Institutionen för systemteknik
Department of Electrical Engineering
Examensarbete
A Study on the Impact of Antenna Downtilt on the
Outdoor Users in an Urban Environment
Examensarbete utfört i Kommunikationssystem vid Tekniska högskolan i Linköping
av
Pradeepa Ramachandra LiTH-ISY-EX--12/4585--SE
Linköping 2012
Department of Electrical Engineering Linköpings tekniska högskola
Linköpings universitet Linköpings universitet
Outdoor Users in an Urban Environment
Examensarbete utfört i Kommunikationssystem
vid Tekniska högskolan i Linköping
av
Pradeepa Ramachandra LiTH-ISY-EX--12/4585--SE
Handledare: Mehdi Amirijoo
Ericsson Research, Linköping, Sweden Birgitta Olin
Ericsson Research, Stockholm, Sweden Jan-Erik Berg
Ericsson Research, Stockholm, Sweden Chaitanya TVK
isy, Linköpings universitet Examinator: Danyo Danev
isy, Linköpings universitet Linköping, 14 June, 2012
Division of Communication Systems Department of Electrical Engineering Linköpings universitet
SE-581 83 Linköping, Sweden
2012-006-14 Språk Language Svenska/Swedish Engelska/English Rapporttyp Report category Licentiatavhandling Examensarbete C-uppsats D-uppsats Övrig rapport
URL för elektronisk version
http://www.commsys.isy.liu.se http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-ZZZZ ISBN — ISRN LiTH-ISY-EX--12/4585--SE
Serietitel och serienummer
Title of series, numbering
ISSN
—
Titel
Title
En Studie om Effekterna av Antenn Nedvipp för Outdoor-Användare i Stadsmiljö A Study on the Impact of Antenna Downtilt on the Outdoor Users in an Urban Environment Författare Author Pradeepa Ramachandra Sammanfattning Abstract
Inter-site interference distribution acts as a basic limitation on how much perfor-mance a network service provider can achieve in an urban network scenario. There are many different ways of controlling this interference levels. One such method is tuning the antenna downtilt depending on the network situation. Antenna down-tilt can also be seen as a powerful tool for load balancing in the network.
This thesis work involves a study of the impact of the antenna downtilt in an urban environment, involving non-uniform user distribution. A realistic dual ray propagation model is used to model the path gain from the base station to a UE. Such a propagation model is used along with a directional antenna radiation pattern model to calculate the overall path gain from the base station to a UE. Under such modeling, the results of the simulations show that the antenna downtilt plays a crucial role in optimizing the network performance. The results show that the optimal antenna downtilt angle is not very sensitive to the location of the hotspot in the network. The results also show that the antenna downtilt sensitivity is very much dependent on the network scenario. The coupling between the antenna downtilt and the elevation half power beamwidth is also evaluated.
Nyckelord
Abstract
Inter-site interference distribution acts as a basic limitation on how much perfor-mance a network service provider can achieve in an urban network scenario. There are many different ways of controlling this interference levels. One such method is tuning the antenna downtilt depending on the network situation. Antenna down-tilt can also be seen as a powerful tool for load balancing in the network.
This thesis work involves a study of the impact of the antenna downtilt in an urban environment, involving non-uniform user distribution. A realistic dual ray propagation model is used to model the path gain from the base station to a UE. Such a propagation model is used along with a directional antenna radiation pattern model to calculate the overall path gain from the base station to a UE. Under such modeling, the results of the simulations show that the antenna downtilt plays a crucial role in optimizing the network performance. The results show that the optimal antenna downtilt angle is not very sensitive to the location of the hotspot in the network. The results also show that the antenna downtilt sensitivity is very much dependent on the network scenario. The coupling between the antenna downtilt and the elevation half power beamwidth is also evaluated.
Acknowledgments
It has been a privilege to be guided by Mehdi Amirijoo and words fail me in conveying the admiration that I hold for him. I wish to thank him for providing this opportunity and for the many hours that he spared. His enthusiasm and fascination have been infectious and I will certainly carry them with me forward in life. Thanks are due to Birgitta Olin for guiding me through the simulator and also in giving valuable inputs all along the thesis. I am grateful to Jan-Erik Berg for his patience in answering my long list of doubts in the propagation model. Martin Johansson and Fredrik Gunnarsson have at various point of time answered my doubts and I wish to acknowledge their help.
I am also thankful to Chaitanya TVK for the patience and grace that he has shown during the many times that I have approached him. Thanks are due to Danyo Danev for consenting to be the examiner. I owe immensely to the LinLab team at Ericsson Research, Linköping, for making the workplace very friendly. And lastly, but in no small measure do I wish to acknowledge the benevolence of the invisible and unassuming hands of the Swedish taxpayer who made it possible for me to have this education.
Contents
List of figures . . . xii
List of tables . . . xiii
1 Introduction 1 1.1 Self Organizing Networks . . . 1
1.1.1 Introduction . . . 1
1.1.2 Self-optimization . . . 2
1.2 Antenna Fundamentals . . . 3
1.2.1 Antenna Parameters . . . 4
1.2.2 Directional Antenna Radiation Pattern . . . 5
1.3 Literature Review . . . 7
1.4 Thesis Contribution . . . 8
1.5 Thesis Organization . . . 8
2 Problem Formulation 9 2.1 Aim of the Study . . . 9
2.2 Methodology . . . 9
2.2.1 Simulation Scenario . . . 10
2.2.2 Observability Study . . . 10
2.2.3 Controllability Study . . . 10
2.3 Scope and Limitation . . . 11
3 Simulation Scenario 13 3.1 Basic Simulation Settings . . . 13
3.1.1 Network Scenarios . . . 15
3.2 Antenna Radiation Pattern Modeling . . . 18
4 Observability 21 4.1 Optimization Objective . . . 21
4.1.1 5th Percentile User Throughput . . . . 21
4.1.2 Spectral Efficiency of the Cell . . . 22
4.2 Region of Observability . . . 23 ix
x Contents
5 Controllability 27
5.1 Input towards Self Planning Scenario . . . 28 5.2 Impact of the Location of Hotspot on the Optimality of Downtilt
Angle . . . 30 5.3 Sensitivity Analysis . . . 30 5.4 Comparison with Okumura-Hata Propagation Model . . . 34 5.4.1 Comparison with Urban Propagation Model Results . . . . 35
6 Impact of Elevation Beamwidth on Optimal Downtilt 39
6.1 Lower Elevation Beamwidth - 6.40 . . . . 39 6.2 Higher Elevation Beamwidth - 100 . . . . 40
7 Conclusions and Future Work 43
7.1 Conclusion . . . 43 7.1.1 Optimal Downtilt Angle . . . 43 7.1.2 Sensitivity of Downtilt for Different Hotspot Location . . . 43 7.1.3 Impact of Elevation Beamwidth on Optimal Downtilt Angle 44 7.2 Future Work . . . 44 7.2.1 Introduction of Indoor and Vertical Plane Users . . . 44 7.2.2 Study of the Impact of using a Real Antenna Model . . . . 44 7.2.3 Study of the Impact of Antenna Downtilt for Uplink
Trans-mission . . . 45 7.2.4 Study of the Impact of Horizontal Radiation Pattern . . . . 45 A 5thpercentile user throughput for different initialization seeds and
elevation beamwidth of 80 47
B 5thpercentile user throughput for different initialization seeds and
elevation beamwidth of 6.40 52
C 5thpercentile user throughput for different initialization seeds and
elevation beamwidth of 100 56
List of Figures
1.1 Antenna coordinate system . . . 4
1.2 Directional antenna radiation pattern . . . 6
3.1 Network deployment with wrap around copies . . . 15
3.2 Urban loading scenario . . . 16
3.3 5th percentile user throughput vs downtilt for urban case . . . 17
3.4 Spectral efficiency vs downtilt for urban case . . . 17
3.5 Horizontal antenna radiation pattern of HV model. . . 18
3.6 Vertical antenna radiation pattern of HV model. . . 19
4.1 Changes in path loss while changing downtilt from 110 to 120 . . . 23
4.2 Observability region . . . 24
4.3 Observability region . . . 25
5.1 Investigated hotspot locations . . . 28
5.2 User throughput vs downtilt angle for different hotspot locations . 29 5.3 5thpercentile user throughput vs downtilt angle for different hotspot locations . . . 29
5.4 Cell spectral efficiency vs downtilt angle for different hotspot locations 30 5.5 Particular seed results of 5thSINR values . . . . 31
5.6 Load distribution . . . 31
5.7 Particular seed results of 5thpercentile user throughput values . . 32
5.8 Deployment to understand interference situation . . . 33
5.9 5th percentile user throughput of selected cells. . . 33
5.10 User distribution in selected cells. . . 34
5.11 SINR distribution for the Okumura-Hata propagation model . . . 35
5.12 SINR distribution for the Urban propagation model . . . 36
5.13 5th percentile user throughput vs downtilt angle for OkumuraHata model . . . 36
6.1 5th percentile user throughput as a function of downtilt angle for an elevation beamwidth of 6.40. . . 40
6.2 5th percentile user throughput as a function of downtilt angle for an elevation beamwidth of 100. . . . 41
A.1 Simulations with seed#1 for an elevation beamwidth of 80. . . . . 48
A.2 Simulations with seed#2 for an elevation beamwidth of 80. . . . . 48
A.3 Simulations with seed#3 for an elevation beamwidth of 80. . . 49
A.4 Simulations with seed#4 for an elevation beamwidth of 80. . . 49
A.5 Simulations with seed#5 for an elevation beamwidth of 80. . . . . 50
A.6 Simulations with seed#6 for an elevation beamwidth of 80. . . . . 50
A.7 Simulations with seed#7 for an elevation beamwidth of 80. . . . . 51
A.8 Simulations with seed#8 for an elevation beamwidth of 80. . . 51
B.2 Simulations with seed#2 for an elevation beamwidth of 6.40. . . . 53
B.3 Simulations with seed#3 for an elevation beamwidth of 6.40. . . . 54
B.4 Simulations with seed#4 for an elevation beamwidth of 6.40. . . . 54
B.5 Simulations with seed#5 for an elevation beamwidth of 6.40. . . . 55
C.1 Simulations with seed#1 for an elevation beamwidth of 100. . . . . 57
C.2 Simulations with seed#2 for an elevation beamwidth of 100. . . . . 57
C.3 Simulations with seed#3 for an elevation beamwidth of 100. . . . . 58
3.1 Common network parameter settings. . . 14 3.2 Urban network parameter settings. . . 16
Abbreviation
3GPP Third Generation Partnership Project AWGN Additive White Gaussian Noise CQI Channel Quality Information FTP File Transfer Protocol HV Horizontal-Vertical LOS Line Of Sight
LTE Long Term Evolution OPEX OPerational EXpenditure QoE Quality of Experience
RSRP Reference Signal Received Power SINR Signal to Interference and Noise Ratio SON Self Organizing Network
UE User Equipment
Chapter 1
Introduction
The cellular network has seen an unprecedented growth in the last decade. With a drastic increase in cellular traffic, it has become very challenging to meet all the requirements from both service providers’ and users’ perspective. A large amount of effort has been invested to make these cellular networks capable of handling themselves. It is in this area, an application of Self Organizing Networks (SON) plays a crucial role. There are many parameters in a network that can be used to fine tune the network performance in accordance to the service providers’ requirements. Antenna downtilt is one such parameter. In this chapter, a brief introduction to the SON is provided. Later, some of the basic antenna parameters are discussed and also some previous literature that deals with the impact of antenna downtilt is discussed.
1.1
Self Organizing Networks
A SON can refer to any network that tries to organize itself by looking at the parameters available in its control. A SON network will try to get the best out of its available resources by continually monitoring the variations in the network performance. In this report, SON is used to refer to Self Organizing Networks in cellular use case.
1.1.1
Introduction
With a drastic increase in network traffic, there is a need for the service providers to minimize the human involvement in handling of radio network management. This is where the Self-Organizing Networks (SON) plays a major role. SON also play a key role in reducing the Operational Expenditure (OPEX) of the network operators, improving the overall system performance and in faster adaptation to the changing network conditions.
The SON functionalities can be broadly classified as, 1. Self-planning
2. Self-configuration 3. Self-optimization 4. Self-healing Self-planning
Self-planning involves the planning of the location of a new base station and de-ciding its radio and transport parameters. As an example, it basically specifies the number of sectors required, suitable antenna and power settings for these sectors. This is based on the capacity or coverage goals of the network, estimated traffic forecast and some predefined limitations on the available resources from a service provider. The parameters that are set during the planning phase will be default parameters and they can be optimized during the self-optimization process. Self-configuration
Self-configuration involves the procedures required for bringing up the newly planned eNodeB. They involve hardware installation, transmission setup, node authenti-cation, automatic software download to the eNodeB and also self testing.
Self-optimization
The radio network parameters are continuously optimized based on the network operators’ objective. The objective can be improving the network capacity or the network coverage or it can be in terms of improving the user experience. More details regarding self-optimization are covered in section 1.1.2.
Self-healing
Self-healing mainly involves software updating for the system maintenance and also cell outage detection and cell outage compensation.
1.1.2
Self-optimization
Self-optimization involves changing the radio network parameters to enhance the network performance. The network performance metric can be evaluated in terms of network capacity, network coverage or the quality of experience (QoE) as per the users in the network. As an example for the capacity based objective, one can consider improving the cell throughput in terms of achieved bit rate. Coverage can be enhanced by improving the 5th percentile spectral efficiency of the users
in the cell. The objective can be power efficiency oriented, where reduction of transmission power is considered.
The self-optimization process can be categorized as: 1. Off-line optimization
1.2 Antenna Fundamentals 3
The following subsections will brief about the off-line and the on-line optimiza-tion process.
Off-line Optimization
Off-line optimization is an open-loop optimization process. As the optimization process is open loop, the impact of the changes made during the optimization will not be fed back. Therefore, one has to be careful while changing the parameters as it might have a large negative impact on the network performance. In order to do so, the network needs to be observed for a long duration so that one has confidence on the knowledge of network working condition. By observing for a longer duration, generally one can estimate the network parameters, like loading on the network and user density, with a smaller variance. This knowledge along with similar knowledge from the neighboring cells will help in choosing the parameters of the optimization algorithm. The impact of these changes is initially estimated by using some propagation models and this will be used as reference for choosing the right parameters for optimization. Based on the confidence in the knowledge of the network, a large degree of freedom is provided in choosing the parameters.
On-line Optimization
On-line optimization is a closed-loop optimization process. Here, the radio param-eters are changed and its impact on the network is continuously monitored. Using this as the feedback, the radio parameters are retuned to optimize the network performance and this process will continue indefinitely, adapting to the varying network conditions or for a predefined number of retuning iterations. Here, the knowledge about the network might not be accurate enough to explore the entire dynamic range of the parameters to be tuned and therefore, a careful evaluation of the impact of changing these parameters on the network performance needs to be considered.
The knowledge of the behavior of a particular objective function while tuning a given parameter will help in developing an efficient optimization algorithm. In this thesis work, the analysis with respect to the impact of antenna downtilt in a realistic scenario is considered. The antenna downtilt will in general control the distribution of inter-site interference in the network. By varying the antenna downtilt, one controls the boundary of the cell with the neighboring sites. Un-derstanding how the antenna downtilt will impact the users within its coverage area and also understanding the interference caused to the neighboring cell is very important. The results of thesis work will act as a pre-study for the on-line optimization of antenna parameters in a Long Term Evolution (LTE) network scenario.
1.2
Antenna Fundamentals
An antenna is used to convert the guided waves in a feeder cable or a transmission line into a radiating wave traveling in free space. The pattern in which the
radi-
Figure 1.1. Coordinate system for antenna calculations.
ating wave travels in the free space can be controlled by using different antenna parameters.
1.2.1
Antenna Parameters
Some of the terminology related to the antennas is used repeatedly throughout this thesis. A brief explanation of these parameters is given in this sub-section. The diagram indicating the coordinate system for the antenna calculations is given in the Figure 1.1.
Antenna radiation pattern is the plot of radiated power from an antenna per unit solid angle. In other words, it is a plot of the power radiated from an antenna per radiation intensity of the antenna, represented by U. This radiation intensity of the antenna is given by [1],
U = r2 1 2EθH ∗ φrˆ = Z0 2 kI(0)L 4π sin2θ (1.1) where:
• U - radiation intensity of the antenna,
• r - radius of the sphere on which the measurement is done, • Eθ - electric field component along the Poynting direction,
• Hφ - magnetic field component along the Poynting direction,
1.2 Antenna Fundamentals 5 • Z0- wave impedance = 120π,
• k - wave number = 2π λ ,
• I(0) - input current to the antenna, • L - length of the wire,
• θ - polar angle.
The antenna parameters that define the radiation pattern are explained very briefly below.
• Poynting vector describes the magnitude and direction of the power flow carried by the wave per square meter of area, and is measured in watts per square meter.
• Directivity of an antenna is the ratio of radiation intensity of the antenna in a particular direction to the radiation intensity of an isotropic antenna radiating with same total power.
• Side lobe level is the amplitude of the biggest side lobe expressed in decibels relative to the peak of the main lobe.
• Front-to-back ratio is the ratio of the peak amplitudes of the main lobe and back lobes expressed in decibels.
• Antenna downtilt is the direction of the main lobe in the vertical direction with positive values for the down side tilting of the main lobe. The down-tilting of the antenna radiation pattern can be done either by mechanical downtilting or by electrical downtilting. In case of mechanical downtilting, with changes in the downtilt values, there will be a variation in the horizon-tal radiation pattern of the antenna. In electrical downtilting, only vertical antenna radiation pattern is affected. In this report, downtilt and electrical downtilt terms are used interchangeably.
• Antenna azimuth orientation is the direction of the main lobe in the hor-izontal direction with positive values for the clockwise measurements from the horizontal axis.
• Half power beamwidth is the angle subtended by the half power points of the main lobe.
1.2.2
Directional Antenna Radiation Pattern
In order to reuse the available spectrum efficiently and also to improve the path gain in a particular direction in comparison with other directions, modern radio networks use directional antennas. A directional antenna radiation pattern can be obtained by using antenna arrays. By increasing the number of elements in an antenna array, one can achieve highly directional radiation patterns. As an
Figure 1.2. An example radiation pattern indicating some of the antenna parameters.
example, the antenna radiation pattern obtained by a seven element linear antenna array is shown in Figure 1.2.
Ideally, the total instantaneous radiated power at a point is found by inte-grating the instantaneous Poynting vector over the spherical surface. In order to minimize the involved calculations, a simplified antenna radiation pattern model is used as suggested by 3GPP model [8] and is given by:
A(θ, φ) = −min [− (AH(φ) + AV(θ)) , Am] , (1.2)
where, AH(φ) and AV(θ) are the horizontal and the vertical antenna radiation
patterns respectively. The vertical and the horizontal antenna radiation patterns are given by:
AV(θ) = −min " 12 θ − θtilt θ3dB 2 , SLAv # , (1.3) AH(φ) = −min " 12 φ − φori φ3dB 2 , Am # . (1.4) where:
• θ - elevation angle from the user’s current location to the cell, • φ - azimuth angle from the user’s current location to the cell, • θtilt - antenna downtilt at the base station cell,
1.3 Literature Review 7 • θ3dB - vertical half power beamwidth,
• φ3dB - horizontal half power beamwidth, • SLAv - side lobe attenuation = 20dB,
• Am- front to back ratio = 25.
This approximation of the antenna radiation pattern will help in reducing the mathematical operations while taking antenna directivity losses into account.
1.3
Literature Review
Fair amount of effort has been invested in optimizing the antenna parameters, an-tenna downtilt in particular, to suit the changing network demands. Brief overview of some of these existing literature is provided in this section.
A detailed analysis of the vertical antenna downtilt optimization for LTE base stations is carried out by Eckhardt et al. [2]. In this work, the authors discuss the multi-cell optimization technique for a three-sector LTE network by considering spectral efficiency as the optimization objective. Here, the cell-edge users are given higher priority in order to improve the coverage of the cell. Up to 16 cells are considered for optimization at a given time instance. The authors illustrate the effect of optimization with respect to the utility per sector and also its impact during a cell outage. The results show that the average spectral efficiency increases by 10% and the spectral efficiency at the cell edge (5thpercentile) will increase by
up to 100% after the optimization.
Impact of the combined optimization of antenna downtilt and horizontal and vertical half power beamwidths is given by Yilmaz et al. [3]. In this work, the au-thors discuss the multi-variable optimization for cell coverage and capacity under uniform user distribution for full buffer traffic. The behavior of these parameters is considered for two different inter-site distance scenarios, namely 500m and 1732m. The results show that the optimal downtilt angle is independent of the horizontal half power beamwidth in both the cases. The results also show that, in an interfer-ence limited scenario, a wider horizontal beamwidth and a greater downtilt angle will provide a better coverage and capacity performances. In the noise limited scenario, the downtilt angle and horizontal beamwidth will not have much impact to a large range of values. The results also show that the optimal parameters will vary for noise limited and interference limited scenarios. All results are given in further detail in [4].
A detailed study of the antenna parameter optimization involving the antenna downtilt and the elevation beamwidth is carried out by Athley et al. [5] . In this work, the authors propose a default antenna downtilt setting depending on the inter-site distances and the antenna elevation beamwidth. The results also show the coupling between the optimal antenna downtilt and the antenna elevation beamwidth.
1.4
Thesis Contribution
In all the existing literatures that addresses the antenna downtilt related issues, the propagation model used do not consider the direct and the reflected rays separately. In these literatures, a Okumura-Hata like propagation model is considered. In a real network scenario, the propagation will be more complex with many buildings in an urban scenario. There will be some locations experiencing Line Pf Sight (LOS) connection with a base station and also some regions that experience deep shadowing. Presence of the buildings will also impact the path between the base station and a User Equipment (UE) and there by changing the understanding of the optimal downtilt angle. In this thesis work, a more realistic urban dual ray propagation model is considered for studying the impact of antenna downtilt on the network performance. The ideas behind this propagation model are explained in [6] and [7].
1.5
Thesis Organization
The report is divided into further siz different chapters. In the chapter 2, the prob-lem that is being addressed in the thesis is discussed and the approach adopted during the thesis is presented. Chapter 3 covers the modeling of the antenna radia-tion pattern and also discusses many other parameter settings. Chapter 4 explains the method used to understand the impact of the antenna downtilt changes on the network performance is presented. In the chapter 5, the results of changing the antenna downtilt are presented. In the chapter 6, the relation of the antenna down-tilt to the antenna elevation beamwidth is discussed. In the concluding chapter 7, the conclusions that are drawn from the thesis and the possible future works in this field are discussed.
Chapter 2
Problem Formulation
In a densely loaded urban network, antenna downtilt can act as a crucial parameter in improving the network performance. This section provides the aim of the study and also the methodology involved in deriving the results.
2.1
Aim of the Study
The aim of this study can be listed as follows:
1. Incorporate relevant changes in the existing propagation model in the simu-lator that suits the urban environment and also the ones which help in un-derstanding the impact of antenna parameters on the network performance. 2. Study the impact of different locations of hotspot in the network on the
optimal antenna downtilt angle.
3. Study the impact of the elevation HPBW on the network performance and also study how it affects the optimal downtilt angle.
4. Derive the conclusions from the study that can be used for the self planning and self optimization processes of a SON network.
The above mentioned tasks need to be carried out in a systematic way for which the following methodology is adapted.
2.2
Methodology
Antenna downtilt plays an important role in shaping the inter-site interference dis-tribution. As the traffic load varies continuously with time and with geographical location, it is important to understand how the antenna downtilt can be tuned accordingly. A systematic approach for studying the impact of antenna downtilt involves several key steps. They are briefly explained in the following sub-sections.
2.2.1
Simulation Scenario
The first step in carrying out this study is to establish a simulation set up that depicts an urban environment. After that, parameters related to an urban network like loading on the network and default antenna parameter settings of all the cells in the network also needs to be derived. By the end of this study, one should be able to simulate an urban environment that can be used for antenna parameter related studies.
2.2.2
Observability Study
The operators will try to optimize the performance of the network by trying to fine tune the parameters available at their disposal. One such tunable parameter is antenna downtilt. From an operator’s point of view, it is important to decide what needs to be improved or optimized in the network. Once the optimization objec-tive is decided then its relation to the antenna downtilt needs to be understood. Optimization objectives can be cell throughput based (as an example, improving the average spectral efficiency in the cell) or cell coverage based (as an example, improving the 5th percentile user spectral efficiency in the cell) or it can just be improving the user experience (as an example, Signal to Interference and Noise Ratio (SINR) improvement).
There is also a need to establish a region of observation in which one can observe the changes in the network performance while tuning the antenna parameters. This will help in understanding the extent of the impact of changing the antenna parameter. Considering a too small region might show large improvement or degradation during the modification of the antenna downtilt value but it might not cover the impact on the overall network. Whereas, considering a large region might not help in observing the impact of changes as it might average out the results and therefore, acts as a poor feedback. Therefore, a region of observation needs to be decided before trying to understand how antenna downtilt is affecting the network performance.
2.2.3
Controllability Study
This study will involve understanding of the impact of vertical radiation pattern of the antenna on the network performance. This study can be split into following sub studies.
1. Study the impact of different locations of the hotspot in a cell on its optimal downtilt angle. Study the relation between the position of the hotspot and the optimal downtilt angle.
2. Study the impact of smaller and larger elevation HPBW on the optimal downtilt angle.
3. Draw conclusions from the above mentioned studies. This study will also list all the observations that can be seen in the results and how can they be used in the future works in this field.
2.3 Scope and Limitation 11
2.3
Scope and Limitation
Not all the network scenarios are considered for evaluation in this thesis. A very specific set of simulation parameters is used for analysis. The parameters are chosen carefully to depict a realistic network scenario. Some of the scope and limitations of the thesis work are,
• Only downlink direction of transmission is considered for evaluation. The optimal downtilt value can be different for downlink and uplink direction of transmission.
• Users considered in the simulator are static in nature i.e. the users will arrive at a particular location and they continue to remain at the same location until the fulfillment of their request.
• Users in the network are always assumed to be outdoor users. No indoor related losses are added in the simulations. Also, as all the users in the network are outdoor users, they are assumed to be at the ground level. Some of the limitations imposed here will help in reducing the variances in the measured values which can potentially make it difficult to see the impact of antenna downtilt. Such reduction in variances will help in easier understanding of the impact of the antenna parameters.
Chapter 3
Simulation Scenario
A Matlab based simulator is used to carry out the simulations during the thesis work. The simulator depicts the entire network in a time-dynamic fashion, i.e., it tries to mimic a real network environment for every specified interval of time.
Different network parameters can have different effects on the optimality of the antenna downtilt. For example, if we consider intersite distance as a network parameter, then the optimal antenna downtilt value decreases with the increase in the intersite distance. This is due to the larger coverage requirement per cell with the increase in the intersite distance. Also, mean height of the buildings in a city will also have an impact on the optimal downtilt angle. Here, a larger mean building height will force a smaller downtilt angle to be the optimal downtilt value. Therefore, it is important to declare the simulation settings that are used in the simulations to study the impact of the antenna downtilt.
3.1
Basic Simulation Settings
Some of the common network settings that are used for all the network scenarios studied during this thesis work are given in Table 3.1.
The wrap around functionality will ensure that the region that is far away from the central cell will also experience the effects of changing the antenna parameter in the center cell. A plot indicating the network deployment with part of its wrap around copies is shown in the Figure 3.1. The blue colored cells are the corresponding center cells of each of the wrap around copies. The gray colored cells belong to the center network region that is being investigated. In the remaining parts of the report, only central region shown in gray color is used for illustration, but the simulations are carried out with the wrap arounds.
All the base station antennas are initialized with the default antenna parameter settings given in Table 3.1. Only downlink is considered for the study during the thesis work. All the users arriving in the network will be expecting a download of 80,000 bits of data in the file transfer mode. The propagation model used is explained in [6] and [7].
Table 3.1. Common network parameter settings.
Number of macro sites 7
Number of sectors per macro site 3 Number of wrap around copies 7
Network bandwidth 10 MHz
Traffic type FTP - small packets (80000 bits)
Scheduling Round-robin
Transmission power of base station 20 W
Propagation model Urban model
Transmission mode Downlink
Base station antenna height 28.5 m
UE antenna height 1.5 m
Number of transmit antennas 2
Number of receive antennas 2
Default azimuth orientation 00 Default horizontal half power beamwidth 650
Default vertical half power beamwidth 80
Side lobe level 17 dB
Front to back ratio 25
The parameters given in Table 3.1 can be used to derive different network per-formance parameters. For example, one can use Table 3.1 to calculate the number of bits transmitted by the base station to a particular UE during a given time instant. This will depend upon the SINR experienced by the UE in the previous time instant and this information will be fed back to the base station through control signaling (Channel Quality Information (CQI) information). Using these parameters, the number of bits transmitted by the base station to a particular UE is given by,
N umBits = N umBitsperRE ∗ N umREperSubband ∗
N umSubband ∗ (1 − ControlOverhead) (3.1) where,
• ControlOverHead = 0.15, is the different control signaling overhead involved in terms of bandwidth utilization,
3.1 Basic Simulation Settings 15
Figure 3.1. Network deployment with parts of the wrap around copies being shown.
• N umREperSubband= 14*12 = 168, where 14 is the number of time slots and 12 is the number of sub carriers.
The value of control overhead is pre calculated and its value is used during all calculations although the actual control signaling is not simulated.
3.1.1
Network Scenarios
Before starting the antenna parameter tuning related study, the network scenario under which the simulations are carried out needs to be defined. In the following sub-sections we define the set of network scenarios that are investigated during the thesis work.
In an urban network, there will be high traffic density because of the presence of large number of users in the network. This will demand for a denser deployment of the base stations and therefore the inter-site distance in an urban network is small. The user distribution in the network will not be uniform and it is necessary to have hotspots while simulating an urban network. As the antenna parameter variations in only one cell is considered for the analysis, the hotspots in the network are simulated to be present in the cell which has the reconfigurable antenna setting. The density of arrival of users in the hotspot will determine how crowded a hotspot is in comparison with the neighboring area.
to have an impact because of the previous antenna parameter settings for a brief span of time. This time span is called as queue settling time. Queue settling time should not be considered for performance evaluation of the new antenna parameter settings. The value of queue settling time is calculated by observing the time it takes for the resource block utilization in a cell to reach a steady state value when the downtilt is changed from one value to another value.
The duration of observation of the network that captures all the impact of a change in the antenna parameter is called sampling time. This is the observed duration of the network for a given antenna parameter setting. This duration should be long enough to average out any event that causes instantaneous large variation of the network performance. Such events can be related to the sequence of user arrival or the location of user arrival. Some of these specific network settings used with respect to the urban scenario are given in Table 3.2.
Table 3.2. Urban network parameter settings. Inter-site distance 500 m Radius of circular hotspot 40 m
Hotspot density 10
Number of hotspots in center cell 1 Queue settling time 10 s
Sampling time 90 s
Load on the network 145 Mbits/s/km2 Default downtilt angle 100
0 50 100 150 200 0 0.2 0.4 0.6 0.8 1 Load in Mbits/s/km2
Resource block utilization
RB utilization vs Load
Figure 3.2. Resource block utilization as a function of the load on the network, measured
3.1 Basic Simulation Settings 17 0 2 4 6 8 10 12 14 16 18 20 0 1 2 3 4 5 6x 10 6
Down tilt angle in degrees
5th percentile user throughput in bits/s
Figure 3.3. 5th percentile user throughput as a function of common downtilt angle of all the cells in the urban network scenario.
0 2 4 6 8 10 12 14 16 18 20 0.5 1 1.5 2 2.5
Downtilt angle in degrees
Cell spectral efficiency in bits/s/Hz
Figure 3.4. Spectral efficiency as a function of the downtilt angle value of all cells in
the urban network scenario.
Load on the network is set by considering the percentage of resource block utilization while sweeping the load in small steps. In an urban network scenario, as more traffic is expected, the load value which gives 50% resource block utilization is considered for evaluation. This is obtained from the Figure 3.2. The load value that corresponds to 50% resource block utilization is 145 Mbits/s/km2.
The value for the default downtilt angle in all the cells is calculated by sweep-ing the downtilt angle from 00 to 200 for all the cells simultaneously in steps of 10. The 5th percentile user throughput is used as a performance metric. The
obtained results are plotted as shown in the Figure 3.3. The results show that a downtilt of 100 can be chosen as the common downtilt in the network. As a
further conformance, the spectral efficiency of all the cells is plotted against the common downtilt angle in the Figure 3.4. The result from this figure will also point towards having 100as the default downtilt angle.
3.2
Antenna Radiation Pattern Modeling
Different antenna models are supported in the simulator including the 3GPP model [8] and the Omni-directional model. The antenna model considered in this thesis work is Horizontal-Vertical model (HV model). This model’s antenna gain val-ues in horizontal and vertical directions are given in Figure 3.5 and Figure 3.6 respectively. 0 50 100 150 200 250 300 350 400 −10 −5 0 5 10 15 20
Horizontal angle in degrees
Horizontal antenna directivity gain
Horizontal antenna directivity gain
3.2 Antenna Radiation Pattern Modeling 19 0 50 100 150 200 250 300 350 400 −45 −40 −35 −30 −25 −20 −15 −10 −5 0
Vertical angle in degrees
Vertical antenna directivity gain
Vertical Antenna Directivity
Chapter 4
Observability
Network service providers will continuously try to get the best out of the resources available at their disposal. A service provider initially needs to decide what to op-timize in the network. This optimization objective might be cell capacity centered or cell coverage centered or it even might deal with improving the user experi-ence in the network. Therefore, it is important to initially define the objective of optimization. Once the objective of optimization is decided, the optimization algo-rithm can be initialized. In order to provide efficient feedback for this optimization algorithm, it is important to know the region of network that will get impacted while changing the parameters of a given cell. These two important aspects are discussed in the subsequent sub-sections.
4.1
Optimization Objective
The performance of a network can be measured in many different ways. In this thesis work, two different optimization objectives are considered to analyze the impact of changes in antenna downtilt on the performance of the network. They are:
1. 5thpercentile user throughput
2. Spectral efficiency of the cell
The 5thpercentile user throughput is used as the primary criterion for
evalua-tion and when the results need more analysis, spectral efficiency is also considered.
4.1.1
5
thPercentile User Throughput
The path gain from the base station to the UE will depend on many factors. By looking at obtained path gain values from each cell, the serving cell is chosen to be the one with the highest path gain to the UE (transmission power for all cells is same as mentioned in Table 3.1). The SINR value as experienced by the UE is
the ratio of received power from the serving cell to the interference caused by the power levels from all the other cells. It is given by,
SIN RU E = PtGs N +P i6=s(PtGi) (4.1) where,
• SIN RU E - SINR as experienced by the UE,
• Pt- transmission power of the base station as given in Table 3.1,
• N - thermal noise power,
• Gs- total path gain from the serving cell to the UE,
• Gi -total path gain from the non-serving ith cell to the UE.
The total path gain used in the above calculations is the sum of the path gain from the ithcell to the UE (P G
i) and the antenna directivity gain (ADi) and is
given by,
Gi= (P Gi+ ADi) dB (4.2)
The SINR value so obtained is fed back to the base station not directly but in the form of Channel Quality Information (CQI). In the simulations, SINR value is directly used to calculate the number of bits that can be sent through the channel in the next scheduled time slot for the UE. Depending on the total path gain as observed by the user, the user throughput will vary. The user throughput is the ratio of the number of bits requested by the user for downloading and the time taken by the base station to full fill the user’s request. It is given by,
U serT hroughput = N umber of bits requested
LeavingT ime − ArrivalT ime (4.3)
In this equation, the LeavingT ime is the time instant when the last part of the download was fulfilled and ArrivalT ime is the time instant when the user requested for the download.
4.1.2
Spectral Efficiency of the Cell
Spectral efficiency is the number of bits transmitted per second per unit bandwidth and is measured in bits/s/Hz. As the total path gain to the UEs gets worse, one needs more and more resource blocks to transmit the same amount of data and this will bring down the overall spectral efficiency of the cell. Therefore, one can use this as a measure of the performance of the network as well.
4.2 Region of Observability 23 −800 −600 −400 −200 0 200 400 600 800 −600 −400 −200 0 200 400 600 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Distance in meters Distance in meters
Figure 4.1. Bin positions that experience a change in the SINR value when the downtilt
angle of the center cell is changed from 110 to 120 in an urban network scenario.
4.2
Region of Observability
Changing the value of the antenna downtilt in a given cell will not only affect its own performance but also the performance in the neighboring cells. This can be illustrated with the help of the Figure 4.1. The brown and blue colored dots in the figure indicate the bin positions that experienced a change in SINR value when the downtilt of the center cell is changed from 110 to 120. The blue colored dots indicate the bin positions that experienced an improvement in the SINR and the brown colored bin positions indicate the degradation in SINR. This change is because of the variation of antenna directivity gain to those bin positions. The bin positions on the left side of the Figure 4.1 are affected because of the wrap around phenomenon. There are also bin positions very close to the central base station but in the opposite direction of its antenna propagation direction, and these are affected by the back lobe of the vertical antenna radiation pattern as shown in the Figure 3.6. Therefore, it is necessary to establish a region in the network to look for changes while changing the parameters in the center cell.
There are many different ways in which one can choose the region of observ-ability. Selecting a large number of cells in the network will ensure that all the effects of changing the parameters in the center cell are captured but it comes at a cost. When we include a large region for evaluation, we include the locations which will hardly see any change in their path gain values but they still continue to be included in the evaluation process. These values might overshadow the actual impacted region. Therefore, considering a large area will not help in having an effective feedback mechanism for the optimization process. A very small region will not be sufficient to capture the significant changes that might take place in
Figure 4.2. Region of Observability - All the colored cells are considered for evaluation.
different regions of the network.
In order to choose the region of observability, a study of the impact of antenna downtilt of the center cell to all the cells is carried out individually. The 5th
percentile user throughput in each cell is considered as the performance metric. Using this data, the region of observability is decided and is shown in the Figure 4.2. In the figure, the wrap-around cells are also considered. In order to simulate the observability region using a single copy of the network, different set of cells (but having the same effect after wrap around) is considered. The region of observability for the single copy of the network is shown in the Figure 4.3.
The center cell is denoted by orange color and the antenna downtilt of this cell is changed during evaluation. The cells that share immediate geographical border with this cell are shown in green color. The cells that are part of observability region and share geographical boundaries with green colored cells are shown in yellow color. The cells that are part of observability region and share geographical boundaries with yellow colored cells are shown in gray color. For yellow and gray colored cells, one has to look at the network with wrap around in mind (for wrap around, refer to the Figure 3.1).
Some of the key observations in this region of observability study are:
• There will be variations in the 5thpercentile user throughput in all (orange,
green, yellow and gray colored) cells belonging to observability region while changing the downtilt of center cell in the range 00 to 80.
• The impact of variations in the 5th percentile user throughput is limited
only to orange, green and yellow cells while varying the downtilt values in the range 80 to 140.
4.2 Region of Observability 25
Figure 4.3. Region of Observability - All the cells that are colored other than blue color
are considered for evaluation.
• Only orange and green cell’s 5thpercentile user throughput will get impacted
while changing the downtilt values in the range 140 to 200.
These observations are based on the changes in the 5thpercentile user
through-put values of each individual cell while sweeping the downtilt from 00 to 200 in steps of 10. When a cell continues to have the same 5thpercentile user throughput
value even while changing the downtilt of the center cell, then one can say that the particular cell has become insensitive to the downtilt of the center cell. The reasoning behind this observation is that, the vertical antenna radiation pattern will hit the flooring value (referring to Figure 3.6) for distant users. This will make these users insensitive to higher downtilt values of the center cell.
The region of observability as shown in the Figure 4.3 will take care of the entire downtilt range of 00 to 200. This region can be reduced based on the above observations while having a smaller range of downtilt values of the center cell. In other words, one can choose the region of observability based on the dynamic downtilt range that is available during optimization based on the above observations.
Chapter 5
Controllability
With the details of the simulation settings and the observable region already being established, in this chapter, a detailed study of the impact of antenna downtilt on the performance of the network and also the sensitivity analysis of antenna downtilt is carried out. The range in which the measured value is within 95% of its optimum value is calculated. The width of this interval can be used as a measure of sensitivity under the given network scenario. In this study, all the antenna parameters of the center cell other than the antenna downtilt are kept constant as per Table 3.1. The downtilt of the center cell is swept from 00 to 200 in steps of 10.
The optimization objective considered for evaluation is the 5thpercentile user
throughput. The Figure 5.1 indicates the different hotspot location scenarios that are investigated. Three subplots along the first row indicate the scenarios where in, the hotspot is located close to the center cell’s deployment position. Three subplots along the second row indicate the scenarios where in the hotspot is located at a distance of 150 to 200 m away from the center cell deployment location. The subplots along the final row indicate the scenarios where in hotspot is located at the cell edge.
The result of sweeping the downtilt angle of the center cell is shown in the Figure 5.2. Along with the 5thpercentile user throughput, 50thand 95thpercentile user throughput are also plotted in the Figure 5.2. These results are in sync with respect to the position of hotspots as shown in the Figure 5.1. More results with different seeds are given in the Appendix A. Here, the seed refers to the seed used by the Matlab random generator. By running multiple simulations with different seeds and having all other settings the same, one can ensure that the results are not influenced by the instantaneous values (for example, user arrival into the network, location of arrival of new users, and assignment of path gain based on the propagation model calculations).
In all the subplots of the Figure 5.2, the 95th percentile user throughput will
always be the same and also for all downtilt angles. This is due to the type of traffic that is being used in the simulations and the quantization effect due to the allocation of number of bits to per resource element depending on the observed
Figure 5.1. Investigated hotspot locations for the urban network scenario.
channel quality.
As the 5th percentile user throughput is used as the network performance
measuring criterion, it is separately plotted in the Figure 5.2. Here, the blue colored curves indicate the value of 5thpercentile user throughput for each of the
simulation seeds and the red curve indicates the average value of 5thpercentile user
throughput. This red colored curve can be taken as an input for the self-planning scenarios. The green colored horizontal line indicates the sensitivity region for the given hotspot scenario. It is the five percent below the highest 5thpercentile user throughput value. The individual values for each of the blue curves can be found in the Appendix A.
In order to cross verify the results, the spectral efficiency of the cells that are a part of the region of observability is shown in the Figure 5.4.
Some of the observations from the above results are mentioned in the following subsections.
5.1
Input towards Self Planning Scenario
In the self-planning case, it is a safe bet to initialize the default downtilt angle for a new base station to 100 in an urban scenario that has the network parameters as mentioned in Table 3.1 and Table 3.2. The optimal downtilt value is always in the range 90 to 130 (including the simulations with different seeds as mentioned in Appendix A) with 100 being within the 95thpercentile value of the highest 5th
percentile user throughput.
5.1 Input towards Self Planning Scenario 29 0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7
User throughput in bits/s
0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7 0 10 20 0 2 4x 10 7
Down tilt angle in degrees
5th percentile 50th percentile 95th percentile
Figure 5.2. User throughput as a function of downtilt angle for different hotspot
sce-narios (the results are shown in the same order as the Figure 5.1 with respect to the locations of hotspots).
Figure 5.3. Only 5th percentile user throughputs as a function of downtilt angle.
and the optimal load balancing will happen at different downtilt angles. In a highly loaded network, it is important not to further overload any of the cells.
In the Figure 5.5, the 5thpercentile of the SINR is shown as a function of the
downtilt of the center cell. In the Figure 5.6, the load distribution in terms of number of users served per sampling time by each cell is shown as a function of the downtilt of the center cell. In the Figure 5.7, the corresponding 5thpercentile
0 10 20 2 2.5 0 10 20 1.5 2 2.5 0 10 20 1.5 2 2.5 0 10 20 2 2.2 2.4 0 10 20 2 2.2 2.4 0 10 20 2 2.2 2.4 0 10 20 2 2.2 2.4 0 10 20 2 2.2 2.4 0 10 20 2 2.2 2.4
Down tilt angle in degrees
Cell spectral efficiency
Figure 5.4. Spectral efficiency of cells for different hotspot scenarios.
be seen, not all the time we observe the optimal SINR to be at 100. During some of the cases, when the optimal downtilt is less than 100 (80 for the 8th subplot),
the corresponding load distribution shows that the center cell will be overloaded. This will pull back the user throughput as users will start to spend more time in the queues. Therefore, the optimal 5th percentile user throughput will result in between the optimal downtilt for the SINR values and the one that corresponds to the optimal load distribution in the network.
5.2
Impact of the Location of Hotspot on the
Op-timality of Downtilt Angle
The location of hotspot will have a very small impact on the optimal downtilt value as smaller downtilt angles (90 to110) are suitable for hotspots at the cell edge and larger downtilt angles (100to 130) are suitable for hotspots that are close to base station.
5.3
Sensitivity Analysis
The sensitivity of downtilt value is highly dependent on the particular scenario of path gain values that the hotspot users are experiencing. From the results of Figure 5.3 (and also from the Appendix A), it is clear that the sensitivity level differs from one scenario to the other (it is not just the location of the hotspot that matters but the path loss distribution to the center cell and the highest interferes for the hotspot users). The observations with respect to sensitivity of antenna downtilt are:
5.3 Sensitivity Analysis 31 0 10 20 0 5 0 10 20 0 5 0 10 20 0 5 0 10 20 0 5 5th percentile SINR in dB 0 10 20 0 5 0 10 20 0 5 0 10 20 0 5 0 10 20 0 5 0 10 20 0 5
Downtilt angle in degrees
Figure 5.5. 5th
percentile SINR values in dB as a function of the downtilt of the centre cell.
Figure 5.6. Load distribution in terms of number of users served per sampling duration
0 10 20 0 5x 10 6 0 10 20 0 5x 10 6 0 10 20 0 5x 10 6 0 10 20 0 5x 10 6
5th percentile user throughput in bits/s
0 10 20 0 5x 10 6 0 10 20 0 5x 10 6 0 10 20 0 5x 10 6 0 10 20 0 5x 10 6
Downtilt angle in degrees 0 10 20
0
5x 10
6
Figure 5.7. 5th percentile user throughput as a function of the downtilt of the centre
cell.
1. When the hotspot is located very close to the base station (within 100 m), the downtilt is not very sensitive to higher downtilt angles.
2. If the hotspot users have multiple base stations as large interferers then after the handover of hotspot users (at a particular downtilt) to one of these base stations, the overall 5th percentile user throughput will degrade drastically.
The performance degradation will continue to happen even when the downtilt is large enough to have very small interference to the hotspot users. The degradation of 5th percentile user throughput is due to the fact that the new cell has a different base station as its main interferer and changing the downtilt value of the center cell will not have much impact on these users SINR value. Such a result can be better explained using Figure 5.8, Figure 5.9 and Figure 5.10.
Figure 5.8 illustrates the network situation when the hotspot users (shown in a light brown square) are connected to the center cell (orange color). These users are experiencing a large interference from the neighboring cells, 21st
cell (green color) and 17thcell (blue color). When the downtilt of the center
cell is increased, the hotspot users will get handed over to 21st cell. Figure 5.9 shows the 5thpercentile of the user throughput of each cell as a function of the downtilt angle of the center cell. As can be seen in the figure, from 100 downtilt onwards, the 5thpercentile user throughput of the 21stcell will get degraded because of the handover of the hotspot users. There is also some degradation in the 5th percentile of user throughput of 17th cell. Also, the number of users served by each cell during the sampling duration is given in Figure . It clearly shows that the degradation in quality is due to the large additional load that the 21st cell is serving. The cell acting as its largest
interferer is 17thcell.
5.3 Sensitivity Analysis 33
Figure 5.8. Network deployment showing the interferers to the hotspot users. Orange
colored cell is the center cell. Hotspot users will get handed over to green cell and then blue cell will act as the largest interferer.
Figure 5.9. 5thpercentile user throughput of selected cells.
cells, it is important to see the loading condition in the neighboring cell. By handing over a large set of users to a loaded cell, the probability of
Figure 5.10. Number of users served by individual cells as a function of downtilt of the
center cell.
degradation in the network performance increases.
5.4
Comparison with Okumura-Hata Propagation
Model
A study with a different propagation model will help in comparing the impact of the antenna downtilt on the network performance. Therefore, an Okumura-Hata like model is taken as the reference propagation model to see how it impacts the results. This Okumura-Hata like model is based on the free space path loss calculations and log normal fading losses associated with the UE. The total path loss model used here is,
T otal path loss = min(Distance loss + Lognormal f ading loss + k,
F ree space loss) (5.1)
where,
• k - a constant which depends on the network scenario (rural/urban), • Distance loss - is log10(dα), with d representing the distance of UE from the
base station and α representing the path loss exponent, • Lognormal loss - is Sigma ∗ ((√Ra∗ (randn)) + (randn ∗ (
√
1 − Ra))), with Sigma being the standard deviation and Ra is a constant.
5.4 Comparison with Okumura-Hata Propagation Model 35 −500 0 500 −600 −400 −200 0 200 400 600 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Distance in meters Distance in meters −6 −4 −2 0 2 4 6 8 10
Figure 5.11. Spatial SINR distribution (in dB scale) for Okumura-Hata propagation
model.
• F ree space loss - is −20log10 4πf c − 10log10 d2+ (h BS− hU E) 2 , with f being the operating frequency, c being the speed of light in vacuum, hBS
and hU E being the height of base station and UE respectively.
From the equation, it is clear that two UE’s located next to one another are more probable to have similar path gain values towards a particular base station.
5.4.1
Comparison with Urban Propagation Model Results
A plot of SINR values as experienced by the bin positions is given in the Figure 5.11 and the Figure 5.12 for the Okumura-Hata model and Urban propagation model respectively. The downtilt angle of the center cell in both the cases is 100. For the Okumura-Hata model, as shown in 5.11, there is always a smooth variation of SINR in the spatial domain. For the Urban propagation model, as shown in 5.12, there are some clear boundaries, especially visible in the hotspot region, where the smoothness of variation of SINR in the spatial domain is impacted. It is clear from this plot that the SINR distribution is different from Okumura-Hata model and therefore previous study results cannot be directly compared with the current results.
With all the parameters set similar to the values given in Table 3.1 and Table 3.2, and only changing the propagation model to Okumura-Hata, the downtilt was swept from 00 to 200 in steps of 10 and the changes in the 5th percentile user
throughput is plotted in the Figure 5.13.
In case of Okumura-Hata model, diffraction is not taken into account. There-fore, in an ideal scenario, the optimal downtilt angle in the Urban propagation
−500 0 500 −600 −400 −200 0 200 400 600 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Distance in meters Distance in meters −6 −4 −2 0 2 4 6 8 10
Figure 5.12. Spatial SINR distribution (in dB scale) for Urban propagation model.
0 10 20 5 6 7 8x 10 6 5 10 15 20 5 6 7 8x 10 6 5 10 15 20 5 6 7 8x 10 6 0 10 20 5 6 7 8x 10 6 0 10 20 5 6 7 8x 10 6 0 10 20 5 6 7 8x 10 6 0 10 20 5 6 7 8x 10 6 0 10 20 5 6 7 8x 10 6 0 10 20 5 6 7 8x 10 6
Down tilt angle in degrees
5th percentile user throughput
Figure 5.13. 5th
percentile user throughput while sweeping the downtilt from 00to 200
5.4 Comparison with Okumura-Hata Propagation Model 37
model would have been smaller as the direct ray will have a smaller optimal angle in comparison to the Okumura-Hata model. Although smaller downtilt angle pro-vides a better path gain to the UE’s located close to the central base station, it also introduces a large interference in the network and also the center cell will be covering a large region in the network. Therefore, smaller downtilt angles will re-sult in a worse performance. As this explanation is true for every cell, the overall performance will be lesser in Urban propagation model compared to Okumura-Hata model. This can be seen in the results as the optimal downtilt angle for Okumura-Hata model is in the range 6 to 8 Mbits/s. Whereas, similar value in Urban propagation model will be in the range 3 to 4 Mbits/s.
Chapter 6
Impact of Elevation
Beamwidth on Optimal
Downtilt
As per the previously carried out studies by Athley et.al. [5] and Yilmaz et.al. [4], elevation beamwidth plays an important role in the resulting optimal downtilt value in an urban scenario. Therefore, it is relevant to see how a realistic prop-agation model impacts the previously observed results. Ideally, a large value of elevation beamwidth, will introduce higher interference to the neighboring cells. A smaller elevation beamwidth is suitable for pointing towards a particular region like hotspot. In this chapter, all the simulations are carried out as per the network settings in Table 3.1 and Table 3.2.
6.1
Lower Elevation Beamwidth - 6.4
0The default elevation HPBW of all the cells is changed from 80to 6.40. This will have an impact on the coverage area of the cell. When the elevation HPBW is decreased, the beam will get narrower and the region where the vertical antenna radiation pattern (as shown in the Figure 3.6) gets the flooring value will be closer to the main beam direction.
With all the settings similar to the ones mentioned in Table 3.1 and Table 3.2 (except for elevation HPBW which is set to 6.40), the downtilt of the center cell is swept from 00 to 200 in steps of 10. It should be observed that the antenna gain is increased from 18 dBi to 18.6 dBi to have the same input power. The results are plotted in the Figure 6.1. The individual results for each seed are given in the Appendix B.
As can be seen in the Figure 6.1, all the figures seem to be a shifted version of the Figure 5.3. This is in sync with the observations from the study of Athley et al. [5]. In that study, a relation between elevation HPBW and the antenna downtilt was derived and it was observed that a smaller elevation HPBW will
