Switched Antenna Array for Maritime Communications

FACULDADE DE ENGENHARIA DA UNIVERSIDADE DO PORTO Switched Antenna Array for Maritime Communications Diogo Durães Moreira MASTER IN ELECTRICAL AND COMPUTER ENGINEERING Supervisor: Prof. Dr. Henrique Manuel de Castro Faria Salgado Co-supervisor: M. Eng. Hugo Miguel Guedes Pereira dos Santos July 15, 2018 c Diogo Moreira, 2017 Switched Antenna Array for Maritime Communications Diogo Durães Moreira MASTER IN ELECTRICAL AND COMPUTER ENGINEERING July 15, 2018 Abstract In this dissertation work, different antenna topologies are evaluated to find the best design capable of overcoming the difficulties of communications at high seas. The impact of multiple design variables is studied, and their effects explained resorting to theoretical principles. Furthermore, a switching matrix of eight sectors is also designed, so the antenna beam direction can be automatically reconfigured with the movement of the vessel. The antenna and both the feeding and switching networks were designed and validated resorting to Finite Element Method in HFSS and the Method of Moments in ADS, respectively. The final antenna element experimental results show a fractional impedance bandwidth of 22.3%, an isolation between inputs above 45 dB and a gain of 7.5 dBi at the centre frequency of 5.15 GHz. The measured results of the switched antenna array show that the horizontal polarization is matched between 4.4 and 5GHz, and the vertical polarization between 4.4 and 5.47GHz. Finally, the experimental measurements were compared with the simulated results of the unit cell and the entire system. Keywords: Antennas, Dual-polarization, MIMO, Parasitic Patch, Patch, Switched Antenna, WiFi. i ii Acknowledgements Hoping that no one is left behind, I would like to give special thanks to Professor Henrique Salgado for giving me the opportunity of working in this project and for supervising this dissertation. M. Eng. Hugo Santos for his guidance and support, fundamental to complete this work, and for his friendship, that was extremely important to overcome the difficulties that were faced. All my work colleagues, especially Joana, Erick and Ricardo, for providing a good work environment and for the many funny moments during lunch time. INESC TEC for providing the needed technical resources. MareCom project for providing the financial resources. My family, especially to my parents, Helena and Paulo, for giving me the opportunity of pursue my dreams. My brother, Tiago, for supporting me and for being with me in the most difficult moments. Daniela, Filipa and Rita, for making me believe in myself and for their unconditional friendship. Luís Costa, for the unconditional support. All my master’s colleagues and friends, specially to Bruno Correia, Diogo Fonseca, Eduardo Rodrigues and Rafael Kraemer, for their friendship and for having so many moments of fun during this period. Diogo Durães Moreira iii iv “God runs electromagnetics on Monday, Wednesday, and Friday by the wave theory, and the devil runs it by quantum theory on Tuesday, Thursday, and Saturday. ” Lawrence Bragg v xii LIST OF FIGURES 4.12 Simulated insertion loss of the microstrip and CBCPW lines, tuned to a characteristic impedance of 50Ω. .............................. 70 4.13 HFSS model of the edge mount connector and its land pattern. . . . . . . . . . . 71 4.14 Simulated reflection coefficient of the edge mount connector for different values of the microstrip line width. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 4.15 HFSS model of the edge mount connector and its land pattern with an additional linesection...................................... 72 4.16 Simulated reflection coefficient of the edge mount connector for different values of the extra microstrip line width. . . . . . . . . . . . . . . . . . . . . . . . . . . 73 4.17 Simulated insertion loss of the edge mount connector for an extra microstrip line width of 0.7mm. .................................. 73 4.18 Provided S-parameters for the capacitor 04023J2R2BBSTR. . . . . . . . . . . . 74 4.19 Layout of the footprint for the switch IC. . . . . . . . . . . . . . . . . . . . . . . 74 4.20 Simulation result of the impedance seen from the grounding pad. . . . . . . . . . 75 4.21 Routing of the digital and supply pins of the switch IC. . . . . . . . . . . . . . . 76 4.22 Schematic of the optimization setup employed in ADS for a single antenna. . . . 77 4.23 Mean value of the reflection coefficient of the H-pol switching matrix, with a confidenceintervalof95%.............................. 78 4.24 Mean value of the reflection coefficient of the V-pol switching matrix, with a confidenceintervalof95%................................ 79 4.25 Final layout of the H-pol switching matrix. . . . . . . . . . . . . . . . . . . . . . 79 4.26 Final layout of the V-pol switching matrix. . . . . . . . . . . . . . . . . . . . . . 80 5.1 Manufactured MIMO stacked parasitic patch antenna with differential probe feed. 81 5.2 Setup used for the measurement of the antenna unit cell. . . . . . . . . . . . . . 82 5.3 Horizontal polarization reflection coefficient as a function of frequency. . . . . . 83 5.4 Vertical polarization reflection coefficient as a function of frequency. . . . . . . . 83 5.5 Horizontal polarization reflection coefficient as a function of frequency for five differentantennas................................... 84 5.6 Vertical polarization reflection coefficient as a function of frequency for five differentantennas.................................... 84 5.7 Isolation between polarizations as a function of frequency. . . . . . . . . . . . . 85 5.8 Simulated (line) and measured (circles) co-polarization E-plane and H-plane radiation patterns, with horizontal polarization excited. . . . . . . . . . . . . . . . . 85 5.9 Simulated (line) and measured (circles) co-polarization E-plane and H-plane radiation patterns, with vertical polarization excited. . . . . . . . . . . . . . . . . . . 86 5.10 Manufactured switched antenna array. . . . . . . . . . . . . . . . . . . . . . . . 86 5.11 Setup used for the measurement of the switched antenna array. . . . . . . . . . . 87 5.12 Mean value of the reflection coefficient of the H-pol switching matrix, with a confidenceintervalof95%.............................. 88 5.13 Mean value of the reflection coefficient of the V-pol switching matrix, with a confidenceintervalof95%................................ 88 List of Tables 3.1 Final dimensions of the patch antenna with ILA probe feed. . . . . . . . . . . . . 34 3.2 Final dimensions of the stacked parasitic patch antenna with probe feed. . . . . . 44 3.3 Dimensions of the NTL transformer obtained through a random optimization process.......................................... 58 3.4 Final dimensions of the NTL transformers obtained through an optimization process. 59 3.5 Power measured at each input port of the antenna when a power of 0dBm is injected at the input of the feeding network. . . . . . . . . . . . . . . . . . . . . . 59 4.1 Analysis of the total effort of using each switching topology. . . . . . . . . . . . 66 4.2 Physical characteristics of Rogers4003c and FR4 substrates. . . . . . . . . . . . 67 4.3 Dimensions fo the NTL sections in Rogers and FR4. . . . . . . . . . . . . . . . 68 4.4 Truth table of the switch IC. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 4.5 Final dimensions of the NTL transformers obtained through an optimization process. 78 xiii xiv LIST OF TABLES Abbreviations and symbols ACP Aperture-Coupled Patch AF Array Factor DGS Defected Ground Structure FEM Finite Element Method FNBW First-Null Beam Width FTBR Front To Back Ratio HPBW Half-Power Beam Width IoT Internet of Things ILA Inverted L Antenna LHCP Left-Hand Circularly Polarized MIMO Multiple Input and Multiple Output RHCP Right-Hand Circularly Polarized SLL Side-Lobe-Level TE Transverse Electric TEM Transverse Electromagnetic TM Transverse Magnetic εElectrical permittivity µMagnetic permeability λWavelength CCapacitance DDirectivity eEfficiency fFrequency GGain ICurrent LInductance PPower RResistance vVelocity xv Chapter 1 Introduction This introductory chapter presents the motivation for this thesis as well as its objectives. In addition, the structure of the document is also presented. 1.1 Motivation Currently, the exploration of ocean is a main goal for both Portugal and the European Union. In addition to modern activities, such as deep sea mining and environmental monitoring with autonomous vehicles, the traditional activities, like fishing, maritime transport, rescue operations and tourism, are considered significant in marine environment. The support of those activities requires the use of long range wireless communications to serve mobile and fixed floating platforms located in remote areas of the sea, that typically are out of reach of terrestrial networks. On the other hand, with the convergence to an all-IP and IoT (Internet of Things) paradigm, the access to terrestrial private IP networks and even to the global Internet will bring an increased value to all the referred activities. This way, humans and systems can have access to applications and IP services available on land as well as new applications and services more suitable for maritime activities. However, communications at sea are limited to the use of HF/VHF, which in many cases are still analogue and only with voice support. Another alternative for communicating when in remote ocean locations is satellite links, which are expensive to maintain and suffer from narrow channel bandwidth. For example, Iridium, a satellite operator, has a service with a monthly signature of 1700 $, 1 GB of traffic and a bit rate of 126 kbit/s, being that the cost of a receiver equipment is between 4999 $ and 17995 $. Wideband services, such as GPRS/UMTS/LTE and Wi-Fi, are only available close to shore. To answer this problem, the main goal of the MareCom project is to develop an alternative communication system with high availability, wideband and low cost, based on IP protocol and off-the-shelf hardware combined with new antennas. This will permit to serve communities operating in marine environment and the convergence with communication systems already used in terrestrial environment. To do so, it is necessary to overcome the challenges imposed by the new 1 2Introduction environment, namely the development of antenna technologies which can cope with the rapidly changing channel due to vessel oscillations and mobility. When dealing with moving vessels, the use of omnidirectional antennas is usually preferred. However, this has an impact on the maximum transmission distance since these antennas are characterised by a low gain. For this reason, the use of directional antennas is more desirable, yet since the oscillations of the boat may lead to those antennas pointing in the wrong direction, intelligent solutions to overcome this limitation are addressed in this work. Signal attenuation and multipath originated at the sea surface is also an issue to consider. With the aim of increasing the reliability of the communication link and overcoming the aforementioned impairments, polarization diversity MIMO is proposed. To that aim, a dual polarized antenna with an adequate HPBW (Half-Power Beamwidth) and isolation can be used, so that MIMO processing can be made, resulting in a more reliable channel. With MareCom technology, it would be possible to use boats and other floating platforms as access points to the terrestrial networks, as seen in Figure 1.1. Figure 1.1: Illustration of MareCom concept 1.2 Objectives This thesis is aligned with the goals of the MareCom project, and its objective is to develop a new antenna topology capable of overcoming the difficulties of maritime RF communications. Therefore, the antenna must fulfil the following characteristics: 1. Wide bandwidth, from 4.4 GHz to 5.9 GHz, so it can transmit in the Wi-Fi band and also in a NATO licensed band. 2. An half power beam width greater than 45◦in the horizontal plane, so the oscillations of the vessel do not break the communication link. 3. A gain as high as possible that is able to fulfil the HPBW restriction. Therefore, the radiation and mismatch efficiencies shall be maximized to comply with gain and HPBW restrictions. 4. Support MIMO communications with polarization diversity over the same antenna structure, reducing the occupation of the electromagnetic spectrum. 1.3 Contributions 3 5. Automatic reconfiguration of the beam direction, allowing to direct all the power to the direction of interest. 1.3 Contributions In the course of this work several technical challenges had to be overcomed that led to relevant contributions in the area of antenna system design, which can be summarized as follows: 1. Development of a novel two stackup parasitic dual-polarization patch antenna with differential probe feed and high isolation. 2. Selection of the best antenna parameters based on an analysis of modal significance. 3. Development of equivalent circuit models of stacked patch antenna with differential probe feed. 4. Impedance matching of the balun to the antenna using multiple sections of non-uniform transmission lines through an optimization procedure. 5. Solution for overcoming the orientation of the vessel based on multiple antenna selection driven by a 1 by 8 switching matrix. 1.4 Document Structure This report is divided in six chapters. This chapter presents the motivation for the research, its objectives, its main contributions in the area of antenna design and the structure of the document. Chapter 2provides the fundamental concepts on antennas and arrays theory and a related work review on other researches. Chapters 3and 4present a detailed description of the simulation procedure performed to develop the antenna unit element and the switching matrix, respectively. Experimental procedure and results are detailed in Chapter 5. Furthermore, a comparison between these and the simulation results is also presented and discussed. Finally, the most relevant conclusions reached through this research are shown in Chapter 6. Additionally some guidelines and recommendations for future developments on the same topic will be given in this chapter. 4Introduction Chapter 2 State-of-the-Art In this chapter the fundamentals on antennas are given and the special cases of patches and arrays are specially addressed. Furthermore, recent work related with the thesis to be developed is also analysed in this chapter. 2.1 Antenna Fundamentals An antenna element is a transitional structure between free-space and a guided device. In other words, it converts electrical energy in free-space electromagnetic waves and the other way around. As a transducer, it is defined by certain parameters, such as directivity, gain, efficiency, among others. In this section those parameters and the general basis on antenna theory are described. The presented contents follow the work of Balanis, in which the fundamental parameters of antennas are addressed, [1]. 2.1.1 Radiation Mechanism Radiation occurs from the movement of electrical charges in a material. For this reason, the most suitable type of material for an antenna is a conductor. Radiation may occur from a single wire, a pair of wires or a wire in proximity with ground, which can be seen as two wires if one resorts to image theory. According to Balanis, it only occurs when there is a time-varying current or an acceleration in the electrical charges in the conductor, [1, p. 9]. For a very thin wire with infinite conductivity and surface current I, the following expression can be written: dI dt=qdv dt(2.1) where qis the unit charge and vis the charge velocity. From this equation, it is clear that a timevarying current can be obtained by achieving a velocity shift along the wire, creating radiation by this mean. Such can be accomplished by curving the wire, introducing a discontinuity in its diameter, making abrupt bends, placing it in proximity with ground or truncating it, as shown in 5 12 State-of-the-Art 2.1.6 Polarization The polarization of an antenna has to do with how the electric field magnitude varies over time for a given point in the far field space. This parameter can be divided in three types: linear, circular or elliptical polarization. If the vector that describes the electric field of the antenna in a point of the far field over time is always directed along the same line, the polarization is called linear. Usually, it is classified as horizontal or vertical, depending on the axis on which it is aligned. Although it is the most simple to implement, it is the one that faces more adversities, such as polarization mismatch. That is, if the two antennas are misaligned, the power budget of the link will suffer a penalty by the factor of cos(ψ), where ψis the angle between the two antennas. Circularly and elliptically polarized antennas are very similar, since in both cases, the electric field magnitude vector rotates along an axis with the course of time. This can be seen as the superposition of two linearly polarized out-phased sources. The difference between them is that in circular polarized antennas, the electric field magnitude is always constant, while in the elliptically polarized case, it has an higher magnitude over an axis than the other. The best outcome of a circularly polarized antenna is the immunity to misalignment between the transmitter and the receiver, as they are able to receive electric fields which vector is pointing in any direction. Furthermore, an antenna that receives right-hand circularly polarized waves (RHCP) doesn’t receive left-hand polarized waves (LHCP). Since the polarization of a wave shifts from RHCP to LHCP when it is reflected, they will not be received by the antenna, therefore, there is less multipath interference. In Figure 2.6 a circularly polarized wave is represented, since its E-field rotates in clockwise, it is RHCP, otherwise it would be LHCP. ωt ωt 0 0 2π π 6π 6π 4π 2π x x y Figure 2.6: Circularly polarized wave, [1, p. 67]. 2.2 Patch Antennas 13 2.2 Patch Antennas Patch antennas are of special interest on the work under development, either for their reduced dimensions, low manufacturing cost and the simplicity of implementing dual polarization over a single patch structure. For this reason, these antennas are specially addressed in this section, following the work of Balanis in [2]. 2.2.1 Patch Antenna Structure These antennas are commonly composed of three layers. The bottom layer is the ground plane, a thin metallic sheet. On the top of this, is placed a non-conducting substrate, which is used as a stable support for the patch and also to reduce its size, if a high permitivity material is used, [2, p. 160]. However, there is a trade off between the size of the patch and the antenna’s radiation efficiency, since a high permittivity substrate confines the electric field inside it, increasing the dielectric losses and reducing the air-side radiation. The radiating element is printed over these two layers, and is usually made by thin copper foil that can have different shapes (rectangular, circular, etc.). The stack-up of a patch antenna can be seen in 2.7. A patch antenna is a resonant structure, therefore it can be modelled as an RLC circuit. However, it is characterised by a narrow bandwidth, which is a disadvantage in the context of this thesis. Ground plane Substrate L W r Patch h Figure 2.7: Example of a patch antenna structure, [1, p. 786]. 2.2.2 Radiation Mechanism In order to design a microstrip antenna, it is necessary to have methods of analysis, in order to better understand the system and compute initial values for the antenna CAD design software. Then, these dimensions can be optimized to achieve the desired input impedance, bandwidth, radiation patterns, etc. The two most common methods of analysing a patch antenna are the transmission-line circuit model and the multimode cavity model. 14 State-of-the-Art 2.2.2.1 Transmission-Line Circuit Model Balanis states that a patch, operating at its fundamental mode, is basically a λ 2-long microstrip line with a low characteristic impedance, thus, it can be represented by its equivalent circuit. This way, a rectangular patch antenna can be seen as a long microstrip line with two open edges, where most of the radiation is generated. This way, the patch can be described by the equivalent circuit of Figure 2.8, where Gand Bare the conductance and the susceptance that characterize de slot aperture. Although this model is simple, intuitive and computationally fast, its accuracy is very limited, since it doesn’t account the radiation from the non radiating edges of the patch and doesn’t consider mutual coupling between the two radiating slots. Furthermore, with this model it is easy to analyse rectangular patches, but for other shapes, it is less accurate. Figure 2.8: Equivalent circuit of a microstrip patch antenna, [2, p. 162]. By using this method it is possible to calculate the values of the width, W, and the length, L, of the patch antenna using Equations (2.15) and (2.16). In these equations, the effective permittivity constant of the dielectric for the operation frequency and the effect of the fringing fields on the length of the patch are already taken into account, [1, pp. 789-791]. W=1 2fr√µoεor2 εr+1=vo 2fr 2 εr+1(2.15) L=1 2fr√εre f f √µoεo−2∆L(2.16) where fris the resonant frequency of the dominant mode TM010, given by Equation (2.17), εre f f is the effective dielectric constant at the operating frequency, given by Equation (2.18), and ∆Lis the patch length extension, caused by the fringing fields, given by Equation (2.19). fr=1 2L√εr√µoεo =vo 2L√εr (2.17) εre f f =εr+1 2+εr−1 21+12 h W−1 2 (2.18) ∆L h=0.412(εre f f +0.3)W h+0.265 (εre f f −0.258)W h+0.8(2.19) 2.2 Patch Antennas 15 2.2.2.2 Multimode Cavity Model As stated in section 2.1.1, a microstrip patch antenna can seen as a waveguide radiator if one considers a cavity bounded by the patch itself and the ground plane, with magnetic walls along the perimeter of the patch, to simulate an open circuit, [1, p. 798]. This cavity supports multiple discrete modes. At the open ends, the current distribution in the patch is null, since it is a very high impedance point, therefore, the magnitude of the voltage is maximum at those points, as can be perceived from Figure 2.9. On those edges, the fringing E-fields add up in phase, producing radiation. On the other hand, in the middle of the patch, the E-field is almost null, which enables one to use two orthogonal polarizations over the same structure, without much cross-talk between the two probes, [2, p. 162]. Balanis concludes that the resonant frequency of the cavity is given by Equation (2.20), [1, p. 803]. From that he also concludes that the dominant mode for all microstrip antennas is the TM010 with a resonant frequency given by Equation (2.21). However, these considerations are valid only for antennas with L>W>h. (fr)mnp =1 2π√µε rmπ h2+nπ L2+pπ W2 (2.20) (fr)010 =1 2L√µε =vo 2L√εr (2.21) Feed y x Figure 2.9: Fundamental mode electric field configuration underneath a rectangular patch, [2, p. 163]. 2.2.3 Feeding Methods Microstrip antennas can be fed in many different configurations, however, the four most common are the microstrip line, coaxial probe, aperture coupling and proximity coupling, as shown in Figure 2.10, [1, p. 785]. 16 State-of-the-Art Ground plane (a) Microstrip line feed Substrate L W r Patch h (b) Probe feed Circular microstrip patch Dielectric substrate r Ground plane Coaxial connector r1 r2 (c) Aperture-coupled feed Patch Slot Microstrip line r1 r2 (d) Proximity-coupled feed Patch Microstrip line Figure 2.10: Typical feeds for microstrip antennas, [1, p. 786]. (a) Microstrip line (b) Probe (c) Aperture-coupled (d) Proximity-coupled Figure 2.11: Equivalent circuits for the typical feeds of microstrip antennas, [1, p. 787]. 2.2.3.1 Microstrip Line This feeding method consists of connecting the patch directly to the input port using a microstrip line. The main advantages of this design is to be easy to manufacture and the matching can be easily done by controlling the inset position of the line. Despite that the feed can be easily 2.2 Patch Antennas 17 modulated as a simple inductance, as shown in Figure 2.11, this configuration has a great inconvenient on the scope of this project. As the substrate thickness increases, the surface waves and spurious feed radiation also increase, which limits the bandwidth to 2 −5%, [1, p. 786], making this configuration the least interesting. 2.2.3.2 Coaxial Probe This feeding technique is very similar to the microstrip line, as it is even modulated by the same equivalent circuit. In this case, the outer conductor of the coax is attached to the ground plane, while the inner conductor crosses it and connects directly to the patch. Adjusting its position a good match can also be achieved. In this work, the coax probe feed is advantageous, since it has a low spurious noise, it can achieve a slightly bigger bandwidth. However, if the substrate thickness increases much (h> 0.02λo), it can no longer be modulated as an inductor. Additionally, since it has some asymmetry, higher order modes can be excited, which can produce cross-polarized radiation, thus compromising the isolation between the two feeds. 2.2.3.3 Aperture-Coupled The aperture-coupled feed uses two substrates, separated by a ground plane. In the lower substrate, a microstrip line is coupled to the patch, placed on the other substrate, through a slot on the ground plane. In this case, the system has three resonant frequencies, one for the patch, one for the slot and another for the set patch-slot. This can be perceived from Figure 2.11, where the feed is represented by another RLC circuit. This way, the bandwidth of the patch can be increased, optimizing the feed line width, the slot dimensions and its position. However, to manufacture a MIMO antenna with this method, multiple substrate layers could be necessary, which would make the design more difficult and more expensive. Additionally, the radiation generated in the feeding slot could originate a back lobe on the radiation pattern, decreasing the gain of the antenna in the desired direction. 2.2.3.4 Proximity Coupling This feed method consists in using two stacked substrates with a common ground plane. In the first substrate layer is printed a microstrip feeding line, while in the second is printed the patch antenna. This method is characterized by the largest bandwidth of all the described methods (up to 13), [1, p. 787], which would be favourable. Meanwhile, the impedance matching is obtained by adjusting the width and length of the feed line, which could lead to an undesired approximation between the two feeding lines that could compromise the isolation between ports. 18 State-of-the-Art 2.3 Antenna Array Theory Patch antennas are characterized with a relatively low gain. For this reason, ways of increasing the gain of those antennas must be studied. In this section the array fundamentals are exploited, as a solution for this problem. An array of antennas is the organization of the single antenna elements in an electrical and geometrical configuration. This way, the total E-field of the array is result of the sum of the field vectors of each element. To achieve the desired gain, the field of the elements has to be in phase in the desired direction and to cancel each other in the other directions. To do that, there are many variables, that can be adjusted, such as the geometrical configuration of the array (linear, circular, etc.), the distance between the elements, the amplitude and phase of the excitation currents and the patterns of the individual elements, [1, p. 285]. The easiest way of forming an array is to place equal elements along a line, adjusting only the distance between them. This simple case of uniform amplitude and spacing is the one discussed below. A simple way of determining the field pattern of an array is through its AF (Array Factor), which is a function of the geometry of the array and its excitation phase. For an array of N uniform and isotropic elements, the array factor is given by AF (θ) = N ∑ n=1 ej(n−1)ψ(θ)(2.22) with ψ(θ) = kd cosθ+β(2.23) where kis the wave number, dthe distance between elements and βthe current phase shift between consecutive elements. Therefore, varying dand βleads to a change in the AF, hence, in the overall radiation pattern. Since the AF considers isotropic elements, the radiation pattern of the array is given by the product of the AF and the pattern of the unit element as described in Equation (2.24). This effect is illustrated in Figure 2.12. Earray =E(single element at ref erence point)×AF (θ)(2.24) With some algebric manipulation, the AF can also be written in its compact form as seen in Equation (2.25), [2, p. 535]. Considering only the numerator of this expression it is possible to compute the nulls, maximums and the −3 dB point of the AF, as depicted in expressions 2.26, 2.27 and 2.28. AF ="sinN 2ψ sin1 2ψ#(2.25) 2.4 PIN Diodes and Control Circuits 19  180 150 120 90 0.2 0.4 0.6 0.8 1 60 30 0 Element Factor –30 –60 –90 –120 –150 180 150 120 90 0.2 0.4 0.6 0.8 1 60 30 0 Array Factor –30 –60 –90 –120 –150 180 150 120 90 0.2 0.4 0.6 0.8 1 60 30 0 Total –30 –60 –90 –120 –150 Normalized Power Normalized Power Normalized Power Figure 2.12: Example of the total array pattern calculation, [2, p. 535]. θn=cos−1λ 2πd−β±2n Nπ n=1,2,3,... n6=N,2N,3N,... (2.26) θm=cos−1λβ 2πd(2.27) θh=π 2−λ 2πd−β±2.782 N (2.28) These results can be used to adjust the radiation pattern of the array, changing the directivity of the antenna system. In [2, p. 565], expression 2.29 is given to estimate the gain increment as a function of the number of elements in the array. However, this result does not take into account the mutual coupling between the array elements, which is important when dimensioning an array antenna. When two elements are near each other, part of the energy radiated by one of them can be received by the other. This energy can be rescattered in different directions, behaving as a second transmission and changing the radiation pattern, thus the gain of the antenna. This effect is illustrated in Figure 2.13 G≈10log10 M(2.29) 2.4 PIN Diodes and Control Circuits To implement the switching network of the multiple sectors of the antenna, it will be necessary to use switches. For this reason, in this section, the PIN diodes operation mechanism is briefly discussed. Pozar presents the concept of a PIN diode as an alternative to mechanical switches implemented in waveguide or coaxial form, which can handle high powers but are very slow, [3]. On 20 State-of-the-Art 4 5 2 3 3 1 1 z Antenna m 0 z Antenna n Figure 2.13: Mutual coupling between array elements illustration, [1, p. 475]. the contrary, PIN diodes can be easily implemented in a planar form and are capable of a highspeed operation. As a common diode these are active elements, however, with an intrinsic layer between the p and the nmaterials. This way, when the diode is reverse biased, it is seen as an high impedance, due to the small junction capacitance formed between the doped materials. On the other hand, when it is forward biased this capacitance disappears, which leads to a low impedance state. For this reason, the author of [12, p. 298] states that a PIN diode has a capacitive behaviour under reverse bias and a resistive behaviour under forward bias, as described by the equivalent circuit of Figure 2.14. Pozar describes two types of switches using PIN diodes, as described in Figure 2.15. In the first topology, with a series PIN diode, the switch is ON when the diode is forward biased. On the contrary, the shunt diode configuration is ON when the diode is reverse biased. As seen, both configurations require the use of RF chokes and DC blocks, to create an isolation between the bias and the RF signal. However ideally, a switch should have a null insertion loss when in ON state and an infinite attenuation in the OFF state, real switches haven’t. This happens because, when in forward bias state, the diode has a conduction resistance, as well as in the reverse bias state, the impedance is not actually infinite. Performing a simple circuit analysis, the following expressions were obtained by Pozar, to calculate the insertion loss of a switch in each of the presented configurations. Series Switch : IL =−20log 2Zo 2Zo+Zd (2.30) Shunt Switch : IL =−20log 2Zd 2Zd+Zo (2.31) where Zdis the diode impedance in either forward or reverse bias state. 2.5 Wide Band Dual Polarization Patch Antennas 21 Figure 2.14: Equivalent circuit for a PIN diode. (a) Reverse bias state. (b) Forward bias state. [3, p. 531]. Figure 2.15: Single-pole PIN diode switches. (a) Series configuration. (b) Shunt configuration. [3, p. 531]. 2.5 Wide Band Dual Polarization Patch Antennas This section is used to discuss work already developed on the field of wideband patch antennas and MIMO antennas. Since patch antennas are characterized by a narrow bandwidth, different ways of increasing it are presented. Additionally, ways of reducing the mutual coupling between the two feeds of a polarization-diversity MIMO patch and the cross-talk between the polarizations are also exploited. 2.5.1 Bandwidth Improvement Techniques for Patch Antennas In this subsection, two successful techniques of increasing the bandwidth of a patch antenna are deeply exploited, whiele other similar methods are briefly exposed. 2.5.1.1 Inverted L Antenna Probe Feed In [4], the authors use a feeding method different from the ones explained in 2.2.3. In this paper, a configuration using an L-probe feed, as the one seen in Figure 2.16. Given that the patch and the ILA have their own resonant frequencies, if they are properly combined, a broadband behaviour is obtained, up to 30% of fractional bandwidth. This system equivalent model is shown in Figure 2.17. As expected, the patch is modelized by an RLC resonant circuit and the probe is represented by a transmission line. Since the near-field of the probe coincides with the fundamental mode of the patch, the coupling is represented as 28 State-of-the-Art 180 180 0 0 Figure 2.26: Four probes with phase arrangement for mutual coupling reduction, [2, p. 180]. Using this method a good isolation between the antenna’s ports can be achieved. However, the phase shifter used in the discussed paper had a narrowband behaviour. For this reason, to use this method on the context of this thesis, it would be necessary to search for a wideband phase shifting method. The authors of [17] propose a similar alternative, however, using Γ-shaped balanced stripes, a parallel strip line to microstrip line balun and an hexagonal patch. This way, a greater operational bandwidth was achieved, 41.7% with a centre frequency of 2.4 GHz, as well as an isolation between ports higher than 30 dB. Figure 2.27: Patch antenna with differential Txand Rxoperation using two identical 3dB/180◦ couplers, [10]. 2.6 Matching Networks Using Non-Uniform Transmission Lines An impedance matching network is used to make the impedance seen into the matching network equal to the characteristic impedance of the transmission line, independently of the load. This way, the reflected power in the load is minimized and the power delivered to it is maximized. There are multiple types of impedance matching networks, such as using stubs and multi-section 2.6 Matching Networks Using Non-Uniform Transmission Lines 29 Figure 2.28: Simulated and measured S-parameters for 2.4GHz using a double differential patch antenna, [10]. quarter wave transformers, [3]. However, stubs have a narrowband behaviour, while a multisection transformer requires a several number of sections to achieve a broadband effect, which results in a long matching network. If instead of using multiple sections with a continuous variation of characteristic impedance between them, one could use a tappered line, designed between the load and the transmission line, as seen in Figure 2.29, a passband effect can be achieved. Different types of tapper, such as exponential, triangular or Klopfenstein tappers, give rise to different passband characteristics. z zz z +∆z 0 (a) (b) L ZL Z(z) Z0 ZZ + ∆Z ∆Γ Figure 2.29: Tapered line. (a) The tapered transmission line matching section. (b) Model for an incremental step change of taper characteristic impedance, [3]. Although a wideband impedance matching can be achieved using a tapered line, the theory behind these methods considers a real load that doesn’t vary with frequency, which, most of the times, isn’t true for an antenna. In [18], an impedance matching method for complex loads using NTLs (Non-uniform Transmission Lines) is presented, as well as a numerical method to calculate 30 State-of-the-Art the dimensions of such lines. In opposition to the other methods, these lines widths don’t vary uniformly as well as their length, as can be perceived from Figure 2.30. Figure 2.30: Non-uniform transmission line example. 2.7 Summary In this chapter an overview of the state-of-art was provided. Base parameters of antennas and the particular case of patch antennas were presented. Furthermore, a brief analysis on switching networks and wideband impedance matching techniques was also presented. Finally, some recent papers on topics of relevance for the development of this thesis were discussed and analysed, and the major difficulties of this project were identified. Chapter 3 Unit Cell As it can be perceived from the state-of-art review in the previous chapter, patch antennas have some interesting characteristics to solve the problem in hands. Namely, their directive characteristics and the possibility of implementing dual-polarization over the same patch structure. However, there is a compromise between the bandwidth of these antennas and the isolation between the inputs, which is their main problem. In this chapter, two different antenna topologies are fully described and the simulation results are presented and compared, in order to choose the best unit cell. Additionally, the results are also analysed resorting to the theoretical expectations. 3.1 MIMO Patch Antenna With Inverted L Probe Feed In this section, the technique described in [4] is replicated to the desired operation frequency. Furthermore, the dual polarization was also implemented using this feeding method. Finally, multiple approaches were tried to solve the problems of this topology. 3.1.1 Stackup The antenna studied in this section is a patch antenna with an ILA (Inverted L Antenna) probe feed. It can be seen from the stackup of Figure 3.1, that the patch is placed over a thick air substrate and the probe crosses the ground plane, feeding the patch through magnetic coupling. To tune this antenna for the desired input impedance there are multiple variables that can be optimized, such as the thickness of the substrate (airH), the width of the patch (patchW), the inset of the probe (D) and its dimensions (Lh and Lv). Despite the degrees of freedom that were previously mentioned, as stated in [4], the total length of the feeding probe should be approximately a quarter-wavelength 31 32 Unit Cell Air D airH patchW Lv Lh patchH d Figure 3.1: Structure of the patch antenna with inverted L probe feed. 3.1.2 System Analysis In order to understand the behaviour of the system, an HFSS model of the antenna was created and a parametric analysis of the effect of the design variables on the system performance was performed. This procedure consists of sweeping the patch size, the substrate height and the feed inset position, independently, and evaluating their effect in the impedance bandwidth of the antenna. In Figure 3.2 the reflection coefficient of the antenna for different values of the patch size is shown. As expected, the patch width has an optimum value, which was found to be 20mm. The value of the width controls the fundamental resonance of the patch, which when coupled to the feeding probe results in different bandwidths and different center frequencies. Frequency (GHz) 3 3.5 4 4.5 5 5.5 6 6.5 7 Reflection Coefficient Magnitude (dB) -35 -30 -25 -20 -15 -10 -5 0 patchW = 16 mm patchW = 18 mm patchW = 20 mm patchW = 22 mm patchW = 24 mm Figure 3.2: Simulated reflection coefficient of the patch antenna with ILA probe feed for different values of the patch width, with feed inset set to 7mm and substrate height to 9mm. The result depicted in Figure 3.3 shows that, when increasing the air height the radiating slot increases, decreasing the antenna quality factor due to higher radiation losses, which results in a bandwidth increase. However, if the probe height remains the same, the coupling between it and the patch begins to fade, resulting in a narrower bandwidth. 3.1 MIMO Patch Antenna With Inverted L Probe Feed 33 Frequency (GHz) 3 3.5 4 4.5 5 5.5 6 6.5 7 Reflection Coefficient Magnitude (dB) -35 -30 -25 -20 -15 -10 -5 0 airH = 7 mm airH = 8 mm airH = 9 mm airH = 10 mm Figure 3.3: Simulated reflection coefficient of the patch antenna with ILA probe feed for different values of the substrate height, with feed inset set to 7mm and patch width to 20mm. The last analysis refers to the impact of the feed inset position in the bandwidth of the system. As it can be seen, this design variable does not have a significant impact in the antenna’s performance. For this reason, the feed inset should be as high as possible, so when a second feed is introduced in the system, they do not couple, compromising the isolation. Frequency (GHz) 3 3.5 4 4.5 5 5.5 6 6.5 7 Reflection Coefficient Magnitude (dB) -35 -30 -25 -20 -15 -10 -5 0 D = 0 mm D = 2 mm D = 4 mm D = 6 mm Figure 3.4: Simulated reflection coefficient of the patch antenna with ILA probe feed for different values of the feed inset, with substrate height set to 10mm and patch width to 20mm. 3.1.3 Impedance Matching The initial values for this simulation were obtained through theoretical expressions. For the L probe, a total length of λ 4is recommended in [4], while for the patch, the length given by the 34 Unit Cell expressions already described in Chapter 2. However, in this case, the patch has to be square to enable the implementation of a dual polarization, for this reason, the the length and the width of the patch have the same value. Adjusting these parameters, as well as the inset of the probe and the air heigh, the best result was obtained for the values described in Table 3.1. As denoted from Figure 3.5, with these values, a 50Ωfractional impedance bandwidth of 41.6% at the central frequency of 5.05GHz was obtained, fulfilling the bandwidth requirement with a large margin. Table 3.1: Final dimensions of the patch antenna with ILA probe feed. Parameter Value(mm) patchH 0.1 patchW 20 d 1 airH 9 Lv 6.6 Lh 9 D 7 Frequency (GHz) 3 3.5 4 4.5 5 5.5 6 6.5 7 Reflection Coefficient Magnitude (dB) -35 -30 -25 -20 -15 -10 -5 0 Figure 3.5: Simulated reflection coefficient of the patch antenna with L probe feeding. This structure can also be described by the equivalent circuit model of Figure 3.6. As shown, the probe is modeled by two microstrip lines, one of them, magnetically coupled to patch, which is represented by the usual RLC resonant circuit. The inductance L2represents the inductance of the probe and the capacitances are parasitics generated between the two resonant elements (C2and C5) and between each element and the ground plane (C3,C4and C6). In Figure 3.7, it is proved that the behavioural model of the antenna matches the frequency response of the real antenna. Once the design was settled using a single feed, an orthogonal probe was inserted in the system, as it can be seen in the HFSS model of Figure 3.8. As predictable this addition had no influence on the the bandwidth of the antenna and the reflection coefficient of the antenna remains 3.1 MIMO Patch Antenna With Inverted L Probe Feed 35 Figure 3.6: Equivalent circuit of the patch antenna with ILA probe feed, [4]. Frequency (GHz) 4 4.5 5 5.5 6 Reflection Coefficient Magnitude (dB) -40 -35 -30 -25 -20 -15 -10 -5 0HFSS Simulation Equivalent Model Figure 3.7: Reflection coefficient of the patch antenna with ILA probe feed and its equivalent circuit. equal to the one shown in Figure 3.5 for both input ports. However, the biggest problem of this configuration was in the cross-talk between the two feeds, as shown from the isolation level of Figure 3.10, that should be above 20dB. Figure 3.8: HFSS model of the patch antenna with two orthogonal ILA probe feeds. 36 Unit Cell 3.1.4 Isolation Improvement The isolation problem of this topology may arise from multiple factors. For example, current sharing through the ground plane or due to magnetic coupling between the probes. 3.1.4.1 Current Sharing Reduction In order to minimize the current sharing over the ground plane, different DGS structures were simulated in order to achieve a good isolation. Some of these structures are illustrated in Figure 3.9. L W (a) Rectangular DGS. L W1 W2 (b) Tapered DGS. L1 W1 L2 W2 (c) Two Rectangular DGS with different dimensions. Figure 3.9: Examples of DGS simulated. 3.1 MIMO Patch Antenna With Inverted L Probe Feed 37 In Figure 3.10 is represented the isolation level between the two ports of the antenna using these structures. As it can be seen, all of them are characterized by a narrow bandwidth response, which can not cover the desired band. The best result was obtained using a single rectangular slot with a width of 1mm and a length of 20mm. Although it did a great improvement in the system performance, the isolation requirement is only satisfied for the range of frequencies between 4.7 and 5.55GHz. For these frequencies, the slot behaves as an open circuit between the probes, reducing the current sharing. Frequency (GHz) 3 3.5 4 4.5 5 5.5 6 6.5 7 Isolation Magnitude (dB) 0 10 20 30 40 50 without DGS 1 rectangular DGS 2 different rectangular DGSs tappered DGS Figure 3.10: Simulated isolation of the MIMO patch antenna with L probe feeding with different DGSs. This slot can be modeled as an RLC band-reject circuit placed between the references of each probe, as depicted in Figure 3.11. Using the optimization tool in ADS, it was possible to model the slot as an RLC with the values shown in the same Figure. Using this equivalent circuit, it is shown from the graphics of Figure 3.12 that the system has a similar behaviour as when the slot is employed. Two alternatives were tried in order to make this rectangular slot operate in a wider bandwidth. The first alternative was to draw two rectangular slots with different dimensions in parallel. Doing this, it was expected that the two slots would behave as two series RLC circuits, with different reject bands. The second was to draw a tapered slot, that was meant to behave as an open circuit in two different frequencies, leading to a broader bandwidth effect. However, both these attempts failed, which proved that a slot or a group of slots, behave as a single resonator with narrowband characteristics. Another interesting result has to do with the influence of the slot in the bandwidth of the antenna. Since it is relatively long, it starts to act as a complementary dipole, according to the Babinet principle. As shown in Figure 3.13, the slot behaves exactly like a slot dipole, since the current distribution has a maximum in the edges and a minimum in the center. Since it is radiating, it couples to the patch, changing the antenna input impedance, thence its bandwidth, as it can be 44 Unit Cell Modal Significance 34567 Air height (mm) 3 3.5 4 4.5 5 5.5 6 6.5 7 Frequency (GHz) 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Figure 3.20: Modal significance as a function of frequency and air height with driven patch size and parasitic patch size set to 20mm. Table 3.2: Final dimensions of the stacked parasitic patch antenna with probe feed. Parameter Value(mm) airH 5.55 driven_patchW 17.19 parasitic_patchW 18.39 feed_in 2 d 0.7 3 3.5 4 4.5 5 5.5 6 6.5 7 Frequency (GHz) 0 0.2 0.4 0.6 0.8 1 Modal Significance Figure 3.21: Modal significance as function of frequency using the dimensions of Table 3.2. input impedance of 50Ω, for different values of the feed inset variable. As shown, it is not possible to achieve the desired bandwidth only by changing this variable. 3.2 MIMO Stacked Parasitic Patch Antenna With Differential Probe Feed 45 In an usual patch structure, when the feed position gets closer to the center of the patch, the input impedance is reduced, since the voltage has a minimum in that point and the current has a maximum. However, in this case, because of the mutual coupling between the two patches, it does not work that way. As shown in Figure 3.23, the real part of the antenna impedance is around 100Ωfor both feed inset values of 2 and 5mm. Nonetheless, its imaginary part becomes more inductive with the increase of this variable, since the current has to go through a longer path in the patch with the increasing feed inset, which degrades the impedance bandwidth of the antenna. Since the imaginary part of the antenna impedance is close to 0Ωwhen using a feed inset of 2mm, a possible method to increase the bandwidth is to match it to 100Ω, transforming this impedance to 50Ωin the feeding network. In Figure 3.24, one can observe the return loss of the antenna when matched to a port of 100Ω, for different feeding positions. As expected, a degradation of the bandwidth is observed with the increasing feed inset value, because of the imaginary part increment. For this reason, the best result was obtained using a feed inset of 2mm, achieving a fractional bandwidth of 28.35% at the central frequency of 5.15GHz. Frequency (GHz) 4 4.5 5 5.5 6 6.5 7 Reflection Coefficient Magnitude (dB) -40 -35 -30 -25 -20 -15 -10 -5 0 feedIn = 2mm feedIn = 3mm feedIn = 4mm feedIn = 5mm feedIn = 6mm Figure 3.22: Simulated reflection coefficient referred to 50Ωof the stacked patch antenna with probe feed for different values of the feed inset position. 3.2.3.2 Ground Plane Hole Another problem that was faced during the process of matching the antenna, had to do with the probe passing through the ground plane. It was observed that, if the size of the gap in the ground plane was too small, it would have a negative impact in the impedance bandwidth of the antenna. This effect can be seen in the parametric analysis of Figure 3.25, where the reflection coefficient is shown for different values of the radius of the ground plane hole. In order to study the behaviour of this hole, an ADS schematic was created using the Sparameters of the antenna with a reference plane before the hole. In Figure 3.26 the schematic used in this simulation is shown. One can observe that the probe feed is represented by two inductances 46 Unit Cell Frequency (GHz) 4 4.5 5 5.5 6 Impedance (Ω) -50 0 50 100 150 200 250 300 Re(Z) (feedIn=2 mm) Re(Z) (feedIn=5 mm) Im(Z) (feedIn=2 mm) Im(Z) (feedIn=5 mm) Figure 3.23: Simulated real and imaginary parts of the input impedance of the antenna for different values of the feed inset position. Frequency (GHz) 4 4.5 5 5.5 6 6.5 7 Reflection Coefficient Magnitude (dB) -30 -25 -20 -15 -10 -5 0 feedIn = 2mm feedIn = 3mm feedIn = 4mm feedIn = 5mm feedIn = 6mm Figure 3.24: Simulated reflection coefficient referred to 100Ωof the stacked patch antenna with probe feed for different values of the feed inset position. and, between them, a capacitance is placed connecting to the ground plane, which describes the parasitic effects present in this transition. As can be seen from Figure 3.27, a capacitor of 0.2pF is able to describe the behaviour of a ground hole with a 0.4mm radius. 3.2.4 Dual-Polarization Once the design was optimized using a single feed, an orthogonal probe was inserted in the model. Since both patches have a rotationally symmetric shape, the bandwidth of the antenna was 3.2 MIMO Stacked Parasitic Patch Antenna With Differential Probe Feed 47 Frequency (GHz) 4 4.5 5 5.5 6 Reflection Coefficient Magnitude (dB) -30 -25 -20 -15 -10 -5 0 radius=0.37 mm radius=0.41 mm radius=0.4 mm radius=0.5 mm radius=0.6 mm Figure 3.25: Simulated reflection coefficient of the stacked patch antenna with probe feed for different values of the radius of the ground plane hole. Figure 3.26: Schematic of the equivalent circuit that describes a ground plane hole with a 0.4mm radius. not affected, and remained the same for both polarizations. However, as it can be observed in Figure 3.28, the isolation between them is very low. 3.2.4.1 Coupling Between the Probes As seen in the case of the patch antenna with ILA probe feed, the biggest problem for the reduced isolation came from the probes being too long, leading to magnetic coupling between them. In the present case the probes are much smaller, having a total length of 3.48mm. However, 48 Unit Cell Frequency (GHz) 4 4.5 5 5.5 6 Reflection Coefficient Magnitude (dB) -16 -14 -12 -10 -8 -6 -4 Equivalent Model HFSS Simulation Figure 3.27: Simulated reflection coefficient of the stacked patch antenna with probe feed for a ground hole with a radius of 0.4mm versus equivalent model created in ADS. to guarantee that the problem does not come from the coupling, a simple simulation setup was developed in HFSS. As can be observed in Figure 3.29, a perfect magnetic boundary was placed below the patch, between the two probes. This boundary forces the tangential component of the magnetic field to have opposite directions on each side of the border, which means that in the boundary, those components cancel each other, leading to no magnetic coupling. Withal, as it can be seen from Figure 3.28, the perfect magnetic boundary has no impact over the system performance, which proves that, in this case, the isolation problem does not arise from the coupling between the probes. Frequency (GHz) 4 4.5 5 5.5 6 6.5 7 Isolation Magnitude (dB) 0 5 10 15 20 25 Without H Boundary With H Boundary Figure 3.28: Simulated isolation of the MIMO stacked patch antenna with probe feed, with and without a perfect magnetic boundary between the probes. 3.2 MIMO Stacked Parasitic Patch Antenna With Differential Probe Feed 49 Figure 3.29: HFSS model of the MIMO stacked patch antenna with probe feed and a perfect magnetic boundary between the probes. 3.2.4.2 Differential Feeding Since a very thick substrate is used in this topology, the two fundamental modes of the patch, TM100 and TM010, are mixed into a TM110 mode that creates diagonal current distributions in the patches. This ultimately leads to a higher cross-talk between the two probes, which degrades the isolation between polarizations. Furthermore, since the two patches have different sizes, for higher frequencies, the biggest patch begins to support higher order modes, which also leads to a higher cross-talk. To solve this problem, instead of using a single probe feed for each polarization, two oppositely located feeds with a phase arrangement of 0◦and 180◦were used, as depicted in the HFSS model of Figure 3.30. By doing this, one is creating a boundary condition at each edge of the patch. This way, not only the patch degenerate modes are reinforced, but the higher order modes are also cancelled out, leading to a fractional impedance bandwidth of 52.21%, as depicted in Figure 3.31. Since the higher order modes are cancelled, the radiation patterns of the antenna become very symmetric, as shown in Figures 3.32 and 3.33. Furthermore, since the diagonal current distributions are mitigated, the cross-polarization level of this antenna is also below −20dB, as it can be perceived from Figures 3.34 and 3.35. This structure can also be described by the equivalent circuit of Figure 3.36. As it can be realized, the probes are represented as inductances, the patches are modeled as two RLC resonant circuits coupled through the inductors with a coupling factor k. Lastly, the capacitances are parasitics between the two patches and between each patch and the ground plane. As shown in Figure 3.37, the behavioural model of the antenna has a frequency response similar to the real antenna, simulated in HFSS. 3.2.4.3 Phase-Shifting As explained above, the differential feeding requires that the two opposite probes have a phaseshift of 180◦between each other. However, achieving this with a lossless phase-shifter would lead 50 Unit Cell Figure 3.30: HFSS model for the MIMO stacked parasitic patch antenna with differential probe feed. Frequency (GHz) 3 3.5 4 4.5 5 5.5 6 6.5 Reflection Coefficient Magnitude (dB) -20 -15 -10 -5 0 Figure 3.31: Simulated reflection coefficient of the MIMO stacked parasitic patch antenna with differential probe feed. to a large design and it would be harder to reach such a wide bandwidth. For this reason, a commercial wideband balun phase-shifter (TDK - HHM1595A1), characterized by the S-parameters of Figure 3.38, was used. 3.2 MIMO Stacked Parasitic Patch Antenna With Differential Probe Feed 51 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (a) E-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (b) E-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (c) E-plane 5.9GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (d) H-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (e) H-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (f) H-plane 5.9GHz. Figure 3.32: Simulated co-polarization E-plane and H-plane radiation patterns, with horizontal polarization excited. 3.2.5 Wideband Impedance Transformation The described balun has to see an impedance of 50Ωin its balanced outputs. Yet, the antenna was matched to an impedance of 100Ω. For this reason, it is necessary to transform the impedance of the antenna, which is not simple to achieve in such a wide bandwidth resorting to the most simple impedance matching techniques, such as quarter wavelength transformers. Furthermore, most of these matching techniques are used to match loads that do not vary with frequency, which is not the case of the stacked parasitic patch antenna. To do this, non-uniform transmission lines where deployed between the ports of the antenna and the balanced ports of the balun, as shown in Figure 3.39, where the balun is represented by a 1:2 transformer. According to [18], it is possible to algebraically determine the optimal impedances in a nonuniform transmission line matching network. However, these calculations become intricate and time demanding if the load impedance is a complex high order equivalent circuit as it is in this case. As such, a random optimizer was employed in Keysight’s ADS, to determine line lengths and widths that would match the antenna to the wideband baluns. 52 Unit Cell 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (a) E-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (b) E-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (c) E-plane 5.9GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (d) H-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (e) H-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (f) H-plane 5.9GHz. Figure 3.33: Simulated co-polarization E-plane and H-plane radiation patterns, with vertical polarization excited. 3.2.5.1 Vertical Mount SMA Connector In order to properly match the antenna, it was necessary to simulate the launcher of the vertical edge mount SMA connector that was used (Cinch Connectivity Solutions - 142-0711-201). This simulation was performed using the HFSS model of Figure 3.40 and the return loss seen from the coaxial input was optimized by changing the width of the microstrip line. In Figure 3.41, one can observe a parametric analysis of the reflection coefficient of the connector when sweeping the width of the microstrip line in which the connector lands. The optimum result was obtained for a line width of 0.4mm and the insertion loss of the connector is depicted in Figure 3.42. 3.2.5.2 Balun Grounding Since the manufacturer of the balun does not give proper guidelines on how the grounding of the IC should be performed, it was done in order to reduce the impedance from the grounding pad to the reference plane in the best way possible. To do that, the maximum number of vias with the smallest diameter enabled by the manufacturing process (0.3mm), were used to make the connection. 3.2 MIMO Stacked Parasitic Patch Antenna With Differential Probe Feed 53 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -50 -40 -30 -20 (a) E-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -50 -40 -30 -20 (b) E-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -50 -40 -30 -20 (c) E-plane 5.9GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -50 -40 -30 -20 (d) H-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -50 -40 -30 -20 (e) H-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -50 -40 -30 -20 (f) H-plane 5.9GHz. Figure 3.34: Simulated cross-polarization E-plane and H-plane radiation patterns, with horizontal polarization excited. The footprint of the balun can be seen in Figure 3.43 and is characterized by a low inductance and a resistance close to 0Ω, as it can be perceived from Figure 3.44. 3.2.5.3 Optimization and Layout The first approach to the impedance matching problem was done using the ADS optimization tool to optimize the electric length and the characteristic impedance of each NTL section, from a total of six sections. Firstly, the optimization used a random algorithm to avoid convergence to a local minimum. Then a gradient algorithm was applied to this solution, to refine the previous result. Since all the ports of the antenna are equal, the NTL transformer is equal independently of the input it is matching. The optimization was performed taking into account the simulated S-parameters of the connector and the antenna, as well as the S-parameters provided by the manufacturer of the balun, as described in the schematic of Figure 3.45. The dimensions of the lines that resulted of this first optimization process are shown in Table 3.3. In [19], the authors discuss the fact that bending a microstrip line has an impact on its characteristic impedance and a way of computing the characteristic impedance of a curved line is also shown. Taking this result into account, it was necessary to perform a tune of their width so the 60 Unit Cell Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Isolation Magnitude (dB) 0 10 20 30 40 50 60 70 80 Figure 3.48: Simulated isolation of the complete antenna system. Figure 3.49: Final layout of the non-uniform transmission lines matching network. requirements, its gain is 1dB higher. The worst problem of this structure is its cost, since the substrates that were used are expensive. Taking this into account, it is possible to perceive that this topology can be used in the context of this dissertation work, since it meets all the requirements of the system. 3.3 Summary In this chapter two antenna topologies were evaluated in order to choose the more suitable solution for this problem. The simulation results of both topologies were presented and discussed. 3.3 Summary 61 Furthermore, the MIMO stacked parasitic patch antenna was chosen as the unit cell to be used in this dissertation work, and its final layout was also presented. 62 Unit Cell Chapter 4 Switching Matrix In this chapter, different topologies of switching matrices are evaluated in order to find the best solution for the problem, in terms of cost, difficulty and efficiency. Then, the chosen topology is dimensioned and simulated resorting to the Method of Moments in ADS. 4.1 Topology Evaluation One of the main objectives of this dissertation work, is to build a sectoral antenna in which the beam direction is automatically reconfigured with the moving vessel. To enable this, a switching matrix had to be developed. Since each antenna unit has an HPBW of 60◦, the use of six antenna sectors would be required. However, it was decided that eight sectors would be employed so it was guaranteed that the antenna would not be transmitting below −3dB of the maximum gain in the azimuth plane. The switching matrix is a PCB, with the shape of a regular octagon, with an antenna connected to the middle of each edge. The switching mechanism is placed in the center of the octagon, to make the design more symmetric, as represented by the schematic of Figure 4.1. The switching mechanism is controlled by an Arduino which, based on data provided by two GPS modules, computes the heading and the position of the boat, and from this information determines which of the eight antenna sections should be turned on. In this section, three different topologies of switching matrices, using distinct RF switches integrated circuits, are analysed and discussed taking into account three important factors: the cost of the system, the difficulty of routing the chips and the number of control signals required to control it. Furthermore, the impact of the type of transmission line and substrate to use, in the cost and the efficiency of the system, is also analysed. 4.1.1 Switching Options As already explained, the switching mechanism has a common RF input and eight RF outputs that are activated by controlling the switch. To do so, different alternatives using distinct commercial RF switches were evaluated. The first topology uses seven 1:2 RF switches, displayed in 63 64 Switching Matrix Figure 4.1: Schematic of the switching matrix a tree shape, as depicted in the block diagram of Figure 4.2. Since each switch needs a control bit to choose between the two possible outputs, a total of seven control signals would be required. Furthermore, the typical insertion loss of the switch is 0.67dB, which makes a total loss of 2.01dB in the path between the RF input and an output. The use of so many ICs, would require a huge routing effort to make the connections between the chips and each antenna. RFC Ctrl1 RF1 RF2 RF3 RF4 RF5 RF6 RF7 RF8 Ctrl3 Ctrl2 Ctrl4 Ctrl5 Ctrl6 Ctrl7 Figure 4.2: Block diagram of topology 1, using seven 1:2 switches. The second topology is a bit simpler than the previous one. As seen in from Figure 4.3, in this topology, a 1:2 switch is used to choose between two different paths, each one connecting to a 1:4 switch. In this case, the number of ICs is reduced to three, which makes the routing of the system easier and uses only five control signals to choose between the eight RF outputs. Compared to the first topology, this one is more expensive, however, since the typical insertion loss of the SP4T switch IC is 1dB, the total insertion loss is 1.67dB, which is lower than the previous topology. Lastly, the third topology uses a simple 1:8 RF switch, as the one depicted in the block diagram of Figure 4.4. In this scenario, one would require only three control signals to choose between the 4.1 Topology Evaluation 65 RFC Ctrl1 RF1 RF2 RF3 RF4 RF5 RF6 RF7 RF8 Ctrl2 Ctrl3 Ctrl4 Ctrl5 Figure 4.3: Block diagram of topology 2, using one 1:2 and two 1:4 switches. eight outputs and, since only one IC would be used, the routing effort is very low. Nonetheless, this is a more expensive alternative, since each switch costs ¤22.2, and the insertion loss of the chip is also higher, typically around 2.2dB. RFC Ctrl1 RF1 RF2 RF3 RF4 RF5 RF6 RF7 RF8 Ctrl2 Ctrl3 Figure 4.4: Block diagram of topology 3, using one 1:8 switch. This analysis is summarized in Table 4.1. Here, the total effort of each topology is evaluated taking into account its price, the number of control signals it uses, the routing effort (classified as 0, 0.5 or 1) and its typical insertion loss. The expression for the total effort is given by Equation (4.1). As it can be inferred, the total effort of topology 1 is the highest, which makes this less interesting. However, the other two topologies are very similar in terms of effort. The trade off between them relates to the price of the topology and its routing effort. Since this is being used in a wide bandwidth, the use of impedance matching techniques will be required to match the antennas to the switch output ports. For this reason, an easy routing would facilitate the use of such matching techniques, making the design simpler. Furthermore, considering that the matching has to take into account the S-parameters of the ICs, the error introduced by 66 Switching Matrix the S-parameters provided by the manufacturer is lower if one uses less chips. Total E f f ort =Price max(Price)+#Control signals max(#Control signals) +Routing E f f ort +Insertion Loss max(Insertion Loss) (4.1) Table 4.1: Analysis of the total effort of using each switching topology. Topology Price #Control Signals Routing Effort Insertion Loss (dB) Total Effort 1 11.27 7 1 2.01 3.42 2 15.47 5 0.5 1.67 2.67 3 22.2 3 0 2.2 2.42 The switch chosen for this effect is the SP8T (Analog Devices Inc. - HMC321a), which is characterized by the S-parameters of Figures 4.5,4.6 and 4.7. As depicted, the return loss of the IC is different depending on the chosen output port. For this reason, the impedance matching lines will have to be different for each sector of the switching matrix. From Figures 4.6 and 4.7, one can conclude that the insertion loss of the IC is around 2.1dB and the isolation to the disconnected ports is above 40dB for the desired band of frequencies. Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Reflection Coefficient Magnitude (dB) -15.5 -15 -14.5 -14 -13.5 -13 -12.5 -12 RF1: ON Other ports: ON Figure 4.5: Reflection coefficient of the switch for different output ports seen at the RF common port. 4.1.1.1 Substrate Another important consideration to make, is the best substrate to use in the switching matrix PCB. The two options that will be considered further, are the Rogers4003c and the FR4, with the characteristics shown in Table 4.2. The main advantage of using FR4 is its cost, which is much 4.1 Topology Evaluation 67 Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Insertion Loss Magnitude (dB) -2.25 -2.2 -2.15 -2.1 -2.05 Figure 4.6: Insertion loss from the input of the switch to a connected output port. Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Isolation Magnitude (dB) 39 40 41 42 43 44 45 46 Figure 4.7: Isolation from the input of the switch to a disconnected output port. lower. Yet, since it is thicker and has a higher electrical permittivity, wider transmission lines are required to achieve the same impedance as in the case of the Rogers substrate, which can lead to higher radiation losses. Moreover, since the loss tangent is approximately ten times bigger in FR4, the losses in this substrate can be so high that it can become unviable. Table 4.2: Physical characteristics of Rogers4003c and FR4 substrates. Substrate Thickness (mm) Copper Thickness (µm) εrtan(δ) Rogers4003c 0.350 18 3.55 0.0022 FR4 1.6 34 4.4 0.017 To evaluate these effects, a simple test was performed in Keysight’s ADS. Firstly, using the 68 Switching Matrix random optimizer, an NTL impedance transformer with six sections of equal electrical length was dimensioned to match the antenna to the switch, using both substrates. This test environment is depicted in Figure 4.8 and the dimension of each section in both substrates is exposed in Table 4.3. Figure 4.8: Schematic of an NTL transformer of six sections between the switch and the antenna. Table 4.3: Dimensions fo the NTL sections in Rogers and FR4. Substrate W1(mm) W2(mm) W3(mm) W4(mm) W5(mm) W6(mm) Rogers4003c 0.26 0.26 0.64 0.64 0.24 1 FR4 0.61 0.61 3.4 0.83 0.5 1.13 By applying a power source of 0dBm in port 1 and measuring the power that reaches the antenna, it is possible to compute the losses in both cases. As it can be seen from Figure 4.9, although the difference in both substrates is below 1dB in almost all the bandwidth, near 4.4GHz this difference is close to 3dB, which would have an impact of 3dB in the realized gain of the antenna. Since the presented results are not from an EM simulation, they do not take into account the radiation losses in the lines. Given that in FR4 the lines are much wider, the largest has a width of 3.4mm, the radiation losses will be greater than in the Rogers case, where the largest line width is 1.13mm. Taking these results, it was possible to conclude that, besides the higher cost, the Rogers alternative would be the best to solve the problem because it has lower losses, leading to a higher efficiency of the entire system. However, since the substrate is very thin, it was stacked on an FR4 substrate to give extra mechanical robustness to the PCB. 4.1.1.2 Transmission Lines The last analysis performed while evaluating the best topology for the switching network, was the type of transmission line to use for the matching network (between the antenna and the switch). The two considered options were the simple case of microstrip lines and the more complex CBCPW lines, which have a more complex design. To evaluate this, a model of each type of line was built in HFSS, as can be seen from the models of Figure 4.10. Both lines have a length of 15mm and were tuned to 50Ω. Such leads to a microstrip line with a width of 0.7mm and a CBCPW line with 0.56mm of width and a gap of 0.2mm between the edge and the ground plane. Given these values, the return loss of Figure 4.11 shows that both lines have a good match to 50Ω, since the reflection coefficient magnitude is below −41dB. Figure 4.12 shows the insertion loss of the two lines, and from this, one can conclude that the losses within them are very similar. Since in the biggest part of the bandwidth, the microstrip line has slightly lower losses, its use is preferred, also because they are easier to design and tune. 4.1 Topology Evaluation 69 Frequency (GHz) 4 4.5 5 5.5 6 Power Received (dBm) -7 -6 -5 -4 -3 -2 Rogers4003c FR4 Figure 4.9: Power received in the antenna when the input of the NTL transformer is connected to a power source of 0dBm. (a) Microstrip line model. (b) CBCPW line model. Figure 4.10: HFSS models of the microstrip and CBCPW lines, both tuned to 50Ω. 4.1.2 Layout and Simulation Results In this section, the dimensions of the PCB are revealed, the simulation results of an edge mount connector are discussed and the final layout of the switching matrix is also presented, as well as its simulation results. 4.1.2.1 PCB Dimensions As already stated, the board will have the shape of a regular octagon and an antenna will be connected to the mid point of each of its edges through an edge mount connector. Since the unit cell has a width of 5cm, it was decided that the octagon’s edges would have the same dimension. This way, when the antennas are connected to the edge mount connector, there will be some space left between them, which can be used to pass the cables that will connect to the switching matrix. 76 Switching Matrix Figure 4.21: Routing of the digital and supply pins of the switch IC. microstrip mode, instead of the CBCPW mode. Such simulation was performed in HFSS, and it was perceived that when the gap between the line and the ground plane was 2mm, the line could be considered a microstrip line. 4.1 Topology Evaluation 77 Figure 4.22: Schematic of the optimization setup employed in ADS for a single antenna. 78 Switching Matrix Table 4.5: Final dimensions of the NTL transformers obtained through an optimization process. Parameter H-Pol Dimension (mm) V-Pol Dimension (mm) Parameter H-Pol Dimension (mm) V-Pol Dimension (mm) W10.21 0.23 W34 0.35 0.47 W20.35 0.47 W35 0.70 0.65 W30.70 0.56 W36 1.61 1.23 W41.61 1.33 W37 0.79 0.60 W50.79 0.58 W38 0.48 0.79 W60.48 0.76 W39 0.30 0.48 W70.30 0.48 W40 0.91 1.26 W80.91 1.25 W41 0.21 0.25 W90.21 0.23 W42 0.35 0.47 W10 0.35 0.47 W43 0.70 0.65 W11 0.70 0.56 W44 1.61 1.23 W12 1.61 1.33 W45 0.79 0.60 W13 0.79 0.58 W46 0.48 0.79 W14 0.48 0.76 W47 0.30 0.48 W15 0.30 0.48 W48 0.91 1.26 W16 0.91 1.25 W49 0.21 0.23 W17 0.21 0.23 W50 0.48 0.47 W18 0.35 0.47 W51 0.74 0.56 W19 0.70 0.56 W52 1.38 1.33 W20 1.61 1.33 W53 0.54 0.58 W21 0.79 0.58 W54 0.40 0.76 W22 0.48 0.76 W55 0.36 0.48 W23 0.30 0.48 W56 0.94 1.25 W24 0.91 1.25 W57 0.21 0.25 W25 0.21 0.25 W58 0.35 0.47 W26 0.35 0.47 W59 0.70 0.65 W27 0.70 0.65 W60 1.61 1.23 W28 1.61 1.23 W61 0.79 0.60 W29 0.79 0.60 W62 0.48 0.79 W30 0.48 0.79 W63 0.30 0.48 W31 0.30 0.48 W64 0.91 1.26 W32 0.91 1.26 W65 0.68 0.68 W33 0.21 0.25 Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Reflection Coefficient Magnitude (dB) -40 -35 -30 -25 -20 -15 -10 -5 0 Figure 4.23: Mean value of the reflection coefficient of the H-pol switching matrix, with a confidence interval of 95%. 4.1 Topology Evaluation 79 Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Reflection Coefficient Magnitude (dB) -30 -25 -20 -15 -10 -5 0 Figure 4.24: Mean value of the reflection coefficient of the V-pol switching matrix, with a confidence interval of 95%. Figure 4.25: Final layout of the H-pol switching matrix. 80 Switching Matrix Figure 4.26: Final layout of the V-pol switching matrix. Chapter 5 Experimental Evaluation In this chapter, the experimental setup that supports the theory and simulations of the previous chapters, is presented. Firstly, the experimental results of the fabricated unit cell are exposed and discussed. Then, the results of the complete system, composed by the antennas and the switching matrices, are also evaluated. 5.1 Unit Cell 5.1.1 Manufacturing In Figure 5.1 the produced unit cell is shown. As shown in the picture, the parasitic patch is stacked above the Neltec substrate by using four teflon screws. Furthermore, it is also shown which of the two input ports is considered to be the vertical and the horizontal polarization. Figure 5.1: Manufactured MIMO stacked parasitic patch antenna with differential probe feed. 5.1.2 Measurements To validate the theoretical predictions and the simulation results, the measurement setup of Figure 5.2 was mounted in the anechoic chamber. The evaluation of each of the antenna’s polar81 82 Experimental Evaluation izations was performed individually, while the other port was terminated with a broadband load of 50Ω. One of the ports of the VNA was connected to the antenna through a coaxial cable terminated in an SMA connector. The other port of the VNA was connected to an horn antenna with a well known radiation pattern within the desired bandwidth. (a) Manufactured antenna. (b) Horn antenna. (c) Measurement system. Figure 5.2: Setup used for the measurement of the antenna unit cell. 5.1.2.1 Results and Discussion In Figures 5.3 and 5.4 the reflection coefficient as a function of frequency is presented for the horizontal and vertical polarizations, respectively. The measured reflection coefficient magnitude of five different antennas is also demonstrated in Figures 5.5 and 5.6. The isolation between the two polarizations is shown in Figure 5.7. Furthermore, the co-polarized radiation patterns for each polarization are depicted in Figures 5.8 and 5.9. Cross-polarization measurements were not made due to excessive measurement noise. From the exposed results, one realizes that the fabricated antenna is not a perfect match with the simulation results. The horizontal polarization of the antenna has a mismatch between two different frequency ranges (4.4-4.7GHz and 5.45 - 5.7GHz), and its reflection coefficient magnitude goes up to a maximum of −6dB. However the result in the vertical polarization is better, 5.1 Unit Cell 83 Frequency (GHz) 4 4.5 5 5.5 6 Reflection Coefficient Magnitude (dB) -40 -35 -30 -25 -20 -15 -10 -5 0 Simulated Measured Figure 5.3: Horizontal polarization reflection coefficient as a function of frequency. Frequency (GHz) 4 4.5 5 5.5 6 Reflection Coefficient Magnitude (dB) -40 -35 -30 -25 -20 -15 -10 -5 0 Simulated Measured Figure 5.4: Vertical polarization reflection coefficient as a function of frequency. it also has a mismatch between 4.5 and 4.85GHz and a maximum of −8.4dB of reflection coefficient. Although the antenna is mismatched in some frequency ranges, this result is similar in all the fabricated unit cells, which makes the design reproducible. Furthermore, the isolation between the two polarizations meets the requirement with a comfortable margin. The radiation patterns of the antenna are almost a perfect match with the simulation, nonetheless, the realized gain is 1dB lower, which may be caused by the already referred mismatch. The reason for the difference between the simulation and the measured results was identified. Since the balun manufacturer does not give proper guidelines on how to connect the chip, the S-parameters that are provided are not reliable for every footprint. Since the NTL transformers are optimized taking into account these S-parameters, the system performance is compromised by 84 Experimental Evaluation Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Reflection Coefficient Magnitude (dB) -50 -40 -30 -20 -10 0 Antenna 1 Antenna 2 Antenna 3 Antenna 4 Antenna 5 Figure 5.5: Horizontal polarization reflection coefficient as a function of frequency for five different antennas. Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Reflection Coefficient Magnitude (dB) -35 -30 -25 -20 -15 -10 -5 0 Antenna 1 Antenna 2 Antenna 3 Antenna 4 Antenna 5 Figure 5.6: Vertical polarization reflection coefficient as a function of frequency for five different antennas. this imprecision. 5.2 Switching Matrix 5.2.1 Manufacturing In Figure 5.10 one can observe a picture of the complete system, that incorporates two switching matrices (for horizontal and vertical polarizations), with eight antennas, one connected to each port of the switching matrix. Furthermore, a flat cable connects to eight headers in each switching 5.2 Switching Matrix 85 Frequency (GHz) 4.4 4.6 4.8 5 5.2 5.4 5.6 5.8 Isolation Magnitude (dB) 0 10 20 30 40 50 60 70 80 Simulated Measured Figure 5.7: Isolation between polarizations as a function of frequency. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (a) E-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (b) E-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (c) E-plane 5.9GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (d) H-plane 4.4GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (e) H-plane 5.15GHz. 0° 15° 30° 45° 60° 75° 90° 105° 120° 135° 150° 165°180° -165° -150° -135° -120° -105° -90° -75° -60° -45° -30° -15° -20 -10 0 10 (f) H-plane 5.9GHz. Figure 5.8: Simulated (line) and measured (circles) co-polarization E-plane and H-plane radiation patterns, with horizontal polarization excited. matrix. This cable is then attached to an Arduino controller, that will provide the power to the IC, as well as the control signals of the switching matrix. 92 BIBLIOGRAPHY [12] R. Ludwig and P. Bretchko, RF Circuit Design: Theory and Applications, N. Prentice Hall, Upper Sadle River, Ed., 2000. [13] Li, Lai, Luk, and Lau, “A wideband patch antenna with cross-polarization suppression,” IEEE Antennas and Wireless Propagation Letters, vol. 3, pp. 211–214, 2004. [14] Y.-X. Guo, K.-M. Luk, and K.-F. Lee, “Broadband dual polarization patch element for cellular-phone base stations,” IEEE Transactions on Antennas and Propagation, vol. 50, no. 2, pp. 251–253, Feb 2002. [15] H. Guodong, C. Lixin, G. Chong, and L. Ning, “Dual-polarized and multiple-layers microstrip patch antenna for satellite mobile communication,” in 2013 IEEE INTERNATIONAL CONFERENCE ON MICROWAVE TECHNOLOGY COMPUTATIONAL ELECTROMAGNETICS, Aug 2013, pp. 227–231. [16] D. Wolanský, P. Všetula, J. Puskely, and Z. Raida, “Broadband small patch antenna array for Ka-band application,” in 2013 7th European Conference on Antennas and Propagation (EuCAP), April 2013, pp. 907–910. [17] J. Zhang, X. Q. Lin, L. Y. Nie, J. W. Yu, and Y. Fan, “Wideband dual-polarization patch antenna array with parallel strip line balun feeding,” IEEE Antennas and Wireless Propagation Letters, vol. 15, pp. 1499–1501, 2016. [18] G. Xiao and K. Yashiro, “Impedance matching for complex loads through nonuniform transmission lines,” IEEE Transactions on Microwave Theory and Techniques, vol. 50, no. 6, pp. 1520–1525, Jun 2002. [19] A. M. J. S. Roy, 0. R. Poddar and S. K. Chowdhur, “Characteristic impedance of a curved microstrip trannsmission line,” Microwave and Optical Technology Letters, vol. PP, no. 99, pp. 1–1, 1998.