted sequence of pulses, *i*(*t*), is generated based on Equation (5), with *B<sup>k</sup>* = *I<sup>k</sup>* , where *I<sup>k</sup>* is the *k*th element of ITS. The received signal *yi*(*t*) = *h*(*t*) ∗ *i*(*t*) is the convolution of PDP with the emitted sequence. The CIR estimation, ˆ *h*(*t*), is achieved by the cross-correlation of the received signal *y*(*t*) with the Ipatov symbol *I*. Figure 5 shows that the CIR estimation, estimated PDP, is close enough to the true PDP. By successive transmission of ITSs contained in the preamble, the estimations are accumulated and averaged, resulting in SNR reduction and CIR estimation improvement.

()

ℎ()

ℎ()

() = ℎ() ∗ ()

() ()

 

ത Λ Γ ௦

() =

−

ℎ(),

() ⋆ () = ℎ() ∗ () ⋆ () = ℎ() ∗ ூ,ூ()

ூ,ூ(0) = N; ூ,ூ() = 0 for ≠ 0

**Figure 5.** CIR estimation for channel with multipath propagation.

This CIR estimate will be used in the following paragraphs, for analysis of timestamp estimation in channels with multipath propagation. This is a classical channel model, more detailed models are being published in the recent research [33,34].

#### *4.3. Timestamp Estimation in Channel with NLOS and Multipath Propagation*

The EiR specifies that STS consists of 32, 64, 128 AES-128 sequences, successively transmitted, but in this section only one AES sequence is considered for analysis the impact of multipath propagation.

The STS is generated based on Equation (5) with random spread out, and for propagation simulation it is convoluted with *h*(*t*).

In the usual way, the received signal is passed through an equalizer filter and cross-corelated with the STS reference *sr*(*t*). In this article, an easier way for timestamp detection is proposed, that is, to generate a virtual propagated STS reference by convolution of the STS reference with the estimated CIR, ˆ *h*(*t*), and cross-correlation of received sequence *ys*(*t*) with locally generated replicas ˆ *yr*(*t*) as depicted in Algorithm 1.


Figure 6 shows that by cross-correlating the received STS, *ys*(*t*), with the STS reference, *sr*(*t*), the timestamp is not detectable, (see the correlation with STS reference trace), and that the cross-correlation of received signal with the locally generated replica, ˆ *yr*(*t*), the side lobes are strongly attenuated (see the correlation with STS reference convoluted with PDP trace).

()

**Figure 6.** Timestamp estimation in channel with NLOS and multipath propagation.

#### *4.4. Timestamp Estimation in Noisy Channel with NLOS and Multipath Propagation* () <sup>௦</sup> ();

ℎ()

ℎ() <sup>௦</sup>

()

() <sup>௦</sup>

()

௦

()

() = ℎ() ∗ ()

() = ℎ() ∗ ()

௦() = () ⋆ <sup>௦</sup>

௦

()

() ()

= 0.0228 V

()

()

௦()

The EiR standard specifies that the typical range of radio is 100 m. For timestamp estimation in noisy channel we consider that the transmitter is situated at 22 m and the noise source is situated at reference distance of 1 m. () = ℎ() ∗ () ∈ [0, − 1] :

− The transmitter emits with maximum compliant power (Figure 2a) and the noise source emits with floor noise level (i.e., −41 dBm/MHz) [35]. In order to mitigate the effect of noise, consider that the receiver has a 1 GHz passband filter on the input. The standard deviation of noise is σ = 0.0228 V and the signal to noise ratio on the receiver is *SNR* = −38 dB. In this situation, the timestamp is undetectable from only a single AES-128 sequence (Figure 7b, trace 1 STS), so STS will consist of multiple AES sequences as the EiR standard specifies. <sup>௦</sup> () = ℎ() ∗ () + () ௦ () = 1 <sup>௦</sup> () ேିଵ ୀ ௦() = () ⋆ <sup>௦</sup> () ()

The timestamp detection is detailed in Algorithm 2 and the results are shown in Figure 7. ௦()

**Figure 7.** Timestamp estimation in noisy channel with multipath propagation.

**Algorithm 2**: Timestamp estimation in noisy channel

**Inputs**: STS reference (AES-128 sequence), *sr*(*t*); Received STSs, *y i s* (*t*); Number of STS, *N*;

12 dBm = 0.13 () () () exp (2 1. Generate locally replica of STS: <sup>ˆ</sup> *<sup>y</sup>r*(*t*) <sup>=</sup> <sup>ˆ</sup> *<sup>h</sup>*(*t*) <sup>∗</sup>*sr*(*t*) 2. Compute the received sequences: *f or i* ∈ [0, *N* − 1] *do* : *y i s* (*t*) = *h*(*t*) ∗*s*(*t*) + *n i* (*t*) 3. Average the received sequences: *ys*(*t*) = <sup>1</sup> *N* P*N*−<sup>1</sup> *i*=0 *y i s* (*t*) 4. Timestamp estimation: *rrs*(*t*) = ˆ *yr*(*t*) ⋆ *ys*(*t*) Where *n i* (*t*) is the noise with standard deviation σ **Output**: return *rrs*(*t*)

#### *4.5. UWB Channel with Frequency Selective Fading*

Depending on antenna design, or antenna position relative to external objects, it is possible to encounter frequency selective fading [35,36]. To analyze the impact of frequency fading in timestamp detection, consider UWB channel 9 with the spectral mask profile power spectral density, illustrated in Figure 8, having a selective fading of 12 dBm at carrier frequency and frequency decaying factor of κ = 0.13. The channel frequency response, *H*(*f*), is achieved by Kolmogorov factorization (Appendix A).

**Figure 8.** Psd mask and channel response.

() (); () = ଶగ<sup>௧</sup> ∙ () To estimate the CIR, *H* ˆ (*f*), the Ipatov symbol, *i*(*t*), is shifted on to the carrier, *fc*, by complex modulation with exp(*j*2π*fct*). The cross-correlation is performed in the frequency domain, and to avoid the circular correlation, the series is padded with zeros. The propagation of STS in the faded channel is simulated by the shifting of the STS on to the carrier and by the convolution with *H*(*f*). The timestamp estimation is detailed in Algorithm 3.


Figure 9a, psd at the emission, shows that Ipatov symbol has a uniform distribution over the entire bandwidth, that it follows the compliant spectral mask (Figure 2a), and that after propagation, psd after fading, it borrows CIR spectral mask.

**Figure 9.** Timestamp estimation in channel with selective fading.

The propagation in a channel with frequency fading leads to a small error in timestamp estimation, as shown in Figure 9b, correlation with STS reference, and this error is cancelled if the received STS is correlated with the locally generated replica, correlation with STS convolved with H.

It should be highlighted that the spectral leakage in the FFT leads to the introduction of nonuniformity in the estimated CIR spectrum, Figure 10a, resulting in incorrect results regarding timestamp estimation, Figure 10b.

estimation of timestamp.

#### **5. Results**

NA**Figure 10.** Spectral leakage in timstamp estimation: (**a**) Nonuniformity in CIR estimation; (**b**) Erroneous The NESP shows the usage efficiency of the allocated bandwidth. Table 3 displays the occupied bandwidth and NESP for the pulse synthesized from entire spectral mask which is compared with previously published related works.

**β**

NA

**Table 3.** UWB pulse parameters.


The uncertainty in range estimation for noisy channel with multipath propagation is computed based on Equation (4). The effective bandwidth is computed based on Equation (3). The leading lobe in Figure 7b is extracted, the samples series is completed with zero and then transformed in frequency domain. The effective bandwidth is nearly constant β = 234 MHz.

The propagation loss is determined considering outdoor LOS channel with path loss exponent *n* = 1.76,

$$L = 10 \cdot n \cdot \log\left(\frac{d}{d\_o}\right) \tag{10}$$

where *d*<sup>0</sup> = 1 m is reference distance.

The noise has psd at floor level *P<sup>n</sup>* − 41 dBm and 7 GHz bandwidth.

The SNR and minimum uncertainty are depicted in Table 4.

**Table 4.** CRLB for timestamp estimation by STS in noisy channels and multipath propagation.


#### **6. Discussion**

The synthesis of UWB impulse from a complete psd mask (Figure 2a) results in a risetime reduction, amplitude increasing, and the attenuation of side lobes relative to the impulse proposed by EiR (Table 1). Furthermore, the pulse psd fully complies with the standard spectral mask. Therefore, this impulse is advisable as a UWB reference pulse.

The random spreading of STS leads to an attenuation of lateral lobes of cross-correlation, but also leads to an increase in the probability of intra-symbol interferences. Thus, more investigations should be performed for testing these interferences in noisy channels with multipath propagation.

The cross-correlation in a noisy channel with multipath propagation displays a wide main lobe, which leads to a decrease in the accuracy of the timestamp estimation. This result seems to be due to the fact that the AES-120 sequence does not have uniform spectral distribution in the entire frequency band. The whitening of the AES sequence or of the entire STS by additional randomization could lead to the main lobe narrowing, and this is a subject of interest in our research.

**Author Contributions:** Conceptualization, I.D. and T.P.P.; methodology, E.P.; software, I.D.; investigation and writing, I.D., E.P. and A.P.; visualization, A.P.; supervision, T.P.P.; funding acquisition, T.P.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** This study was accomplished within the Centre of competence for wireless intrasatellite technologies (IntraSAT-Tech), Technical University of Cluj-Napoca.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A**

Kolmogorov Factorization


#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*
