Outdoor Small Cell Deployment with Complementary Spectrum Authorizations, Licensed (LSA) and Unlicensed (LAA): Techno-Economic Analysis

(1)

IN

DEGREE PROJECT ELECTRICAL ENGINEERING, SECOND CYCLE, 30 CREDITS

STOCKHOLM SWEDEN 2017,

Outdoor Small Cell Deployment with Complementary Spectrum Authorizations, Licensed (LSA) and Unlicensed (LAA)

Techno-Economic Analysis FIKRI FIRMAN

KTH ROYAL INSTITUTE OF TECHNOLOGY

SCHOOL OF INFORMATION AND COMMUNICATION TECHNOLOGY

(2)

(3)

Abstract (English)

The signicant increase in mobile data trac has put a considerable load on the wireless mobile networks. In the current highly competitive market, Mobile Network Operators (MNOs) have to strive to provide additional capacity of their network, by also considering the cost factor to make their business sustainable. Along with advances in spectrum-ecient technologies, small cells deployment have provided cost-ecient methods to provide additional capacity for indoor and outdoor subscribers.

The gain of better spectrum utilization and opportunistic spectrum access have motivated the deployment of wireless networks utilizing below 6 GHz spectrum, where there are opportunities for mobile networks to access the spectrum by co-existing with incumbent users and technologies. Two emerging complementary spectrum authorizations have attracted industry and academia, Licensed Shared Access (LSA) and License Assisted Access (LAA).

In this thesis, the techno-economic aspects of operating under individual authorization (LSA) and general authorization (LAA) regimes are investigated and compared. The dynamics of operating under unlicensed spectrum are represented considering the scenario of two MNOs co-existing following the regulatory requirements.

The results show that choosing the appropriate channel selection mechanism is of high importance when operating under the unlicensed regime (LAA). The results indicate that LAA can be an alternative for cost-ecient deployment method in some scenarios, for example when there is a low or moderate availability of LSA bandwidth. For the future work, we suggest an optimized user association to the small cells to provide a better load- balancing mechanism.

i

(4)

Abstract (Swedish)

Den avsevärda ökningen av den mobila datatraken har skapat stor be- lastning på de trådlösa mobilnäten. I den nuvarande mycket konkur- ren- sutsatta marknaden, måste mobiloperatörerna (MNO) sträva efter att skapa ytterligare kapacitet i deras nätverk, samtidigt som de måste tänka på kost- nadsfaktorer för att göra sin verksamhet hållbar. Tillsammans med fram- steg inom spektrumeektiv teknik och driftsättning av små basstationer, har man fått fram kostnadseektiva metoder för att öka kapaciteten för inom- och utomhusanvändare.

Fördelen av bättre spektrumanvändning för frekvenser under 6 GHz och opportunistiska tillgång av spektrum, har motiverat utbyggnaden av trådlösa nätverk. Detta möjliggör för mobila nätverk att använda spektrumet genom att samexistera med etablerade användare och tekniker. Två nya komplet- terande spektrumtillstånd har lockat industrin och den akademiska världen, Licensed Shared Access (LSA) och License Assisted Access (LAA).

I denna avhandling, har de tekno-ekonomiska aspekterna av LSA och LAA regimer undersökts och jämförts. Dynamiken av drift i olicensierat spektrum representeras i scenariot av två mobilnätsoperatörer samexisterar och följer lagkraven.

Resultaten indikerar att valet av lämplig mekanism t.ex. val av rätt kanal är av stor betydelse vid användning av olicensierad regim (LAA). Resultaten tyder på att LAA kan vara ett alternativ för kostnadseektiv distributions- metod i vissa scenarier, till exempel när det nns en låg eller måttlig till- gång på LSA bandbredd. För det framtida arbetet, föreslår vi en optimerad användarassociation till de små cellerna för att ge en bättre lastbalansering mekanism.

ii

(5)

List of Figures

1.1 Mobile Data Trac Forecast [2] . . . 2 2.1 LTE 3GPP Releases, adapted from [10] . . . 6 2.2 Spectrum authorization schemes for wireless networks, adapted

from [20] and [21] . . . 10 3.1 Base Stations deployment realization under hexagonal grid

and point process . . . 15 3.2 System model and the workow of the thesis . . . 24 4.1 Network dimensioning for LAA system for low activity factor 26 4.2 SINR for dierent channel selection mechanisms, 25.2 GB

monthly usage, λB = 19, low activity factor . . . 26 4.3 Downlink throughput for dierent channel selection mecha-

nism, 25.2 GB monthly usage, λB = 19, low activity factor . . 27 4.4 PDF of SINR for random channel selection, 25.2 GB monthly

usage, λB = 19, low activity factor . . . 28 4.5 PDF of SINR for deterministic channel selection, 25.2 GB

monthly usage, λB = 19, low activity factor . . . 28 4.6 PDF of downlink throughput for random channel selection,

25.2 GB monthly usage, λB = 19, low activity factor . . . 29 4.7 PDF of downlink throughput for deterministic channel selec-

tion, 25.2 GB monthly usage, λB = 19, low activity factor . . 29 4.8 SINR for dierent channel selection mechanisms, 210.9 GB

monthly usage, λB = 53 . . . 30 4.9 Downlink throughput for dierent channel selection mecha-

nisms, 210.9 GB monthly usage, λB = 53 . . . 30 4.10 Network dimensioning for LAA system with high activity factor 31 4.11 Network dimensioning for LSA system with low activity factor 32 4.12 Network dimensioning for LSA system with high activity factor 32 4.13 Incremental TCO of network deployments with dierent cost

assumptions . . . 33 4.14 Normalized NPV for dierent deployment type and cost factor 34

iii

(6)

List of Tables

2.1 Spectrum holding of an MNO in Sweden [11] . . . 7 2.2 LSA incumbent actors and uses and bandwidth availability in

several European countries [23] . . . 11 3.1 Mobile data usage in Swedish market [38] . . . 19 3.2 Average population densities for dierent deployment types [40] 20 3.3 System Parameters . . . 24 4.1 Cost value and normalized cost . . . 33 4.2 Description of deployment types in gure 4.14 . . . 35

iv

(7)

Introduction

1.1 Background

The success of GSM (Global System for Mobile Communications) has paved the way for developing the next generations of mobile cellular networks. 3GPP (The 3rd Generation Partnership Project) was established with the goal of developing the standards and specications related to cellular telecommunications technologies [1]. Each release introduces new features and improved functionality by oering higher data rates, better Quality of Service (QoS) and cost-ecient techniques.

Mobile data trac has increased signicantly and will continue growing for the next few years. Ericsson in its annual mobility report projected that the mobile data trac by 2021 would increase 10-fold with the expected 45%

compound annual growth rate (CAGR), as illustrated in Figure 1.1 [2].

Ericsson stated that higher adoption of smartphones and tablets, increase in mobile broadband subscription and higher monthly mobile data usage are the main factors leading to the growth in global mobile data trac. Cisco in its Visual Networking Index (VNI) projected that on 2020 the mobile data trac would increase eightfold compared to 2015 with 53% CAGR [3].

Ericsson cited that most mobile video trac came from mobile access to video sharing platform such as Youtube, and also noted the increase of video content in online application of news, advertisement, and social media.

Internet of Things (IoT) has been taking o for the last couple of years, notably by the introduction of wearable devices and any objects equipped with sensors and software that are interconnected and able to collect and send data. Cisco predicted that by the year 2020, there would be 3.1 billion of wearable devices and Machine-to-Machine (M2M) communications that rely on mobile networks for their connectivity.

The increasing mobile data trac and also new types of communications add signicant load to the mobile network. Mobile operators need to solve the capacity issue either by increasing the capacity of the mobile network or

1

(10)

Chapter 1. Introduction 2

Figure 1.1 Mobile Data Trac Forecast [2]

by trac-ooading. Increasing the capacity of mobile networks often implies increasing capital and operational expenditure. To make their business more sustainable, mobile operators need to nd cost-eective means in handling increasing mobile data trac. The importance of nding such solutions is emphasized by the trend of wireless service revenue. Ericsson reported that revenue growth from mobile services has been slowing down [4]. From 2010 to 2014, revenue for mobile services had been increasing with CAGR of 2.7%. In 2014, the growth of mobile service revenue was only 1.7%, which is a steep decline compared to annual growth of 10% to 12% ten years ago.

Nevertheless, there is a trend of a healthy growth of mobile data revenue with CAGR of 34%. These factors further state the importance of increasing network capacity in the most economical way.

Increasing mobile networks capacity can be accomplished by several methods such as network densication, implementing more spectrum-ecient techniques and by adding more radio spectrum [5, 6]. Each of the methods has its tradeo. Implementing more macro-cell base stations might raise concern with high costs related to equipment, installations, and site acquisitions. Heterogeneous networks emerge as a cost-eective solution in providing additional coverage and capacity by implementing low-power base stations. Adding more spectrum is another straightforward technique to increase mobile network capacity. This method is subject to the spectrum availability, spectrum cost, and regulatory policies. Each country has a regulatory authority that decides how the spectrum divided and allocated to dierent applications and services. The Regulator also determines autho-

(11)

rization schemes, who can utilize particular part of the spectrum within a dened area. Conventionally, spectrum allocation and authorization are implemented in a static manner, and there is a presumption that certain parts of the spectrum are underutilized wherein dierent areas of the spectrum (wireless cellular network band) experiencing congestion [7].

Two complementary spectrum authorizations have been proposed for the mobile network. They are Licensed Shared Access (LSA) and Licensed- Assisted Access (LAA). Licensed Shared Access (LSA) enables primary spectrum licensee to grant part of the spectrum to be used by secondary spectrum licensee with regards to several conditions. Licensed-Assisted Access (LAA) enables utilization of unlicensed spectrum band. The main unlicensed band considered for LAA is the 5 GHz ISM band where the dominant legacy application is WiFi technology.

1.2 Problem

The introduction of complementary spectrum authorization opens opportunities for mobile operators to utilize parts of the spectrum that belong to other non-telecommunication actors. This thesis explores the impacts of

exible spectrum authorization on the strategy of mobile operators in providing capacity. The answer to this question is investigated in this thesis:

Under which condition LSA approach is more cost-ecient than LAA approach for outdoor small cell deployment scenario?

1.3 Purpose

This thesis presents techno-economic study related to the impact of exible spectrum authorization with the strategy of mobile operators. The technical part illustrates gains regarding network capacity and perceived user experience. The economic part will discuss the cost analysis.

1.4 Goal

The goal of this thesis is the techno-economic discussion of the implementation of dierent spectrum authorizations in wireless cellular networks, namely licensed-operation (LSA) and unlicensed-operation (LAA). The goal is divide into following sub-goals:

1. Investigate the network dimensioning under dierent channel selection mechanisms when operating in unlicensed spectrum.

2. Investigate the network dimensioning operating under LSA approach, with dierent bandwidth availability.

(12)

3. Analyze the capacity and cost aspects for LAA and LSA approach.

The result of this thesis is the strategy that can be adopted by mobile operators with regards to the opportunities provided by exible spectrum authorization.

1.4.1 Benets, Ethics and Sustainability

This results of this thesis can be benecial as insight for mobile operators in planning for their networks. The operators can decide which method is more cost-ecient based on the spectrum availability and regulatory frameworks in their respective countries. LSA and LAA enable mobile operators to access the part of the spectrum exibly and opportunistically.

LSA and LAA incorporate the spectrum sharing mechanisms where both approaches have to be able to co-exist with the incumbent users or the legacy applications. The introduction of LTE on the unlicensed spectrum raises the ethical issue where LTE-LAA will aect the performance of the WiFi users.

It is of great importance that LTE-LAA implements co-existence mechanisms that enable fair access to the unlicensed spectrum, between devices of the same or dierent technologies. With adequate mechanisms, LTE-LAA Base Stations (BSs) are acting like the good neighbors to WiFi Access Points (APs) [8].

1.5 Methodology

This thesis adopts Quantitative and experimental research method, with deductive as the chosen research approach [9]. Experimental research method inspects causes and outcomes. This approach investigates variables and builds connections between them [9]. One of the task within the project is to build a system-level simulator. Data will be collected from simulations and analyzed to reach the conclusions.

1.6 Delimitations

This thesis investigates the implementation of exible spectrum allocation in the wireless cellular network. The system model designed implements the necessary mechanism to ensure co-existence with the legacy systems operating on the unlicensed band (WiFi). The performance of such systems is not within the scope of this thesis. This thesis focuses on outdoor environment, where outdoor subscribers are connected to outdoor small cells (Base Stations). Specically, this thesis examines the scenario with two mobile operators co-existing within an area.

(13)

1.7 Outline

Chapter 2 provides the theoretical background and related works. Chap- ter 3 explains the methodology of the project. Chapter 4 presents the results and analysis. Chapter 5 provides the conclusions of the thesis.

(14)

Chapter 2

Background and Related Work

This chapter provides the background on Wireless Cellular Network, its current and the envisioned future architecture, mobile trac consumption trend, and the emerging complementary spectrum authorization schemes.

2.1 Wireless Cellular Networks

Long Term Evolution (LTE) is 4G wireless standards, successor to 2G and 3G standards. The 3rd Generation Partnership Project (3GPP) is a Standard Developing Organization (SDO) with the primary task of developing future mobile communication technologies. 3GPP released the rst version of LTE (Release 8) on 2008. Some of its highlights are new radio in- terface based on Orthogonal Frequency Division Multiple Access (OFDMA) and all-IP network. The 3GPP standards up to 3GPP Release 13 along with their main features are shown at gure 2.1.

Figure 2.1 LTE 3GPP Releases, adapted from [10]

Spectrum is one key ingredient for increasing the capacity of the wireless mobile network. The frequency and the bandwidth of the spectrum being made available by the National Regulatory Authority (NRA) at a time are subject to the availability and the policy of each country. The common

6

(15)

Chapter 2. Background and Related Work 7

practice of spectrum auctions being used as the method to assign spectrum plays a part in the fragmentation of spectrum, especially for wireless mobile networks. This practice leads to spectrum fragmentation, where one operator has the license of spectrum with dierent sizes in dierent bands. As an example, Table 2.1 represents spectrum holding of a wireless operator in Sweden.

Table 2.1 Spectrum holding of an MNO in Sweden [11]

Spectrum Bandwidth (MHz) 800 MHz 2Ö10

900 MHz 2Ö6 + 2Ö5

1800 MHz 2Ö5 + 2Ö20 + 2Ö10 2100 MHz 2Ö19.8 + 1Ö5 2600 MHz 2Ö40

LTE Release 10 introduces Carrier Aggregation (CA) along with LTE- Advanced. CA enables operators to fully utilize chunks of spectrum that they have. LTE-Advanced enables carrier aggregation up to ve component carrier, where one component carrier can have 1.4, 3, 5, 10, 15 or 20 MHz of bandwidth, which allows aggregated maximum bandwidth of 100 MHz [22].

Carrier Aggregation can be implemented on contiguous or non-contiguous component carriers. For example given in Table 2.1, the MNO could improve the performance of the 900 MHz system by implementing CA with other parts of the spectrum that it holds. This way, CA enables better spectrum resource utilization and improving the data rate.

LTE employs Orthogonal Frequency Division Multiple Access (OFDMA) for downlink transmission [12]. OFDMA is multiple access scheme that utilizes Orthogonal Frequency Division Multiplexing (OFDM) modulation technique. OFDM divides the bandwidth into multiple orthogonal sub-carriers.

Information-carrying sub-carriers will be sent simultaneously from the transmitter. Subsets of sub-carriers can be allocated to multiple users.

Previous works study the eect of Carrier Aggregation (CA) to the performance of cellular networks as well as on the cost and revenue sides [13].

Previous works investigate the aspects of deploying 4G LTE together with dierent spectrum authorization schemes, where an MNO can deploy LTE on the licensed band and also on license-exempt bands.

2.2 Methods to Increase Wireless Network Capac- ity

The increasing mobile data trac has lead to many works related to investigating methods to increase the capacity of wireless cellular networks,

(16)

with cost-ecient ways. The growing data trac leads to the goal of providing 1000 more capacity for the future wireless mobile networks compared to the existing wireless cellular networks [5]. In the Introduction, three methods are specied as the means to increase the capacity of wireless mobile networks. This section will provide the current trends of each method, and how they aect the system model that is used in the thesis.

Spectrum Ecient Technology

Spectrum eciency can be described as how big the data rate (bits/second) can be successfully sent per unit spectrum (Hz). Increasing spectrum e- ciency means that more data can be sent for the same amount of spectrum.

Examples that fall into this category are related to the implementing multi- antenna on the transmitter and receivers [10]. One advantage that can be achieved from the multi-antenna technique is the spatial diversity, where the same information is transmitted through multiple antennas. Spatial diversity will provide better protection towards the radio channel and leads to better SINR. Multi-antenna technique called beamforming enables the shaping of the transmitted and received beam for the signal. Beamform- ing will increase received signal gain for the intended user while reducing interference to the other users. The third advantage of the availability of multi-antenna on the transmitters and the receivers is what is called by spatial multiplexing. Under certain condition, spatial multiplexing can increase the channel capacity by a factor of min(NT,NR), where NT and NR are the available numbers of antennas on the receivers and the transmitters. The multi-antenna technique previously described are often covered as Multiple Input Multiple Output (MIMO) concepts.

Network Densication and Heterogeneous Networks

Network densication is a method to increase network capacity by adding more base stations within an area. Network densication has been considered as one key solution to achieve the target of 1000 more capacity for the future wireless networks [5]. Adding more base stations will have two-fold eects towards the network capacity. First, spatial densication will increase the capacity per area by reducing the number of users being served by one base station, allowing the users to receive better performance. Second, spatial densication will bring the users closer to the base stations, and that usually mean better signal strength and better user throughput. Based on their coverage and transmitted power, base stations can be divided into dif- ferent types. A Macrocell is a category of high-powered base stations that able to cover a wide area. It is usually installed on a mast on greeneld sites or at the rooftop of a building. For areas with high trac density (of-

ce parks, downtown area), adding more macrocell as mean to add capacity

(17)

might face some obstacles. The rst obstacle is related to site acquisition, where it might be dicult to nd a space to lease that is appropriate for macrocell deployment [14]. The second is cost related, due to the high- cost deployment of a macrocell, such as the expenses related to civil works.

Heterogeneous Network is a concept of deploying low-powered base stations within the coverage of macrocells. The low-powered base stations, sometimes also called as small cells are installed whether to provide more capacity or to provide better coverage. Small cells can be further divided, usually based on their deployment method and range of coverage. Femtocell or Home eNodeB (HeNB) refers to a small cell that is deployed to provide indoor coverage and capacity. It requires IP connectivity to provide the connectivity to the operator's network. It has low complexity installation eort, making it possible for deployment by users. Picocell and microcell resemble small cells with higher capability and capacity than femtocells, and they are deployed by the operators. Picocell and microcell can be implemented to cover public areas located outdoor and indoor where microcell has bigger transmission power than picocell.

Several factors motivate and promote small cell deployment. It has been mentioned earlier that each 3GPP standard release brings new features and enhancements, some have improved the performance and simplied the deployment of small cells. Features such as Inter-cell Interference Coordination (ICIC) and enhanced-ICIC (eICIC) can alleviate the interference experienced by cell-edge users of the small cells, due to the co-channel operation of the macro-cells and the small cells. The other factor that drives the deployment of small cells is its low deployment cost compared to the deployment of macrocells [15].

Access to More Spectrum

The third method to increase the capacity is by utilizing more radio spectrum. The spectrum range between 300 MHz and 3 GHz is considered as the

`sweet spot' for wireless networks, due to the propagation characteristics and the wavelength that enables the antenna to be conveniently designed to t in user equipment [16] [17]. The rst two mentioned methods heavily depend on the strategy of each MNO, but the last method is more complicated to implement, mainly because of lack of available spectrum in these sweet spot range. Obtaining additional bandwidth at similar frequencies might involve moving the incumbent to new operating frequency (refarming), something that is usually considered as costly and time-consuming eort. There are extensive discussions and studies in making better use of spectrum resources that bring new paradigm in spectrum allocation and authorization schemes.

The method and its relation to the theme of the thesis will be explained in the following section.

(18)

2.3 Flexible Spectrum Authorization Scheme

The purpose of provisioning the increasing mobile data trac has led to new schemes related to spectrum policies and implementations of new technical advances implementation on the mobile networks. This section will present earlier works on the subject of the exible use of non-telecommunication actors spectrum and the technology that can pave the way of utilizing them on the current and future radio access network. Each country has a regulatory authority that decides how the spectrum divided and allocated to dierent applications and services. The regulator also determines authorization schemes, who can utilize particular part of the spectrum within a dened area. Conventionally, spectrum allocation and authorization are implemented in a static manner, and there is a presumption that certain areas of the spectrum are experiencing underutilization wherein dierent parts of the spectrum (wireless cellular network) are experiencing congestion [18]. There are two main methods in spectrum allocation and authorization, namely individual authorization and general authorization [19]. In individual authorization, the rights to utilize a part of a spectrum is assigned to one or more actors. In general authorization scheme, the right to use part of the spectrum is granted without any fee, with the requirements that each technology has to co-exist and accessing the channel conforming with the sets of rules. Some examples of the general authorization scheme are the operation on the TV White Space and the ISM Band, including the operation of wireless networks on the 5 GHz unlicensed spectrum (LAA).

Figure 2.2 Spectrum authorization schemes for wireless networks, adapted from [20] and [21]

Individual authorization can be divided further into authorized primary use and authorized secondary use. In authorized primary use, the licensee

(19)

Table 2.2 LSA incumbent actors and uses and bandwidth availability in several European countries [23]

Country Main incumbent appli-

cations Spectrum availability for wireless networks

Finland Wireless cameras,

PMSE 90% time availability for 85 MHz;

15 MHz available within dened France Aeronautical telemetry, area

PMSE 80 MHz for 80% of population

Ireland Aeronautical telemetry,

PMSE 100 MHz for all population

Italy Aeronautical telemetry,

PMSE 85 MHz available nationwide,

15 MHz available within dened Sweden Aeronautical telemetry area

with small utilization 100MHz in all populated area, not applicable in unoccupied area UK Defence applications 40 MHz exclusively available, with additional 20 MHz considered for future sharing

will be able to use the part of the spectrum exclusively with protection from harmful interference from other users. Authorized secondary use scheme enables part of the spectrum that has been licensed to a certain licensee, to be used by other actor or entity, with certain technical requirements to protect the incumbent licensee. This scheme is also known as Licensed Shared Access / Authorized Shared Access (LSA/ASA). LSA is complementary spectrum scheme, where QoS guarantee can be achieved as if in using licensed spectrum. There is a considerable interest within the wireless communication community in using 2.3 GHz band for wireless broadband networks. Incumbent users of 2.3 GHz band within European Conference of Postal and Telecommunications Administrations (CEPT) countries consist of Programme-making and special events (PMSE) applications, telemetry and other governmental use [22]. Authors in [22] conduct feasibility study LSA concept by measuring interference level between LTE and wireless camera in the 2.3 GHz band.

Authors in [24] studied and conducted experimental trials in several locations in London, where availability and capacity that achievable in TV White Space were investigated. The trials were initiated by UK Telecommu- nications Regulator, Ofcom. To increase the spectrum allocated for wireless and mobile services, Ofcom devised plans and strategy for future wireless spectrum. The envisioned strategy was to make almost 80% of new spectrum below 6 GHz as shared spectrum and will be accessed either via LSA or

(20)

opportunistic spectrum access scheme [25]. Capacity, Machine-to-machine communications (M2M), coverage and wireless backhaul were identied as primary drivers of UK future wireless spectrum allocation strategy.

2.4 Related Work

This section explains previous works that are relevant to the topic of the thesis. They are found during the literature review phase of the thesis.

Authors in [26] propose blank sub-frames as co-existence mechanism between LTE-LAA and WiFi system. The studied scenario is for indoor deployment of WiFi, LTE-LAA base stations, and their corresponding users. The proposed mechanisms is a modied version of Almost Blank Sub-Frames (ABS) that was initially designed to overcome interference for co-channel HetNets. The blank sub-frames allow WiFi Access Points to access the channel. The main result shows there is a tradeo between the number of blank sub-frames, with the performance of LTE-LAA. The main issue of this work is that it does not introduce LBT that is required in most regions as the co-existence mechanism in the unlicensed spectrum.

Authors in [8] investigate dierent co-existence mechanisms between LTE- LAA and WiFi systems namely Duty Cycling and Carrier-Sensing Adaptive Transmission (CSAT). The scenario studied is for Heterogeneous Networks where picocells co-exist with WiFi access points. CSAT works by dening a Time-Division Multiplexing (TDM) cycle, where LTE-LAA base stations either taking on or o state. LTE-LAA base stations occupy the channel during on state, while during o state, other technologies can occupy the channel. The authors also show that, by applying co-existence mechanism, LTE-LAA is a good neighbor for WiFi system.

Authors in [27] investigate the performance of two operators deploying their networks under LAA using stochastic geometry. The authors show that by adopting adaptive carrier selection, there is an improvement compared to random channel selection. Authors in [28] investigate the gain of implementing frequency reuse and LBT for a single channel. They show that LBT can increase the performance of LAA deployment, especially in high load situation.

Authors in [29] investigate dierent scenarios for the co-existence between LTE-LAA and WiFi system, indoor and outdoor scenarios are investigated.

The authors adopt LBT as the co-existence scenario and examine multiple- channel utilization and selection methods. They also study the scenario where seven operators deploy outdoor LTE-LAA small cells within an area.

But, for the outdoor multi-operator scenario, the authors do not investigate the scenario where deterministic channel selection is used. It is also deserving to mention that the authors use real base-stations instead of random small

(21)

cells generated by Point Process. The performance reported is for one user associated with each LTE-LAA small cells.

The authors in [30] investigate the performance of 'WiFi-like' system operating on TV White Space (TVWS). TVWS refers to the unused part of the spectrum initially allocated for TV broadcast. The locations of the access points are generated according to Point Process. The paper and [29] provide the model for SINR and throughput for LTE-LAA that will be adopted in this thesis.

We identify the existing research gaps that motivate the topic of the thesis. Most of the previous works focus on nding the optimum co-existence mechanism for LTE-LAA and the WiFi system. The following is the rst identied research gap:

No previous work investigates the feasibility of deploying LTE-LAA networks to meet the current and future mobile data demand

The rst contribution of the thesis is the feasibility study of deploying outdoor LTE-LAA small cells network, and the network dimensioning according to the feasible load concept.

The techno-economic discussion related to the opportunity to increase capacity by exploiting additional spectrum through LSA is done in [31]. The authors compare alternatives between implementing LSA and MIMO to nd the cost-ecient method the MNOs can adopt.

Authors in [32] investigate techno-economic aspects of the impact of dierent spectrum pricing and spectrum holding for two distinctive circumstances.

The authors consider India where the spectrum pricing is high and representing high user density, while Sweden is chosen representing condition where spectrum price relatively low and low user density.

The following is the second research gap that motivates the theme of the thesis.

There is lack of work investigating the techno-economic discussion in operating LTE networks on two dierent licensing regimes, namely LTE-LAA as unlicensed regime and LSA as licensed regime

The operation of LAA that being discussed and standardized is based on the complementary and opportunistic mechanism with the licensed spectrum. But, in this thesis, the performance that is investigated is for the unlicensed band as if operating in a 'stand-alone.' The motivation is to compare the performance of an unlicensed operation, with the licensed (LSA) mode of operation. This thesis will specically study the market condition in Europe. This thesis will adopt the LSA bandwidth availability and the regulatory requirements regarding the operation of mobile networks in the unlicensed 5 GHz band for European countries.

(22)

Chapter 3

Methodology

This chapter will explain methodology adopted for the thesis. The rst part of the degree project is to make a quantitative analysis of multi-operator small cells network deployment when selecting LAA and LSA spectrum access method. The primary task for the rst part is to build system-level simulator using MATLAB. Section 3.1 explains about stochastic geometry and how the thesis utilizes it for generating Base Stations and User Equip- ment locations. Section 3.2 describes the channel propagation model that is adopted in this thesis. Section 3.3 provides the information regarding feasible load concept and target throughput. Section 3.4 explains the method to do network dimensioning. Section 3.5 and 3.6 describe the channel selection method and the interference modeling in the unlicensed spectrum respectively. Finally, section 3.7 provides the cost model that will be used in the techno-economic discussion. In this thesis, the performance of LTE operating in unlicensed spectrum (LAA) is investigated under harsh environment, where there are two MNOs with each of them deploying outdoor small cells to serve their outdoor subscribers. The chosen scenario also represent the case where mobile operators are deploying their networks to cover outdoor subscribers in a dense urban area.

3.1 Base Stations and Users Distribution

Stochastic Geometry is a mathematical tool that investigates the prop- erties of random placements of objects within an area. It has been used exhaustively in the eld of wireless networks to provide analysis of some key network components such as capacity regarding user throughput, or connectivity regarding Signal to Noise and Interference Ratio (SINR). Although this thesis does not provide mathematical analysis, this thesis will adopt key aspects of stochastic geometry which is the point process or random point pattern. Point process has been used to model placements of Base Stations (BSs) and mobile users within wireless networks. It provides more realistic

14

(23)

Chapter 3. Methodology 15

(a) Hexagonal Grid (b) Point Process

Figure 3.1 Base Stations deployment realization under hexagonal grid and point process

modeling of Base Stations where Base Station's placements practically are often not ideal like the concept of grid-based hexagonal or square lattice. In heterogeneous networks, dierent base stations (macrocells, microcells, picocells, femtocells, and small cells) dier from each other regarding coverage, inter-site distance, antenna height, etc. In this project, the outdoor small cells are modeled by Poisson Point Process (PPP), which is the most com- monly used Point Process to model base station placement, with minimum inter-distance from each other. Whereas, the mobile users are independently generated and according to homogeneous PPP.

Red circles on gure 3.1 represent the location of the BSs under dierent realizations.

3.2 Channel Model

METIS channel model provides various Propagation Scenarios (PSs) for dierent Test Cases (TCs) that are considered to be the feasible scenarios for the implementation of 5G networks. Extensive measurements developed the channel model in various environments [33].

This thesis implements METIS channel model for Urban Micro Outdoor- to-Outdoor Propagation Scenario for Line of Sight (LoS) scenario. The channel model itself is a modication of Urban Micro (UMi) path loss model of IMT-Advanced [34]. For the Non-Line of Sight (NLoS) scenario, the UMi path loss model based on a hexagonal grid is utilized. The probability of LoS or NLoS condition will be based on UMi LoS Probability. Both channel models support frequency range between 0 and 6 GHz, making both eligible to be utilized in the thesis.

Breakpoint distance is the distance where the path loss shift from one

(24)

regime to the other. It is calculated by the following equation:

d⁰_BP= 0.87 exp

−log₁₀(_{1 GHz}^f^c ) 0.65

4 h⁰_BSh⁰_UE

λ (3.1)

Where fcis the carrier frequency and λ is the corresponding wavelength, h⁰_BSand h⁰_UEare the eective height of the Base Station and the User Equip- ment. In the system model, the actual height of the BS h_BS is 10 m, whereas the actual height of the UE h_UE is 1.5 m. The eective height of the BS and the UE can be calculated as follows:

h⁰_BS= h_BS− h_env, h⁰_UE= h_UE− h_env (3.2) where h_env is the environment height in urban case scenario, which will be assumed to be 1.0 m.

For the LoS scenario, the propagation model can be divided based on the distance between UE and the BS d compared to the breakpoint distance d⁰_BP:

P L_LOS(d)|dB= 10 n1 log₁₀ d

1 m + 28.0 + 20 log₁₀ fc

1 GHz + P L_1|dB (3.3) when 10 m < d < d⁰_BP, and

P L_LOS(d)|dB= 10 n2 log₁₀ d

d⁰_BP + P L_LOS(d⁰_BP)|dB (3.4) when d⁰_BP < d ≤ 500 m, with n1 = 2.2 and n2 = 4.0.

P L_1|dB is path loss oset added to the channel model to improve the consistency of the channel model with the measurement. Path loss oset P L_1|dB is calculated as follow:

P L_1|dB = −1.38 log₁₀ f_c

1 GHz + 3.34 (3.5)

This thesis adopts the formula in [34] for Urban Microcell (UMi) scenario to generate LoS probabilities P_LOS between transmitter and receiver.

P_LOS= min(18/d, 1).(1 − exp(−d/36)) + exp(−d/36) (3.6) Equation 3.6 shows that if the distance between transmitter and the receiver less or equal than 18 m, there will be LoS condition. The greater the distance d, the smaller the probability of LoS will be, which will result in greater path loss. For NLoS condition, we will adopt the path loss model for hexagonal cell layout provided in [34].

P L = 36.7 log₁₀(d) + 22.7 + 26 log₁₀(fc) (3.7)

(25)

3.3 Feasible Load Concept and Target Throughput

The system model implemented on this thesis is assuming an operator deploying small cells within an area to provide capacity and coverage for its subscriber. A load of cell i ρi can be considered as the resource utilization of cell i which is associated with the power needed to activate the Resource Block (RB). A load of a particular cell will be inuenced by the number of user and the service requested by users. For example, video streaming services will result in higher cell load compared to web browsing, assuming the same number of active users within a cell. The load vector of the network ρ = (ρ₁, ρ₂, . . . , ρ_N) constitutes the load of all cells within the network, where N is the number of cells in the network. Link gain gij is the path loss between user i and cell j. For user i that is associated or served by cell k, the corresponding signal to noise plus noise ratio (SINR) for user i is calculated as follows.

γ_i(ρ) = g_ikP_k P

j6=kρ_jg_ijP_j+ σ² (3.8) where ρj is the cell load of the interfering cell j. Pkis the power transmission of the serving cell k and Pj is the transmission power of the interfering cell j.

The additive noise power is represented by σ². Cell load ρj also represents the probability of transmission. In this sense, the co-channel interference experienced by one link is not aected by all co-channel transmissions but is subject to statistical interference. To calculate the downlink user throughput, this thesis assumes modied Shannon formula for Single Input Single Output (SISO) system as follow [35]

b_i(γ_i(ρ)) = W η log₂(1 + β γ_i(ρ)) (3.9) where W is the system bandwidth, η is the system bandwidth eciency and β is the SINR eciency. This thesis assumes SISO system with the system bandwidth eciency η and the SINR eciency β to be 0.57 and 0.8.

Dierent services and applications will require dierent number of resources. User utilization is the number of resource needed to be allocated, for the requested data rate. The following equation shows the utilization of user i connected to cell j, when requesting data rate of Ω [36] [37]

n^j_i = Ω

b_i(γ_i(ρ)) (3.10)

Cell load ρj is the aggregation of users utilization associated with cell j ρ_j = X

i∈φj

n^j_i (3.11)

where i ∈ φj is the set of users, associated with cell j. Authors in [37] introduces the Feasible Load Concept and Feasible Load Problem. The solution

(26)

for the Feasible Load Problem is by nding the load vector that balances utilization of resources in all cells with the corresponding resource demands.

The Feasible Load Concept also states that cell load can take value less or equal to ρmax, the maximum load allowed for each cell.

The primary goal of the technical part of the thesis is to get the minimum number of required base stations, which can meet the required throughput target Ω.

3.4 Network Dimensioning

Network planning is a challenging task faced by the operators in deciding where to deploy base stations and how big the capacity needed for each base stations to achieve the Key Performance Indicator (KPI) targets.

Establishing the right location to deploy base stations ensures the optimal coverage for subscribers. The proper network dimensioning, deciding the optimal resources (transceivers, bandwidth) for each base station, provides the cost-eective method of providing capacity. This process establishes the minimum number of base stations and the minimum resources needed for elements in the network. Proper network dimensioning ensures Mobile Net- work Operators adopt the most economical method in fullling the targeted KPI. In this thesis, the KPI of the network is the 5th percentile of downlink user throughput. The objective is to nd the number of BS required when during busy hours at the maximum fth percentile of users do not achieve the target throughput. For each scenario, the number of BS densities needed for dierent user throughput targets is investigated. The downlink user throughput target is deduced from the mobile data consumption trend and forecast, which will be explained in the following section.

3.4.1 Mobile Data Consumption

Previous works have illustrated the method of dimensioning mobile networks with regards to the mobile data consumption. In this thesis, mobile data forecast and reports on mobile data usage and pattern are used to derive the required user throughput.

This thesis assumes the mobile data consumption in Sweden. The Swedish Post and Telecom Authority (PTS) on May 2016 published the Swedish Telecommunication Market report [38]. The report contains data such as mobile market subscriptions, market share, and mobile data usage for the year of 2015. Private and corporate segments are distinguished in the report for mobile data usage. Table 3.1 shows the mobile data usage for dierent segments.

In the report for mobile data usage sections, PTS divides the mobile data usage into three categories, namely mobile broadband as a standalone service, mobile broadband as additional service and subscription for call

(27)

Table 3.1 Mobile data usage in Swedish market [38]

Subscription

Type (GB) Data usage per month for Private User

Data usage per month for Corpo- rate User

Average data usage per month Mobile broad-

band as a standalone service

11.8 5.9 9.9

Mobile broad-

band as an

additional service

3 1.3 2.5

Subscription for call and data services

0.5 0.5 0.5

and data services. The rst type is the subscriber who uses wireless dongle or router to access mobile data network. The dongle is then attached to laptops or desktops to enable connecting to the Internet. This category is specic for data trac, with no support for voice service. The second category is for type of subscription where at least 1 GB of mobile data credit is purchased. The mobile data credit can be purchased as an add-on or comes with the subscription. This category is the most common subscription for smartphones. The last category is mobile subscription with or without data or voice services. The rst category is not suitable to be considered considering the scope of the thesis. Hence, the monthly mobile data usage for the year 2015 is assumed to be 3 GB, which corresponds to the mobile data usage for individual segments.

Section 1.1 briey explains the trac forecast made by Ericsson and Cisco. This thesis will assume the optimistic scenario regarding mobile data usage. It is assumed there will be an annual growth rate of 53% that corresponds with the forecast made by Cisco [3]. Based on these assumptions, for the year 2016, it is assumed that the mobile data usage in Sweden will be approximately 4.6 GB per month per subscriber.

Energy Aware Radio and Network Technologies (Earth) Project is a con- sortium of academia and industry that studies energy ecient wireless system [39]. Earth Project divides deployment areas by population densities of population per square km, as shown on Table 3.2.

In this thesis, we will assume the scenario of High Dense Urban with 5000 population per square km, with 100% mobile penetration and each subscriber utilizes one User Equipment (UE). In network planning and dimensioning, it is important to study where the trac is originating. The common denominator made by telecommunication vendors and consultants is that 80% of data trac are generated from indoor subscribers [41][42][43].

(28)

Table 3.2Average population densities for dierent deployment types [40]

Deployment Type Average Population Density (citizen/km²)

Dense Urban 3000

Urban 1000

Suburban 500

Rural 100

Sparsely Populated & Wilderness 25

Previous works have assumed the same value in the eort of investigating network dimensioning for dierent scenarios [44][45].

This thesis considers that 20% of the subscriber are located outdoor. The required user data throughput for each year (Ω in Equation 3.10), will be estimated based on the mobile trac forecast. Moreover, two values of user activity will be used. Each of them represents low and high activity factor.

For the low activity factor, each user will utilize its subscription during 6 hours each day. Whereas in high activity scenario, each user will utilize its data subscription during 3 hours each day.

3.5 Distributed Channel Selection in LAA

There are 11 non-overlapping channels, each with 20 MHz bandwidth, available for outdoor implementation in LAA in Europe by utilizing the spectrum within the range of 5470 - 5725 MHz [46, 47]. Previous works have investigated the performance of LAA system with dierent channel selection methods when assuming dierent numbers of available channels. Impact on implementing dierent channel selections (random or deterministic) method will be investigated in this thesis. The motivation of such investigation is to explore the opportunity to increase further the performance of the LAA system, and the additional cost that might be raised due to the increasing hardware or software complexity. The algorithm for the deterministic channel selection adopts the algorithm proposed in [48].

The algorithms proposed on [48] are considered for the thesis. The algorithms are as follows

Algorithm 1

1. Calculate temperature parameter: T = _log(2+t)^K

2. For BS j, measure the local energy detected on all channels c:

Aj(c) = Nj+ 2 X

k∈B;ck=c

Pk(j)

(29)

3. Calculate the probability, for all channels c P(c) = exp⁻^{Aj (c)}^T /X

c∈C

exp⁻^{Aj (c)}^T

4. Generate a random value according to probability P and select a channel according to the random value

Where Nj is the thermal noise detected on BS j on a certain channel, and Pk(j) represents the level of the signal received by BS j from transmission of BS k. B comprises of the sets of BSs. Also, t is the age variable, and K is a constant chosen during initialization.

Algorithm 2

Select channel c when c = arg minc∈C(A_j(c))

In words, Algorithm 2 chooses the channel with the minimum detected energy. The paper explained that Algorithm 2 could be considered as the specic case of Algorithm 1 when the temperature is set to 0 during the initialization step. The model for the thesis will use Algorithm 2 because the estimation provides good approximations with less computational complexity.

The algorithm being discussed requires the capability of each base station to measure and store the signal received from other base stations on a particular channel. The paper states that the new feature could be realized through a software update. It is important to mention that the algorithm being discussed is designed for WLAN devices, where they are intended to be able to operate within the range of the unlicensed spectrum (2.4 GHz and 5 GHz). To transport these capabilities for the operation of LTE on the unlicensed spectrum might require additional costs. These costs are related to the necessary hardware or software upgrade to enable LTE-LAA base stations operate conforming with the regulatory frameworks (LBT, Dy- namic Frequency Selection). The potential additional cost will be factored into the economic model, where we will compare the total deployment cost of LTE-LAA and LSA.

3.6 Interference and Throughput Calculation in LAA

This thesis will adopt the required co-existence mechanisms for the operation of LTE in the unlicensed 5 GHz band (LAA), the Listen Before Talk (LBT) mechanism and channel selection mechanism. This thesis will investigate the co-existence of two MNOs deploying their outdoor small cells utilizing LAA.

(30)

The mechanism of LBT emulates the mechanism of CSMA/CA of WiFi systems. The main idea is that each BS or WiFi Access Point (AP) have to make clear channel assessment before transmitting data. The channel is considered idle if the BS / AP detect that it has lower energy than Energy De- tection (ED) threshold. For multiple BS deployments, a contention domain may exist consisting sets of BSs where they can sense others transmission with the received power greater than Energy Detection threshold. In this scenario, each BS has the chance to access the channel inversely proportional to the number of BSs within the contention domain, adopting the approach dened in [29] and [30]. Assuming perfect channel sensing and LBT access scheme, the downlink throughput experienced by user i associated with BS j can be considered as:

Ri,j = Mjbi(γi) (3.12) Where Mj is the channel access time for BS j when accessing the channel, and bi(γ_i) is the downlink user throughput as the function of SINR experienced by user i.

3.7 Economic Model

Estimating the cost of equipment in making techno-economic analysis is a challenging task. One factor is because pricing is considered to be a business secret for vendors. The pricing is sensitive to several factors, for example, the agreements between vendors and operators and it also vary between countries. Finding a reference price for technology at its infancy is an arduous task, which is the case for LTE-LAA. There are previous works that can be used to get the price estimation from existing equipment that has similar capabilities. In [49, 15] the author breaks down the cost for dierent wireless network deployments. The data are gathered from interviews with vendors, operators, and independent research. The author in [15] estimates the price for outdoor LTE small cells and LTE small cells with built-in WiFi module. The author assumes the cost of an outdoor LTE base station equipment with WiFi built-in module to be approximately 20%

more expensive than the same equipment without WiFi module. In [50], the authors assume that the cost for cognitive radio equipment to be twice the price of legacy equipment. The authors mention the factors that drive the high cost, namely the sensing capability, wide bandwidth operation, database capability and small scale production.

Total Cost of Ownership for Radio Access Network

One aspect that will be investigated in this thesis is the economic discussion that will point out, under the taken assumptions, which deployment

(31)

method (LAA or LSA) is more cost ecient. This subsection will give information of the expenditures involved to build and operate wireless networks, with the focus of radio access networks.

The cost of radio access network can be broken into three components [49]:

Capital Expenditure (CAPEX): Capital Expenditure is one-time investment cost made by the MNOs to deploy the network. It covers the cost of base station equipment, backhaul connection, and also accessories such as the antenna, battery and power supply.

Implementation Expenditure (IMPEX): Implementation Expen- diture covers the expenses to build the network. It covers the cost of site acquisition, the civil works, initial planning and network optimiza- tion.

Operational Expenditure (OPEX): Operational Expenditure in- volves all ongoing and recurring expenses that aroused during the life- time of the network. Some examples are site rental fee, electricity cost, operation, and maintenance fee of the base station, backhaul, and site.

The values for the IMPEX and OPEX of small cell networks in the thesis will adopt the values for low-cost outdoor microcell in [49]. Whereas for the CAPEX for the equipment, we will utilize sensitivity analysis due to the uncertainty of the cost for LTE-LAA equipment. We will assume additional cost φ for LTE-LAA equipment relative to the cost of low-cost microcell equipment in [49]. The value of φ will take the value between 0 and 1 (0

≤ φ ≤1).

3.8 System Parameters

The parameters implemented in the thesis are shown on Table 3.3.

In this thesis, each scenario is iterated 1000 times. During each iteration for each scenario, the locations for the UEs and BSs are generated, and the path loss, SINR, and UE downlink throughput are calculated based on the methods explain in previous sections. The data is gathered by sampling the users located within the 100 m x 100 m area on the origin/center, to reduce the edge eects [51].

Figure 3.2 shows the overall workow of the thesis, where the main result is the deployment cost regarding Total Cost of Ownership (TCO) and the Net Present Value (NPV) for each scenario.

(32)

Table 3.3 System Parameters

Parameter Value

Carrier Frequency LSA flsa 2.3 GHz

Carrier Frequency LAA flaa 5.6 GHz [46, 47]

Bandwidth LSA WLSA 30;40;50 MHz per operator

Bandwidth LAA WLAA 20 MHz per channel; 11 non-overlapping channels

Thermal Noise -174 dBm

Propagation Model METIS modication ITU-R UMi pathloss [33]

LOS & NLOS ITU-R UMi LOS & NLOS Probability [34]

Energy Detection (ED) Threshold -62 dBm [46]

System Area 1 km x 1 km

Outdoor User Density 500 subscribers/km²/Operator

eNB Tx power 30 dBm

BS Antenna Type Omnidirectional [46]

Figure 3.2 System model and the workow of the thesis

(33)

Chapter 4

Results

This chapter presents the results obtained from applying the models explained in the previous sections. The rst step is to nd Base Stations densities λB needed to provide the capacity each year, for each scenario (LSA and LAA). The second step provides the techno-economic analysis of the data gathered from the rst step. The output is the recommendation of the most cost-ecient method to be adopted by the MNO.

4.1 LAA

This section presents the required Base Stations (BSs) densities required, in the scenario of two MNOs deploying their network on unlicensed 5 GHZ band. The required BS density is investigated when implementing random channel selection method and deterministic channel selection method as previously explained on subsection 3.4. In both cases, both mobile operators (Operator A and Operator B), implement the symmetric strategy. Either both of them perform random channel selection or deterministic channel selection.

4.1.1 Impact of Channel Selection Mechanism on LAA with Low Activity Factor

Figure 4.1 shows the minimum BS densities needed to deploy by each MNO to provide the trac demand each year.

We can see that by utilizing deterministic channel selection, the number of base station required to meet the trac demand can be reduced. Figure 4.1 also shows the gain of implementing deterministic channel selection is more signicant in higher trac demand that requires more BS to be deployed.

The less number of required BSs can be explained by examining the SINR and downlink throughput distribution on gure 4.2 and 4.3.

25

(34)

Chapter 4. Results 26

Figure 4.1 Network dimensioning for LAA system for low activity factor

Figure 4.2 shows the improvement of SINR when implementing deterministic channel selection for the fth year scenario (25.2 GB monthly usage), for BS density λB = 19. Deterministic channel selection results in the 1.82 dB increase of SINR from -1.8494 dB to -0.0233 dB for the cell-edge user.

Figure 4.2 SINR for dierent channel selection mechanisms, 25.2 GB monthly usage, λB = 19, low activity factor

Figure 4.3 shows the Cumulative Distribution Function (CDF) of user

(35)

downlink throughput with the same scenario as mentioned earlier. For random channel selection, the cell-edge user that is represented by the 5th percentile can not be served. Deterministic channel selection mechanism improves the cell-edge user downlink throughput to 3.83 Mbps.

Figure 4.3 Downlink throughput for dierent channel selection mechanism, 25.2 GB monthly usage, λB = 19, low activity factor

Figure 4.4 and gure 4.5 shows the Probability Density Function (PDF) of SINR for random channel selection and deterministic channel selection respectively.

The two following gures, gure 4.6 and gure 4.7 show the PDF of downlink user throughput for random channel selection and deterministic channel selection.

The following two gures show the SINR and downlink throughput distribution for the nal year within the scope of the thesis. The scenario represents the target throughput of the cell-edge user of 2.6 Mbps. The scenario is when the projected user monthly usage of 210.9 GB and the BS density λBis 53. Figure 4.8 shows that by implementing deterministic channel selection, the SINR of the cell-edge user can be improved from 3.3 dB to 9.2 dB.

Figure 4.9 indicates the improvement in the downlink throughput for two channel selection mechanisms. Deterministic channel selection achieves 3.58 Mbps downlink throughput for the cell-edge user, whereas, for the random channel selection, the cell-edge user can not be served.

(36)

Figure 4.4 PDF of SINR for random channel selection, 25.2 GB monthly usage, λB = 19, low activity factor

Figure 4.5 PDF of SINR for deterministic channel selection, 25.2 GB monthly usage, λB = 19, low activity factor

4.1.2 Impact of Channel Selection Mechanism on LAA with High Activity Factor

In this subsection, we will analyze the required base station density to be deployed for the case of high user activity. Figure 4.10 shows that for the high trac activity, the number of base stations needed to meet the demand is greater for each case of monthly subscriber usage compared for the case

(37)

Figure 4.6 PDF of downlink throughput for random channel selection, 25.2 GB monthly usage, λB = 19, low activity factor

Figure 4.7PDF of downlink throughput for deterministic channel selection, 25.2 GB monthly usage, λB = 19, low activity factor

of low activity factor. Figure 4.10 also shows the similar trend as in gure 4.1 for the case of low trac activity. For there is a higher incentive in using deterministic channel selection as the subscriber load increases, due to the higher reduction gain of the number base station.

(38)

Figure 4.8 SINR for dierent channel selection mechanisms, 210.9 GB monthly usage, λB = 53

Figure 4.9Downlink throughput for dierent channel selection mechanisms, 210.9 GB monthly usage, λB = 53

4.2 LSA

This section presents the network dimensioning for the case of LSA. In the case of LSA, the available bandwidth is dierent in the various countries.

In this thesis, we assume three types of spectrum availability in the 2.3 GHz

(39)

Figure 4.10Network dimensioning for LAA system with high activity factor

band. Low, moderate and high availability represent where the channel available for LSA operation equal to 60 MHz, 80 Mhz, and 100 MHz. In this thesis, we assume two MNOs are utilizing the LSA spectrum. Thus each MNO will have access to 30 MHz, 40 MHz and 50 MHz of spectrum for each scenario. Figure 4.11 and gure 4.12 show the required base station density for low and high activity factor. From gure 4.12 we can see that acquiring more spectrum will result in the reduction of the number base stations need to be deployed. The gain is more evident in the case of high activity factor and high usage. Moderate (40 MHz) and high (50 MHz) spectrum holding will result in the reduction of number base station by 30% and 40% compared to low spectrum holding (30 MHz).

4.3 Techno-Economic Discussion

This section will investigate the Total Cost of Ownership (TCO) for each scenario. The ultimate result of this thesis is the recommendation under the assumptions being made, which deployment type is the most cost-ecient.

The method to calculate TCO for the radio access networs will adopt the method used in [49]

T CO = CAP EX + IM P EX + OP EX (4.1) The CAPEX for deploying base station in the case of utilizing LSA can be shown as

CCAP EXLSA = NBS(CBS+ CBH) (4.2)

(40)

Figure 4.11 Network dimensioning for LSA system with low activity factor

Figure 4.12Network dimensioning for LSA system with high activity factor Where NBS is the number of base station deployed, CBS is the cost of a base station equipment, and CBH is the cost for the backhaul connection. For LAA, we will use sensitivity analysis due to the unknown cost of a small cell LTE-LAA as has been described on section 3.7. For LAA we can state the CAPEX as follows

C_{CAP EX}_LAA = N_BS C_BS(1 + φ) + C_BH

(4.3) Both LSA and LAA will use the same backhauling type with the same

(41)

Table 4.1Cost value and normalized cost

Parameters Cost Value (Euro) Normalized Cost

LAA LSA LAA LSA

Base Station Equipment CBS 4000 (1+φ) 4000 (1+φ) 1

Microwave Backhaul CBH 3500 0.875

IMPEX 2500 0.625

OPEX 2000 0.5

cost value. Table 4.1 shows the breakdown of the cost adopted in the thesis.

For the case of LSA deployment, we will also add the spectrum licensing fee Cspectrum. The spectrum licensing fee will be divided and distributed during the whole network operation. The value of the spectrum we will assume that the cost should be lower than the standard license paid for the 2600 MHz band. The auction result in Sweden will be used as a baseline, where the spectrum license fee is Euro 0.3/MHz/pop [52]. We will vary the values for the LSA license fee, high spectrum cost, and low spectrum cost. For high spectrum cost, we will assume the cost for the LSA spectrum to be 80% of the baseline value, which is Euro 0.24/MHz/pop. For low spectrum cost, we will assume the price of the LSA spectrum to be Euro 0.09/MHz/pop. The total cost (normalized) to meet the demand for each year in the case of high activity factor is shown in gure 4.13. From gure

Figure 4.13 Incremental TCO of network deployments with dierent cost assumptions

4.13 we can see the incremental cost for each deployment. In the case of LSA with high bandwidth availability (50 MHz), will always provide the most cost-ecient deployment choice, regardless the price of the spectrum. For the