Antenna Characterization with Autonomous UAV and Software Defined Radio
Adam Alenius
Lucas Wennerholm
Teknisk- naturvetenskaplig fakultet UTH-enheten
Besöksadress:
Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0
Postadress:
Box 536 751 21 Uppsala
Telefon:
018 – 471 30 03
Telefax:
018 – 471 30 00
Hemsida:
http://www.teknat.uu.se/student
Software Defined Radio
Adam Alenius, Lucas Wennerholm
A measurement equipment with the purpose of measuring the radiation pattern of antennas in the frequency interval 30-300 MHz was constructed. To perform the necessary measurements the equipment needs to be mounted on a UAV, a necessity that demands a low weight from the measurement equipment. These kinds of measurements are today done with equipment that is mounted on helicopters, making the equipment smaller and fitting it on an UAV will save cost for the persons or companies that need to utilize this service.
To ensure that the system is easy to use for anyone who wants to characterize an antenna efforts were made to make the software application user friendly. The system visualizes measurement results in 2D diagrams that are simple to analyze.
Since the equipment has size restrictions the computer in the system needs to be small and light. The single board computer used has computational limitations and therefore the digital signal processing must be carefully designed to both be fast and generate good measurement data. To verify the performance of the system tests and theoretical simulations where performed and compared. The tests were performed both in an echo free antenna chamber and in realistic outdoor environments with an UAV. The finished system performed well and the measurement results showed clear similarities with the theoretical simulations. The outdoor environment clearly influences the shape of an antennas radiation pattern and the need to characterize antennas in a realistic environment became clear.
ISSN: 1401-5757, UPTEC F 18035 Examinator: Tomas Nyberg Ämnesgranskare: Mikael Sternad Handledare: Daniel Brogren
effekt och strålningsmönster utvecklats. När mätningar på antenner med lång våglängd utförs behöver utrustningen vara på ett betydande avstånd från antennen då det elektromagnetiska fältet kring en antenn inte får sitt karakteristiska utseende förrän på cirka tre gånger våglängdens avstånd. Detta problem löses genom att fästa utrustningen på en drönare vilken autonomt flyger runt antennen.
Det faktum att utrustningen ska fästas på en drönare medför krav på dess ut- formning. Den kan inte vara för stor för då går den inte att montera och den kan inte heller vara för tung eftersom en drönare då inte kan lyfta den. Krav måste även ställas på mjuvkarans prestanda eftersom den signalbehandlingskedja som imple- menteras behöver vara tillräckligt snabb för att kunna generera många mätvärden med god kvalitet.
För att möjliggöra detta system behövs: En antenn för att motta signalerna, en mjukvaruradio som kan läsa av de inkommande signalerna och omvandla dem från analog till digitalt, en enkortsdator som implementerar en signalbehandlingskedja för att beräkna resultatet och möjlighet för användaren att på avstånd kommuni- cera med och kontrollera utrustningen.
Under utvecklingen måste hårdvarukomponenterna individuellt karakteriseras och kalibreras. Detta för att säkerställa att det slutgiltiga systemets uppmätta mätvärden faktiskt stämmer överens med den mottagna signalen. Mjukvaran måste genomgående uppfylla prestandakraven så att den inte blir för processortung och långsam.
Efter att systemet var framtaget så testades dess egenskapar för att kontrollera att det uppför sig som förväntat och presenterar rimliga resultat. För detta utfördes mätningar i en antennkammare som är ett rum skärmat från extern radiostrålning och har dämpande väggar, golv och tak som förhindrar reflektioner. Mätningarna visade på goda resultat som låg nära de teoretiska värdena.
När systemet var redo testades det i fält med drönare. Utomhus blev resultatet inte lika tillförlitligt som i antennkammaren men detta var väntat. Utomhusmätningarna visade avvikelser från de simulerade värdena eftersom de elektromagnetiska vågorna blir starkt påverkade av omgivningen.
Systemet lämnar i slutändan lite att önska, det kan inte mäta på hela det önskade frekvensspannet eller rita upp figurer i tre dimensioner. Sammantaget såg dock mätresultaten bra ut och utrustningen borde med lite modifikation kunna användas till sitt syfte: att mäta upp strålningsmönstret hos olika antenner och presentera detta för användaren på ett lätt och smidigt sätt.
i
FFT . . . Fast Fourier Transform
FIR filter . . . Finite Impulse Response filter GRC . . . GNU Radio Companion SDR . . . Software Defined Radio SNR . . . Signal-to-Noise Ratio UAV . . . Unmanned Aerial Vehicle VSWR . . . Voltage Standing Wave Ratio
ii
1 Introduction 1
1.1 Preface . . . 1
1.2 Project Description . . . 1
1.3 Project Goals . . . 1
1.4 Background . . . 2
2 Theory 3 2.1 Electromagnetism . . . 3
2.1.1 Maxwell’s equations . . . 3
2.1.2 Electromagnetic waves . . . 3
2.1.3 Polarization . . . 4
2.1.4 Reflection and interference . . . 4
2.2 Antennas . . . 4
2.2.1 Basics . . . 4
2.2.2 Measuring units and fundamental parameters . . . 6
2.2.3 Field regions . . . 7
2.2.4 Polarization . . . 8
2.2.5 Radiation pattern . . . 8
2.2.6 Infinitesimal small dipole antenna . . . 9
2.2.7 Finite length dipole antenna . . . 9
2.2.8 Finite length dipole antenna above ground . . . 9
2.3 Properties of Antenna Systems . . . 10
2.3.1 Transmission lines and two port networks . . . 10
2.3.2 Reflection and transmission . . . 11
2.3.3 Impedance matching . . . 11
2.3.4 Polarization matching . . . 11
2.3.5 Antenna factor . . . 12
2.3.6 Link budget . . . 12
2.4 Digital Signal Processing . . . 12
2.4.1 Sampling rate and bandwidth . . . 13
2.4.2 Quadrature signals . . . 13
2.4.3 Noise . . . 14
2.4.4 Minimum detectable signal . . . 14
2.4.5 Decimation . . . 14
2.4.6 Finite impulse response filters . . . 15
2.4.7 Decimating polyphase FIR filter . . . 17
2.4.8 Fast Fourier transform . . . 17
2.4.9 Windowing functions . . . 20
2.4.10 Aliasing . . . 20
2.5 Software Defined Radio . . . 21
2.5.1 Basic layout . . . 21
2.5.2 Heterodyne principle . . . 21
2.6 Reference System Transformation . . . 22
iii
3.1.2 RTL-SDR . . . 24
3.1.3 LimeSDR . . . 25
3.2 Single-board Computer . . . 25
3.2.1 Odroid-XU4 . . . 25
3.2.2 Raspberry Pi 2 . . . 26
3.3 Antennas . . . 26
3.3.1 HP 11955A . . . 26
3.3.2 Schwarzbeck FSH3D . . . 26
4 Software 27 5 Development Procedure 28 5.1 Development Method . . . 28
5.2 Hardware . . . 28
5.2.1 Software defined radio . . . 29
5.2.2 Low noise amplifier . . . 29
5.2.3 Single board computer . . . 29
5.2.4 Antenna . . . 30
5.2.5 Positioning system . . . 30
5.3 Software . . . 30
5.3.1 Software requirements . . . 30
5.3.2 Receiving the quadrature data . . . 32
5.3.3 Decimation process . . . 32
5.3.4 Anti aliasing filter . . . 32
5.3.5 FIR low-pass filter . . . 33
5.3.6 Fourier transform . . . 33
5.3.7 Power spectral density . . . 33
5.3.8 Frequency identification . . . 33
5.3.9 Error correction . . . 34
5.3.10 Critical parameters . . . 34
5.4 System Tests . . . 36
5.4.1 Antenna chamber . . . 36
5.4.2 Outdoors . . . 37
5.4.3 Electrical properties of the components . . . 37
5.5 System Simulations . . . 38
5.5.1 FIR filter . . . 38
5.5.2 Fourier transform . . . 38
5.5.3 Antenna chamber . . . 38
5.5.4 Dipole above ground . . . 39
5.6 Unexpected Error Handling . . . 39
6 Calibration measurements 41 6.1 Error and Impedance Measurements of SDR . . . 41
iv
6.1.4 Frequency offset in the internal oscillator . . . 43
6.2 Response Time of Schwarzbeck Antenna . . . 44
6.3 Antenna Chamber Measurements . . . 45
6.3.1 The chamber . . . 45
6.3.2 Measurement method . . . 45
6.3.3 Results . . . 48
6.4 Pre-flight Measurements . . . 49
6.4.1 The environment . . . 49
6.4.2 Method of measuring . . . 49
6.4.3 Results . . . 50
7 Results 54 7.1 Hardware . . . 54
7.2 Signal Processing Chain . . . 55
7.3 Recording Chain . . . 56
7.4 Time and Frequency Analysis . . . 57
7.5 Oscillation Phenomenon Investigation . . . 61
7.6 Simulations . . . 61
7.6.1 FIR and FFT . . . 62
7.6.2 Antenna chamber . . . 65
7.6.3 Dipole above ground . . . 65
7.7 Real Case Measurements . . . 68
7.7.1 169,375 MHz . . . 68
7.7.2 69,6 MHz . . . 70
7.7.3 50 MHz . . . 72
8 Discussion 75 8.1 Final System . . . 75
8.2 Oscillation Phenomenon . . . 75
8.3 Measurement Results . . . 76
8.3.1 Antenna chamber . . . 76
8.3.2 Pre-flight measurements . . . 77
8.3.3 Flight measurements . . . 77
8.4 Measurement on Unknown Antenna . . . 78
8.5 Future Work . . . 78
9 Conclusions 79 9.1 System . . . 79
9.2 Measurements . . . 79
v
1 Introduction
1.1 Preface
This master thesis was written by two students from the Engineering Physics program at Uppsala University. The project itself has been done at ÅF Digital Solutions AB in Uppsala.
This project has been a collaboration between Adam Alenius and Lucas Wennerholm where the students have had different responsibility areas. The project was mainly divided into two parts, an analog and a digital part.
Alenius was responsible for the digital software parts. This includes the design of the digital signal processing, the theory on signal processing and the user application that controls the system.
Wennerholm was responsible for the analog electrical parts. This includes the design of the hardware system, the theory behind antennas and electromagnetism, calibration of hardware components and the simulations of the system and reference antenna.
The students collaborated to merge the two parts into one complete system. The design of the measurements and the analysis of the results were done as a joint effort. The work presented in this report is the result of a collaboration.
1.2 Project Description
This project is supposed to develop measuring equipment for characterizing antennas with the frequency span 3-300 MHz. The measurement equipment should characterize the antennas by plotting their antenna diagrams in both 2 and 3 dimensions. To make satisfactory measurements for antennas with these wavelengths it is necessary for the measuring equipment to be airborne. In this project the equipment should be lifted by a Unmanned Aerial Vehicle (UAV). The measurement equipment should be divided into two distinct parts: an analog part and a digital part. The finished product should not depend on the choice of UAV provided it is powerful enough to carry the system.
1.3 Project Goals
The goal of this project is to construct measurement equipment with some chosen qual- ities. It should be possible to mount on a reasonably sized UAV while being as light as possible in order to increase the flight time. It should be able to detect the output of the measured antenna in different polarizations. A single-board computer should be used as control unit for the equipment.
The software for the equipment should be easy to use. It should have the possibility to plot radiation pattern in both 2 and 3 dimensions for different polarizations, plot
the measured signal over time and should be able to change important measurement parameters on-the-fly.
In order to make the measurements the UAV must follow a flight path. The best path for these measurements is a spiralling pattern on the edge of a half-sphere where the measured antenna is placed in the centre of the sphere. It should then be possible to plot a 3D radiation pattern if the flight is successful.
1.4 Background
Antennas are the basis of all wireless communication. It is critical to have a good know- ledge about how they radiate if they are to be used in real world applications. To gather information about an antenna one must make measurements on it. These measurements must be made where the field around the antenna is stable which is approximately three wavelengths away. The wavelength is inversely proportional to the frequency, a small fre- quency yields a large wavelength and vice versa. Therefore, antennas with low frequencies can have wavelengths of tens, hundreds or thousands of meters which makes them very hard to characterize. Today when measuring low-frequency antennas the measurement equipment must be attached to a helicopter which is then flown around the antenna in order to make the measurements. This project is about constructing a simpler, cheaper alternative in order to make these measurements with equipment that can be mounted onto a UAV.
2 Theory
2.1 Electromagnetism
In the analysis of antennas and electromagnetic radiation it is important to understand the theory behind electromagnetism. This section will lay out the necessary details that are of importance to understand this report.
2.1.1 Maxwell’s equations
Maxwell’s equations describe all classic electromagnetic interactions. In equations 1 to 4 the four Maxwell equations are displayed [1].
∇ · D = ρv (1)
∇ · B = 0 (2)
∇ × E = −∂B
∂t (3)
∇ × H = J + ∂D
∂t (4)
The Maxwell equations describe the relationships between the following physical phe- nomenons and physical constants:
• H The magnetic field
• E The electric field
• D = E The electric flux density
• B = µH The magnetic flux density
• ρv The electric volume charge
• The permittivity of a medium
• µ The permeability of a medium
In this project the third and fourth Maxwell equations are of most importance since they describe the nature of electromagnetic waves.
2.1.2 Electromagnetic waves
A direct consequence of Maxwell’s equations is the electromagnetic wave. The third and fourth Maxwell equation describe how a magnetic and electric field can continue to propagate and oscillate after the fields have been created in an electric circuit [1].
If one solves these two equations for either H or E the Helmholtz equation or the wave equation will be found as seen in equation 5 and 6.
∇2E + ω2µE = 0 (5)
∇2H + ω2µH = 0 (6)
The medium propagation constant or wave number k is defined as k = ω√
µ. This constant describes how the wave behaves in different media. The electric and magnetic field of a propagating wave oscillate perpendicular to each other and travel with the speed of light in vacuum.
2.1.3 Polarization
The polarization of a wave describes the orientation of the electric field given the direction of travel and a reference coordinate system [2].
The two most common polarizations are linear/plane polarization and elliptic polariz- ation. A linearly polarized wave has a constant direction of the electric and magnetic fields. This project will only work with linear polarized waves. In an elliptical polar- ized wave the direction of the electric and magnetic field rotates around the direction of travel with respect to time and the magnitude varies to create an elliptical shape.
Circular polarization is a special case of elliptical polarization where the magnitude is constant.
2.1.4 Reflection and interference
An electromagnetic wave has the ability to reflect against a surface of some medium. In the general case the reflection occurs when a wave is changing propagation medium.
Electromagnetic waves follow the rules of superposition. This means that constructive and destructive interference will occur when two waves meet. This is also the case when a wave is reflected against a surface and interferes with itself as the reflection causes a 180 degree phase shift on the incident wave [1].
2.2 Antennas
2.2.1 Basics
Antennas are electrical conductors that are adapted specifically to emit electromagnetic radiation. When connected to an oscillating current source the antenna will create an oscillating electric and magnetic field around itself resulting in an electromagnetic wave.
The electric and magnetic fields (also called E-field and H-field respectively) are ortho- gonal to each other and they are both orthogonal to the Poynting vector [1], as shown in figure 1. The Poynting vector corresponds to the S-vector in the figure below.
time E
H S
Figure 1: Propagation of an electromagnetic wave
The antennas of most importance for this report is:
• Isotropic antennas: A point source antenna with a spherical radiation pattern.
This means that they radiate with the same power in all directions. This type of antenna is theoretical and can not be physically constructed but are often used as a theoretical reference point when computing the gain of real antennas. The radiation pattern of an isotropic antenna can be seen in figure 2 [2].
x y
z
Figure 2: Radiation pattern of a isotropic antenna
• Dipole antennas: One of the most simple and common antenna types. It consists of two straight conductors mounted in a straight line with connectors in the center.
Dipole antennas are also known as a linear wire antennas. The radiation pattern and shape of the dipole antenna can be seen in figure 3 [2].
x y z
Figure 3: Radiation pattern of a dipole antenna
• Loop antennas: Consists of a conductor in a circular coil. These antennas are used in various applications but they are generally poor radiators and more suited be receivers. The ideal loop antenna has the same radiation pattern as an ideal dipole antenna. [2]
2.2.2 Measuring units and fundamental parameters
To understand the inner workings of antennas and radio systems there are some important concepts and measuring units that must be defined according to convention. Most units are expressed in a logarithmic unit proportional to milliwatts or microvolts since antenna measurements often are in those order of magnitudes.
Units [2]:
• V/m Unit of electric fields
• A/m Unit of magnetic fields
• W/m2 Unit for the radiated power of an antenna.
• dBµV/m Unit of electric fields in decibel
• dBmW/m2 Milliwatt per square meter in decibel
• dBmW Milliwatts in decibel usually expressed as dBm, the unit used to measure transmitted and received power
• η0 = 377Ω The intrinsic impedance of free space
The parameters of importance in this report is all expressed in spherical coordinates in three dimensions and polar coordinates in two dimensions.
Parameters [1]:
• F The radiation pattern
• U = ηk2
32π2|F | Radiation intensity
• pt= U
r2 Radiation power density
• Prad =Rπ 0
R2π
0 U dΩ Total radiated power
• UI = Prad
4π Isotropic power density
• D = U
UI Directive gain
• Dmax = Umax
UI Directivity
• G = 4πU
Prad Power gain
• g = G
Gmax Normalized power gain
When measuring signal strength of a transmitting antenna the result varies several mag- nitudes depending on the input power and how close the measurement equipment is to the transmitting antenna. This means that a linear scale of measurement is very hard to read, therefore a log-scale is used instead. Decibels (dB) is the ratio between two measurements of the same unit which means that it is effectively unitless. The ratio can be arbitrarily chosen, but it is common to compare the power at the main lobe (i.e.
the largest measurement value) with all other points. This is however not an absolute measurement and it is impossible to compare two antennas just by their dB values alone using this method. To solve this problem the dBm unit is used. What this means is that the measurement values are put in relation to 1 milliwatt [2]. It is calculated according to equation 7.
dBm = 10 · log10( P
1mW) (7)
Where P is the measurement value in W. If the antenna outputs 1 milliwatt at a certain point one can also say that it has 0 dBm in that point in space. Since this method uses a fixed reference point it is possible to compare antennas with each other.
2.2.3 Field regions
The radiating field surrounding an antenna is divided into three principal regions based on the distance R from the antenna, the maximum linear dimension D and the chosen radiation wavelength λ. [2]
The three different regions are:
• Reactive near field
• Radiating near field also known as the Fresnel region
• Far field also known as the Fraunhofer Region
Table 1: Table over the field regions surrounding an antenna
Reactive near field Radiating near field Far field Condition 1 0 < R < 0.62q
D3
λ 0.62
qD3
λ < R < 2Dλ2 R > 2Dλ2
Condition 2 - - R >> D
Condition 3 - - R >> λ
In this report the most important field region is the far field since the radiation pattern is defined in this region. Conditions of the division are given in table 1.
In the reactive near field region the field the E-field components are in time-phase and the HΦ-field are in time-phase quadrature with the E-field.
In the radiating near field region the field the E-field components lose their in time- phase condition and approach time-phase quadrature. Since the different components of the E-field does not have the same magnitude the polarization the polarization becomes rotating with an elliptical shape.
In the far field region the E and H components becomes perpendicular to each other and in transverse to the direction of propagation as seen in figure 1. [2]
2.2.4 Polarization
As mentioned in section 2.1.3 the electromagnetic wave has a property called polarization.
The polarization of a wave generated by an antenna is dependent on the physical shape of the antenna. The electromagnetic wave emitted by a dipole antenna will have an electric field oriented in the same direction as the linear wires of the antenna. [2].
2.2.5 Radiation pattern
When performing measurements on an antenna one plots the final result in a so called antenna diagram. The antenna diagram is used for easy visualization of how the an- tenna radiates in different directions, i.e. the radiation pattern. Most often there are two plots to one antenna diagram, one showing the antennas radiation pattern over the azimuth angle, and the other shows the radiation pattern over the elevation angle. It is possible to combine these measurements into one 3D-plot if the radiation pattern is symmetrical.
2.2.6 Infinitesimal small dipole antenna
The infinitesimal small dipole antenna is an important concept that is used to describe how the ideal dipole radiation pattern looks. As seen in figure 3 the dipole-radiation is known and it can be built in practice but only for very small dipoles where the length of the dipole is l < 50λ and the conductor is assumed to be infinitesimally thin. The radiation pattern of a dipole antenna in the far field can mathematically be calculated using Maxwell’s equation and takes the the form seen in equation 8 [2].
F = sin2(θ) (8)
The radiation pattern of the dipole is not uniform and the power radiated in a certain direction changes with θ. The average energy flux in a certain direction at a distance r by the infinitesimal dipole is seen in equation 9 [2].
Wavg(θ) = η 2
kI0l 4π
2 sin2(θ)
r2 (9)
Where I0is the input current, l is the longest dimension of the antenna i.e. the length.
2.2.7 Finite length dipole antenna
The infinitesimal small dipole is a poor approximation for most real dipole antennas. A better approximation is the finite length dipole. The average energy flux for the finite length dipole can be seen in equation 10 [2]. Unlike equation 9 equation 10 is valid for any length of the dipole antenna.
Wavg(θ) = η |I0|2 8π2r2
cos(kl2 cos(θ)) − cos(kl2) sin θ
!2
(10)
2.2.8 Finite length dipole antenna above ground
The radiation pattern of the dipole seen in figure 3 is calculated with the condition that the antenna is in a reflection free environment. In reality this is not feasible and there will always be objects and matter surrounding the antenna. This causes the radiation pattern to not behave as the theoretically ideal pattern. A common way of dealing with this problem is to simulate the antenna in a non ideal environment. Considering an antenna mounted at a certain hight above the ground a better estimation of the radiation pattern can be calculated using equation 11. The equation describes a finite length dipole antenna mounted a certain distance from a flat surface with a certain set of material properties [2].
Eφ= jηI0e−jkr 2πr
cos(kl2 cos(θ)) − cos(kl2) sin θ
!
[ejkh cos θ + Rve−jkh cos θ] (11) Where Rv is calculated according to equation 12.
Rv = η0cos θi − η1cos θt
η0cos θi+ η1cos θt (12)
Where I0 is the input current, l is the longest dimension of the antenna i.e. the length and h is the height above ground.
2.3 Properties of Antenna Systems
When using antennas in a system where they are connected to electrical devices it’s important to understand how these systems behave. In this section the important parts of this theory is presented.
2.3.1 Transmission lines and two port networks
The equipment used in this project has the properties of a two port network and in high frequency applications it is important to understand how these networks influence an electrical signal.
The two port network is graphically explained in figure 4.
Figure 4: Illustration of the two-port network
The four parameters seen in figure 4 are called S parameters [1] and their properties are described in table 2.
Table 2: Table over the different S-parameters
S-parameter Description
S11 Incident reflection parameter S12 Incident transmission parameter S21 Outgoing transmission parameter S22 Outgoing reflection parameter
2.3.2 Reflection and transmission
When electrical equipment is used for transmitting and receiving radio frequency signals it is important to know how the reflection and transmission parameters in all parts of the system behave. A large reflection coefficient is likely if the impedance of the different parts in the system are poorly matched. [1].
2.3.3 Impedance matching
The input and output terminals of electrical components often have varying impedance as a function of the frequency. When two components are connected to each other and the output impedance of one does not match the input impedance of the other then some of the transmitted power will be reflected. This is somewhat analog with a mechanical wave traveling through a rope with a fastened or completely free end. When the end is fastened (impedance mismatch) then the entire wave will be reflected and vice versa [1].
When quantifying how much of the wave that is reflected one uses the reflection coefficient Γ which is defined as equation 13.
Γ = Vr
Vf = S11 (13)
Where Vf is the complex amplitude of the forward traveling wave, and Vr is the complex amplitude of the reflected wave. However, most often the voltage standing wave ratio is used for characterizing a system, and that is given by equation 14.
V SW R = Vmax
Vmin = 1 + |Γ|
1 − |Γ| (14)
The Γ or reflection parameter is the same as the S11 parameter. The VSWR is a measure of how well the equipment is impedance matched. A low VSWR means a well matched system.
2.3.4 Polarization matching
In a system composed of a transmitting and receiving antenna it is important that the polarization of the antennas are matched. As mentioned in section 2.2.4 an antenna has a polarization that varies with it’s orientation. In the case of the dipole antenna its important that the receiving and transmitting antennas are oriented in the same way [1].
If the two dipole antennas are not oriented in the same manner a polarization mismatch occurs. When this happens the receiving antenna will not receive all of the incoming wave since it lies in the wrong plane. If the polarization of the incoming wave is known
it is possible to measure all polarizations and add the components to get the full power.
Equation 15 describes how this is done.
Ptot =p P12
+ P22
+ P32 (15)
2.3.5 Antenna factor
The antenna factor is a way of combining mismatch losses in a receiving or transmitting antenna including the free space impedance into one convenient factor. It describes how strong an incident electrical field has to be to generate 1 volt at the terminal of a receiving antenna. The antenna factor is defined according to equation 16.
AF = s
4πηf2
c2GZLoad (16)
2.3.6 Link budget
When measuring an antenna it is crucial that all the loss and gain throughout the system are known so they can be compensated for when plotting the antenna diagram. The complete model of a system is called a link budget and it’s described in equation 17 in log terms. The link budget is derived from Friis transmission formula [2].
PRX = PT X+ GT X+ GRX− LT X− LRX − LF S− LM (17) Where:
• PRX, PT X is the received power at the receiving antenna and transmitted output power of the transmitting antenna respectively.
• GT X, GRX is the antenna gain at the transmitting resp. receiving antenna.
• LT X, LRX are the losses at the transmitting, resp. receiving antennas due to con- nectors, ohmic losses etc.
• LF S is the path loss, and LM are miscellaneous losses due to e.g. polarization mismatch and the like.
If the link budget is not calculated correctly then the measurements at the receiving antenna cannot be guaranteed to describe the true field strength of the transmitting antenna.
2.4 Digital Signal Processing
Digital signal processing (DSP) is used everywhere a signal needs to be processed digitally instead of through analog processes. DSP is a wide term and encompasses many subjects.
It includes, but is not limited to: sampling, digital filters, digital transforms etc. This section describes the theory behind the DSP techniques used in this project.
2.4.1 Sampling rate and bandwidth
When sampling it is crucial that the sample rate is chosen carefully. It has to be in proportion to the frequency of the signal that is to be sampled otherwise the sampled signal may not represent the true signal as it should. When sampling, the sampling rate of the system should always be chosen such that it is at least double the frequency of the signal to be measured to avoid aliasing. [3] The frequency that is half the sampling frequency is called the Nyqvist frequency, named after Harry Nyqvist (1889-1976). The Nyquist theorem consists of two criteria related to the minimum sampling frequency fs
and the bandwidth B, equation 18 and 19 describe the relationship.
fs ≥ 2 · f (18)
B < fs/2 (19)
2.4.2 Quadrature signals
In all applications that use RF signals in communication there is a need for reliability when extracting information from a signal at the receiver. In DSP this is usually done through so called IQ-data. The equation of the most general sinusoid can be seen in equation 20.
V (t) = A · sin(2πtf + φ) (20)
Where A is the amplitude, t is time, f is frequency and φ is the phase shift.
The key to IQ-modulation is the possibility to represent very complex phase modulation of a signal by only using amplitude modulation on two waves. This is possible through combining two sinusoid to create a resulting wave. By changing the component waves’
amplitude in relation to each other the phase will shift in the resulting wave.
In practical implementation two signals that are shifted π/2 radians in respect to each other are used. Two sinusoids that are shifted π/2 radians by definition are cosine waves and sine waves. This is where the I and Q in the name IQ-data comes from. A wave is said to be in quadrature with another if it is shifted exactly π/2 radians. Therefore the I-data (In-phase data) will affect the cosine wave and the Q-data (quadrature data) will affect the sine wave. If now the cosine amplitude is changed, then the resulting wave will be phase shifted more towards 0. By doing this any phase can be constructed, and thus it is possible to convey phase modulation through IQ-data. If both the cosine and the sine are changed in amplitude by the same amount, the phase will not shift, but the amplitude will.
2.4.3 Noise
When dealing with real physical systems it is impossible to avoid noise. Noise can stem from many different sources.
Instead of worrying about what every source of noise is separately, in practical applic- ations these unknown noise signals can be combined into a noise floor. This is the magnitude of all combined noise that is present at all times in a given system. The signal that is supposed to be measured has to be clearly separated from the noise floor. The difference in level between the signal to be measured and the noise floor is called the signal-to-noise ratio (SNR). To be able to trust some measurements the SNR has to be sufficiently large. If it is not it can be very hard to distinguish real measurement values from the noise. If the SNR is low, there are of course steps that can be taken to improve it. Filtering is one example. If the noise is very high in frequency, a low pass filter can be applied to the system and this will result in a better SNR. One example of this can be the 50 Hz noise that exists in all landline telephones due to the AC that runs through it. This would be extremely annoying for any telephone user, but instead a bandstop or high pass filter is applied and this makes it much easier to hear the person at the other end.
2.4.4 Minimum detectable signal
There is a theoretical limit to what level of signal that a given system can detect. This is theoretically defined as the thermal noise:
N oise f loordBm= 10 · log10(k · T0· 1000) + N F + 10 · log10(BW ) (21) where k is Boltzmann’s constant = 1, 38 · 10−23, T0 is the temperature of the receiv- ing system in Kelvin, NF is the receiver noise figure and BW is the bandwidth of the system.
What this limit really says is that it is impossible for the system to detect any signals below this level. In practice it is likely that the noise floor is above this theoretical limit since the theory cannot take into account random disturbances. This limit can however be used as a good indicator of what the SNR will be in a relatively non-noisy measurement.
2.4.5 Decimation
Decimation (sometimes known as downsampling) is the processing of extracting a certain number of samples from a discrete sampled signal. This is effectively the same process as lowering the sample rate, but lowering the sample rate outright is not always a possibility.
[3] In a system with a fixed sample rate decimation is an important tool for making the system usable for hardware with limited performance. When decimating it is important to make sure that no vital information about the signal is lost and that the Nyquist
criterion seen in equation 18 is met. If the Nyquist criterion is not fulfilled the output of the decimation process might suffer from aliasing discussed further in section 2.4.10.
The sample rate after decimation becomes fsM = Mfs where M is the decimation factor.
The new Nyquist criterion with decimation is described in equation 22.
fsM ≤ 2 f
M (22)
The decimation process can be implemented as a digital filter where the output signal y(n) is the input signal x(n) decimated with a factor M. This is done according to equation 23 [3].
y(n) = x(nM ) (23)
Decimation is often done in combination with other DSP processes. When decimation is used in practice the signal is first passes through a low-pass filter to reduce high-frequency signal components and reduce aliasing. Then the downsampling itself is performed where only every M:th sample is kept.
2.4.6 Finite impulse response filters
Finite impulse response (FIR) filters are used in many DSP applications. A general FIR filter is defined by:
y(n) =
K
X
i=0
bix(n − i) = b0x(n) + b1x(n − 1) + b2x(n − 2) + ... + bKx(n − K) (24)
where y(n) and x(n) are the output and input at time n respectively, bi are the filter coefficients and K + 1 is the length of the FIR filter, also known as the filters number of taps [4].
FIR filters can be implemented as: low-pass, high-pass, band-pass and band-stop filters.
What kind of filter the FIR-filter becomes is completely dependent on the design of the magnitude frequency response of the FIR-filter. This magnitude frequency response is chosen so that it matches the magnitude frequency response for one of the four filters mentioned above. The complete derivation for this can be found in section 7.2 of [4]
To calculate causal FIR-filters coefficients in table 3 and shift it to the right with M number of samples. This leads to the transfer function:
H(z) = b0+ b1z−1+ b2z−2+ ... + b2Mz−2M (25) where bn = h(n − M ), n = 0, 1, ..., M.
Table 3: Table over the ideal impulse responses for different kind of filters.
Type of filter: Ideal impulse response h(n) (non-casual FIR-coefficients) Low pass: h(n) =
(Ωc
π , n = 0
sin(Ωcn)
nπ for n 6= 0, −M ≤ n ≤ M High pass: h(n) =
(π−Ωc
π , n = 0
−sin(Ωnπcn) for n 6= 0, −M ≤ n ≤ M Band pass: h(n) =
(Ω
H−ΩL
π , n = 0
sin(ΩHn)
nπ −sin(ΩnπLn) for n 6= 0, −M ≤ n ≤ M Band stop: h(n) =
(π−Ω
H+ΩL
π , n = 0
−sin(ΩnπHn) +sin(ΩnπLn) for n 6= 0, −M ≤ n ≤ M
where Ωc is the normalized cutoff frequency given by:
Ωc= 2πfcTs (26)
where fc and Ts is the cutoff frequency and sample time respectively. ΩH, ΩL is the high and low cutoff frequencies in the case of band-pass and band-stop filters where there are two cutoff frequencies.
FIR filters can be implemented together with windowing functions so that oscillation in the passband and stopband will be dampened as much as possible. Windowing functions can be read about in section 2.4.9
When designing a FIR filter there are several important parameters to take into account in relation to the expected performance. The performance parameters sets a requirement on the minimum amount of filter taps necessary [3]. The parameters can be seen in table 4.
Table 4: FIR filter design parameters
Description Notation Cut-off frequency ωp Start of stop band ωs Transmission width ωt= ωs− ωp
Pass band ripple δp Stop band attenuation δs
To make a simple estimate of the number of taps M necessary to reach a specific demand on the filter the formula developed by Kaiser can be used [3]. The formula can be seen in equation 27.
M = −20log10(pδpδs) − 13
14.6(ωs− ωp)/2π + 1 (27)
2.4.7 Decimating polyphase FIR filter
The decimation process is often used in combination with FIR filters. Perfoming these processes one at a time i.e. first filtering and then decimating is an inefficient way of achieving the wanted result. To make the process more efficient one can implement the polyphase filter structure to decimate and implement a FIR filter in one combined process [4].
The polyphase structure divides the FIR filter taps into M sub filters where M is the decimation factor. The subfilter is denoted Er, the FIR-taps h(i) and the input signal zn. This is done according to equation 28. And the complete filter structure H(z) is described in equation 29 [3].
Er =
M N
X
n=0
h(M n + r)z−n, 0 ≤ r ≤ M − 1 (28)
H(z) =
M −1
X
r=0
z−rEr(zM) (29)
2.4.8 Fast Fourier transform
The Fast Fourier transform (FFT) is commonly used in applications related to signal processing. FFT transforms a signal from time-domain to frequency-domain and the inverse FFT does the opposite. The FFT is a collection name for different algorithms that all compute the discreet Fourier transform in a way that saves computing power compared to the regular Descreet Fourier transform (DFT). [3]. This is very useful when decomposing a wave into sinusoids, so that one can exactly know the frequency composition of the measured wave.
A problem that can occur when using a FFT is spectral leakage. Spectral leakage is the phenomenon when frequencies that are not present in the input signal is present in the FFT. This is due to the nature of how the discreet Fourier transform works. First, it samples a signal in a fixed time-window and then it will repeat that signal to imitate a infinite signal. If the original signal is sampled for a non-integer number of periods, then there will be a discrepancy between the repeating sequences. The transform will then interpret this sequence of repeated samples as a continous sine wave. This results in a completely different signal which when transformed will show a number of sinusoids that are not present in the input signal. This discrepancy phenomenon is displayed in figure 5.
time Original sample Repeated samples
magnitude
sampling rate
magnitude time
Figure 5: Above is the sampled signal. Below is the same signal as the FFT will interpret it.
To reduce spectral leakage windowing functions (discussed in detail in section 2.4.9) are commonly used. Windowing functions supresses the sampled signal at the beginning and end of the sampling window leading to smaller discrepancies when the sampled signal is repeated. The effect of window functions on a sampled signal can be found in figure 6 which is borrowed from [4]
Figure 6: Illustration of how the window function affects the sampled signal.
Another important aspect when using the FFT is the resolution of bins. A bin is a section or sample of the frequency spectrum, that is given by equation 31.
nb = ns
2 (30)
rb = fmax
nb (31)
where rb is the resolution of a single bin, fmax is the maximum frequency where the sampled signal can contain information about the original signal, and nb and ns is the number of bins and number of samples respectively. If rb is too coarse then frequency information of the original signal can be lost. E.g. if the original signal contains two frequencies at 1001 Hz and 1005 Hz but the resolution of the bin is 100 Hz/bin then there is a large possibility that both the frequencies are contained in the same bin. This will cause a single spike in the frequency spectra. If rb is 1 Hz there will be two separate spikes in the frequency spectra, one at 1001 Hz and the other at 1005 Hz, which of course is desirable.
When using the Fourier transform it is in many cases desirable to show the energy content of the frequencies. This is called the power spectral density (PSD) and is defined according to equation 32 [5]. Where F is the Fourier transform vector, F∗ is the complex conjugate and L is the length of the vector.
P SD = F · F∗
L (32)
2.4.9 Windowing functions
When sampling it is very common to apply a windowing function to the measurements.
This is done to avoid discrepancies in the sampled signal which introduces unwanted sinusoids into the sampled signal. Windowing functions are functions that suppress the start and end values of the sampled signal in the time spectrum, this can be seen in figure 6. The window can be chosen as different trigonometric functions with parameters.
Different windowing functions has different attributes that makes them more or less fitting for the specific task at hand. When looking for a very specific frequency in a measurement range with other frequencies close to the one of interest then a window with large roll off for the closest side lobes is a good idea. Some of the most common windows are Hamming, Hanning and Blackman-Harris. These are all so called cosine-sum windows.
• Hann window: Moderate window function, not extreme to either end. Given by:
w(n) = 0.5(1 − cos
2πn N − 1
) = sin2
πn N − 1
(33)
• Hamming window: Optimized to minimize the nearest side lobe, but otherwise not that extreme either. Given by:
w(n) = α − β cos
2πn N − 1
(34) Where α = 0.54, β = 1 − α = 0.46
• Blackman-Harris: Even more optimized to minimize side lobes, but does this by additional computional load. Given by:
a0 = a1cos
2πn N − 1
+ a2cos
4πn N − 1
− a3cos
6πn N − 1
(35) Where a0 = 0, 35875; a1 = 0, 48829; a2 = 0, 14128; a3 = 0, 01168
2.4.10 Aliasing
Aliasing is a variant of information loss due to poor choice of sampling rates for a specific frequency. Aliasing occurs when a signal is sampled with a sample rate that has is less than double the minimum frequency in the original signal. If a signal that is reconstructed from samples has frequency components that is not present in the original continuous signal then aliasing has occurred. This can happen because the sample frequency is below the Nyqvist frequency (section 2.4.1).
2.5 Software Defined Radio
The radio unit used in this project is a SDR short for Software Defined Radio. A SDR has in comparison to a conventional radio very few electrical components and most parts of the signal processing is done digitally.
2.5.1 Basic layout
The software defined radio has no precise definition since there are many varieties of SDRs.
The main concept is however the same for every SDR, the radio signal is digitalized in an early stages of the system after the antenna receives the signal. The main stages in a SDR commonly are: RF processing, down converting, IF processing and a digitization stage [6]. All stages may include some form of low noise amplifier. An important feature of the SDR is that the different stages are reconfigurable at runtime making the device flexible and useful in many different applications.
The RF processing stage refers to the part where the electrical signal from the antenna is received by the SDR. This stage commonly contain an analog band pass filter where the frequency content of the incoming signal is filtered to only contain frequencies that the following stages can manage.
The down converting stage refers to the process where the signal incoming from the previous is mixed with an intermediate frequency. This utilizes the heterodyne principle to reduce the frequency of the signal without losing any information contained in the signal, further discussed in section 2.5.2.
The IF stage refers to the part of system that processes the intermediate frequency signal.
The signal is commonly filtered with an bandpass filter and in some cases the signal is converted to a quadrature signal depending on the application.
The last step is the digitization of the signal. This is done with an analog to digital converter (ADC). After the signal has been digitized it is processed digitally. All DSP is done in software.
2.5.2 Heterodyne principle
The heterodyne principal is commonly used in SDRs [6]. The principle is founded in trigonometry and explains how a multiplication of two sinusoids can be written as the sum of two other sinusoids. This is explained in equation 36. To get the original signal the modulated carrier frequency ΩC is mixed with the ΩIF frequency to recover the original signal.
cos(ΩC) · cos(ΩIF) = 1
2(cos(ΩC + ΩIF) + cos(ΩC− ΩIF)) (36) Using this principle the SDR can receive a modulated high frequency signal without running into problems because of low sample rate. The principle is implemented on the
analog signal in the SDR but it can also be preformed digitally.
2.6 Reference System Transformation
When working with GPS the only point of reference is the Earth’s center. This is not always desirable since it can be very hard to compare values and understand how they correlate to each other. Therefore it is simpler to convert the global spherical coordinates to local cartesian. The Earth has an elliptic shape which means that the radius is not the same at every location. At the equator the radius is Req = 6378, 1370 km and at the poles Rpoles = 6356, 7523 km. If the Earth is approximated as a theoretical ellipse then the radius at any given latitude is given by:
Rloc= 3
2· Rpol· (1 − e)
(1 − e · sin φ)2 (37)
where e is the eccentricity of the ellipsoid given by e = 1 − (RRpoleq )2.
When the local radius has been calculated it is time to calculate the cartesian coordinates.
This is done through:
X = r · sin θ · cos φ Y = r · sin θ · sin φ Z = r · cos θ
(38)
When the global coordinate system is converted to cartesian coordinates the local co- ordinate system can be found based on the current position on Earth. This is done by the following calculations:
• Generate a local orthogonal coordinate system that is parallel to the normal vector of the origin point in global coordinates
• Generate a transfer matrix to convert from global coordinates to the local coordinate system
The result is a local reference system at Rloc distance from the Earth’s core. This is displayed in figure 7 where the xyz reference system is the Earth-centric one and x0 y0 z0 is the local reference system.
x
y z
R
φ θ x0
y0
z0
Figure 7: An illustration of the coordinate transformation from global coordinates to local coordinates.
3 Hardware
3.1 SDR
3.1.1 Airspy R2
In this project Airspy R2 was the choice of SDR. It is a relatively cheap (around 200 USD at the time of publication) dongle that has sufficient performance for this project. Another advantage is that the Airspy online community is active and eager to help with problems that can arise. The Airspy has a frequency range of 24-1800 MHz. For measuring frequencies lower than that one can connect a Spyverter R2 dongle to the Airspy. The Spyverter is an upconverter add-on that enables the Airspy to have the frequency range of 1 kHz to 60 MHz.
Specifications:
All specifications are taken from Airspy’s homepage [7]
Airspy R2
• Frequency range: 24-1800 MHz
• Output of 10 MS/s (2,5 MS/s for low power devices such as the Odroid used in this project)
• 12 bit ADC
• 3,5 dB noise figure between 42-1002 MHz Spyverter R2
• Frequency range: 1 kHz - 60 MHz
• Conversion loss: 5,2 dB typical
• Max RF power: 10 dB
3.1.2 RTL-SDR
RTL-SDR was one of the first SDRs that got widely available at a low cost. It is based on a TV tuner dongle chipset that was modified to receive other types of signals than just TV-signals. It still is very cheap at around $25. It is however not as advanced as the other SDRs considered for this project.
Specifications:
All specifications are taken from the RTL-SDR homepage [8]
• Frequency range: 24 - 1766 MHz
• Bandwidth: 3,2 MHz
• ADC resolution 8 bits
• Sample rate: 2.4 megasamples per second (It is possible to tune it to 3,2 MS/s but it will drop packets).
3.1.3 LimeSDR
LimeSDR is manufactured by Lime Microsystems which is a UK-based company. It is relatively expensive at $299 at the time of publication. It has a on-board FPGA so it has good performance that is more than sufficient for this project. The frequency range covers more than what is needed here, without any add-ons.
Specifications:
All specifications are taken from Lime microsystems homepage [9]
• Frequency range: 100 kHz - 3.8 GHz
• 256 MBytes SDRAM
• Bandwidth 61,44 MHz
• Power output up to 10 dBm
3.2 Single-board Computer
3.2.1 Odroid-XU4
The Odroid-XU4 is a small but powerful single-board computer and was the choice of single-board computer in this project. It has two processor chips that are working with the ARM big.LITTLE-architecture. It is capable of running various flavours of Linux. It is also compatible with some add-on boards such as the XU4 Shifter Shield which gives the possibility to run all GPIO pins from 3,3 or 5 V and is using a conventional pin size for the GPIO pins. All specifications are taken from the Hardkernel website [10]
Specifications:
• Processors: Samsung Exynos5422 Cortex-A15 2Ghz and Cortex-A7 Octa core
• 2 USB 3.0 ports and 1 USB 2.0 port
• 1 Ethernet port
• 2 Gigabytes RAM
• 30 GPIO pins
3.2.2 Raspberry Pi 2
Raspberry is a very common single-board computer that is used for many different types of applications. It can be used with a variety of different operating systems but is shipped with Raspian wich is an operating system based on Debian Stretch that is especially made for Raspberry Pi. All specifications are taken from Raspberry Pi’s homepage [11]
Specifications:
• 900 MHz ARM Cortex-A7 processor
• 1 GB RAM
• 4 USB 2.0 ports
• Ethernet port
• 17 GPIO pins
3.3 Antennas
3.3.1 HP 11955A
The HP 11955A is a biconical dipole antenna. Its theoretical radiation pattern can be found in figure 3 in section 2.2.1. Since the characteristic of a dipole antenna is well defined the HP 11955A will be used in this project as an reference antenna.
3.3.2 Schwarzbeck FSH3D
The Schwarzbeck FSH3D contains three individual loop antennas that are aligned ortho- gonally respectively to each other. This makes it possible to measure every polarization regardless of the orientation of both the transmitting and the receiving antennas. When combining the measurements of all three loop antennas the Schwarzbeck can be approx- imated as a single isotropic antenna.
4 Software
GNU Radio is a free tool that can be used to control the input of software-defined radio on a computer. The accompanying GUI is called GNU Radio Companion, or GRC for short. GRC uses graphical signal processing blocks in a flowchart to process the data.
The blocks are used as placeholders for a specific processing action, the behavior of the block is designed by the user to meet the requirements of the system.
When running a program in GNU Radio, several Python files are generated and run based on the flow chart the user has designed. This makes it fairly easy to write own signal processing blocks in GNU Radio if a custom block is needed. The infrastructure behind GNU Radio is however not built in Python but rather in C++. GNU Radio is used in many applications that are in need of signal processing software for SDR, for hobbyists, companies and academics alike. The complete description and specifications of GNU Radio can be found on their website [12].
GNU Radio provides the possibility to include custom built blocks. One set of custom blocks used in this project is the grOsmoSDR blocks developed by Osmocom [13]. These blocks are used to enable GNU Radio to connect with and control software defined radios via USB. The complete documentation of the grOsmoSDR blocks can be found on their website wiki at [13].
5 Development Procedure
This section presents the methods used to develop a system that meets the project de- scription and goals.
5.1 Development Method
The method used to develop a system that fulfill the project description is based on the goal and the available resources. At the start of the project several components where already available. The project members chose to investigate and compare the available components to avoid additional project costs. Only minor and missing components where purchased during the project. The list below describes the general procedure used when deciding on what hardware to use and how to design the software:
1. Specify the components needed to build the complete measurement system 2. Identify the individual requirements of the hardware components
3. Investigate available hardware and decide on suitable components based on price and availability
4. Specify the requirements of the software based on the hardware limits 5. Identify a suitable DSP chain to meet project goals
6. Develop initial software and DSP chain 7. Identify critical system parameters
8. Measure critical parameters and calibrate the software and hardware 9. Perform simulations and calculate expected results
10. Preform tests to verify expected results 11. Analyze results
12. Correct eventual errors
13. Repeat (10-12) until the system has satisfying performance
5.2 Hardware
The project description specifies that a software defined radio should be used and that the complete system is to be mounted on a UAV to do the measurements. The weight of the system should be kept as low as possible in order to maximize the flight time. To keep the system light it should use as few hardware components as possible. The list below specifies the minimum amount of components needed to meet the goal:
• The system weight must not exceed 2 kg for the available UAV to be able to carry the system
• A receiving antenna with frequency range 3-300 MHz that can measure in all po- larizations
• An on-board GPS positioning system
• A SDR with a minimum frequency range of 3-300 MHz
• A low noise amplifier
• A single board computer with sufficient performance to use the SDR and implement a DSP chain.
• Cables, power supplies, and additional fastening equipment needed to connect, mount and use the system
5.2.1 Software defined radio
At the beginning of the project three different SDRs where available, the Lime SDR, Air- spy and RTL-SDR. In section 3 the specifications of the different SDRs can be found.
The Airspy was the most suitable to use because of its compatibility with GNU Radio, high sample rate, resolution and size. It also has a very strong online community support which is a valuable resource when running into problems or questions. The RTL-SDR did not fulfill the frequency range requirement and had a lower sampling rate and resolution.
The Lime SDR is not compatible with GNU Radio and is twice the size of the Airspy.
The Lime SDR also has limited documentation making it hard to use.
5.2.2 Low noise amplifier
When the system is used to characterize antennas with low frequencies the measuring system has to be at a large distance from the antenna under test. The receiving antenna also decrease in efficiency with lower frequencies. This means that the signal at the input of the SDR will be weak. To amplify the signal a 20 dB low noise amplifier is needed.
The internal gain stages of the Airspy has a limit of 10dB and that is not enough to make measurements at large distances with high precision.
5.2.3 Single board computer
At the beginning of the project two different single board computers where available, the Raspberry Pi 2 and Odroid XU4.
The performance of the Odroid is four times that of the Raspberry Pi but at the cost of higher power consumption. The Odroid also has two USB 3.0 ports making it ideal for high data rates from a SDR. It is heavier than the Raspberry Pi but is still more suitable to use.
5.2.4 Antenna
At the beginning of the project only one antenna was available, the Schwarzbeck kFSH3D.
This antenna is especially built for measuring E and H fields with high precision in three polarization. The cost of buying a better antenna was far beyond the project budget and the kFSH3D fulfills all necessary requirements even though it is on the verge of being too big and heavy.
5.2.5 Positioning system
To measure a radiation pattern the UAV must circulate around the antenna. This makes it important to have a positioning system to determine where a specific measurement is made. The GNSS (GPS, Galileo, GLONASS, BeiDou) positioning system was chosen because of its ease of use. It lacks in precision but building a custom positioning system lies outside of the project description and is too time consuming to undertake. In com- bination with a barometer to measure height the GNSS system was deemed sufficient as a positioning system. The GNSS used is the UBLOX M8N and the barometer used is the BME280.
5.3 Software
The development of the software used in the project is heavily dependent on the choice of hardware. The DSP chain was designed as a result of hardware performance and the project goals. The user application was designed with ease of use and performance requirements as main concerns.
5.3.1 Software requirements
The hardware components chosen for the project sets limitations on the software design.
The project description states that the antenna is to be characterized for a specific fre- quency. In practice the antenna under test will be connected to a signal source with known output power and frequency. If the input power is known the total transmitted power, eventual losses and the radiation pattern can be calculated. The DSP process that runs on the measurement system has the ultimate goal of identifying the received power at the chosen characterisation frequency.
The Airspy utilizes the heterodyne principal to down shift the incoming signal with help of an internal oscillator. The frequency of this oscillator is set in the software. However the exact precision of the oscillator frequency is not known. Therefore it is important that the DSP chain has a built in safety margin for small frequency offsets.
The system is required to detect and filter out a signal with a specific frequency. The fre- quency of this signal has to be within the detectable range. To avoid detecting unwanted signals a FIR low-pass filter is used to remove as much of the surrounding frequencies as
possible. To calculate the power of the received signal a FFT is utilized to get the power spectrum.
The system is supposed to measure three different polarizations and the Schwarzbeck antenna has the possibility to switch between 3 perpendicular loop antennas. The DSP process then must keep track of what samples is pertaining to a specific antenna loop.
This can be resolved by having the process take in one set of samples at a time for a specific loop.
Since the Odroid XU4 has a limited computational power it is important that the signal processing is not too demanding. In section 3 it can be seen that the sample rate of the Airspy is fixed at either 2,5 or 10 MS/s with a bit depth of 12 bits represented as a 16 bit integer. When data is in quadrature each sample corresponds to two values, one I-value and one Q-value. The resulting data rate from the Airspy then at minimum becomes
2·16·2.5·106
8·106 [MB/s] = 10 [MB/s].
The Odroid is not fast enough to apply a computationally heavy FIR filter nor a FFT at that data rate. The problem can be solved by using a decimating process that reduces the sample rate.
The generalized DSP chain necessary to meet the demands and limits then becomes:
1. A process that keeps track of what samples belongs to what loop 2. An anti aliasing filter that prepares the signal for decimation 3. A decimation block that reduces the sample rate
4. A lowpass FIR filter to remove most unwanted frequencies
5. A Fourier transform that converts the time domain signal to frequency domain 6. A block that converts the frequency domain signal to a power spectral density 7. A block that detects and stores the power data of the wanted frequency
8. A block that applies correction factors due to non linearities in the Airspy and the antenna factor of the Schwarzbeck antenna
Besides the signal processing of the incoming signal the software must also control the other features of the system:
• Receive the data from the Airspy
• Control the what loop is active in the Schwartzbeck antenna
• Receive GPS and altitude data and transform it to a local coordinate system
• Respond to user input
• Receive data from the UAV
• Store the measurements, UAV and positioning data
• Plot and present the data.
