Vehicular-based Antenna Architectures for Two-tier 5G ... · Vehicular-based Antenna Architectures for Two-tier 5G Wireless Systems by SamerHenry Athesissubmittedinconformitywiththerequirements

Vehicular-based Antenna Architectures for Two-tier 5G Wireless Systems

by

Samer Henry

A thesis submitted in conformity with the requirementsfor the degree of Master of Applied Sciences

Graduate Department of Electrical and Computer EngineeringUniversity of Toronto

© Copyright 2016 by Samer Henry

Abstract

Vehicular-based Antenna Architectures for Two-tier 5G Wireless Systems

Samer Henry

Master of Applied Sciences

Graduate Department of Electrical and Computer Engineering

University of Toronto

2016

This thesis considers the two-tier wireless network consisting of primary and secondary nodes placed in a car

as an array antenna. Many antenna configurations are possible such as linear, circular, 2D square grids, etc. We

consider a MIMO system as a link from the primary to secondary nodes. Antenna element spacing determines

correlation between the elements. We determine the capacity and coverage gains over state-of-the-art vehicular

antennas for different environments.

This thesis also studies the effect of antenna element tilt on the performance. Using an 8× 8 configuration, our

results yield a four-fold enhancement of capacity in urban areas, and a two-fold gain in highways. We are able to

decrease the load on primary nodes by up to 60 per cent. We also note that for 8 × 8 MIMO, the fading handling

capabilities note that it does not matter what angle the vehicle antenna is oriented at.

ii

Acknowledgment

I would like to thank my supervisor Professor Elvino Sousa for his support to the completion of this thesis. I would

also like to thank Dr. Ahmed Al-Sohaily for his helpful comments and informative discussions.

iii

Dedication

I would like to dedicate this thesis to my late Mother Nagwa, who through her perseverance and battle with cancer,

taught me the true reason behind research.

iv

Contents

1 Introduction 1

1.1 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.2 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

2 Antenna Configurations 10

2.1 Past Antenna Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.1.1 Antenna Types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.1.2 Locations on Vehicle to Place Antennas . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.2 Novel Antenna Configurations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.3 Increasing the number of antenna elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

3 System Model 25

3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.1.1 Capacity Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27

3.1.2 Modelled Environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

3.1.3 Coverage Calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

3.2 Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

3.2.1 Antenna array processing with a non-uniform set of element positions . . . . . . . . . . . 46

4 Capacity and Service Availability Improvement Results 54

4.1 Capacity gain via Vehicle Mounted MIMO antennas . . . . . . . . . . . . . . . . . . . . . . . . . 54

4.2 Improved Service Availability via Vehicle Mounted MIMO . . . . . . . . . . . . . . . . . . . . . 73

5 Summary, Conclusion, and Future Work 80

Bibliography 82

v

List of Tables

2.1 Spacing Table for arbitrary antenna configuration with n elements . . . . . . . . . . . . . . . . . 18

2.2 Coupling factor table for arbitrary antenna configuration with n elements . . . . . . . . . . . . . . 19

3.1 Throughput Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

3.2 Inter-Antenna-Element spacing for 2 × 2 MIMO configurations on figure 3.10 . . . . . . . . . . . 45



4.1 Default Channel Parameters and suggested values for Urban Scenario . . . . . . . . . . . . . . . 55

4.2 Default Channel Parameters and suggested values for Suburban and Highway Scenario . . . . . . 74

4.3 Results- 2λ spacing- 25 Km x 25 Km Urban Area . . . . . . . . . . . . . . . . . . . . . . . . . . 77



vi

List of Figures

1.1 Classic hexagonal cell structure layout [2]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Two-tier networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.3 Vehicles that can benefit from two-tier networks [34]. . . . . . . . . . . . . . . . . . . . . . . . . 3

1.4 First cellular system in 1947 [29]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.5 2 × 2 vehicle antenna configurations (from [16]). . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.6 Throughput for each configuration(from [16]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

1.7 Outer structure of a shark-fin antenna on a BMW [36]. . . . . . . . . . . . . . . . . . . . . . . . 8

2.1 Monopole whip antenna on a car [36]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2.2 Structure of the shark-Fin antenna [36]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.3 Shark-fin antenna mounted on vehicle [34]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2.4 Mounting antennas on rear-view mirrors of BMW vehicles [6]. . . . . . . . . . . . . . . . . . . . 13

2.5 Patch antenna [36]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.6 On-glass antenna [36]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2.7 Fractal antenna [36]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.8 Diagram of fractal antenna [36]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.9 Locations to place antennas on vehicles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

2.10 Non-linear antenna element configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

2.11 Configuration of the state of the art shark-fin antenna. . . . . . . . . . . . . . . . . . . . . . . . . 20

2.12 A Proposed 8 × 8 MIMO antenna configuration on a car rooftop. . . . . . . . . . . . . . . . . . . 20

2.13 Proposed 8 × 8 MIMO on car windshields. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.14 Optimum antenna placement model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2.15 8-antenna choice for side-direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.16 8-antenna Choice for back direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.17 RSSI for 1.5 GHz LTE for different passenger car positions. . . . . . . . . . . . . . . . . . . . . . 24

vii

3.1 Multipath for a vehicle in an urban area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.2 Primary and secondary nodes in an urban model. . . . . . . . . . . . . . . . . . . . . . . . . . . 29

3.3 Primary and secondary nodes in a suburban model. . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.4 Multipath for a vehicle in a sub-urban area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

3.5 Primary and secondary nodes in a highway model. . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.6 Multipath for a vehicle in a highway area. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.7 Ray-based method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.8 Received signal (blue) and LTE receiver cutoff power (red). . . . . . . . . . . . . . . . . . . . . . 36

3.9 1 meter squared available area for antenna placement on car roof. . . . . . . . . . . . . . . . . . . 43

3.10 2 × 2 MIMO optimum antenna placement on roof. . . . . . . . . . . . . . . . . . . . . . . . . . . 43

3.11 4 × 4 MIMO optimum antenna placement on roof. . . . . . . . . . . . . . . . . . . . . . . . . . . 44

3.12 8 × 8 MIMO configurations for antenna placement on roof . . . . . . . . . . . . . . . . . . . . . 45

3.13 Rays arriving at a non-uniform antenna element configuration . . . . . . . . . . . . . . . . . . . 46

4.1 A 3-D representation of capacity Vs. antenna elements at the primary node and Vehicle at 1.9 GHz. 56

4.2 Number of vehicle antenna elements Vs. capacity at urban area. . . . . . . . . . . . . . . . . . . 57

4.3 Number of vehicle antenna elements Vs. capacity at suburban area. . . . . . . . . . . . . . . . . 58

4.4 Number of vehicle antenna elements Vs. capacity at highway. . . . . . . . . . . . . . . . . . . . . 59

4.5 Channel capacity gain Vs. vehicle antenna element spacing for urban areas. . . . . . . . . . . . . 60

4.6 Channel capacity gain Vs. vehicle antenna element spacing for suburban areas. . . . . . . . . . . 61

4.7 Channel capacity gain Vs. vehicle antenna element spacing for highways. . . . . . . . . . . . . . 62

4.8 Downlink channel capacity Vs. primary node antenna spacing- Urban area. . . . . . . . . . . . . 63

4.9 Downlink channel capacity Vs. primary node antenna spacing- suburban area. . . . . . . . . . . . 64

4.10 Downlink channel capacity Vs. primary node antenna spacing- highways. . . . . . . . . . . . . . 65

4.11 Channel capacity gain Vs. frequency for 2 × 2 MIMO. . . . . . . . . . . . . . . . . . . . . . . . 66



4.14 Theoretical and simulated correlation amplitude Vs. antenna tilt. . . . . . . . . . . . . . . . . . . 69

4.15 Channel capacity Vs. vehicle antenna polarization Angle for SISO. . . . . . . . . . . . . . . . . . 70

4.16 Channel capacity Vs. vehicle antenna polarization angle for 2 × 2 MIMO. . . . . . . . . . . . . . 71



4.19 Service availability Vs. primary node vehicle element separation distance of 2λ spacing. . . . . . 74

viii

4.20 Service availability Vs. primary node vehicle element separation distance for 4λ spacing. . . . . . 75

4.21 Service availability Vs. primary node vehicle element separation distance for 6λ spacing. . . . . . 76

4.22 Urban cell radius optimization Vs. antenna spacing at 97 per cent service availability. . . . . . . . 78

4.23 Suburban cell radius optimization Vs. antenna spacing at 97 per cent service availability. . . . . . 78

4.24 Highway cell radius optimization Vs. antenna spacing at 97 per cent service availability. . . . . . 79

ix

List of Abbreviations

MIMO: Multiple-Input-Multiple-Output

LTE: Long-Term-Evolution

ITU-R: International Telecommunication Union-Radio

3GPP: 3rd-Generation Partnership Project

BS: Base-Station

MS: Mobile-Station

OFDMA: Orthogonal-Frequency-Division-Multiple-Access

GSM: Global System for Mobile Communications

CDMA: Code-Division Multiple Access

TDMA: Time-Division Multiple Access

CoMP: Coordinated Multipoint

EM: Electromagnetic

AM: Amplitude Modulation

FM: Frequency Modulation

ALU: Alcatel-Lucent

UE: User Equipment

VHF: Very-High Frequency

WLAN: Wireless Local Area Network

GPS: Global Positioning System

SDARS: Satellite Digital Audio Radio Services

x

Chapter 1

Introduction

Over the previous decade, 3G as well as 4G LTE wireless systems sought to enhance the architecture of wireless

networks. We had the classic hexagonal cell layout with a primary node in the center, as shown in figure 1.1.

Figure 1.1: Classic hexagonal cell structure layout [2].

To meet increasing capacity requirements, two-tier networks were introduced. In two-tier networks, the primary

node could connect to a terminal through a secondary node such as a middle antenna. This middle antenna should

1

Chapter 1. Introduction 2

be strong to accommodate the connectivity to the primary node. This thesis considers a MIMO link from primary

to secondary node and discusses the different methods in which to position this secondary node as an array of

elements in order to satisfy the capacity requirements for a vehicle. This thesis will introduce an architectural and

simulation framework for vehicles in 5G cellular systems.

Figure 1.2: Two-tier networks.

Primary nodes in two-tier networks are typically basestations within the classic cell layout that are fixed points

of communication. The primary node is connected to an antenna (with one or multiple antenna elements) that

receives and transmits the signals in the cellular network. Secondary nodes connect to primary nodes and can

range from basestations to handheld terminals as shown in figure 1.2. In the case where secondary nodes act as

repeaters to the signal, they are known as middle antennas because they are in-between the primary node and a

terminal device. This middle antenna should be able to accommodate the connectivity to the primary node. There

are many use cases depending on the deployment of the secondary node, one of which is for use in vehicles. This

thesis discusses the different optimal methods in which to position this middle antenna on a vehicle. Examples of

vehicles that can benefit from two-tier networks are shown in figure 1.3. Street cars and trains have strict routes,

while cars follow different paths.


Figure 1.3: Vehicles that can benefit from two-tier networks [34].

Before two-tier networks, the first cellular system antenna was mounted on vehicles as shown in figure 1.4. The

transmitter and receiver were placed on the car, leading to the inauguration of mobile service. Car phones were

developed later, but the technology development slowed down greatly with the introduction of pocket handheld

devices. Nowadays, mounting an antenna on a vehicle allows us to have a connected vehicle. The term ’Connected

vehicles’ is an expression according to [10] that signify high data rates in the automotive environment while

connected to primary nodes or secodary nodes within the 2-tier network. They are vehicles that are equipped with

Internet access to share this access to other devices in and out of the vehicle.


Figure 1.4: First cellular system in 1947 [29].

The study of connecting vehicles led to an unprecedented demand for various infotainment applications. Ex-

amples of these implementations include safety applications such as collision notification, roadside assistance,

remote door unlock, stolen vehicle tracking, remote vehicle slowdown, optimized navigation and infotainment ap-

plications. However, the methods in which these deployments are currently implemented are not largely practical.

Current vehicle implementations still resemble that of handheld user equipment (i.e.: smartphones). The reasons

will be outlined in the following sections.

The implementations for connected vehicles would not be possible without the Long Term Evolution (LTE)

standard. Over the previous decade, LTE has had important characteristics over its 3rd-generation predecessor. It

can easily adapt to severe channel conditions without the need for complex channel equalization algorithms being

employed. Also, LTE has support for MIMO, and hence higher spectral efficiency, and more efficient bandwidth.

The LTE standard used in 4G communications uses Orthogonal-Frequency-Division-Multiple-Access which is

pivotal in this thesis to allow trivial multiple antenna implementations found in MIMO. LTE supports deployment

on different frequency bandwidths. The current specification outlines the following bandwidth blocks: 1.4MHz,

3MHz, 5MHz, 10MHz, 15MHz, and 20MHz. Frequency bandwidth blocks are essentially the amount of space a

network operator dedicates to a network. Depending on the type of LTE being deployed. An operator may choose

to deploy LTE in a smaller bandwidth and grow it to a larger one as it transitions subscribers off of its legacy

networks (GSM, CDMA, etc.). LTE uses two different types of air interfaces (radio links), one for downlink (from


tower to device), and one for uplink (from device to tower). By using different types of interfaces for the downlink

and uplink, LTE utilizes the optimal way to do wireless connections both ways, which enhances the network and

power efficiency on LTE devices.

For the downlink, LTE uses an OFDMA (orthogonal frequency division multiple access) air interface as

opposed to the CDMA (code division multiple access) and TDMA (time division multiple access) air interfaces.

OFDMA (unlike CDMA and TDMA) mandates that MIMO (multiple in, multiple out) is used. Having MIMO

means that devices have multiple connections to a single cell, which increases the stability of the connection and

reduces latency as well as increasing the total throughput of a connection. There is a significant disadvantage

though. MIMO works better the further apart the individual carrier antennae are. On smaller phones, the noise

caused by the antennae being so close to each other will cause LTE performance to drop. Recent releases of

the LTE standard, referred to as LTE-Advanced, further enhanced the performance of LTE by supporting the

aggregation of multiple radio frequency carriers, Coordinated Multipoint Processing (CoMP) in addition to higher

order Multiple Input Multiple Output (MIMO) antenna configurations for up to 64 elements.

One of the studies in this thesis discusses the effect of amultiple-element antenna tilt in a vehicle on performance

gains. To study this, we need to understand polarization. Polarization is a property of waves that can oscillate

with more than one orientation. In an electromagnetic (EM) wave, both the electric field and magnetic field are

oscillating but in different directions; by convention the polarization of EM waves refers to the polarization of the

electric field. EM waves which can be approximated as a plane waves in free space or in an isotropic medium

propagates as a transverse wave. Both the electric and magnetic fields are perpendicular to the wave direction of

travel. The oscillation of these fields may be in a single direction (linear polarization), or the field may rotate at

the optical frequency (circular or elliptical polarization). In that case the direction of the field rotation, and thus

the specified polarization, may be either clockwise or counter clockwise [30].

These basic concepts will assist in our main objective to find the optimal architecture for vehicular antennas.

We take advantage of 4G and 5G advancements to optimize system performance, namely optimizing capacity and

service availability in urban areas, suburban areas, and highways. It could potentially contribute in decreasing the

load on base stations and saving operational cost and space for major phone companies.


1.1 Literature Review

Not many research publications have addressed the challenge of connecting vehicles using LTE MIMO antennas.

There are different kinds of antennas that have been mounted in vehicles. Chronologically, monopole antennas

are the first and most popular type of antenna, as they are very easy to handle in AM/FM radio applications, easy

to manufacture, easy to implement to the vehicle and very cheap [30]. They are typically called ’whip antennas’

on vehicles. More advanced antennas that can connect to primary nodes use a shark fin antenna which contain

2 closely spaced monopole antennas that form a 2 × 2 MIMO antenna and are strapped onto the roof. The patch

antenna [31] is a flatted monopole. On-glass antennas [36] can be placed on windshields. Glued-foil antennas are

flat structures placed on a metallic foil which is glued to any non-metallic part of the vehicle. Fractal antennas are

also flat but repeat their structure leading to multiple resonances of the antenna [36].

We now look into how effective 2 × 2 MIMO antennas have performed in the research literature. Bell Labs

performed a number of experiment involving different 2×2 antenna configurations and measured throughput. The

number of configurations tested are listed in the following table; there are 13 configurations.

Figure 1.5: 2 × 2 vehicle antenna configurations (from [16]).

The results that they obtained are in figure 1.6. On the horizontal axis, each point represents a configuration,

on the vertical axis we see the throughput for that. For example, you can see that the worst configuration is number

5 with 2 dipoles on the car roof.


Figure 1.6: Throughput for each configuration(from [16]).

The graph implies that although the maximum data throughput exists for some configurations, there is a low

data throughput (< 15Mbps) with a 90 per cent certainty for all configurations which yet again proves there is a

problemwith the existing configuration. The maximum number of antennas that have experimentally been used for

vehicle-mounted MIMO is a 2×2 configuration which performs betters than existing UE levels [16] but Hagerman

et al. notes that,depending on the environment, 2 × 2 MIMO antennas with a spacing of 0.4λ perform poorer

than single antennas at 700 MHz in certain experiments due to the effect of the environment [14]. To resolve that

issue, Thiel et al. introduced an optimized 2 × 2 MIMO antenna mounted on the car rooftop that supports all 3

frequency bands with dimensions of 18.5 cm length, and 6.5 cm height, and was later on developed to be coined

as the shark-fin antenna [6].


Figure 1.7: Outer structure of a shark-fin antenna on a BMW [36].

Figure 1.7 on page 8 shows the outer structure of a shark-fin antenna on a BMW car and figure 1.4 on page 6

shows state-of-the-art realizations of the shark-fin antenna that are basically equivalent to sticking such UE on the

roof of the vehicle. As seen, the blue unit is mounted on top of the car and the antennas inside it are close although

we do not have the space limitation in UE. The shark-fin antenna was proven to perform equal to 2 × 2 antennas

of higher spacing which indicated that optimizing the downlink by increasing the distance than 0.4λ would be

irrelevant to enhancing performance at 746-756 MHz band [3]. However, Sturman et al. indicate in [12] that

shark-fin antennas already contain antennae to support GNSS, FM, and Digital Radio, so careful system design for

RF co-existence modelling within this spatial capacity should take place. New releases of LTE have increased the

number of antenna elements that can be used for MIMO to 64 elements hence we have a spatial limitation when

considering the shark-fin antenna [2]. In [7], [8], an important observation is made concerning the relationship

between antenna spacing and transmit power. With high reference base station transmit power as stated by 3GPP

standards [2], the antenna spacing for lower correlation should be in the range of 10-20λ, so an increasing number

of elements and increasing power will be needed for the vehicle-mounted antenna. The characteristics of a wireless

signal changes as it travels from the primary antenna to the secondary antenna. These characteristics depend upon

the distance between the two antennas, the path(s) taken by the signal, and the environment (buildings and other

objects) around the path. The profile of the received signal can be obtained from that of the transmitted signal

if we have a model of the medium between the two. This model of the medium is called channel model. There

are multiple channel models, one of which was chosen for the simulation in this thesis. The simplest channel is

the free space line of sight channel with no objects between the receiver and the transmitter or around the path

between them. Other empirical path loss models include the Hata Model and its extension the COST 231 Model.

In addition to the COST 231-Hata model, the COST 231 group also proposed another model for micro cells and

small macro cells by combining models proposed by Walfisch and Ikegami [2]. This model considers additional


characteristics of the urban environment, namely, heights of buildings roof, width of roads, building separation,

and road orientation with respect to the direct radio path. The Erceg model characterizes the terrains into three

categories. Category A is a hilly terrain with moderate-to-heavy tree density and has a high path loss. Category C

is mostly a flat terrain with light tree density and has a low path loss. Category B is a hilly terrain with light tree

density or flat terrain with moderate-to-heavy tree density. Another channel model, the Stanford University Interim

model, is a set of 6 channel models representing three terrain types and a variety of Doppler spreads. Another

commonly used set of empirical channel models is that specified in ITU-R models. The recommendation specifies

three different test environments: Indoor office, outdoor-to-indoor pedestrian, and vehicular high antenna. Since

the delay spread can vary significantly, the recommendation specifies two different delay spreads for each test

environment: low delay spread and medium delay spread. The number of multipath components in each model is

different. We choose a ITU-R vehicular model called the ’Ray-based propagation’ model because it is relevant to

the study of vehicular antenna configurations.

1.2 Outline

Over the next chapters of this thesis, we outline antenna configurations in chapter 2 in which we discuss the

properties of car antenna types in detail, as well as different points to position antenna elements on a vehicle.

In chapter 3, we introduce the ITU vehicular model used in simulations and define the environments. It gives a

detailed description of the channel model between primary and secondary. This model is a ray-based model for

MIMO consisting of rays, each having an angle of departure and arrival. Chapter 4 contains results for capacity

and coverage. We place the capacity gain over the shark-fin antenna on the y-axis, and vary antenna element

spacing and frequency on the x-axis, while studying 2 × 2, 4 × 4, and 8 × 8 MIMO. Next, we also study the

coverage as a function of the distance between the primary and secondary node, for different frequencies and

antenna element spacing for 2, 4, and 8 element MIMO. We observe an improvement in capacity and coverage

with more elements and better element spacing. We also observe that lower frequencies are suitable for highways,

and higher frequencies show better performance in urban areas. We conclude the thesis and outline future work

on this topic in chapter 5.

Chapter 2

Antenna Configurations

In this chapter, we study past antenna configurations for vehicles by studying the type of vehicular antennas as

well as available locations on a vehicle to place these elements without interfering with its aesthetics. We then

present a novel antenna configuration for 5G vehicular antenna architectures and answer the question: What can

we do with a large number of elements?

2.1 Past Antenna Configurations

Antenna configurations for vehicles assume two types of antennas according to their location with respect to the

vehicle: Internal antennas and external antennas. Internal car antennas are located inside the trunk, dashboard,

or the windshield of a car. These antennas receive better protection from weather and accidental damage, but

their reception generally is not as clear as external antennas. Installing the antenna inside the car or using an

antenna with a built-in amplifier may help reception performance. External antennas, made of metal or fiberglass,

are typically installed near the hood or trunk of the car. External antennas receive better reception but are more

susceptible to breakage or weather damage. External antennas sometimes come with mechanical retractors that

pull the antenna into the car to protect it when not in use.

10

Chapter 2. Antenna Configurations 11

2.1.1 Antenna Types

Figure 2.1: Monopole whip antenna on a car [36].

Antennas can be placed on the vehicle on many places, but only a few positions are ideal. To define which

position is ideal, this must be considered for the service and frequency band as well as in conjunction with selected

antenna technology. Here we give a short overview about the most popular antenna technologies defined in the

introduction. Firstly, monopole antennas are the most popular antenna type, as they are very easy to handle, easy

to manufacture, easy to implement to the vehicle and very cheap. A monopole antenna consists of an antenna foot

and the rod of some length as shown in the car in figure 2.1. Placing the antenna on the roof of a car requires just

a hole in the metallic structure for the antenna foot and the cable to be attached. The rod can be a stiff metallic

stick of a quarter wavelength (0.25×Wavelength) or longer. There are rods which are telescopic or even flexible.

Monopoles provide a broad range of applications in every frequency band, from VHF sound broadcasting up to

car-2-car communication at 5.9 GHz [29]. The monopole can be placed in the roof center, on an edge of the

roof or on the frame, or even on a bumper or fender. This antenna technology is quickly implemented to nearly

anywhere on the vehicle. In conjunction with a very competitive pricing due to quick development and easy

manufacturing, monopoles are the first choice in the automotive industry. The monopole antenna requires a direct

connection to ground, meaning the metallic structure of the vehicle. The monopole will not work efficiently when

the metallic ground is small in comparison to the wavelength. A minimum of 0.25×Wavelength is required for

the ground plane to neglect effects [30]. However, aesthetically speaking, the rod antenna on a roof or fender

does not satisfy modern design emotions, so car designers try to avoid simple rod antennas on their cars. Espe-

cially for premium car manufacturers, rod antennas are meant to be avoided inmodern car antenna system concepts.


Figure 2.2: Structure of the shark-Fin antenna [36].

Figure 2.2 shows the shark-fin antenna, which is basically treated as a user equipment and is one unit that

contains 2 vertical antenna elements, with 3 cm spacing strapped onto the roof. Figure 2.3 outlines how the

shark-fin antennas are mounted on a passenger vehicle rooftop.

Figure 2.3: Shark-fin antenna mounted on vehicle [34].


Figure 2.4: Mounting antennas on rear-view mirrors of BMW vehicles [6].

Figure 2.4 shows an implementation by BMWwhich shows how theymounted antennas on one of the rear-view

mirrors.

Figure 2.5: Patch antenna [36].

Another type of antenna is called a patch antenna. In principal, a patch antenna is a flattedmonopole. Figure 2.5

shows the principal structure of a patch antenna [31]. A metallic plate of about half wavelength (0.5×Wavelength)

is placed over a metallic ground plane, which are separated either by air or a low loss dielectric material. The

RF is ideally fed directly to the radiating patch. This antenna type is becoming more and more popular in the

automotive industry. At first just for high frequency applications beyond 1 GHz, such as for GPS reception at

1.5 GHz or SDARS satellite broadcasting radio reception at 2.3 GHz, WLAN and Car-to-Car Communication at

5-6 GHz, but nowadays increasingly seen for telephoning antennas for GSM and CDMA at 800/900 MHz. As

this antenna type is unobtrusively flat and can be implemented into the vehicles structure, e.g. behind a fender

or bumper or in a plastic trunk cover. Patch Antennas require a large metallic ground structure, preferably flat of

minimum 1×Wavelength around the antenna. If the vehicle cannot fulfil this requirement at the position with its

surrounding metallic structure, then the patch antenna would provide its own ground plane in the antenna structure.


Figure 2.6: On-glass antenna [36].

On-glass antenna type became popular in the premium car industry in the 1980, when design topics became

increasingly important. With on-glass antennas it becomes possible to hide a number of antennas for the regular

user eye. As most vehicles have glass windows, an antenna structure can be implemented there. The most straight

forward antenna technique for on-glass follows a slot antenna principle as shown in figure 2.6. Here the metallic

car structure becomes part of the antenna and the glass (windscreen, side window or rear window for example) is

used as the slot in the metallic structure. Coupling this slot with a thin conductor, the slot resonates and can be used

as antenna. The feeding position is of some importance. With a clever selection of the feeding position, different

modes can be activated, which can help to excite certain polarities or radiation behaviour [36]. As defrosting

conductor elements are already placed to the rear window, these conductors can be applied for such on-glass

antenna structures. Here, a simple filter is used to separate the DC-current for defrosting from the radio frequency

(RF). The development of such on-glass antennas can take some time, as car structure and glass window holes in

the metallic structure are both parts of the antenna jointly, which influence each other. So changing the shape of

the car will change the on-glass antenna characteristics. But once a suitable structure is found, the manufacturing

process is very cheap.

Similar to slot antennas and patch antennas, glued-foil antennas can be placed on metallic foil, which is glued

to a plastic element. So very flat antenna structures can be produced, which can be placed into rear-view mirrors,

side-mirrors, spoilers, bumper and fenders or any non-metallic parts of the vehicle. However there are some

disadvantages to be mentioned, which are the very narrow usable bandwidth and the problem to connect coaxial

cables to the flat foil antenna. In addition, moisture and spray water is avoided. The connection problem is often

solved by coupling the signal to the structure, but losses must be taken into account. The structure shape is virtually


unlimited and can suit different locations in vehicles.

Figure 2.7: Fractal antenna [36].

A very modern field of antenna technology is summarized as fractal antennas. It has been found that not only

some structures repeat their resonance with n×Wavelength, but also when repeating the structure in itself. This

fractal breakdown of a wire structure or a patch antenna structure leads to multiple resonances of the antenna.

When tuning the structure that many frequency ranges can be used, only one antenna could be used to serve most

of the required services. As the benefit is that only a few antennas need to be placed, which is good for small cars,

the disadvantage is the very limited usable bandwidth where the structure resonates. Figure 2.7 shows an example

of a fractal antenna. Figure 2.8 shows a fractal antenna diagram of operation [36].

Figure 2.8: Diagram of fractal antenna [36].

In order for any of the above antennas to have a good reception in or out a vehicle, some prerequisites must

be fulfilled. The antenna must receive from any direction around the car. If this requirement cannot be satisfied

with one antenna only, then an antenna array (two or more antennas) are considered. In general, an antenna


should be as high over ground on the vehicle as possible. The higher the antenna, the better it can receive and

transmit. The antenna must be integrated into the car easily. The surrounding of the antenna may influence the

antenna performance severely. Hence the materials and distances around the antenna must be considered. As a

rule of thumb, a box of 3 times antenna size around the antenna should be unobstructed. That means for VHF

antennas with 75 cm length, this requirement can never be fulfilled on regular cars, but can be easily achieved

with Telephone or GPS-antennas above 1 GHz [26]. The engine generates spurious noise which can disturb

reception, therefore the antenna should be placed as far away from the engine as possible, but taking all other

requirements mentioned before into account. So in total there would be a trade-off between height over ground, om-

nidirectional reception and reducing spurious emissions influences. Sometimes the polarization of the antenna is

of some importance, as some signals are transmitted specially polarized, e.g. SDARS is left hand circular polarized.

2.1.2 Locations on Vehicle to Place Antennas

We study the structure of a passenger car and determine the different locations to place antennas inside or outside

the vehicle. Figure 2.9 displays the best practices to place antennas to vehicles.

Figure 2.9: Locations to place antennas on vehicles.

Most car manufacturers use the roof to place an antenna. This has an obvious reason, because the roof of

a car is the high above ground and unobstructed. This provides a good reception into every direction. Mostly,

omni-directional reception is required anyway, so placing the antenna on the roof is one of the best options for


most of the vehicles, except convertibles. Some sports cars exhibit a spoiler for better down force on higher speeds.

When the spoiler is made of plastic, it can be used to place antennas inside. Racing cars use this technique for

telemetry communication. Regular hatchback cars can have a little spoiler, in which a number of antennas can

be implemented. This method works exceptionally well and is the second preferred place, whenever a spoiler

is present. Placing antenna structures into windshields, side windows or rear-windows has become very popular

for premium car manufacturers since 1980. As most cars have glass windows, to which an antenna structure can

be applied, it is a cheap but effective method. Here the antenna structure can be either on the glass or along

the frame. The glass itself is usually big enough to inherit large antenna structures or many different antennas.

On-glass antennas require a larger engineering effort but can be manufactured with low costs once the structure

is developed. Especially when the design of a car is of importance and roof antennas would not suite aesthetic

aspects, then on-glass antennas is the way forward. Some of the fenders or bumpers are made of plastic, which

suite for placing the antennas behind as they can offer enough free room. However, special care must be taken for

easy repair, as bumper and fenders can crash. Also take into consideration that engine noise and low height above

ground might degrade antenna performance.

Alternatively to the roof, the trunk cover is a suitable position to contain a number of antennas. Especially

for convertible vehicles, where the roof and screens can be hidden, the trunk cover is advised to place antennas

into. However, the trunk cover must be made of plastic or a double-layer structure with metal frame and plastic

cover on top. Light trucks and sport utility vehicles (SUV) offer huge side mirrors in comparison to normal cars.

The shell of the mirror is mostly plastic. Inside these mirrors a number of higher frequency antennas can be

placed. Some truck side mirrors are large enough to inherit a combination of FM-receiving antenna at VHF-band,

Telephoning antenna from GSM and CDMA systems, a GPS navigation antenna and on top a SDARS satellite

broadcasting antenna. Regular vehicle side mirrors can offer some space for higher frequency services above 1

GHz, e.g. Car-2-Car communication at 5.9 GHz [28].

Antennas receive from the sky optimally, which limits the placements to be 360o unobstructed to sky. Hence

the roof and exposed spoilers are suitable positions [33], in some occasions mirror housings as well. Convertible

cars can utilize the trunk cover or the upper edges of the fender. When looking at various vehicles on the market,

there will be seen a number of antenna solutions on the vehicle structure. The most popular antenna placement is

a monopole rod antenna on the roof of the car, as this method fulfils the most important basic requirements, high

above ground and unobstructed. The radiation characteristic is quasi omni-directional, when the metallic roof is

large enough. For frequencies greater than 100 MHz, which refers to FM-broadcasting reception, TV-reception

and mobile telephoning systems, monopole antennas on vehicle roof are perfectly suited. In addition, they are

very easy to implement and cost less. These are the reason why the majority of the vehicle manufacturers apply


monopole rod antennas.

2.2 Novel Antenna Configurations

Taking 4 arbitrary points of the form Pi = (xi, yi ) representing a non-linear antenna element configuration in

figure 2.10, we calculate the distance ‖di j ‖ between 2 points Pi and Pj .

Figure 2.10: Non-linear antenna element configuration.

This distance between points Pi and Pj is calculated as follows:

‖di j ‖ =

√(x j − xi )2 + (y j − yi )2 (2.1)

For all elements within any antenna configuration, we have a set of inter-element distances from the above

equation, and we construct table2.1 for an arbitrary antenna configuration with n elements:

Table 2.1: Spacing Table for arbitrary antenna configuration with n elements

. P1 P2 Pn

P1 0 ‖d12‖ ‖d1n ‖

P2 ‖d12‖ 0 ‖d23‖

Pn ‖d1n ‖ ‖d2n ‖ 0

The spacing between every 2 elements ‖di j ‖ will help us estimate the antenna coupling factor Ci j of every


element to every other element in the configuration. This in turn would help us to theoretically rate the effectiveness

of each vehicular configuration. From [7], we know the correlation coefficient of elements in a primary node. For

a vehicular configuration operating at a wavelength λ, we translate each ‖di j ‖ on the x-axis to a Ci j on the y-axis

and can construct a coupling factor table2.2 similar to table2.1:

‖di j ‖

λ=⇒ Ci j (2.2)

Table 2.2: Coupling factor table for arbitrary antenna configuration with n elements

. P1 P2 Pn

P1 0 ‖C12‖ ‖C1n ‖

P2 ‖C12‖ 0 ‖C23‖

Pn ‖C1n ‖ ‖C2n ‖ 0

For a incoming wave Xi with frequency f0 arriving at an arbitrary point Pi with phase φi , we have a signal

Xi = cos(2π f0t + φi ) assuming no antenna coupling. With antenna coupling between n elements, we have:

Xi = cos(2π f0t + φi ) −n∑j,i

Ci jcos(2π f0t + φ j ) (2.3)

With that, we are able to differentiate between different configurations based on the inter-element sapcing. In

the next chapter, we will continue our antenna array processing algorithms by calculating the phase difference φi

for each element in the configuration.

After outlining a simple fundamental approach to non-linear antenna configurations, we place these configu-

rations on vehicles. This thesis looks at how to optimize the middle antenna in a 2-tier network. As discussed,

we introduce a better antenna element architecture without the spacing limitation as shown in figure 2.11 which

models the positions of the antenna elements in a regular shark fin antenna. There are many ways to accomplish

and optimize on the current technology. We we look at some of these configurations on a car roof and simulate its

effectiveness.


Figure 2.11: Configuration of the state of the art shark-fin antenna.

Figure 2.12 shows 8 antenna elements which are the maximum number of elements that are simulated in this

thesis.

Figure 2.12: A Proposed 8 × 8 MIMO antenna configuration on a car rooftop.

Sufficient antenna spacing, in addition to low channel and antenna correlation at different antenna elements, is

required to realize the gains of higher order MIMO techniques. 2D array configurations are being standardized in

LTE REL-13 and that we can have up to 64 elements.


Figure 2.13: Proposed 8 × 8 MIMO on car windshields.

Ofcourse, there are other ways to connect the elements even further apart: figure 2.13 shows an 8×8 once more

but placed on the front and back windshields to provide an added dimension of decreasing antenna correlation

and increasing received power. For the remainder of this thesis, we consider the roof configuration because it has

a smaller area and with that, we will prove that the satisfactory results can be further optimized with different

configurations. Although other factors such as aerodynamics and antenna shielding have an effect, we will solely

prove that the strongest effects are to be observed once spacing, antenna tilt, and the number of mounted antenna

elements are optimized.

Sometimes the car designer does not like antennas on the structure. Then hidden antenna concepts must be

used. A very commonway is to place the antennas into a plastic spoiler. As the spoiler is usually high above ground

and unobstructed for good airflow, it fulfils basic requirements for good reception. Racing cars use the spoiler

structure for telemetry communications. Sometimes regular hatchback cars contain a spoiler, into which VHF

antennas for sound and TV broadcasting, GPS-antennas and nowadays Car-2-Car Communication antennas can be

easily integrated. Of cause it is also possible to apply telephoning and satellite broadcasting SDARS antennas into

spoiler structures. When there is no sportive spoiler but the design aspects require invisible antennas, the antennas

can be placed into the screens. Here the slot antenna concept is used, which requires some development but once

the structure is found, it is easy and cheap to manufacture. Usually the rear window is used when the engine is

in front, offering the maximum distance from spurious emission noise. In rare occasions where the engine is in

the backside of the car, e.g. Porsche 911 model, then the windscreen is used. However, the antenna structure

shall not effect visibility then. Alternatively sidewindows can be used for antenna structures but often they are too

small for multiple antennas and VHF-antennas. As foil- and fractal antennas became popular for mobile phone

antennas, this concept has also been applied to the car industry. Today we find fractal foil glued antenna structures


in rearview mirrors, e.g. garage door opener, car entry systems and Bluetooth- / WLAN antennas [34]. Light

trucks and SUVs comprise some large side mirrors, in which even low frequency antennas can be placed, such as

for FM-reception on VHF bands and mobile telephoning systems. The only disadvantage is that in a case of an

accident with damaging the side mirror, the wireless service can be damaged as well or reception becomes poor.

In order to avoid a complete loss of service, either both side mirrors can be equipped or a combination with other

antenna locations can be used. Multiple antennas in this case can provide better reception. This method is known

as diversity reception. We can also place 2 antennas in the rear-window, one is vertically dominated polarized, the

other horizontally dominated polarized. Both rear-window antennas provide good reception, as nearly all phase

components of the multipath signal can be received. However, having both antennas in the rear of the vehicle,

the reception to the front side can be shadowed. This can be overcome with a third antenna, either a monopole

on rooftop or a structure in the spoiler (if the vehicle has one) or using a structure in the front fender. For higher

frequency ranges, the mirrors can be used, e.g. fractal foil structure glued to the plastic housing. Combining

rear-window and side-mirror antennas, a fully decoupled but omni-directional reception is achieved where most

of the multipath signal components can be processed.

2.3 Increasing the number of antenna elements

With a large number of elements, we can do many things such as beamforming, higher order SDMA, MIMO, and

diversity combining. Beamforming may dictate a certain configuration. The more the beamforming, the higher

the gain, meaning that we can go further for the same power. For higher order SDMA, we can reduce interference.

With more elements, we can incorporate MIMO and perform spatial diversity. Finally, with more elements we can

achieve diversity combining which is a method to reduce fading by the channel by switching to the best antenna(s).

Figures 2.14, 2.15, and 2.16 shows a possible application of diversity combining where different antenna element

types mounted on various points of the car turn on and off according to the cell tower location.

Figure 2.14: Optimum antenna placement model.


Figure 2.15: 8-antenna choice for side-direction.

There is a trade-off. With diversity combining you want them as far as possible but with beamforming you

need optimally half a wavelength and good configurations.

Figure 2.16: 8-antenna Choice for back direction.

cos Since there are no commercially available LTE vehicular antennas except on the car roof, I performed

a series of field trials to measure received LTE signal strengths in different sections of the passenger car in a

parking lot in Downtown Toronto near College and Bathurst. Using the Telus Network at 1.5 GHz, we obtained

the following RSSI shown in figure 2.17.


Figure 2.17: RSSI for 1.5 GHz LTE for different passenger car positions.

By rotating a stationary vehicle, we notice that the roof of the car shown in the 2nd and 3rd histograms are not

the best place to place antennas. We can make use of points all around the car for placing antennas that harmonize

with the car aesthetics.

Chapter 3

System Model

In this chapter we study the MIMO link and model it as an ITU-R Ray-based propagation channel model. We

use this model to compute channel capacity and service availability (coverage). We also outline the parameters

in the model which are varied throughout the simulation. These parameters are mainly the number of antenna

elements, antenna element spacing, polarization angle, and cross-polarization discrimination which is a property

of the channel.

3.1 Background

The link between the base station and the array antenna at the vehicle is a MIMO link. Since we are considering

the downlink, the transmitter is located at the base station and the receiver is located at the vehicle. Let nt be the

number of antenna elements at the transmitter and nr be the number of elements at the receiver. In a MIMO system

independent signals are transmitted at the different elements, and as a result of the multi-path propagation the

signal received at each of the receiver antenna elements depends on the transmitter antenna element from which it

was transmitted. The overall channel, involving all the transmitter and receiver antenna elements is described by

a matrix H = (hi j ) . The channel is modeled by a ray model. From each transmitting element there is a ray and

at each receiving antenna element there are multiple arriving rays. Each of the rays has a transmission delay that

depends on the path length and a different Doppler frequency shift that depends on the velocity of the vehicle and

the arrival angle. The characteristics of the delays will depend on the environment in terms of the signal reflecting

structures including terrain and man-made structures. We refer to these characteristics in terms of delay spread.

Depending on the ray arrival angles we can characterize the signal at the receiver in terms of Doppler spread.

Generally depending on the Doppler spread the above path gains hi j will be time varying. However over a short

time interval, which depends on vehicular speed, they can be modeled as being constant. The receiver signal over

25

Chapter 3. System Model 26

one symbol can be represented by a vector which we denote by y. The transmitted signal is denoted by the vector

x. We can then write the received signal as y = H x + w, where w is a noise plus interference vector. Generally

these vectors, path gains, are represented by complex numbers and the these vectors are vectors in a complex

space, i.e. x ∈ Cnt , y ∈ Cnr , and w ∈ Cnr . We assume that w is a complex noise vector modeled as a complex

Gaussian vector w ∼ CN (0, N0Inr ), where Inr is the identity matrix nr × nr . The channel matrix is also an nr × nt

matrix over the complex numbers. We write H ∈ Cnr×nt . Taking a variable S to be the number of primary node

antenna elements, U to be the number of secondary node elements, and N as the number of multipaths, we use

the following equation for the total channel gain for each time instant as a sum of the multipaths, followed by the

channel matrix.

HS,U (t) =N∑n=1

HS,U,n (t, τn ) (3.1)

H (τ, t) =

*..........,

h1,1(t) h1,2(t) · · · h1,nt (t)

h2,1(t) h2,2(t) · · · h2,nt (t)...

.... . .

...

hnr ,1(t) hnr ,2(t) · · · hnr ,nt (t)

+//////////-

(3.2)

Here is where we present the ray based ITU propagation model [1]. The channel gain for every element in the

above matrix is obtained with the following equation:

hu,s,n (t) =√

Pn

M

∑Mm=1

G(v) (θn,m,AoD)

G(h) (θn,m,AoD)

T

×

exp( jΦv,vn,m )

√r1exp( jΦv,h

n,m )√

r2exp( jΦh,vn,m ) exp( jΦv,v

n,m )

×

G(v) (θn,m,AoA)

G(h) (θn,m,AoA)

×

exp( j kdssin(θn,m,AoD)).exp( j kdusin(θn,m,AoD)).exp( j kvcos(θn,m,AoD − θv )t) (3.3)

where:

Pn : nth path power

N : number of paths or clusters

M: subpaths per-path

S: number of Base-station linear array elements

U: number of Base-station linear array elements

r1: power ratio of waves of nth path leaving the base-station in the vertical direction and arriving at the Vehicle in

the horizontal direction (vertical-horizontal) to those leaving in the vertical direction and arriving in the vertical

direction(vertical-vertical)

r2: power ratio of waves of nth path leaving the base-station in the horizontal direction and arriving at the Vehicle


in the vertical direction (horizontal-vertical) to those leaving in the vertical direction and arriving in the vertical

direction(vertical-vertical)

Φ(x,y)n,m : phase of mth subpath of nth path between x component (horizontal:h or vertical:v) of the Base-station

element and the y-component (horizontal:h or vertical:v) of the Vehicle antenna element

θn,m,AoD: Angle of Departure for mth subpath of nth path

θn,m,AoA: Angle of Arrival for mth subpath of nth path

G(x)BS (.): antenna gain for each Base-station antenna element in the x-direction (horizontal:h vertical:v)

G(x)MS(.): antenna gain for each Vehicle antenna element in the x-direction (horizontal:h vertical:v)

ds is the distance of antenna element from the reference Base-station antenna element.

du distance of antenna element from reference Vehicle antenna element

v: magnitude of the vehicle velocity

θv angle of vehicle velocity

There are a number of variables of interest in the above equation. du is the separation distance of the antenna

element index u in the vehicle MIMO array. By using different configurations in different points of the vehicle,

we are able to study du as a variable. In this equation, it is assumed that the surface is completely flat, neglecting

shielding and fading as a result of metallic surfaces. The 2 variables r1 and r2 are reciprocals of the cross

polarization discrimination which is a measure of polarization purity of an electromagnetic wave and is a property

of the channel. These 2 variables differ from one environment to the other. Dense urban area structures affect the

wave polarization more than highways. The third variable of interest is G(x)MS(.) which is the antenna gain for each

vehicle antenna element tilted at an angle α with the vertical axis. This will help us to measure the capacity gain

for antennas mounted on the windshield at an angle compared with the flat vehicle rooftop antenna elements which

could be flat on the surface at an angle of 90 degrees with the vertical axis. We will use the above model equation

varying these 3 sets of variables to calculate the channel capacity for different vehicular antenna configurations.

3.1.1 Capacity Calculation

To compute the channel capacity, it is important to note that we choose the time slot for scheduling to be less than

the static channel coherence time which is defined (in an order of magnitude sense) as the interval over which the

channel gain changes significantly.

This Ray-based model is a time-discrete model where noise is i.i.d. from a zero-mean and variance N Gaussian

distribution such that: if the noise variance is zero, the receiver correctly receives the transmitted symbol and

theoretically has an infinite capacity [20] as shown here as the Information capacity of a Gaussian channel with


power P and noise N :

C = maxp(x):E[X2]≤P

I (X ;Y ) =12log(1 +

PN

) (3.4)

The mean capacity for SISO Systems using ρ as the average Signal-to-Noise (SNR) ratio at the receiver and h11

as a Rayleigh chi-squared distribution is:

C = EH [log2(1 + ρ |h11 |2)] (3.5)

MIMO channels increase capacity without altering the transmit power or bandwidth by utilizing nt transmit

antennas and nr receive antennas. MIMO capacity is calculated via the channel impulse response matrix H

between the j th transmit antenna and ith receive antenna from equation 3.2

By diagonalizing the product of the channel matrix and its Hermitian transformation from [21] we get:

HHH = EΛEH (3.6)

where E is the eigenvector matrix with orthonormal columns and Λ is the eigenvalue diagonal matrix. Hence

MIMO Capacity is expressed as:

C = EH

{log2[det(Inr +

ρ

ntEΛEH )]

}(3.7)

3.1.2 Modelled Environments

We will define the environment used in our simulations. In an ideal environment, we would have beamforming,

and much better ways to handle multipath. But when we consider the practical effects of multipath. We outline our

cellular communication system and the associated environments used in this research. We simulate 3 environments

for testing our proposed model to connect vehicles optimally and investigate whether each environment scenario

has an optimum configuration of its own, or if there is one configuration that fits all environments.


Figure 3.1: Multipath for a vehicle in an urban area.

Urban Environment: We start by defining the urban environment. This environment is characterized by the

classical hexagonal cellular layout with primary nodes in the middle of each cell and vehicles moving within the

vicinity of the cell with a multipath shown in figure 3.1.

Figure 3.2: Primary and secondary nodes in an urban model.

We have multiple structures that create a number of paths and sub-paths and decreases chances of a line-of-

sight. For urban areas we have about 30 active users and we use a proportional fair scheduling algorithm [25] to

be able to model the average throughput per user (vehicle). The Urban area is also characterized by having a 1

km radius as stated in the ITU model which we use in this research. We,however do not account for the dense

traffic areas and assume an equal distribution of active vehicles within the cell radius. Figure 3.2 shows the urban

structure layout.


Figure 3.3: Primary and secondary nodes in a suburban model.

Suburban Environment: This environment shares the same classical hexagonal layout as in the urban environ-

ment case, but has a 3 km cell radius instead and within this radius, there are about 20 active vehicles that spawn

this high surface area as shown in figure 3.3.

Figure 3.4: Multipath for a vehicle in a sub-urban area.

We do not take the line of sight into our considerations in our calculations as per the Spatial Channel 3GPP

model requirements and assume a multi-path shown in figure 3.4. We do not expect a different behavior by

changing the model however. It is important to note that due to lesser scattering from multipath than in urban

areas, it is expected to observe higher activity from the 2 levels of interfering primary nodes.


Figure 3.5: Primary and secondary nodes in a highway model.

Highway Environment: This environment, as portrayed in figure 3.6 and structurally in figure 3.5.

Figure 3.6: Multipath for a vehicle in a highway area.

We have a highway environment where there is no rigid cellular stricture, but primary nodes are placed on the

side of the road, and provide service availability regularly to a 3 km radius. Hence, primary nodes are approxi-

mately placed 6 km away from each other and we have tentatively about 5 active vehicles within each cell radius.


Figure 3.7: Ray-based method.

As mentioned previously, we use a ray-based propagation model which is shown in figure 3.7 and is charac-

terized as follows. The Angle of Departure and Angle of Arrival follow a uniform distribution and model the

angles of arrival and departure that are formed with respect to the vehicle and primary node antenna broadsides

respectively. The broadside of an array of antennas is the perpendicular line to the axis containing the array

elements. In our case due to the presence of 4 × 4 and 8 × 8 configurations that at times form a rectangular 2D

configuration, the broadside is the axis along the length of the vehicle for both angles. We will be using both

of these angles when we calculate the effect of antenna tilt on capacity and service availability. The number of

multipath rays are another characterization as defined by the ITU model. There are 6 paths or rays according to

the model that all share the same power and travel different distances and hence, have different random phases.

Finally, the number of subrays or subpaths in whih we have 20 subpaths in this model that stem from each of the 6

multipaths and share the same phase and Angle of Departure/Arrival and are summed at the receiver side. Hence,

to model the spatial diversity and multiplexing capabilities of MIMO antennas in vehicles traveling at different

velocities in different scenarios and frequencies, this simulation framework that utilizes the parameters in SCM

3GPP assume a Non-Line of sight ray-based method in which, for each drop over a number of time frames assumes

a top-view 2-D scenario where we neglect the effect of shielding on the car roof. We take into consideration the

Cross-polarization and the Co-polarization of the antennas and the fact that the primary and secondary nodes are

both assumed to have ideal tilted cross-polarized, co-located dipoles forming dipole pairs, doubling the number

of antenna elements.

The spectral efficiency of our system is defined as η = R/W(bits/s/Hz). R is the effective throughput and


W is the Bandwidth. The effective throughput is also defined by the number of active users R = Users ×

Throughput Per User. We use an proportional fair scheduler throughout the thesis within the simulations such

that users are subsequently distributed within the cell radius, and the throughput per user is calculated. This

scheduling mechanism is explained in the following subsection which will outline why we chose the proportional

fair scheduling for the users.

It is essential to define scheduling. In the medium access control (MAC) layer of the primary node, the

functionality of the scheduler is to distribute the radio resources among secondary nodes served by a given cell and

represents methodology for radio resource assignment. The throughput of each secondary node and the throughput

of the entire cell area are affected by the methodology selected by the scheduling algorithm.

In 3GPP LTE networks, there are three basic scheduling algorithm types. They can be easily compared on

the basis of fairness and overall throughput. One of the simplest scheduling algorithms is a Round Robin (RR)

scheduling. RR provides fairness and identical priority among all secondary nodes within a cell. It assigns the

radio resources in equal time slots and in an ordered manner. RR schedules resources fairly, regardless of taking in

consideration of the channel state conditions experienced by different secondary nodes. However, it is less efficient

in providing a high data rate to secondary nodes. Consequently, it wastes some resources because it schedules

resources from/to secondary nodes while the secondary nodes are suffering from severe deep fading and less than

the required threshold.

An opportunistic scheduler such as the Maximum Rate (MR) scheduling algorithm, on the other hand, prior-

itizes secondary nodes which have favorable channel state condition. In other words, this scheduling algorithm

schedules the secondary nodes that have higher signal to interference plus noise ratio (SINR) above the required

SINR threshold whereas it does not schedule those secondary nodes which experience sever channel fading. As

a result, the MR scheduling algorithm provides higher capacity and throughput than any other kind of scheduling

algorithms. However, it completely ignores fairness among secondary nodes within a cell. It is well known

in wireless cellular systems that secondary nodes located in different distances have different fading conditions.

Consequently, scheduling the secondary nodes that have high SINR leads to unfair resource allocation amongst

secondary nodes

A Proportional Fair scheduling algorithm (PF) provides balance between fairness and the overall system

throughput. It was first presented in code-division multiple access high data-rates (CDMA-HDR), but is now

used extensively in OFDMA based systems as well. The algorithm tries is provide fairness among secondary

nodes while maximizing the system capacity. This is achieved by means of exploiting the multiuser diversity over


temporally independent channel fluctuations.

The PF algorithm functions as follows: first, the primary node obtains the feedback of the instantaneous

channel quality condition (CQI) for each secondary node k in time slot t in terms of a requested data rate Rk,n (t).

Then, it keeps track of the moving average throughput Tk,n (t) of each secondary node k on every physical resource

block (PRB) n within a past window tc length. The parameter tc controls the system latency, that is, if tc is

large, the scheduler approaches MR algorithm; if tc becomes small, the scheduler becomes RR algorithm. The

scheduling mechanism gives a priority to the secondary node k∗ in the t th time slot and PRB n that satisfy the

maximum relative channel quality condition:

k∗ = argmaxk=1,2, ..,K[Rk,n (t)]α

[Tk,n (t)]β(3.8)

Different parameters yield different algorithms as shown below:

• If α = 1, β = 1, it describes the PF algorithm.

• If α = 1, β = 0, it describes the MR algorithm.

• If α = 0, β = 1, it describes the RR algorithm.

We assume a user density of:

• 30 active vehicles/cell for urban areas.

• 20 active vehicles/cell for suburban areas.

• 5 active vehicles/cell in highways.

Assuming no special cases of opportunistic beamforming and a frequency reuse factor of 1, let us understand the

peak data rate for LTE throughput with the following LTE system configuration:

Table 3.1: Throughput Configuration

Bandwidth of channel 20 MHzResource Elements per sub-frame 16800Number of bits per modulation symbol 6 bits

With this configuration, the theoretical Peak Data rate is the number of resource elements per subframe times

the number of bits per modulation symbol.


3.1.3 Coverage Calculation

To compute the service availability (coverage), firstly we define the received power Pr of an EM signal which is

derived and modified from the Friis Free Space Equation Pr = PtGtGrλ2

(4πd)2 .

where:

Gt and Gr are transmit and receive antenna gains(no units).

λ is the wavelength

d is the T-R separation

Pt is the transmit power (dBm)

Pr is the received power (dBm)

To compute the service availability of the baseline (SISO) antenna and MIMO antennas we assume 2 environ-

ments: an urban area and a highway suburban area, and use the Macro/Micro cell propagation models respectively

for 3GPP for buildings of nearly uniform height which is shown as follows (Figure of building). These models

differ from the free space path loss because it takes scattering, reflection, and refraction of waves of different areas

into consideration. We will use suitable parameters for the highway scenario in the following section on simulation

parameters, where an optimized service availability is modelled. Using this equation for calculating Pr we obtain

Pr = Pt − PL(dBm)


Figure 3.8: Received signal (blue) and LTE receiver cutoff power (red).

To obtain the service availability following calculation of the received power, the signal cutoff times are

measured as shown in figure 3.8. From [2], the floor signal that LTE detectors can receive is -106 dBm [1],

the service availability will be calculated as the percentage of time when the signal is > −106 dBm to < −106

dBm as shown in the below equation: Service Availability = 100 − (Cutoff TimeTotal Time × 100). To further understand

how to measure the received signal quality, RSRP and RSRQ should be mentioned. In cellular networks, when a

mobile moves from cell to cell and performs cell selection/reselection and handover, it has to measure the signal

strength/quality of the neighbor cells. In an LTE network, a secondary node measures two parameters on reference

signal: RSRP (Reference Signal Received Power) and RSRQ (Reference Signal Received Quality). A secondary

node also measures two parameters on reference signal:

RSSI or Received Signal Strength Indicator:The carrier RSSI (Receive Strength Signal Indicator) measures

the average total received power observed only in OFDM symbols containing reference symbols for antenna port

0 (i.e. OFDM symbol 0 and 4 in a slot) in the measurement bandwidth over N resource blocks. The total received

power of the carrier RSSI includes the power from co-channel serving and non-serving cells, adjacent channel

interference, thermal noise, etc... Total measured over 12-sub-carriers including RS from serving cell.

RSRP or Reference Signal Received Power: RSRP is a RSSI type of measurement. It is the power of the LTE

Reference Signals spread over the full bandwidth and narrowband. A minimum of -20 dB SINR (of the S-Synch


channel) is needed to detect RSRP/RSRQ

RSRQ or Reference Signal Received Quality: Quality considering also RSSI and the number of used Resource

Blocks (N)RSRQ= (N×RSRP) /RSSImeasured over the samebandwidth. RSRQ is aC/I type ofmeasurement and

it indicates the quality of the received reference signal. The RSRQ measurement provides additional information

when RSRP is not sufficient to make a reliable handover or cell reselection decision. In the procedure of handover,

the LTE specification provides the flexibility of using RSRP, RSRQ, or both.

3.2 Parameters

In this section, we fundamentally outline the parameters used in the simulation. During the Primary Node-Vehicle

link there are a large number of phenomena that take place and affect the signal reception like the path loss, the

shadow fading, the fast fading, the Doppler shift etc. For that reason, there are many parameters and variables that

should be taken under consideration during a simulation in order for it to produce reliable results for each scenario.

The number of primary node antenna elements is essential to study the MIMO link from primary to secondary

nodes. The Primary nodes (Base Stations) referred here are fixed base stations using directional antennas to reduce

unwanted signals. According to the current simulation layout, there are two environments: in the urban or city

area where the base station will have other interfering primary nodes around it to provide service availability and

capacity to a densely populated area. The effect of increasing the number of primary node antennas shows a linear

relationship. Increasing the number of antennas is a result of the strict delay constraint and a large coherence time

that may prevent us to exploit time diversity and its complementary interleaving and coding over several coherence

time periods. Hence, antenna diversity, which is also known as spatial diversity, is achieved by placing multiple

antennas at the transmitter which is the primary node in this case. The antennas placed sufficiently far apart

can enhance the channel gains between different antenna pairs which fade more or less independently, creating

independent signal paths.

The required antenna separation depends on the local scattering environment, whether in urban or highway

environments, as well as the carrier frequency. The concept of interest here is transmit diversity when we increase

the number of primary node antenna elements and understand how that achieves diversity and realizes channel

gains. We will discuss the simplest, and yet one of the most elegant schemes for this explanatory model: the

so-called Alamouti scheme. This is the transmit diversity scheme which was proposed in several third-generation

cellular standards. The Alamouti scheme is designed for two transmit antennas and can be generalized to more

than two antennas as seen below.


Alamouti Scheme

The Alamouti scheme is a scheme where we transmit from 2 antennas and receive at a single antenna. Taking the

input x with channel gain h and noise w, there is a symbol y at the receiver at index m and the relationship is this

equation. With flat fading, the two transmit, single receive channel is written as:

y[m] = h1[m]x1[m] + h2[m]x2[m] + w[m] (3.9)

where h1 is the channel gain from transmit antenna 1. The Alamouti scheme transmits two complex symbols u1 and

u2 over two symbol times: at time 1, x1[1] = u1, x2[1] = u2,and at time 2: x1[2] = −u∗2, x2[2] = u∗1. If we assume

that the channel remains constant over the two symbol times and set h1 = h1[1] = h1[2],h2 = h2[1] = h2[2], then

it can be written in matrix form:

[y[1] y[2]] = [h1 h2]

u1 −u∗2

u2 u∗1

+ [w[1] w[2]] (3.10)

Since the interest is in detecting u1, u2, so the equation is rewritten as:

[y[1] y[2]∗]T =

h1 h∗2

h∗2 −h∗1

[u1 u2]T + [w[1] w[2]∗] (3.11)

It is observed that the columns of the square matrix are orthogonal. Hence, the detection problem for u1, u2

decomposes into two separate, orthogonal, scalar problems. We project y onto each of the two columns to obtain

the sufficient statistics:

yi = | |h| |ui + wi, i = 1, 2 (3.12)

where h = [h1, h2]T and w1 ∼ CN(0, N0) and w1, w2 are independent. Thus, the diversity gain is 2 for the

detection of each symbol. Compared to the repetition code, two symbols are now transmitted over two symbol

times instead of one symbol, but with half the power in each symbol (assuming that the total transmit power is the

same in both cases).

The Alamouti scheme works for any constellation for the symbols u1, u2 but suppose now they are BPSK

symbols, thus conveying a total of two bits over two symbol times. In the repetition scheme, we need to use

4-PAM symbols to achieve the same data rate. To achieve the same minimum distance as the BPSK symbols in

the Alamouti scheme, we need five times the energy per symbol. Taking into account the factor of 2 energy saving


since we are only transmitting one symbol at a time in the repetition scheme, we see that the repetition scheme

requires a factor of 2.5 (4 dB) more power than the Alamouti scheme. Again, the repetition scheme suffers from

an inefficient utilization of the available degrees of freedom in the channel: over the two symbol times, bits are

packed into only one dimension of the received signal space, namely along the direction [h1, h2]T . In contrast,

the Alamouti scheme spreads the information onto two dimensions âĂŞ along the orthogonal directions [h1, h∗2]

and [h2, −h∗1]T

Extending the Alamouti Scheme to Multiple Transmitters

Consider a space-time code as a set of complex codewords Xi , where each Xi is an L by N matrix. Here, L is the

number of transmit antennas and N is the block length of the code. For example, in the Alamouti scheme, each

codeword is of the form:

u1 −u∗2

u2 u∗1

(3.13)

with L=2 and N=2. Each codeword in the repetition scheme is in the form:

u 0

0 u

(3.14)

Generally, any block length L time diversity code with codewords xi translates into a block length L transmit

diversity code with codeword matrices Xi , where:

Xi = diag{xi1, ..., xiL } (3.15)

For convenience, the codewords are normalized so that the average energy per symbol time is 1, hence SNR =

1/N0. Assuming that the channel remains constant for N symbol times, we can write:

yt = h∗X + wt (3.16)

where:

y :=

y[1]...

y[N]

h :=

h∗1...

h∗L

w :=

w[1]...

w[N]

To bound the error probability, consider the pairwise error probability of confusing XB with XA, when XA

is transmitted. Conditioned on the fading gains h, we have a vector Gaussian detection problem where we are


deciding between the vectors h∗XA and h∗XB under additive circular symmetric white Gaussian noise. A sufficient

statistic is<{x × y}, where v := h × (XAâĹŠXB). The conditional pairwise error probability is:

PXA → XB = E[Q(

√SNR h × (XA − XB)(XA − XB) × h

2)] (3.17)

The matrix (XA − XB)(XA − XB)∗ is Hermitian and can be distinguished by a unitary transformation meaning

that it can be written as XA − XB)(XA − XB)∗ = UΛU∗ where U is a unitary matrix and Λ = diag{λ21, . . . , λ2L }.

The λL are the singular values of the codeword difference matrix XA − XB . So, the pairwise error probability can

be written as:

PXA → XB = E[Q(

√SNR

∑Ll=1 | h̄L |

2λ2L2

)] (3.18)

where h̄ := U × h. In the Rayleigh fading model, the fading coefficients hl are i.i.d. CN (0, 1) and then h̄ has the

same distribution as h, we can bound the average pairwise error probability as:

PXA → XB ≤

L∏l=1

11 + SNR λ2L/4

(3.19)

If all the λ2L are strictly positive for all the codeword differences, then the maximal diversity gain of L is achieved.

Since the number of positive eigenvalues λ2L equals the rank of the codeword difference matrix, this is possible

only if N ≥ L. If all the λ2L are positive then:

PXA → XB ≤4L

SNRL∏Ll=1 λ

2l

=4L

SNRLdet[(XA − XB)(XA − XB)∗](3.20)

and a diversity gain of L is achieved. The coding gain is determined by the minimum of the determinant

det[(XA − XB)(XA − XB)∗] over all codeword pairs which is called the determinant criterion.

It is important to note that the Alamouti transmit diversity scheme has a particularly simple receiver structure.

Essentially, a linear receiver allows us to decouple the two symbols sent over the two transmit antennas in two time

slots. However the purpose of the above section is to outline the effects of increasing the number of base station

antennas while transmitting in the downlink to one or more vehicles. We will observe this fleeting behaviour in

the results section.


Parameters used in simulation

The number of vehicle antenna elements is an important parameter to determine whether increasing the number of

elements at the primary and secondary node would result in significantly better performance. These are the number

of vehicle antenna elements on a car roof-fender-windshield-etc. In this simulation, these antenna elements would

represent the receive antenna and would be placed primarily on the car roof although it was mentioned that there

are potentially different combinations and positions in which to place these antennas. These antennas can be

integrated antennas or multiple monopole antennas.

Through this simulation, we compare Single-Input-Single-Output (SISO) antennas with 2, 4, and 8 element

MIMO antennas. As we will see in the results section, the effect of increasing the number of vehicle antenna

elements results in a linear increase to the capacity and a bigger Service Availability. We can model the linear

relationship in the form: y = k x where y is the capacity and x is the number of vehicle antenna elements. To

understand the significance of increasing the number of receive antennas, let us redefine what receive diversity

means.

Receive Diversity

In a flat fading channel with 1 transmit antenna and L receive antennas, the channel model is written as:

yl [m] = hl [m]x[m] + wl [m] l = 1, . . . , L (3.21)

where the noise wl [m] ∼ CN (0, N0) and is independent across the antennas. We would like to detect x[1] based

on y1[1], . . . yL[1] This is exactly the same detection problem as in the use of a repetition code and interleaving

over time, with L diversity branches now over space instead of over time. If the antennas are spaced sufficiently

far apart, we can assume that the gains hl [1] are independent Rayleigh, and we get a diversity gain of L.

With receive diversity, there are actually two types of gain as we increase L, this is seen via the error probability

of BPSK conditional on the channel gains:

Q(√2| |h| |2SNR) (3.22)

We can break the total received SNR conditioned on the channel gains into a product of two terms:

| |h| |2SNR = L.SNR.1L| |h| |2 (3.23)

The first term corresponds to a power gain (also called array gain): by having multiple receive antennas and


coherent combining at the receiver, the effective total received signal power increases linearly with L: doubling

L yields a 3-dB power gain. The second term reflects the diversity gain: by averaging over multiple independent

signal paths, the probability that the overall gain is small is decreased. The diversity gain L is reflected in the SNR

exponent, the power gain affects the constant before the 1/SNRL . Note that if the channel gains hl [1] are fully

correlated across all branches, then we only get a power gain but no diversity gain as we increase L. On the other

hand, even when all the hl are independent there is a diminishing marginal return as L increases: due to the law

of large numbers, the second term in the previous equation:

1L| |h| |2 =

1L

L∑l=1|hl [1]|2 (3.24)

converges to 1 with increasing L (assuming each of the channel gains is normalized to have unit variance). The

power gain, on the other hand, suffers from no such limitation: a 3-dB gain is obtained for every doubling of the

number of antennas. We will show this effect in the results section of this thesis.

primary node antenna element spacing is another parameter which is the distance between every element and

its adjacent element on the primary node. If modelled using Cartesian coordinates, we can regard the reference

antenna as the (0,0) point and the remainder of the elements equally spaced. It is important to note that the spacing

here is in wavelengths. This spacing will change according to the frequency but from an analytical standpoint,

we keep the frequency constant per drop to be able to realize the effect of spacing on the capacity and service

availability.

When the spacing increases, the antenna coupling between the elements decrease. As we increase the spacing,

the signal diversity is optimized, resulting in a better received signal power, enhancing both the capacity and the

service availability. This enhancement is manifested greatly as the spacing increases from 0-10λ. After a spacing

of 15λ, there is not much optimization to both attributes. The concept of space diversity is applied by using

antenna arrays at both the primary node and the vehicle. Every element of each antenna is described by the same

antenna pattern which is selected by the user. We define the arrays broadside to be a vertical line to the line

connecting the array elements.

Vehicle antenna element spacing is the spacing between antenna elements which is mentioned briefly in the

introduction and antenna-type sections. These elements can be on vehicle roofs, back windshields, front wind-

shields, bumpers, fenders, window panes, and rear-view mirrors. For example, if the antenna element spacing is

4λ for a 2 × 2 MIMO, then the available diagonal distance on the roof is 4λ. In all these orientations, the spacing

affects the inter-element interference between antennas which is the phenomena of antenna coupling.


We use this model similarly with vehicles as well as primary nodes because of the high transmit power of the

primary node antenna. We fundamentally deduce that as vehicles spacing increases, capacity increases due to

better spatial capabilities resulting in higher received power and less blocked signals for better service availability.

Taking the vehicle to be a passenger car for simplicity, we assume a 1-meter squared area on the roof indicated by

the blue arrows in figure 3.9. Within the 1 meter squared, single polarized antennas are placed as far apart from

each other and the following calculations are made.

Figure 3.9: 1 meter squared available area for antenna placement on car roof.

Figure 3.10: 2 × 2 MIMO optimum antenna placement on roof.

Figure 3.10 on page 43 shows the optimum positions of 2 antenna elements placed on the vehicle roof. In

terms of wavelengths the separation distances are shown in table 3.2 on page 45:


Figure 3.11: 4 × 4 MIMO optimum antenna placement on roof.

Figure 3.11 on page 44 shows the optimum positions of 2 configurations for 4 antenna elements placed on the

vehicle roof. In terms of wavelengths the separation distances are shown in table 3.3 on page 45:


Figure 3.12: 8 × 8 MIMO configurations for antenna placement on roof

Figure 3.12 on page 45 shows the optimum positions of 4 different orientations of 8 antenna elements placed

on the vehicle roof. In terms of wavelengths the separation distances are shown in table 3.4 on page 46:

Table 3.2: Inter-Antenna-Element spacing for 2 × 2 MIMO configurations on figure 3.10

Configuration/Frequency 700 MHz 1.9 GHz 2.7 GHzLeft Figure 3.3λ 8.96λ 9.1λRight Figure 2.32λ 6.33λ 9.09λ


Configuration/Frequency 700 MHz 1.9 GHz 2.7 GHzLeft Figure 2.32λ 6.33λ 9.09λRight Figure 0.77λ 2.09λ 3λ

As noted, the optimum spacing for increasing number of elements decreases given the limited surface area



Configuration/Frequency 700 MHz 1.9 GHz 2.7 GHzTop-Left Figure 0.93λ 2.53λ 3.64λTop-Right Figure 0.91λ 2.48λ 3.56λBottom-Left Figure 1.04λ 2.84λ 4.08λBottom-Right Figure 0.47λ 1.28λ 1.83λBottom Figure 0.332λ 0.91λ 1.3λ

on the roof for this simulation, and smaller frequencies decrease the optimum spacing as well because it entitles

larger wavelengths.

There is a carrier frequency of the transmitted signal through the downlink. The LTE spectrums of interest are

700-800 MHz, 1.9 GHz, and 2.4-2.9 GHz. As the frequency increases, the wavelength decreases which leads to

the fact that with spacing kept constant, there is a better performance.

3.2.1 Antenna array processing with a non-uniform set of element positions

In various antenna array processing algorithms we need to find the phase difference of the signal at a given element

relative to a reference element where we consider the phase to be zero. We introduce a coordinate system where

the location of the reference element is the origin and the other elements are in arbitrary positions Pi = (xi, yi )

Figure 3.13: Rays arriving at a non-uniform antenna element configuration

As shown in figure 3.13, we note that the distance of the wavefront to the point Pi is di as shown. If di is to

the left, i.e. the wavefront has not reached the point Pi then we take di to be positive, and if the point is to the

right, as Pj then we will take d j to be negative. Let the distance from the origin to the point Pi be Di . Then

we have di

Di= sin(φ − π

2 − θ) = −cos(φ − θ). Hence di = −Dicos(φ − θ) = −(Dicosφcosθ + Di sin φsinθ) =

−(xicosθ + yisinθ) = −(xi, yi ).(cosθ, sinθ), where the "." indicates a dot product of the two vectors. Note that


(xi, yi ) is a vector from the origin to the point in question, and −(cosθ, sinθ) is a unit vector in the direction of

propagation of the ray in question. Hence the distance to the reference point is the dot product of these two vectors,

or di = ~n · ~p where n is a unit vector in the direction of ray propagation and p is the vector from the origin to the

point Pi .

Now consider an incoming wave with frequency f0. At the origin we have the signal cos(2π f0t). At

element i, the signal is cos(2π f0(t − τ)), where τ is the propagation delay from the time that the wave-

front passes the origin to the time that it passes the point Pi . The phase of the signal at element i is then

φi = −2π f0τ = −2π f0( di

c ) = −2πdi

λ = −2πλ ~n · ~p . The signal at point Pi is then cos(2π f0t + φi )

Wavelength λ is inversely proportional to the frequency f as λ = cf . This will affect our spacing limitation

indicated in the spacing parameter below. Using our 1×1 meter small vehicle roof as the largest possible area to

place the antennas elements in specific orientations, we will indicate in subsequent sections on antenna spacing

that there exists minimum spacing between antenna elements where antenna coupling effects is stronger. Also, a

change in the wavelength λ affects the wave number k since k = 2πλ . This subsequently affects the value of k in

Equation 3.3 on page 26. As the frequency increases, the wave number decreases, and subsequently, the channel

model gain decreases numerically from a value of 1.8 at 700 MHz to 1.0 at 2700 MHz.

In addition to the antenna element separation effect, the Doppler effect plays a strong role in the field of small

scale fading. To understand the Doppler shift, we look at the equation of the electric field received at a point u in

space:

E( f , t, (r, θ,Ψ)) =αs (θ,Ψ, f )cos2π f (t − r/c)

r(3.25)

where (r, θ,Ψ) represents a point u in space in which the electric field is being measured, and r is the distance

from the transmit antenna to u, and where θ,Ψ represent the vertical and horizontal angles from the antenna to

u respectively. c is the constant speed of light and αs (θ,Ψ, f ) is the radiation pattern of the sending antenna at

frequency f in the direction (θ,Ψ). The scaling factor is for antenna losses. The phase of the field varies with f r/c

corresponding to the delay caused by the radiation traveling at the speed of light.

In our scenario, we will consider the moving vehicle as a receive antenna as we are interested in the downlink

(aka: treating the base station as our transmitter). In this case, we have a fixed transmit antenna and a moving


receive antenna with a location defined as u(t) = (r (t)), θ,Ψ)) with r (t) = r0 + vt. The free space electric field in

this case would have the following rule:

E( f , t, (r0 + vt, θ,Ψ)) =αs (θ,Ψ, f )cos2π f (t − r0/c − vt/c)

r0 + vt(3.26)

Note that we can rewrite f (t − r0/c − vt/c) as f (1− v/c)t − f r0/c, hence the sinusoid at frequency f is converted

to a sinusoid of frequency f (1 − v/c) and that is what is known as a Doppler shift of − f v/c due to the motion of

the observation point. So, each successive crest in the transmitted sinusoid has to travel a little further before it

gets observed at the moving observation point. The amount of this Doppler shift depends on the frequency f . This

leads us to understanding the Doppler spread which is simply the standard deviation of multiple spectral shifts due

to each of the multipath Doppler shifts.

This example illustrates the above. If the vehicle is moving at 60 km/h and f =900 MHz, the Doppler spread

in this case is 100 Hz. The role of the Doppler spread is visualized when the vehicle is much closer to a reflective

surface as a building than the base station. In this case the attenuation is roughly the same for different paths.

What will be obtained will have this form:

E( f , t) =2αsin2π f [vt/c + (r0 − d)/c]sin2π f [t − d/c]

r0 + vt(3.27)

This is the product of 2 sinusoids, one is at input frequency f which is in the order of GHz usually for LTE and

the other is at f v/c = Ds/2 which could be in the order of 50 Hz. Hence,the response to a sinusoid at f is another

sinusoid at f with a time-varying envelope, with peaks going to zeros around every 5 ms.

Equation 3.29 on page 49 highlights the path loss model used in highways and rural areas, and includes the

frequency variable which causes the following effect: as the frequency increases through the LTE spectrum, the

path loss decreases and the received power increases. Numerically, from 700 MHz to 2900 MHz yields a decrease

in the path loss from 41.5 dB to 39 dB which means an increase in the received power of 2.5 dB.

From the above, we find that the change in the received power as a result of change in frequency will have

the following effects: When frequency increases, the wavelength needed to achieve similar correlation decreases

and we do not have the spacing restriction. When frequency decreases, the wavelength needed to achieve similar

correlation increases and antenna coupling will increase leading to more interference and lesser received signal

power. This change is stronger than the change as a result of the Doppler shift which is in the Hz range.

Fundamentally, for the urban area, we use the microcell non-line-ofsight (NLOS) COST 231 Walfish-Ikegami


model with the following parameters:

-Primary node antenna height: 12.5m

-Building height:12m

-Building to building distance: 50m

-Street width: 25m

-Vehicle antenna height from ground: 1.5m

-Orientation: 30 degrees for all paths

The equation then simplifies to:

L(dB) = −55.9 + 38 × log10(d) + (24.5 + 1.5 × fc/925) × log10(fc ) (3.28)

We use the NLOS model because of the metropolitan nature of these cites in which buildings cause shadowing for

the different multi-paths. As for the highway scenario, we have

L = 40(1 − 4.10−3 × hBS ).log10(R) − 18.log10(hBS ) + 21.log10(f) + 80dB (3.29)

where:

R is the base station-vehicle separation in kilometres.

f is the carrier frequency in MHz which will be used in our theoretical analysis.

hBS is the base station antenna height in metres, measured from the average rooftop level.

An increase in frequency increases the path loss significantly in urban areas ( 30 dB increase) and does not

vary significantly in highways when increasing the frequency from 0.5 to 3 GHz. This difference is due to the

different structural barriers that diminish the line of sight in urban areas.

There is a velocity magnitude of the vehicle moving in the vicinity of the cell radius of the primary node.

There are 2 scenarios in this simulation leading to 2 different velocity magnitudes. For the urban and suburban

areas, we assume an average velocity magnitude of 30 km/h, and for highways , we assume a magnitude of 120

km/h. This difference will affect different parameters in our simulation such as the Doppler shift which is highly

dependent on the speed of the moving vehicle, the channel gain which depends on the fast fading parameters that

come into effect with multi-path.

As we will notice, its effect will be negligible for both service availability and capacity. Due to scheduling, it

is assumed that the channel rate is greater than the duration. This means that the channel variation is slower than

the time slot. Hence, velocity has no effect on capacity.


Another parameter which will be studied is the cross-polar discrimination which is the channelâĂŹs ability

to maintain radiated or received polarization purity between horizontally and vertically polarized signals, due to

scattering. The higher the XPD, the lesser the scattering effect of the channel. It is equal to the following equation:

XPD =Pco-polarized

Pcross-polarized(3.30)

where:

Pco-polarized: is the Power of the Co-polarized component of the signal.

Pcross-polarized: is the Power of the Cross-polarized component of the signal.

A single XPD ratio applies to all sub-paths of a given path. Each path n experiences an independent realization

of the XPD. For each path the realization of the XPD is drawn from the 3GPP distributions in [1]. Since hand-

handled devices have limited dimensions, spatial diversity might be difficult to be applied to them. For that

reason, polarized arrays which use cross-shaped, co-located dipole antenna pairs might be the primary way to

apply polarization diversity but that is future work.

The primary node antenna element tilt represents the vertical inclination of the elements collectively on the

primary node. There are both vertical and horizontal polarized sub-paths propagating inside the channel, hence

it would be more practical to assume arrays of ideal dipoles at both the primary node and vehicle. These dipoles

can be tilted with respect to the z-axis by a common polarization angle αBS and βMS . The antenna gain for the

vertical and horizontal reception will be given by the equality:

*..,

G(v) (θ)

G(h) (θ)

+//-=*..,

cos α

sin α cos θ

+//-

(3.31)

θ is the angle that the sub-ray arrives/departs to/from the dipole and α is the polarization angle. The reason why

this is of importance is to perform spatial and polarization diversity simulations and it gives the user the ability to

choose from different array elements for both the primary node and vehicle separately.

The vehicle antenna tilt parameter is responsible for the tilt of the vehicle antenna elements. As mentioned

previously, both the primary node and vehicle array elements will be assumed (tilted) ideal dipoles or (tilted)

cross-polarized, co-located dipoles, forming dipole pairs. Note that when dipole pairs are used, the number of

the antennas at each array is twice the number of the array elements. This is vital for the feasibility of which we

place antennas on the vehicle for a number of reasons. Firstly, vertical MIMO antennas on the vehicle roof could

raise concerns on the feasibility of utilizing all the allocated vehicle space, whereas, an antenna that is parallel

to the windshield is a smooth integration with no visible change to the vehicle, and would be more welcomed as

a solution. In the latter case, a tilt of approximately 135 degrees to the vertical upwards direction represents an


integrated antenna parallel to the windshield, and a tilt of zero degrees represents a vertical set of antennas parallel

to the vertical axis. An effect of increasing or decreasing the tilt from the vertical axis affects the polarization that

is receivable by the antenna, and depends greatly on the tilt of the primary node antennas. As seen in Table 4.1 on

page 55, we use default values to simulate the existing technology, and in the results section we show the effect of

altering the tilt. The primary node transmit power parameter according to [2] is 43 dBm in the simulation which

is the maximum BS power for Ultra Frequency Division Duplex (FDD) reference primary nodes. This power is

assigned to the 6 paths mentioned in the paths section according to the following equation:

P′n = e(1−rDS )τ′nrDS .σDS .10−ξn/10, n = 1, ..., 6 (3.32)

where ξn (n=1,...,6) are i.i.d. Gaussian random variables with standard deviation σRND = 3dB. DS is the Doppler

Spread.

It is trivial to note that, as the primary node transmit power increases, the available capacity increases and the

service availability increases but up to a certain level due to the effect of antenna coupling and inter-primary node

interference.

The Signal-to-noise-ratio (SNR) is the input SNR taking into consideration that there are 2-tier interfering

primary nodes surrounding the cell of analysis. Hence, we will obtain a signal-to-Interference-and-noise ratio

(SINR) which is equal to the following:

SINR =Pr∑m

i=1 Pri + N(3.33)

where:

Pr is the received power from the primary node to the vehicle in its cell radius.∑mi=1 Pri is the sum of the received powers from m 2-tier primary nodes that interfere with the main signal.

N is the noise variable.

For a given AWGN channel as simulated in both environments, the Shannon bound represents the maximum

theoretical throughput [37] that can be achieved over the channel with a given signal-to-noise ratio (SNR) as given

in the following equation:

Throughput,Thr,bps/Hz =

Thr = 0 for SINR < SINRMIN

Thr = α.S(SINR) for SINRMIN < SINR < SINRMAX

Thr = ThrMAX for SINR > SINRMAX

(3.34)

Where:


S(SINR): is the Shannon bound S(SINR)=log2(1 + SINR) bps/Hz

α: Attentuation factor representing implementation losses

SINRMIN: Minimum SINR of the codeset, dB

ThrMAX:Maximum throughput of the codeset, bps/Hz

SINRMAX: SINR at which maximum throughput is reached S−1(ThrMAX),dB

Note that the parameters α, SINRMIN, and ThrMAX are chosen to represent implementations and link condi-

tions.

The Delay spread is another variable in microseconds which we study that defines the statistics of the path

delays through the random variable τ. the random delays for each of the N multipath components. This tau is

calculated as per the following equation:

τn =Tc

16.floor(

τ′(n) − τ′(1)

Tc

16+ 0.5)n, n = 1, ..., N . (3.35)

where floor(x) is the integer part of x, and Tc is the chip interval (Tc=1/1.2288x106 seconds for 3GPP2)

The primary node per path Angle of departure is a variable that changes with the environment. We assume

that there is no indication of alternate channel behavior during the downlink which is the highlight of this research

as well as the uplink, the Angle of Departure(AoD) and Angle of Arrival(AoA) values. They are identical between

both propagating directions, although this does not apply to the random subpath phases during the downlink which

will be assumed to be uncorrelated. In other words, we assume that the channel is a double directional system

where the primary node and vehicle can both be the receiver and transmitter despite the fact that we refer to the

primary node as the transmitter. This variable defines the statistics of the BS paths Angle-of-Departure through

the random varible δAoD

The correlation between channel parameters is a number of parameters incorporated into the simulation. As a

summary of the section in [1], the azimuth spread is a log normal distribution, correlated to the delay spread and

shadow fading. To make sure that the model used for simulation can reproduce this correlation behaviour along

with the expected probability and range of each parameter, the intra-site correlations between the shadow fading,

delay spread and azimuth spread:

ραβ= correlation between the delay spread and the azimuth spread.

ργβ= correlation between the shadow fading and the azimuth spread.

ργα= correlation between the shadow fading and the delay spread.

The shadow-fading/delay-spread correlation parameter is an important correlation parameter that is a property


of the channel and affects the capacity and coverage. In [37], Greenstein et al, shows a model for correlating delay

spreads with log normal shadow fading. Both are log-normal distribution and are generally correlated. The result

of the negative correlation between log normal shadowing and delay spread is increasingly significant because it

indicates that a larger shadow fading entitles a reduced delay spread, and if the shadow fading is smaller, then the

delay spread is larger.

Chapter 4

Capacity and Service Availability

Improvement Results

Simulation results obtained in this chapter observe the capacity gain on the y-axis twice; once as a function of

antenna element spacing in wavelengths, and as a function of carrier frequency in our 3 different environments. We

increase the number of antenna elements from 2 to 4 to 8 and study the performance gains within a vehicle and for

the primary node to secondary node downlink from the primary node. Results also study the service availability

(coverage) on the y-axis as a function of the distance between primary and secondary nodes, for different antenna

elements and different antenna-element spacing. We show how the load on basestations decreases to less than 50

per cent for 8 × 8 MIMO with 6λ spacing between elements.

The thesis results lastly show the capacity gains on the y-axis while varying the polarization angle between the

antenna elements on the primary and secondary node on the x-axis, while comparing SISO 2 × 2, 4× 4, and 8 × 8

MIMO. We observe that as we increase the number of elements, we obtain similar performance gains regardless

of the antenna element polarization angle.

4.1 Capacity gain via Vehicle Mounted MIMO antennas

In this section we study the capacity gain as we increase the number of antenna elements from 2 to 4 to 8 in

an urban area. Figure 3.2 on page 29 shows an urban configuration in which we assume that the base-station

is communicating with multiple vehicle users moving at 30km/h within its cell boundaries. The vehicle is also

receiving interfering signals Pr from i=18 inner and outer neighbouring base stations which weaken the signal as

well as noisePn . A subsequent equation for the received SINR at the vehicle is:

54

Chapter 4. Capacity and Service Availability Improvement Results 55

SINR =Pr

Pn +∑18

i=1 Pri

(4.1)

Table 4.1 outlines the parameters used for the urban environment.

Table 4.1: Default Channel Parameters and suggested values for Urban Scenario

Parameter Value UnitPer path Azimuth Spread at BS 5o degPer path Azimuth Spread at Vehicle 35o degNumber of Paths 6 ScalarNumber of subpaths (per path) 20 ScalarBS transmit Power 43 dBmσAoD/σAzimuth Spread ratio 1.3 Scalarσdelays/σDelay Spread ratio 1.7 ScalarBS per path AoD Distribution 40o degLog-Normal Shadowing Standard Deviation 10 dBAzimuth Spread at BS µAS = 0.81/εAS = 0.34 ScalarDelay Spread µDS = −6.18/εDS = 0.18 ScalarAzimuth Spread-Delay Spread Correlation 0.6 ScalarShadow Fading-Azimuth Spread Correlation -0.5 ScalarShadow Fading-Delay Spread Correlation -0.5 Scalar

The following results show the number of antenna elements effect on capacity. Figure 4.1 shows a 3-D

illustration of the effect of simultaneously increasing antenna elements for both the primary node and the vehicle.


Figure 4.1: A 3-D representation of capacity Vs. antenna elements at the primary node and Vehicle at 1.9 GHz.

We notice the same behaviour for 700MHz and 2.7 GHz so there is no need for providing redundant results. For

a limited number of primary node antenna elements, there is only a maximum capacity that is reachable even after

increasing the number of vehicle antennas and vice-versa. There is a linear increase in capacity gains till the number

of vehicle antennas match the number of primary node antennas. This observation matches the linear fundamen-

tal relationship between capacity and the number of antenna elements increasing from 2×2MIMO to 8×8MIMO.

Figure 4.2 shows the specific effect of increasing the number of vehicle antenna elements on capacity gains in

urban areas while varying the number of primary node antennas for each curve.


Figure 4.2: Number of vehicle antenna elements Vs. capacity at urban area.

There is a linear increase as explained in the 3-D figure, and there is a gradual increase in capacity gains after

the point where the number of antennas match. The reason of this behavior is because of lesser fading that is well

handled by MIMO antennas as they increase in the number of elements. Capacity gains are affected due to fading

in urban areas. Figure 4.3 shows the similar effect of vehicle number of antennas but for suburban areas.


Figure 4.3: Number of vehicle antenna elements Vs. capacity at suburban area.

There is a similar linear increase till full rank. 2 × 2 MIMO approaches a gain of 2 which is faster than

in urban areas. 4 × 4 and 8 × 8 approach gains of 4 and 6 respectively which are relatively higher than the

urban gains for the same number of antennas. The capacity gain of 6 for 8 × 8 MIMO is obtained due to strong


inter-cell interference from less fading properties of 8 × 8 MIMO at the downlink from all the interfering primary

nodes. Figure 4.4 shows the highway environment with the sole extra observation that in addition to the previous,

the antennas approach maximum capacity limits more rapidly and the 8×8MIMO approaches a capacity gain of 7.

Figure 4.4: Number of vehicle antenna elements Vs. capacity at highway.

The following results show the capacity gain as a result of increasing the available spacing (a car roof as the

example throughout this thesis) in wavelengths. Logically, we assume optimum antenna spacing at the primary

node and no recurring limiting factors as a result of antenna correlation. It also assumes proportional fair scheduling

as discussed in the previous sections. Finally, these results are at a frequency of 1.9 GHz and the behavior is

consistent for 700 MHz and 2.7 GHz as well. Figures 4.5, 4.6, and 4.7 show how a moving vehicle performs

in terms of the capacity gains of its antenna elements over baseline antenna elements in an urban, suburban, and

highway area respectively.


Figure 4.5: Channel capacity gain Vs. vehicle antenna element spacing for urban areas.


Figure 4.6: Channel capacity gain Vs. vehicle antenna element spacing for suburban areas.


Figure 4.7: Channel capacity gain Vs. vehicle antenna element spacing for highways.

All three environments share a set of common behaviors. No matter how many number of antenna elements,

the lowest antenna spacing available on the car roof (<=1λ) causes all orientations to behave as a SISO antenna.

There is a transient behavior for all orientations during the capacity gain increase between 1-4λ. This slightly

unpredictable behavior is primarily due to antenna coupling even as the number of simulations runs surpass 10,000.

Capacity gains are stabilizing beyond 8λ and are considerably less varying indicating that the effect of antenna

coupling decreases at this point and we are maximizing the capability of our antenna orientations. It is also of

importance to note that, in the figure, the larger the number of antenna elements, the longer it takes for the capacity

to stabilize. This is primarily because the inter-antenna distances for the same available spacing decrease with the

increase of elements. 8 × 8 MIMO stabilize beyond 8λ, 4 × 4 after 6λ, and 2 × 2 after 4λ.

The results below show the capacity gains as a result of increasing primary node (basestation) antenna spacing.

It is different from vehicle antenna spacing because the transmission power of a primary node is much higher

and therefore we need more antenna spacing to avoid coupling. Although primary node element spacing is

not a limiting factor when it comes to finding the optimum downlink parameters, its analysis matches existing

performances in terms of a sanity check for the simulation. It is to be noted that these results assume optimum


antenna spacing at the vehicle side and no limiting factors as a result. It also assumes proportional fair scheduling

as discussed in the previous sections. Finally, these results are at a frequency of 1.9 GHz and the behavior is

consistent for 700 MHz and 2.7 GHz as well.

Figure 4.8shows how primary nodes perform in an urban area where 2 × 2 MIMO antennas optimize till

saturation over the baseline capacity by a factor of 1.7, 4 × 4 MIMO antennas by a factor of 2.9, and 8 × 8 MIMO

antennas by a factor of 4.4.

Figure 4.8: Downlink channel capacity Vs. primary node antenna spacing- Urban area.

Figure 4.9 shows a suburban area in which the capacity saturates at a factor of approximately 1.85 for 2 × 2

MIMO, a factor of 3.55 for 4 × 4 MIMO, and a factor of about 7 for 8 × 8 MIMO. The 3 values are considerably

higher than that of the urban area due to less scattering, shadowing, and fading in general.


Figure 4.9: Downlink channel capacity Vs. primary node antenna spacing- suburban area.

In highway environements shown in figure 4.10, the gains are higher than that of suburban areas due to lesser

inter-cellular interference of the irregular highway communication system. Gains reach double for 2 × 2 MIMO,

3.9 for 4 × 4 MIMO, and 7.8 saturation for 8 × 8 MIMO.


Figure 4.10: Downlink channel capacity Vs. primary node antenna spacing- highways.

One of our most important observations which we should take from this illustration is that for all scenarios,

capacity gains increase considerably till 10λ spacing, moderately from 10-15λ spacing, and saturate gradually

afterwards, with 2× 2 MIMO saturating fastest, followed by 4× 4 MIMO, then 8× 8 MIMO. The reason lies in the

inter-element interference which increases with the rise of the density of antenna elements within a given spacing

provided by the x-axis.

It is important to note that varying primary node antenna spacing in reality is not of substantial importance

in this research because it is not a limiting factor. However, we present it here for quantitative research results

by creating a sanity check for the simulation. The following results observe the channel capacity gains as we

increase the frequency. We study the effect of frequency on the channel capacity gain for different environments

and different number of antenna elements but assuming optimum spacing and no antenna coupling. The baseline

in this case is the 3-cm-separation Shark Fin Antenna Choosing the LTE known frequency ranges from 700 MHz

to 2.7 GHz, we notice the following from 2 × 2, 4 × 4, 8 × 8 MIMO implementations on figures 4.11, 4.12, and

4.13 respectively.


Figure 4.11: Channel capacity gain Vs. frequency for 2 × 2 MIMO.





The channel capacity gain for urban areas slightly decreases from 2600MHz and 700MHz and towards 1800

MHz but this decrease is only by a change of 0.1. The reason is because low frequencies offer better propagation and

high frequencies are subject to less interference from he neighboring primary nodes. The in-between 1800MHz

is a transition mixture of both and is lower by a small factor. The channel capacity gain for suburban areas follows

a linear increase in capacity from 700-1400 MHz, followed by a sharper increase to 1600 MHz-1800 MHz, then a

decrease in capacity gain. This behavior shows how strongly interference plays a role in suburban areas with less

shadowing and scattering, but with 2 layers of interfering primary nodes: the lowest capacity gain is at the best

propagation frequency of 700 MHz as we are getting more received signals from 19 adjacent primary nodes. The

low gain at 2.7 GHz is due to the low propagation characteristics of high frequencies that are modelled in the path

loss model for suburban areas, and the increase in 1600-1800 MHz proves that the degradation of signal power

from interfering primary nodes occurs while keeping propagation within the cell. For Highways, we only have 4

interfering primary nodes as per our model that are aligned on the highway at a radius of 3 km each. The signal

is not subject to strong interference at good propagation frequencies. However, due to propagation characteristics

that become weaker as the frequency is increased, the capacity gain decreases for high frequencies but still offer


capacity gains higher than 1. Increasing the number of elements increases the change in channel capacity gain.

Simulated as a sanity check for the quality of the simulator, the best frequencies for highways are 700MHz, 1600

MHz for Suburban, and 2.7 GHz for Urban areas.

The following results show the capacity gain as a result of varying the antenna tilt. Figure 4.14 shows both

the theoretical and simulated correlation amplitude of the received signal as we vary the azimuth angle tilt of the

antenna element from 0o (vertical) to 90o (horizontal).

Figure 4.14: Theoretical and simulated correlation amplitude Vs. antenna tilt.

The theoretical curve follows a cosine equation where a received signal with an angle of 60o with the vertical

axis to the roof for example has a correlation amplitude of 0.5. But if this is true, the received signal should be

0 for horizontally polarized waves that arrive orthogonal to the vertically polarized antenna element(s) on the car

roof. Is that the case?

Simulated resultswith the vehicular ITU channelmodel show that the signal indeed decreases as the polarization

angle difference increases. It does not reach 0, but decreases slightly to a correlation amplitude of 0.4. From this


observation, we can deduce that, although the transmit and receive antennas are orthogonal, not all the received

signal was cross-polarized but due to scattering induced by the channel itself, a part of the signal power changes

its polarization due to the cross-polarization parameter. We look at the results of cross-polarization discrimination

effect on capacity gain. The figures below show the channel capacity in regards to varying the vehicle antenna

elementsâĂŹ angle of polarization for different XPD values for a suburban area (other environments behave

similarly so we explain the behavior once here). We notice the recurring behavior no matter howmuch we increase

the number of elements from 1 to 8 elements. The SISO antenna in figure 4.15 behaves slightly similar to the

cosine curve as we increase the angle of polarization from vertical (0 degrees) to horizontal (90 degrees) for an

XPD of 0. This means that half the waves were able to maintain their polarization.

Figure 4.15: Channel capacity Vs. vehicle antenna polarization Angle for SISO.

In both cases we realize that the best performance occurs at an antenna tilt of 45 degrees. As the XPD

increases increases, more waves match the angle of polarization of the primary node and we obtain an overall

better performance and capacity in which for an XPD of 8 dB the capacity reaches a peak similarily at 45 degrees

with an increase from 1.2 to 1.8 but it then decreases strongly similar to the other XPD values to a value close to

zero when the antenna is horizontal. The reason why it doesnâĂŹt reach 0 is due to random correlation parameters


that model the scattering effect of the environment. 2 × 2 MIMO antennas in figure 4.16 show a close behavior to

SISO.

Figure 4.16: Channel capacity Vs. vehicle antenna polarization angle for 2 × 2 MIMO.

We observe that all XPD values do not drive the capacity to approach 0 when the antennas are horizontal but

have about 28 per cent of the maximum capacity achieved at 45 degrees at XPD=0 dB and 0 degrees (vertical) for

XPD values 8 and 15. The reason we do not only perceive a mere increase in capacity gains is because MIMO

antennas address fading more efficiently. 4 × 4 MIMO antennas in figure 4.17 show a startling fact.



At an XPD of 0, variation across the different angles of polarization are kept to a bare minimum allowing us

to connect antennas in various ways. As for XPD= 8 and 15 dB, we observe a decline in capacity of 30 per cent

for 45 degrees and degraded capacity for high XPD for horizontal antennas. 8 × 8 MIMO antennas in figure 4.18

provide a varying capacity from 16 to 24 bps/Hz for an XPD of 0 as we increase the polarization angle from 0 to

90 degrees.



In this case, both horizontal and vertical orientations yield a less varying capacity which proves that the

knowledge of the vehicle polarization angle is not important. As for a high unrealistic XPD of 15 dB where there

is much less scattering, the optimum antenna angle should be 0 (vertically polarized) with a strong degrading of

capacity approaching 0 for horizontally polarized antennas. This is why the more probable XPD of 0 proves that

it does not matter what our vehicle antenna element polarization angle is.

It is important to note that the effect of XPD and antenna tilt on coverage is consistent with capacity calculations

and simulations above and is not needed.

4.2 Improved Service Availability via Vehicle Mounted MIMO

In this section we observe the service availability (coverage) improvement as we increase the number of antenna

elements, and as we increase the spacing between the antenna elements. Table 4.2 on page 74 tabulates the

parameters used for the suburban and highway environments.

The following results study the antenna spacing effect on coverage. Figure 4.19 shows the service availability


Table 4.2: Default Channel Parameters and suggested values for Suburban and Highway Scenario

Parameter Value UnitPer path Azimuth Spread at BS 2o degPer path Azimuth Spread at Vehicle 35o degNumber of Paths 6 ScalarNumber of subpaths (per path) 20 ScalarBS transmit Power 43 dBmσAoD/σAzimuth Spread ratio 1.2 Scalarσdelays/σDelay Spread ratio 1.4 ScalarBS per path AoD Distribution N/A degLog-Normal Shadowing Standard Deviation 8 dBAzimuth Spread at BS µAS = 0.69/εAS = 0.13 ScalarDelay Spread µDS = −6.80/εDS = 0.288 ScalarAzimuth Spread-Delay Spread Correlation 0.6 ScalarShadow Fading-Azimuth Spread Correlation -0.5 ScalarShadow Fading-Delay Spread Correlation -0.5 Scalar

for the downlink in an urban area with 2λ spacing available on the car roof.

Figure 4.19: Service availability Vs. primary node vehicle element separation distance of 2λ spacing.

Figures 4.20 and 4.21 show the results for a car-roof spacing of 4 and 6λ respectively.


Figure 4.20: Service availability Vs. primary node vehicle element separation distance for 4λ spacing.


Figure 4.21: Service availability Vs. primary node vehicle element separation distance for 6λ spacing.

The results outline that we observe a gradual increase starting from 100 percent to a separation distance

between the primary node and vehicle of 1 km for 97 per cent service availability which is the intended quality

of service this research focuses on. Moving from 2 to 6λ, the service availability optimizes for higher distances.

Upon looking closely at the difference between these 3 figures, the service availability optimization from 2 to 4λ

is smaller than the difference from 4 to 6λ. This matches the earlier observations that antenna coupling with the

transmit power of the car antenna is greatly decreased from 4 to 6λ which falls into the vicinity of the available

allocated area on the car roof. In Figure 4.19, there are similar increases in service availability between 2 × 2,

4 × 4, and 8 × 8 MIMO. Starting figure 4.20, 8 × 8 MIMO has a noticeable increase in service availability, while

the noticeable increase for 4 × 4 MIMO is observed at 6λ spacing in figure 4.21.

Table 4.3 as well as tables 4.4 and 4.5 show the percent radius increase and percent of load decrease on

base stations in a 25x25 Km area as a result of increasing cell radii for 2,4,and 6λ spacing respectively. This

is considered as a natural advantage to increasing the cell radius, to decrease the load on primary nodes without


sacrificing signal quality.

Table 4.3: Results- 2λ spacing- 25 Km x 25 Km Urban Area

MIMO Configuration Percent Radius increase Percentage of Basestations Retired2 × 2 5.6 10.324 × 4 17 278 × 8 23.3 34.2


MIMO Configuration Percent Radius increase Percentage of Basestations Retired2 × 2 11.1 194 × 4 20.1 30.78 × 8 44.4 52


MIMO Configuration Percent Radius increase Percentage of Basestations Retired2 × 2 27.8 38.74 × 4 38.9 48.28 × 8 50 55.6

With 3 separate graphs and their respective tables that outline different spacing limitation effect on service

availability, a more analytical graph that outlines the increase in cell radius as a percentage Versus the available

spacing on the roof was made as shown in figure 4.22 for the urban area, figure 4.23 for the suburban area, and

figure 4.24 for the highway area. The recurring evident observation, ofcourse, is the notable improvement at 4 to

6λ is observed for all 3 environments.


Figure 4.22: Urban cell radius optimization Vs. antenna spacing at 97 per cent service availability.

Figure 4.23: Suburban cell radius optimization Vs. antenna spacing at 97 per cent service availability.


Figure 4.24: Highway cell radius optimization Vs. antenna spacing at 97 per cent service availability.

Chapter 5

Summary, Conclusion, and Future Work

Current connected vehicles show different performance gains when we disconnect the user equipment. Eliminating

the spacing can provide very substantial performance gains. Using 2×2MIMO antennas and increasing the number

of elements to include 4 × 4 and 8 × 8, we aspire to have maximum spacing on vehicle roofs in the future ranging

from passenger cars to public transportation.

In this research, we evaluated a framework to connect vehicles in the best possible way by maximizing the

ability of LTE to include more than 2 antenna elements and use up to 8 elements that are sufficiently spaced on the

vehicle. The thesis discusses a simple theoretical framework to evaluate non-linear configurations in terms of both

antenna coupling between all the elements and phase difference of the incoming wave to a reference point. We also

outline how we use this non-linear structure to place elements in different parts of the vehicle without interfering

with the aesthetics. Different points in the car require different kinds of antennas which are briefly discussed

and compared in terms of feasibility. The simulation part of this research focuses on placing the elements on

the limited spacing area on the roof using 2 elements, 4 elements and 8 elements for 3 simulated environments

with its specific parameters: Urban areas, suburban areas, and highways. We use an ITU channel model that is

supported by 3GPP and we utilize it to compare different effects of tilting, inter-antenna element spacing, number

of antennas, and frequency on the capacity and hence the coverage of the system. The thesis discusses each of the

parameters used in the simulation and its significance with respect to calculating the channel gain accurately. We

reach a set of important contributions: Vertically tilted antennas do not necessarily provide the maximum channel

capacity for all environments. Also, by placing more antenna elements and setting the spacing for the elements to

be much farther from each other, the optimization in service availability means that with a bigger cell radius, we

could decrease the load on primary nodes in a given area.

80

Chapter 5. Summary, Conclusion, and Future Work 81

For future work, I intend to continue our measurement campaign from figure 2.17 on page 24 to determine

the optimum positions for placing vehicular antennas. We will also consider additional antenna architectures

and practically compare the theoretical framework to evaluate non-linear configurations with assembled MIMO

antennas with non-linear configurations. We compared our novel model with the shark fin but we can explore

more types of antennas with hand-on results on their radiation patterns and their gains. Future research aims to

survey and study the best possible placement of antenna element positions for the least fading and antenna coupling

effects that decrease the channel capacity and service availability of the system. A number of ways this can be

tackled in future research is by using dual-polarized elements. in this research, we used single polarized elements

in the antenna. This meant that we would need 4 elements for a 4 × 4 MIMO whereas 2 dual-polarized antenna

elements can act as a 4 × 4 MIMO. Future research will also feature integrated antennas. An integrated antenna

can be mounted on a car windshield and could prove itself to be immensely useful for feasibility in comparison

with mounting vertical antenna elements on the car roof. Integrated antennas have recently been researched and a

number of working models are accessible in terms of its accuracy and performance. Finally, incorporating adaptive

beam-forming techniques in which I intend to study adaptive antenna architectures such that do not disrupt the car

aerodynamics. We do not have to use all antennas simultaneously and we can do adaptive beamforming as shown

in figure 2.15 on page 23 with 8 or more antennas at a time.

Bibliography

[1] 3GPP. TR 25.996 v. 12.1.0 Release 12, Universal Mobile Telecommunications System (UMTS) Spatial

channel model for Multiple Input Multiple Output (MIMO) simulations. Technical report, December 2012.

[2] 3GPP. TR 36.942 v. 12.1.0 Release 12, LTE Evolved Universal Terrestrial Radio Access (E-UTRA) Radio

Frequency (RF) system scenarios Technical report, October 2014.

[3] J. Schaffner, A. Bekaryan, H. Song, T. Talty, D. Carper, E. Yasan and A. Duzdar, A Drive Test Measurement

Approach to Characterize On-Vehicle 2x2 LTE-MIMO Antennas In Antennas and Propagation in Wireless

Communications (APWC), IEEE-APS Topical Conference on. IEEE, 2014.

[4] A. Khan and R. Vesilo, A Tutorial on SISO and MIMO Channel Capacities.

[5] R. Vaughan and J. Andersen, Antenna diversity in mobile communications. In Vehicular Technology, IEEE

Transactions on 36.4: 149-172., 1987.

[6] A. Thiel, L. Ekiz, O. Klemp, and M .Schultz, Automotive Grade MIMO Antenna Setup and Performance

Evaluation for LTE-Communications. In Antenna Technology (iWAT), International Workshop on. IEEE.,

2013.

[7] W. Lee, Effects on Correlation Between Two Mobile Radio Base-Station Antennas. In Communications,

IEEE Transactions on 21.11: 1214-1224., 1973.

[8] K. Linehan and J . Robson, What base station antenna configuration is best for LTE-Advanced? In White

Paper, 2012.

[9] C. Oikonomopoulos, T. Mohammed, and M. Arnold, Characterization of automotive MIMO antenna

configurations through simulations and measurements. In Antennas and Propagation Society International

Symposium (APSURSI), 2012 IEEE., 2012.

82

Bibliography 83

[10] T. Lankes, P. Turban, and F. Mierke, and J . Byers, Evaluation and optimization of LTE MIMO antenna

configurations in automotive environment. In Antennas and Propagation (EuCAP), 2014 8th European

Conference on (pp. 1100-1104). IEEE., 2014.

[11] D. Tse and V. Pramod, Fundamentals of wireless communication. Cambridge university press, 2005.

[12] C. Sturman and C. Fenger, LTE for automotive applications In Whitepaper, u-blox, 2014.

[13] G. Araniti, C. Campolo, and M. Condoluci, LTE for vehicular networking: a survey. In Communications

Magazine, IEEE 51.5: 148-157, 2013.

[14] B. Hagerman, K. Werner, and J. Yang, MIMO performance at 700MHz: Field trials of LTE with handheld

UE In Vehicular Technology Conference (VTC Fall), 2011 IEEE, 2011.

[15] L. Ekiz, A. Thiel, and O. Klmep, MIMO performance evaluation of automotive qualified LTE antennas. In

Antennas and Propagation (EuCAP), 2013 7th European Conference on. IEEE, 2013.

[16] Z. Oikonomopoulos, T. Christos, and A. Matthias, Outdoor channel characterization of MIMO-LTE antenna

configurations through measurements. In Vehicular Technology Conference (VTC Spring), 2012 IEEE 75th.

IEEE, 2012.

[17] J. Ojha, W. Ian, and M. RenÃľ, Technological Trends of Antennas in Cars. INTECH Open Access Publisher,

2011.

[18] K. Korosec, General Motors CarPlay. TIME, May 28, 2015, http://time.com/3899755/apple-carplay-google-

android-auto/.

[19] T. Cover and J. Thomas, Elements of information theory. In John Wiley and Sons, 2012.

[20] B. Holter, On the capacity of the MIMO channel: A tutorial introduction. In Proc. IEEE Norwegian

Symposium on Signal Processing, 2001.

[21] A. Paulraj, R. Nabar, and D. Gore, Introduction to space-time wireless communications. In Cambridge

university press, 2003.

[22] L. Greenstein, V. Erceg, S. Yeh, and M. Clark, A new path-gain/delay-spread propagation model for digital

cellular channels. In Vehicular Technology, IEEE Transactions on 46.2, 477-485 , 1997.

[23] Infineon . AG, One-chip packaged RF Transceiver Chipsets forMobile Backhaul - Infineon Technologies’, In-

fineon.com, 2015. Available: http://www.infineon.com/cms/en/product/promopages/PMM/backhaul/. [Ac-

cessed: 17- Sep- 2015].

Bibliography 84

[24] G. Foschini andM. Gans, On limits of wireless communications in a fading environment when using multiple

antennas. Wireless personal communications 6.3 (1998): 311-335..

[25] T. Emre, Capacity of multi-antenna Gaussian channels. European transactions on telecommunications 10.6

(1999): 585-595.

[26] W. Keusgen, On limits of wireless communications when using multiple dual-polarized antennas. In

Telecommunications, ICT 2003. 10th International Conference on. Vol. 1. IEEE, 2003.

[27] M. Geissler, MIMO performance optimisation of car antennas. In Antennas and Propagation (EUCAP),

2012 6th European Conference on. IEEE, 2012.

[28] D. Plettemeier, G. Fettweis, and O. Eckhard, MIMO system design and field tests for terminals with confined

space-impact on automotive communication. In Antennas and Propagation (EUCAP), Proceedings of the

5th European Conference on. IEEE , 2011.

[29] P. Papadimitratos, Vehicular communication systems: Enabling technologies, applications, and future

outlook on intelligent transportation. Communications Magazine, IEEE 47.11 (2009): 84-95.

[30] D. Kashif, Wireless communication technologies for ITS applications [topics in automotive networking].

Communications Magazine, IEEE 48.5 (2010): 156-162.

[31] M. Phan, R. Rembarz, and S. Sories, A capacity analysis for the transmission of event and cooperative

awareness messages in LTE networks. In 18th ITS World Congress , 2011.

[32] J. Santa, A. MoragÃşn, and A. GÃşmez-Skarmeta, Experimental evaluation of a novel vehicular communi-

cation paradigm based on cellular networks. In Intelligent Vehicles Symposium, 2008 IEEE , 2008.

[33] D. Yuan, A novel dual-band printed diversity antenna for mobile terminals. In Antennas and Propagation,

IEEE Transactions on 55.7: 2088-2096., 2007.

[34] G. Alexander, MIMO Efficiency of a LTE Terminal Considering Realistic Antenna Models. In 2009

International ITG Workshop on Smart Antennas, WSA., 2009.

[35] S. Seymour, A new path-gain/delay-spread propagation model for digital cellular channels. In Antennas and

Propagation, IRE Transactions on 10.5 (1962): 548-557..

[36] N. Koch, Antennas for Automobiles. In Intech New Advances in Vehicular Technology and Automotive

Engineering Ch 7

Bibliography 85

[37] M. Ruoss, V. Erceg, S. Yeh, and M. Clark, The Digitalization of the FM-Band in Europe. In 9th Workshop

Digital Broadcasting. , 2008.

Documents

Vehicular-based Antenna Architectures for Two-tier 5G ... · Vehicular-based Antenna Architectures for Two-tier 5G Wireless Systems by SamerHenry Athesissubmittedinconformitywiththerequirements