Quantitative Analysis on the Feasibility and Benefits of ...

47
IN DEGREE PROJECT ELECTRICAL ENGINEERING, SECOND CYCLE, 30 CREDITS , STOCKHOLM SWEDEN 2017 Quantitative Analysis on the Feasibility and Benefits of Local Licensing GREGORIUS KRISTIAN PURWIDI KTH ROYAL INSTITUTE OF TECHNOLOGY SCHOOL OF INFORMATION AND COMMUNICATION TECHNOLOGY

Transcript of Quantitative Analysis on the Feasibility and Benefits of ...

Page 1: Quantitative Analysis on the Feasibility and Benefits of ...

IN DEGREE PROJECT ELECTRICAL ENGINEERING,SECOND CYCLE, 30 CREDITS

, STOCKHOLM SWEDEN 2017

Quantitative Analysis on the Feasibility and Benefits of Local Licensing

GREGORIUS KRISTIAN PURWIDI

KTH ROYAL INSTITUTE OF TECHNOLOGYSCHOOL OF INFORMATION AND COMMUNICATION TECHNOLOGY

Page 2: Quantitative Analysis on the Feasibility and Benefits of ...

Abstract

In-building connectivity in enterprises and vertical industries will emerge as animportant use case of future 5G mobile services, given the fact that 80% of the traffic isgenerated indoor [1]. However, spectrum solution to this use case, which is unlicensedspectrum, may not be able to secure spectrum availability for large traffic demand as wellas critical use case due to the nature of Listen-before-Talk (LBT) protocol.

To cope with that, several novel spectrum sharing schemes have been discussedrecently. One approach is to grant exclusive access to a spectrum in limited geographicalarea, or so called local licensing. It is motivated to reduce external interference frommultiple networks that operates in a common area. This thesis evaluated such approach interms of QoS improvement compared to unlicensed spectrum. To carry out the evaluation,two type of network deployment were simulated in three different scenarios:

• unlicensed case (LBT-based system in overlapping deployment)• intermediate case (LBT-based system in non-overlapping deployment)• local licensed case (non-LBT system network in non-overlapping deployment)

By simulating these three simulation scenarios, we have been able to compare two vari-ables (usage of LBT and deployment type) one by one.

Comparing the first case (unlicensed) and third case (local licensed), the local li-censed case’s throughput does not decrease as rapidly as unlicensed case’s throughputwhen the traffic load increases. Local licensed case has 130% higher system capacitythan unlicensed case in 10 Mbps QoS requirement. Between LBT and deployment type,LBT has more significant impact on the result. The increase of traffic load in LBT-basedsystem caused a much significant increase in interference compared to non-LBT basedsystem. LBT-based system performed well in low traffic, giving twice as much through-put as the third case, but the throughput also decrease rapidly as the traffic increase.Comparing the first and second case, the throughput gain of separating the deploymentwas not significant in LBT-based system.

All in all, we proved that local licensing is worthwhile in term of QoS improvementcompared to unlicensed case, given that the traffic load is high enough.

i

Page 3: Quantitative Analysis on the Feasibility and Benefits of ...

Sammanfattning (Svenska)

Uppkoppling inomhus pa foretag och vertikala industrier kommer att vara ett viktigtanvandningsomrade for 5G mobiltjanster, da 80% av trafiken genereras inomhus. Medanledning av hur ”Listen-Before-Talk” (LBT) protokollet fungerar kan dagens losning forlicensfritt spektrum emellertid inte sakerstalla tillgangligheten vid stora trafikbehov.

For att klara av trafikbehovet har scheman for spektrumdelning varit ett stortdiskussionsamne. En losning till problemet ar att ge exklusiv atkomst till spektrum inombegransade geografiska omraden, lokallicensering. Det ar fordelaktigt att pa allmannaplatser reducera extern interferens fran andra operatorer i det omradet. Denna Mas-teruppsats tittar pa sadan losning med service-kvalitet som utgangspunkt och jamformed olicensierat spektrum. Utforandet av utvarderingen har tittat pa tva olika natverksom simulerades i tre olika scenarion:

• Olicensierat fall (LBT-baserat system vid overlappande implementering)• Mellanliggande fall (LBT-baserat system i icke-overlappande implementering)• Lokalt licensierat fall (icke-LBT-systemnat i icke-overlappande implementering)

Genom att anvanda dessa tre simuleringsscenarier har vi kunnat jamfora tva vari-abler: anvandning av LBT och implementeringstyp.

Genom att jamfora det forsta scenariot (Olicensierat) och det tredje scenariot (lokaltlicensierat) kn vi se att det lokalt licensierade scenariot inte minskar med samma hasighetsom det olicensierade fallet. System kapaciteten for det lokalt licensierade fallet ar 130%hogre for det lokalt licensierade scenariot jamfort med olicensierade fallet nar systemettestas med 10 Mbps som QoS-krav. Vid jamforelse mellan system med LBT och icke-LBThar LBT anvandning en tydlig paverkan pa resultatet. Trafikbelastningsokningen i LBT-baserade system gav en stor storningsokning (interferens) jamfort med att anvanda icke-LBT system. LBT system presterar bra under lag trafikbelastning och ger da fordubbladkapacitet jamfort med det lokalt licensierade scenariot (tredje), men med kraftig ka-pacitets reducering nar trafikbelastningen okade. Jamfor man det forsta och det andrascenariot (overlappande mot ej overlappande natverk) var kapacitetsokningen inte sig-nifikant.

Sammanfattningsvis visar vi pa att lokal licensiering ar en fungerande teknik foratt oka QoS jamfort med ett olicensierat scenario givet att natverken har en tillrackligthog belastning.

ii

Page 4: Quantitative Analysis on the Feasibility and Benefits of ...

Acknowledgements

I would like to express a sincere gratitude to my industrial supervisor, Du Ho Kang,for his valuable input as well as great discussion while I was working on this thesis. Thediscussions has led me to a better understanding on the spectrum sharing in general aswell as helped me during the result analysis. I would also like to say thank you to myacademic supervisor, Ki Won Sung, as his assistance has helped me carry out a betterthesis project from scientific research point-of-view. Not to forget to say thank you toAnders Vastberg for his valuable input and examination on this thesis project.

Another express of gratitude goes to Ericsson Research, who has given me thechance to do my master thesis there. I had a nice opportunity to meet and talk withgreat people inside.

Most importantly, thank you to my family back in Indonesia, who has always sup-ported me to dream big and go further. Without their support, it would be impossibleto complete my master education here in Sweden.

GregoriusSeptember 2017

iii

Page 5: Quantitative Analysis on the Feasibility and Benefits of ...

List of Acronyms

AP Access Point

CDF Cumulative Distribution Function

CBRS Citizens Broadband Radio Service

CS Carrier Sense (threshold)

CSMA Carrier Sense Multiple Access

DAS Distributed Antenna System

DL Downlink

ED Energy detect (threshold)

FCC Federal Communications Commision

FDMA Frequency Division Multiple Access

HetNet Heterogeneous Network

IEEE Institute of Electrical and Electronics Engineers

ISM Industrial, Scientific, and Medical

ITU International Telecommunication Union

LAP Local Access Provider

LBT Listen-before-Talk

LoS Line-of-Sight

LSA License Shared Access

LTE Long Term Evolution

LTE-LAA LTE-License Assisted Access

MAC Medium Access Control

MNO Mobile Network Operator

QoS Quality of Service

RAT Radio Access Technology

SAS Spectrum Access Systems

SINR Signal-to-interference-plus-noise ratio

TDD Time Division Duplex

TDMA Time Division Multiple Access

UE User Equipment

UL Uplink

iv

Page 6: Quantitative Analysis on the Feasibility and Benefits of ...

List of Tables

4.1 System parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

v

Page 7: Quantitative Analysis on the Feasibility and Benefits of ...

List of Figures

1.1 Three approaches of spectrum management with their QoS level and accessflexibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

1.2 Three dimensions of sharing a spectrum [9] . . . . . . . . . . . . . . . . . 21.3 Research methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

2.1 Three tier of access level in CBRS . . . . . . . . . . . . . . . . . . . . . . 82.2 Example of local licensed scenario. Each building is served by one Local

Access Provider (LAP) and a common spectrum can be shared among theLAPs [15] . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

3.1 Workflow of data collection and analysis . . . . . . . . . . . . . . . . . . 103.2 Simulation scenarios. Deployment example depicts two building with three

floors each. Blue and red dot are access points of two different network. . 113.3 Network utilization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133.4 Load level and their respective offered traffic . . . . . . . . . . . . . . . . 13

4.1 Environment Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164.2 Two non-overlapping networks in the simulator . . . . . . . . . . . . . . 174.3 Two overlapping networks in the simulator . . . . . . . . . . . . . . . . . 17

5.1 Median throughput vs served traffic . . . . . . . . . . . . . . . . . . . . . 205.2 Throughput vs served traffic on cell edge . . . . . . . . . . . . . . . . . . 215.3 System capacity at 5 Mbps and 10 Mbps required throughput . . . . . . 225.4 Downlink interference and SINR CDF on low traffic load . . . . . . . . . 235.5 Downlink interference and SINR CDF on medium traffic load . . . . . . 245.6 Downlink interference and SINR CDF on high traffic load . . . . . . . . . 255.7 Downlink bitrate and throughput on low traffic load . . . . . . . . . . . . 265.8 Downlink bitrate and throughput on medium traffic load . . . . . . . . . 275.9 Downlink bitrate and throughput on high traffic load . . . . . . . . . . . 285.10 Neutral host model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 295.11 Comparison of available non-overlapping channels in 2.4GHz ISM band

and 5GHz U-NII band. Source: [35] . . . . . . . . . . . . . . . . . . . . . 305.12 Process of managing spectrum. Source: [37] . . . . . . . . . . . . . . . . 31

vi

Page 8: Quantitative Analysis on the Feasibility and Benefits of ...

Contents

Abstract i

Sammanfattning (Svenska) ii

Acknowledgement iii

List of Acronyms iv

List of Tables v

List of Figures vi

1 Introduction 11.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Previous Work and Research Gap . . . . . . . . . . . . . . . . . . . . . . 31.3 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.4 Goal and Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.5 Research Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.6 Scope and Limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.7 Benefit, Ethics, and Sustainability . . . . . . . . . . . . . . . . . . . . . . 5

2 Overview of Spectrum Management 62.1 Current approaches of spectrum management . . . . . . . . . . . . . . . 62.2 Spectrum Sharing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.3 Local licensing for in-building use case . . . . . . . . . . . . . . . . . . . 8

3 Methodology 103.1 Research Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103.2 Data Collection and Analysis Method . . . . . . . . . . . . . . . . . . . . 12

4 System Model 164.1 Network Deployment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164.2 Radio Environment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

4.2.1 Propagation Model . . . . . . . . . . . . . . . . . . . . . . . . . . 184.2.2 SINR Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

4.3 System Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18

vii

Page 9: Quantitative Analysis on the Feasibility and Benefits of ...

5 Result and Analysis 205.1 System Throughput vs Traffic Load . . . . . . . . . . . . . . . . . . . . . 205.2 Capacity Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225.3 Impact on Interference and SINR . . . . . . . . . . . . . . . . . . . . . . 235.4 Impact on User Bitrate and Throughput . . . . . . . . . . . . . . . . . . 255.5 Discussion on the Feasibility of Local Licensing . . . . . . . . . . . . . . 28

6 Conclusion and Future Work 336.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

References 35

viii

Page 10: Quantitative Analysis on the Feasibility and Benefits of ...

Chapter 1

Introduction

1.1 Background

Future 5G network will have to cope with unprecedented traffic demand as industrypredicted that mobile data traffic will grow by eightfold during 2015 to 2020 [2]. Moreover,it is important to consider indoor use case since it generates more than 80% of the traffic[1]. This use case includes enterprises and vertical industries, such as campus, retail,hospital, sport venues, etc.

However, we realized an inevitable limitation on usable radio spectrum. Currentspectrum management model divides usable frequency bands into fragments that can beused for particular services, such as mobile network, satellites, broadcasting, military,etc. From spectrum authorization perspective, they can be categorized into two types:licensed and unlicensed spectrum. Licensed spectrum allows the licensee to keep anexclusive use of certain frequency band. This enables unlimited interference managementto ensure that the quality of service (QoS) is satisfied [3]. In bigger perspective, however,research shows that several frequency range on licensed band is underutilized most of thetime due to the static frequency allocation [4].

In another extreme, unlicensed spectrum, such as ISM (Industrial, Scientific andMedical) band, allows any entity to use it freely without prior registration. While itposses virtually no market entry barrier and highly scalable, the QoS is much harderto be satisfied due to uncontrollable interference from another co-located network in thesame spectrum [5].

Figure 1.1: Three approaches of spectrum management withtheir QoS level and access flexibility

The important question is, are these two solutions enough for future mobile trafficdemand? Given the fact of current traffic growth as well as future prediction, a novel

1

Page 11: Quantitative Analysis on the Feasibility and Benefits of ...

spectrum management model is necessary. There has been intensive discussion about theconcept shared spectrum in recent years. In the regulatory level, Federal CommunicationsCommision (FCC) in the U.S., for example, has opened 150MHz of bandwith in the 3,5GHz band for commercial use. Named the Citizens Broadband Radio Service (CBRS), itdoes not require the mobile network operator (MNO) to buy or own the band. Instead,a Spectrum Access System (SAS) will check spectrum availability and assign short termlicense to user in order to minimize interference [6, 7]. Another example is License SharedAccess (LSA) in Europe. It is defined as a regulatory approach to assign band that isalready used by incumbent user. A sharing rules is introduced to ensure both authorizedand incumbent user can maintain required QoS [8].

Figure 1.2: Three dimensions of sharing a spectrum [9]

Back to our focus, which is indoor use case, unlicensed spectrum is still one ofthe most popular solution among venue owner. Nowadays, WiFi availability has been amandatory facility that also attract guest and customer to a venue. However, given theaforementioned limitation in the unlicensed band, WiFi network might not be enough forfuture traffic demand. Licensed spectrum might be needed to secure spectrum availabilityin a venue so that the network will experience less uncontrollable external interference, aswell as for critical use cases. There have been several writing about the concept of neutralhost or micro\local operator that operates in a locally-licensed spectrum [5, 10]. In thisconcept, a venue owner (or a third party) act as a local operator that deploy a neutralhost network to provide coverage in the venue. Since different users can be subscribed todifferent MNO, the neutral network has to be connected to multiple MNO’s network. Alocally-licensed spectrum can be obtained to provide connectivity to the venue.

Since the local licensing grants an exclusive access to a spectrum in limited ge-ographical area, it ensures required Quality of Service (QoS) by reducing interferencefrom another network in the same area. However, a proof-of-concept is still needed.Therefore, this thesis aims to assess quantitative benefit and feasibility of local licensingcompared to current unlicensed spectrum model, which the coexistance mechanism is

2

Page 12: Quantitative Analysis on the Feasibility and Benefits of ...

based on contention-based listen-before-talk (LBT) protocol. Indoor small cell use caseis the main focus of the study.

1.2 Previous Work and Research Gap

A fast and reliable in-building wireless coverage is one of the research fields thatattracts attention from industry and academia. At the moment, unlicensed spectrumis still one of the most popular spectrum solution for in-building coverage (e.g. usingWiFi). However, since it was predicted that traffic demand will continue to grow rapidlyand most of this traffic comes from indoor use cases, several research efforts have alsobeen made to evaluate WiFi’s potential capacity as well as the limitation. Paper [11]provides analytical model to assess 802.11’s throughput and paper [12] provides morespecific evaluation of 802.11 throughput at 5GHz in dense deployment. From industry,[13] provides a real world measurement of WiFi in 5GHz, although the technology thatwas used, 802.11a, is a little bit outdated.

The concept of spectrum sharing has been widely discussed as one of the solutionsto cope with future traffic demand by using the scarce and limited radio spectrum moreefficiently. A report in 2008 ITU Symposium discussed general overview of spectrumsharing from both regulatory perspective as well as technical and practical implementa-tion [14]. Industry players also published several white papers about spectrum sharing,such as [7] which also discuss the 3.5 GHz CBRS band in US.

The limitation of unlicensed spectrum in indoor use case and the interesting benefitof spectrum sharing are two underlying concepts for this thesis. More specifically, wewant to evaluate the quantitative feasibility of local spectrum sharing in indoor use case.Examples of previous works in this field includes paper [10] that discussed the concept oflocal licensing in the wake of neutral host and micro operator as well as possible businessmodel of it. Paper [15] deepened the previous study by providing techno-economic analy-sis and cost model for wireless system in shared spectrum. To the best of my knowledge,however, we still lack of direct comparison between the usage of unlicensed spectrumversus local licensing in an indoor small cell scenario. We also need in-depth analysisof local licensing feasibility from technical perspective for future indoor coverage. Thisthesis aims to fill that gap with the following research question.

1.3 Problem Statement

Based on the background and the research gap above, this thesis answers following highlevel question:

Is local licensing worthwhile in terms of QoS improvement compared tounlicensed spectrum in indoor use cases?

More specifically, the high level question can be divided into three detailed questions:

1. How significant is the effect of LBT?2. How significant is the effect of deployment restriction?3. Considering the effect of both LBT and deployment restriction, can 5th percentile

data rate per operator be improved compared to unlicensed spectrum?

3

Page 13: Quantitative Analysis on the Feasibility and Benefits of ...

1.4 Goal and Contribution

This research aims to see if the improvement in SINR and user throughput peroperator in defined local licensing scenario is worthwhile compared to unlicensed spec-trum. It is also important to consider the disadvantages of local licensing. In general,this research is also motivated to present an alternative approach in spectrum sharingregime as well as its performance to anticipate future demand. In contributes to a betterway of utilizing available radio spectrum.

1.5 Research Methodology

In general, this thesis project is divided into three phases. Following chapter ex-plains the phases in detail.

Figure 1.3: Research methodology

In the first phase, a literature review is needed to learn limitation in current spec-trum management approach to cope with future mobile traffic demand. Different spec-trum sharing approaches have to be studied as well in order to define the needed evaluationscenarios and its parameter, for example the area size and propagation characteristics.

The second step consist of setting up in the defined use case in simulation using aMatlab-based radio system simulator. Simulation is used because a real life experimentis not possible due to the complexity of deploying a real-world use cases of interferenceinvestigation. The simulation results will then presented and analyzed. Detailed expla-nation about the methodology can be found in chapter 3 and Chapter 4.

1.6 Scope and Limitation

This thesis has several limitation from both technical and research scope. First,it focuses only on indoor scenario. Both downlink and uplink were simulated to mimicreal world condition, but only downlink result is analyzed. The RAT of choice in thesimulation is 802.11n and LTE, which is soon will be replaced by 802.11ac or newer aswell as 5G. The simulation used a simple two-dimensional square map where the usersare randomly distributed throughout the area. The simulation map should be extendedto be more realistic simulation that includes multi-storey building with complex shapeand propagation characteristic. The thesis only focuses on homogeneous network withonly WiFi or only LTE at one time. HetNets is out of scope in this thesis.

4

Page 14: Quantitative Analysis on the Feasibility and Benefits of ...

Furthermore, the thesis also focuses on quantitative analysis of local licensing. Ad-ditional analysis of local licensing from perspective of spectrum regulation and businesslandscape is also added, but only in brief.

1.7 Benefit, Ethics, and Sustainability

This thesis promote a smarter way of utilizing spectrum by a means of spectrumsharing, given that spectrum is a scarce natural resources that has to be used as efficientas possible. More efficient usage of spectrum means more network capacity in the sameamount of bandwidth. Considering that the growth of traffic has no sign of stopping,these larger capacity means that we can sustain the growth for longer time.

As the spectrum is shared, the usage can be elastic following the demand in par-ticular area. It eliminates the expensive spectrum bidding process that exists in thecurrent spectrum assignment. If the spectrum ownership can follow economic principalof supply and demand rather than classic rigid ownership, it is believed that it will bemore economically sustainable in the long run. If the infrastructure can be shared aswell, it will means less production of telecommunication hardware and consequently lessenvironmental footprint from production process. However from social aspect, this thesisdoes not really address the access inequality problem that exist between, for example,large city and remote or poor area.

5

Page 15: Quantitative Analysis on the Feasibility and Benefits of ...

Chapter 2

Overview of Spectrum Management

This chapter describes relevant background knowledge for this thesis project. Chap-ter 2.1 explains current method of spectrum management and its history in brief. Spec-trum sharing approach is described in the Chapter 2.2, and Chapter 2.3 discusses theupcoming 5G for indoor use case.

2.1 Current approaches of spectrum management

Wireless communication relies on radio spectrum range of the electromagneticwaves. According to International Telecommunication Union (ITU), this spectrum spansfrom 3kHz up to 3000 GHz and divided into nine bands [16]. IEEE also has its owndivision and nomenclature for these frequency bands [17]. Radio spectrum is a scarcenatural resource that has to be tightly regulated. In the other hand, interference will alsobe a problem if radio spectrum is not regulated properly.

Current practice of spectrum management divides this responsibility to national andinternational level. At international level, the regulatory body, which is ITU, managesspectrum usage beyond national boundaries, such as aviation and satellite. It allocatesthe spectrum for different applications as well as providing guidance for spectrum al-location and assignment 1 for regulatory body on national level[18]. This allocation iswritten in ITU’s International Radio Regulation. The national regulator further managesthe spectrum for nationwide usage. Often, this body holds spectrum auction to awardspectrum license. The licensee, for example TV broadcaster or mobile operator, will ob-tain exclusive rights to use particular radio band. One benefit of this approach is thelicensee will have freedom to manage its network to have least interference possible.

This rigid spectrum assignment dates back to the dawn of radio communication.At that time, interference was a big problem for broadcasters in the same area if thefrequency was not properly managed. To minimize the problem, national regulatorsassigned specific frequency to specific user and application. However, this command-and-control mechanism has proven to be inflexible to market demand. Its lengthy andcomplicated bureaucratic process reduces the efficiency of the spectrum usage as well asthe economic value [19] [20]. Sometimes new technology also have difficulty in gainingaccess to radio spectrum. In the modern world, this mechanism is still the primary wayof assigning radio spectrum. In the other hand, the scarcity of spectrum also makes the

1Allocation means to manage the general use of particular band, whereas assignment means to decidewhich stakeholder that is allowed to use the band [18]

6

Page 16: Quantitative Analysis on the Feasibility and Benefits of ...

licensee have to spend overwhelming amount of money in spectrum auction.So far, we have mostly discussed about only one approach of spectrum licensing,

called licensed spectrum. At another extreme, there is unlicensed spectrum. It does notrequire prior registration to use the spectrum as long as the transmit power is limited. Asthe consequences, there is no exclusive usage in this spectrum and interference problemmay arise. However, the interference can be minimized by utilizing suitable wirelesstechnology. One success story of this technology is, no doubt, Wi-Fi. Based on the IEEE802.11 standard, the number of Wi-Fi devices quickly grew and spread around the worldsince launched. By the begininning of 2015, Wi-Fi Alliance reported 4.5 billion Wi-Ficertified devices were in use [21] and WiFi hotspot nowadays becomes primary necessityin home and public places, such as offices, campuses, sport venue, etc.

2.2 Spectrum Sharing

We can define the term spectrum sharing as a common spectrum usage by multipleentities or operator in a particular region, provided that they do not interfere each other[22, 3]. The spectrum will not permanently owned by an operator, instead the spectrumcan be used using temporary license for a period of time [7]. This ”light licensing”mechanism is also used to minimize the inter-network interference that might occur.Compared to rigid spectrum assignment in licensed spectrum, such mechanism promisesa lower entry barrier, cost saving, and more efficient spectrum usage due to ’on-demand’model. In the other hand, the mechanism also protects the operator from uncontrollableexternal interference which occurs in unlicensed spectrum.

In following chapter, several approaches of spectrum sharing from regulatory as wellas technological point of view is explained.

3.5Ghz CBRS (US)

In April 2016, Federal Communications Commision (FCC) opened up 3.5 GHzCitizens Radio Broadband Service (CBRS) band for commercial use in the US [6, 7].Previously, this band was used by US Department of Defense and fixed satellite serviceproviders. There is 150 MHz of available spectrum from 3550-3700 MHz. Several com-panies, such as Ericsson, Google, Intel, Qualcomm, Federated Wireless, Ruckus Wirelessand Nokia, have shown interest in this band and are planning for further partnership.

MNO does not have to buy nor exclusively own the spectrum in order to use it.Instead, spectrum management on this band will rely on a innovative spectrum sharedscheme. First, users of this spectrum will be divided into three tiers with different priority.Most prioritized top tier consists of incumbent user, such as Dept. of Defense and satelliteproviders. The second tier is Priority Access Licenses (PAL) layer, in which user canobtain short term license 10MHz bandwidth in the 3550-3650MHz range via auction.Lowest priority tier is General Authorized Access (GAA) layer, which provides 80 MHzbandwidth for any FCC-certified 3.5 GHz device.

7

Page 17: Quantitative Analysis on the Feasibility and Benefits of ...

Figure 2.1: Three tier of access level in CBRS

Acting as a frequency coordinator is the Spectrum Access System (SAS). It willcheck spectrum availability and assign frequency band to users. This automated systemensures that lower tier user does not interfere the higher one. Another system calledEnvironmental Sensing Capability (ESC) will also be deployed to monitor if the band isused by federal communication to protect that communication from interference.

LTE-U / LAA

The concept of LTE in unlicensed spectrum (LTE-U) or LTE License Assisted Ac-cess (LTE-LAA) allows LTE technology to run on unlicensed spectrum, primarily 5 GHzband [23, 24, 25]. It is meant to operate as a secondary channel to traditional licensedLTE access, giving a performance boost to UE when needed. To ensure a fair spectrumusage with another devices in unlicensed spectrum, such as Wi-Fi, LTE LAA employsa Listen before Talk (LBT) protocol. This protocol works by choosing least occupiedchannel during start-up as well as limiting channel access time in order to coexist fairlywith Wi-Fi. Despite the LBT protocol, several entities still raised concerns that thistechnology will degrade performance of Wi-Fi devices on the same area.

2.3 Local licensing for in-building use case

By the time of this thesis writing, the term local licensing might sounds unfamiliarto some people. However, the concept of geographical separation to reuse a certainspectrum might be as old as the radio communication itself. In my opinion, even theexisting per-country spectrum allocation and assignment can be seen as a geographicalseparation, as for example Sweden and and other countries can use the same 1800MHzspectrum for mobile communication, border case aside.

Going down to cellular networks, the concept of frequency reuse can actually be seenas a geographical separation. The same frequency band is used in multiple cells, giventhat the reuse distance D is long enough to minimize the risk of co-channels interference.However, we can also reduce the cell size and reuse distance, thus densify the networkand increase the system capacity [26].

8

Page 18: Quantitative Analysis on the Feasibility and Benefits of ...

Figure 2.2: Example of local licensed scenario. Each building is served by one LocalAccess Provider (LAP) and a common spectrum can be shared among the LAPs [15]

However, all of the concept of geographical separation above is part of rigid spec-trum management that exist today, as explained in chapter 2.1. That kind of spectrummanagement does not promote inter-operator spectrum sharing. Recalling chapter 2.2that spectrum sharing has several benefits as well as the fact that 80% of the traffic isgenerated indoor, the spectrum sharing for indoor use case becomes very interesting.

9

Page 19: Quantitative Analysis on the Feasibility and Benefits of ...

Chapter 3

Methodology

Based on the defined background and research question, the method of experimentalresearch was chosen due to the need of dealing with variables and constants [27]. Figure3.1 shows the workflow of data collection and analysis in this thesis.

Figure 3.1: Workflow of data collection and analysis

We started first by defining three simulation scenarios based on the research ques-tions. There are two major differences between each scenario, which are the networkdeployment type and the radio access technology (RAT). The overview and motivationbehind each scenarios are explained in chapter 3.1. Then, these scenarios were imple-mented in a computer simulator. It is an Ericsson’s Matlab-based static radio networksimulator that can be configured to model numerous use cases and system parameters.Chapter 3.2 explains the data collection method using this simulator as well as the anal-ysis method. The detailed implementation of the simulator is explained in chapter 4.Mathematical calculations is provided only as underlying theory to support the simu-lation result, but not for the data collection method itself. The result was analyzed inquantitative approach with additional qualitative analysis regarding spectrum regulationin the end to complement the quantitative one.

3.1 Research Scenarios

Revisiting the research question, we wanted to investigate two inter-network coex-istence mechanism in shared spectrum, which are listen-before-talk and geographicallyrestricted deployment.

10

Page 20: Quantitative Analysis on the Feasibility and Benefits of ...

For geographical deployment type of the network, we coined two terms: overlappingdeployment and non-overlapping deployment. A network itself corresponds to a group ofbase stations and UEs that can communicate each other. In the other words, an UE ofnetwork A can only be connected to base station on network A. In case of overlappingdeployment, multiple network can coexist in the same geographical area. This is thecase on current mobile network deployment, where multiple operators can occupy thesame space. In the other hand, non-overlapping deployment means the first network andthe second one are geographically separated (depicted respectively as red and blue dotin figure 3.2). The listen-before-talk protocol is used in WiFi system, but not in LTEsystem.

To answer the research question, three simulation scenarios have been defined asdepicted in figure 3.2.

Figure 3.2: Simulation scenarios. Deployment example depicts two building with threefloors each. Blue and red dot are access points of two different network.

• Case 1 - Unlicensed spectrumThis scenario represents existing deployment in unlicensed spectrum as a baselinereference. The technology is WiFi with LBT protocol. Multiple WiFi network cancoexist in the same area.• Case 2 - Intermediate case

This is a hypothetical scenario with two WiFi networks in a separated deploymentarea (hence non-overlapping). This intermediate case is needed to compare theresearch variables (LBT and deployment restriction) one by one.• Case 3 - Local licensed

This is the defined local licensing scenario where a geographical separation betweennetwork is motivated to minimize the effect of interference on the same sharedspectrum. The RAT is LTE without LBT protocol.

Comparison between the first and second scenario is intended to see the effect ofdeployment restriction, whereas the effect of LBT can be seen by comparing the resultof second and third scenario. We can see from figure 3.2 that there is a geographicalseparation in the non-overlapping deployment.

11

Page 21: Quantitative Analysis on the Feasibility and Benefits of ...

In other way, we can also say that local licensing is a way to restrict deployment ofother network in an area. The term deployment restriction and geographical separationwill be used interchangeably throughout this thesis. They don’t have exactly the samemeaning by definition, but they are more or less imply a similar thing in this thesis.

The geographical area of the focus will be in the size of a building, such as officeand shopping mall. Thus, instead of giving the access of a frequency band to incumbentMNOs, the facility owner or a building owner can own or lease a spectrum to realize thenetwork connectivity inside their facility [15]. We call the facility owner that deploystheir own in-building network as a local operator. In this way, this means that the samefrequency band can be reused even in every neighboring buildings as shown in figure 2.2.In the context of this thesis, the concept of local licensing can also be seen in one or otherway as a space division multiple access, which is defined as separation of user in spatialdimension in order to reuse the same frequency band [26].

The choice of RAT determines whether LBT protocol was used or not. We considera homogeneous network deployment where only one type of RAT was used in each simu-lation scenario. A medium access control (MAC) protocol is needed to manage multipleterminals that are accessing a common spectrum on the same time in a given physicalmedium. If it is not properly managed, more than one terminals might communicateat the same time and area using the same channel, in which a strong interference mayoccur and degrades the service quality. MAC protocol itself is divided into contention-free and contention-based protocol. In contention-free MAC protocol, a central controlleris needed to allocate resources to the terminals. This controller usually resides inside abase station or access point, depending on the network type. The terminals will thencommunicate using the preallocated network resources, e.g. frequency channel in FDMAor timeslots in TDMA. In the other hand, contention-based network does not have acentral controller to do resource allocation. The terminals in this network transmits datawithout coordination with another terminals. Therefore, additional mechanism is neededto handle interference, for example Carrier Sense Multiple Access (CSMA).

3.2 Data Collection and Analysis Method

The data collection method refers to computational simulation that was used inthis thesis. There are three independent variables. The first two are MAC protocoland deployment type that are explained above. The third variables is the traffic load.Different traffic loads were simulated to capture as broad load range as possible, as theresearch question is to find whether local licensing is worthwhile in high traffic load.

We refer the traffic load here as offered traffic. Offered traffic is the traffic thatcome to be served by the serving node. A portion of this incoming traffic is served by theserving nodes, which we refer to carried or served traffic. Another portion of the trafficget blocked or dropped if the serving node can not handle all the incoming traffic. In thisthesis, the traffic load is measured per network (system-wide).

Observation points

To get the possible traffic load range in the defined simulation scenario, we firststarted by using low traffic load and increased it gradually until the simulated networkreached 90-100% network utilization. The traffic load was the total offered traffic pernetwork, not per user. The only scenario that could be maximized to reach nearly 100%

12

Page 22: Quantitative Analysis on the Feasibility and Benefits of ...

network utilization was only local licensed scenario. On that traffic load point, utilizationof unlicensed and intermediate case were at 70% and 50%. Notice that intermediate caseis better on utilizing channel resource, given that the amount of served traffic on locallicensed and intermediate case is the same on this highest traffic load point. Because oneof the simulation case (local licensed case) has reached nearly 100%, we took this loadlevel point as the highest offered traffic for all scenarios. The network utilization can beseen in figure 3.3.

Figure 3.3: Network utilization

We simulated 21 traffic load level as depicted in figure below. The traffic load istotal offered traffic per network, not per user.

Figure 3.4: Load level and their respective offered traffic

13

Page 23: Quantitative Analysis on the Feasibility and Benefits of ...

Based on this result, three load level were decided to be taken as observation pointfor further analysis:

• Low traffic loadFirst point is defined as low traffic, which corresponds to the second load level pointwith total offered traffic of 16.9 Mbps. This observation point was used and notthe smallest load level point so that we can still see difference in the result betweenthe three scenarios.• Medium traffic load

The second observation point used 7th load level point, or 42.2 Mbps offered traffic.• High traffic load

The third point was taken on 16th load level point, or 99 Mbps load.

Analysis Method

Based on the observation points, we plotted several graphs. The section belowexplain the aims of each result graphs and how they were analyzed. All result focuseson downlink. Only results from network 1 are displayed for the sake of simplicity sincenetwork 1 and 2 have quite similar result.

The result was arrange so that we first see the main finding on this thesis on systemlevel performance. Then, CDFs corresponding to the link level performance (interference,sinr) are provided to strengthen the understanding of the result in system level.

1. System Throughput vs Served Traffic

This result, which can be found in chapter 5.1, is the main finding of the thesis.We use this graph to understand the relationship between the throughput andtraffic load. As the traffic load range from the lowest point to the highest possiblethroughput that a network can achieve, we can see a broader picture of this relation.The graph is presented in two version, median and 5th percentile, to represent thetypical user experience and user on the cell edge, respectively. Served traffic is usedinstead of offered traffic so we can obtain the system capacity that is experiencedby the user, not the system capacity in the ideal condition.

2. System Capacity

From the throughput vs load graph, the capacity graph was plotted in chapter 5.2.The system capacity was obtained by taking the amount of traffic that can be servedwhile maintaining a certain level of QoS. In this case, the QoS level are representedby 5 Mbps and 10 Mbps throughput per user in the cell edge (5th percentile). Theorange line in figure 5.2 helps us to obtain the value for system capacity graph infigure 5.3.

3. Interference and SINR

Chapter 5.3 provides CDF plots of interference and SINR from the three observa-tions points. Interference and SINR graphs is needed to understand the impact ofdifferent MAC protocol as well as deployment type on radio link level performance.We start by the interference and SINR CDF during low traffic load, then mediumand high load. By providing the graphs one-by-one on each load, we can see theclear development of interference and SINR as the traffic load increases.

14

Page 24: Quantitative Analysis on the Feasibility and Benefits of ...

4. Bitrate and Throughput

Different level of SINR and traffic load on user side will have further impact onbitrate and throughput. Bitrate is defined as maximum achievable data rate in acertain channel condition. Adaptive modulation was used and it determined theSINR to bit per radio resource element calculation. Based on Shannon’s channelcapacity formula, the bitrate will increase if the SINR increase.

However, it is not guaranteed that the user will experience the maximum bitratethat the link can achieve. The actual data rate that is experienced or received bythe user can be lower than the ideal bitrate. This usable data rate of succesfultransmission is defined as throughput. The level of throughput is also determinedby the scheduler, which schedule the traffic based on traffic load and schedulingmodel. The throughput CDF provides more information on how much the networkcan utilize the available bitrate as optimal as possible.

Bitrate and throughput graphs and their analysis are provided in chapter 5.4.

5. Discussion on the Feasibility of Local Licensing

After having all the quantitative result and its analysis, the next section providesfurther discussion of local licensing in qualitative manner. We start by providingmotivation of local licensing itself. Then it will be compared with WiFi networkto cope with future traffic growth. Finally, a discussion of local licensing fromspectrum regulation perspective is provided.

15

Page 25: Quantitative Analysis on the Feasibility and Benefits of ...

Chapter 4

System Model

This chapter provides more detailed explanation on how the simulator was config-ured for this thesis project.

4.1 Network Deployment

We consider an in-building network deployment scenario in the simulator. Tworectangular areas with exact size were defined to mimic a single-floor building with, ascan be seen in figure 4.3 and 4.2. The size of each building was 150x300m. There weretwo different network from two different operator with no interference coordination. Eachnetwork had 400 users and 8 access points that covers 75x75m area. These 8 access pointscan be located in a single building (non-overlapping case) or spread on two building with4 APs each. We can see the difference more clearly in figure 4.2 and 4.3. In the figures,the first network’s access points (APs) is depicted with green square, whereas the starmarker represents APs from the second network. Each building had surrounding outerwall, but no inner wall.

Figure 4.1: Environment Model

Figure 4.3 shows the first network deployment type (overlapping deployment). Thisdeployment type was used in scenario 1 and 2. This deployment type demonstratesexisting real-world deployment, where multiple network can coexist in a common area.Figure 4.2 shows the second type (non-overlapping deployment) which simulated the locallicensing in scenario 3. Neutral host use a similar concept of non-overlapping deployment,where a venue owner can deploy the network inside their building to provide connectivityto the users inside.

16

Page 26: Quantitative Analysis on the Feasibility and Benefits of ...

Users of each network were randomly distributed throughout the building. Theheight of the user equipment (UE) was 1.5 meter and AP’s height was 3 metres. Eachbuilding had building entry loss of 12dB [28] due to the outer wall.

Figure 4.2: Two non-overlapping networks in the simulator

Figure 4.3: Two overlapping networks in the simulator

17

Page 27: Quantitative Analysis on the Feasibility and Benefits of ...

4.2 Radio Environment

4.2.1 Propagation Model

The ITU model for indoor attenuation [29] was used as the propagation model fora building or closed area that is enclosed by surrounding walls. The propagation modelscan be used for frequency 900MHz to 5.2 GHz. The building can have up to three floors.The site-general path loss model can be defined as

Ltotal = 20log10f + Nlog10d + Lf (n)− 28 (4.1)

where f is the carrier frequency (MHz), N is the path loss exponent, and d is the distancebetween Tx and Rx in metres. Two building-dependent factor are also included: floorpenetration loss factor Lf in dB and number of floors between Tx and Rx (n). Lf = 0dBfor n = 0.

4.2.2 SINR Model

The following model focuses on downlink case. A user communicates with an APwith received signal power as

Pr = GPt =cPtrα0

(4.2)

as written in [26], where G is the channel gain, Pt is the AP’s transmit power, and r isthe distance between user and AP. Let us consider simple exponential path loss modelfor the time being. As the medium is also used by several other transmitter-receiver pair,interference is unevitable. The signal-to-interference-plus-noise (SINR) ratio for user kcan be written as

Γ =Pr

I + N(4.3)

where N is noise and I is interference. The interference is a sum of all unwanted-but-received signals from other APs. Given that we have k APs that have similar transmitpower and are located within distance r1 to rk, we can define the SINR as

Γ =

cPtrα0∑

kcPtrαk

+ N(4.4)

4.3 System Parameters

Two type of RAT were used, WiFi and LTE. TD-LTE was used as WiFi is a TDD-based system. The WiFi version is 802.11n.

In terms of medium access control (MAC) protocol, the main differences betweenLTE and WiFi is the existence of listen-before-talk (LBT) protocol. While TD-LTE use atime slot allocation for each user, 802.11n use a contention-based Carrier Sense MultipleAccess with Collision Avoidance (CSMA/CA). Before transmitting any signal, a WiFi

18

Page 28: Quantitative Analysis on the Feasibility and Benefits of ...

Table 4.1: System parameters

WiFi LTE

PHY type 802.11n TD-LTECarrier Frequency 5 GHzBandwidth 20 MHzMacro Tx Power 23 dBmUE Tx Power 20 dBmCell reuse 1TDD Subframe config[DL:UL:Special]

6:3:1 n/a

CS threshold n/a -85 dBmED threshold n/a -65 dBmBeamforming n/aMIMO (DL/UL) 2x2 / 1x1

radio assess the medium to determine whether the channel is busy or not. This is knownas Clear Channel Assessment (CCA). CCA is divided into two parts, energy detect andcarrier sense. Energy detect CCA is ability of a 802.11 radio to detects any kind of signalwith power above the Energy Detect threshold, regardless whether the signal is a 802.11signal or not. In the other hand, carrier sense CCA can detect any 802.11 signal bydecoding incoming 802.11 frames. In this simulation, the carrier sense threshold was setto -85dBm while the energy detect threshold was set to -65dBm, which are a little bitlower than the reference threshold of -82 and -62dBm for 20MHz channel spacing [30].Practically, the ED was not used due to no other type of radio than 802.11 were usedwhile the WiFi network was deployed.

Other than the LBT mechanism RAT-specific parameters, most of the system pa-rameters on both RAT were kept as common as possible for all scenarios to limit un-wanted external variable. Both of them used the same 5GHz carrier frequency with20MHz bandwidth. 5GHz was chosen since we already have WiFi standard in 5GHzU-NII band. Thus, the same frequency was chosen for LTE system to achieve similarpropagation characteristic. More importantly, both network in each scenario used thesame carrier frequency. This is why the MAC protocol is very important. Downlinktransmission used 2x2 MIMO, while the uplink was a 1x1 SISO due to limitation in thesimulator. Transmit power was 23dBm in the access point and 28dBm on the UE. Theaccess point used an omnidirectional antenna while the UE used an isotropic one. Thesimulator used ITU indoor propagation model. This model is not entirely accurate inthis simulation since the spacing between bulding should be considered as a free spaceenvironment. Users get scheduled in round-robin fashion, means that all user get thesame amount of radio resource element. The throughput is calculated based on bitrateand offered traffic.

19

Page 29: Quantitative Analysis on the Feasibility and Benefits of ...

Chapter 5

Result and Analysis

We first recall the problem statement that we would like to analyze whether perfor-mance improvement in local licensed deployment is worthwhile compared to unlicensedspectrum. We would also like to understand the effect of deployment restriction and LBTon the results. The comparison between different deployment scenario using the sameRAT can be seen by comparing the red line (unlicensed case, WiFi in overlapping de-ployment) and the green line (intermediate case, WiFi in non-overlapping deployment).Whereas comparison of different MAC protocol can be seen between green line and theblue line (local licensed scenario, LTE in non-overlapping network).

The result figures are arranged in such a way that we see the result in system levelperformance first, then moving down to link level performance.

5.1 System Throughput vs Traffic Load

Figure 5.1: Median throughput vs served traffic

The median throughput graph in figure 5.1 shows the relation between the typicaluser throughput in relation to the increase of traffic load. Here we can see that LBT-

20

Page 30: Quantitative Analysis on the Feasibility and Benefits of ...

based unlicensed and intermediate case are the clear winner in low traffic as it achievednearly twice as much throughput as the non-LBT based local licensed case. This resultproves that contention-based LBT performs better in low traffic load [26]. In LBT-basedsystem, the system can transmit data as soon as the channel is clear, which is usuallyhappens in low load condition. In the other hand, LTE has static scheduling where alluser have been pre-assigned with constant TDD subframe configuration, which does notinfluenced by network utilization at a moment.

Comparing the unlicensed case and intermediate case both cases have roughly thesame amount of throughput when the traffic load is in its lowest. It means that in thisload, geographical separation does not have any significant impact.

As the traffic load increases, we see sharp decline in throughput of LBT-basedcases, whereas the local local licensed case declined less rapidly. The throughput of theLBT-based cases continued to decline rapidly until approximately 43 Mbps traffic load,when the value reached near zero and the decrease slowed down. This result proves thatLBT-based system can be saturated in high traffic load [26]. The DCF (DistributedCoordinated Function) in 802.11 protocol enforce the user to defer transmission if thechannel is busy in a given DIFS (DCF Interframe Space) period. As multiple terminalstry to communicate simultaneously with the AP, the channel will be busy all the time.It continues to have near zero value, but it actually never reach zero. It means that someincoming traffic still get to be served, even on very high traffic load.

To see the effect of deployment restriction, we compare the unlicensed case andintermediate case. As the traffic load increases, both cases have similar trend of rapid de-crease. It proves that the rapid decrease was caused by the LBT and the deployment typedid not have much effect on that. Intermediate case showed a slightly higher throughputas it added network separation on top of LBT. It has less hidden node problem.

Figure 5.2: Throughput vs served traffic on cell edge

The 5th percentile case in figure 5.2 shows the same result with its 50th percentilecounterpart in figure 5.1, except 5th percentile data can be perceived as cell edge case.Unlicensed case is a clear winner on low traffic case, but it degrades rapidly while the locallicensed case degrades more slowly. However, the value of the throughput is now around

21

Page 31: Quantitative Analysis on the Feasibility and Benefits of ...

half of the median result. In the lowest point of traffic load, LBT-based unlicensed casestays at around 35 Mbps, while other non LBT-based system stays at around 19 Mbps.

5.2 Capacity Comparison

Figure 5.3: System capacity at 5 Mbps and 10 Mbps required throughput

Figure 5.3 shows the capacity or offered traffic that the system can serve in arequired amount of throughput as QoS level requirement. Unlicensed case will be thebaseline reference for the comparison. In unlicensed case, the system capacity is 29 Mbpsfor 5 Mbps QoS and 23 Mbps for 10 Mbps QoS level. Adding the geographical separationin intermediate case gave a capacity gain of 10.3% and 10.8% in 5 and 10 Mbps QoS,respectively. The capacity gain is therefore not significant as we separating deploymentin LBT-based system.

On the rightmost bar, there is the local licensed case. It has a significant jumpfrom the intermediate case, with 150% more capacity in 5 Mbps QoS and 107% morein 10 Mbps. We can see that changing the MAC protocol from LBT-based to non-LBTsystem gives more significant impact than deployment restriction on LBT-based system.We proved that the usage of LBT hinder the network to achieve potential limit of systemcapacity. However, the geographical restriction is also important as changing the MAC tonon-LBT in overlapping deployment will cause severe internetwork interference as thereis no means of coexistence between different network, rendering the network unusable onhigh load.

Finally, we can see the most significant increase of capacity when switching fromunlicensed cased to local licensed case. In 5 Mbps QoS, the capacity increased by 175.8%.In higher QoS requirement of 10 Mbps, the capacity increased by 130.4%. While theimprovement in local licensed case compared to unlicensed case is obvious, we did notexpect the difference to be this big. In the QoS requirement that was used, the capacityof the system increased at least twice compared to unlicensed case.

22

Page 32: Quantitative Analysis on the Feasibility and Benefits of ...

5.3 Impact on Interference and SINR

Low traffic load

(a) (b)

Figure 5.4: Downlink interference and SINR CDF on low traffic load

Figure 5.4 presents the downlink interference CDF (a) and SINR CDF (b) fromnetwork 1 in low traffic load. In low traffic, we can see that the user in intermediatecase experienced the lowest median interference with -143.9 dB. The second lowest isunlicensed case with -127.9, then local licensed with -121.7. The low interference produceda much higher median SINR in both WiFi cases compared to the LTE case, or 26-31 dBhigher to be exact.

Comparing different MAC protocol in the same deployment type, we see that LBT-based network (intermediate case) has much less median interference than non-LBT sys-tem (local licensed case) with 22.2 dB difference. We see that in low traffic, LBT worksreally well in reducing intra-network interference. The inter-network interference itselfhas already been reduced by the non-overlapping deployment.

In the LBT-based WiFi system, separating both networks gave 16dB less medianinterference. The separation omitted hidden node problem. It occurs in overlappingnetwork where, for example, node fron network A is located near an AP from networkB. However, since node A can only be connected to AP from network A, the node Areceives strong interference from AP B. This phenomenon might be more visible in higherpercentile interference in both WiFi cases, in which it becomes higher than the LTE case.

One important thing to be noted is that both network 1 and 2 used the samechannel. In practice, inter-network coexistence can be realized easily by using differentfrequency channel to avoid interference [31]. However, it is crucial to remember thatspectrum sharing is all about a common spectrum that is shared among the users.

23

Page 33: Quantitative Analysis on the Feasibility and Benefits of ...

Medium traffic load

(a) (b)

Figure 5.5: Downlink interference and SINR CDF on medium traffic load

Moving on to second part of interference and SINR analysis, we have the result inmedium load as depicted in figure 5.5. Previously, we have figure 5.4 which show the samething but in low load. The difference of traffic load is intended to see the developmentand the trend of the result as the load increase.

From figure 5.5, we can see that now both unlicensed and intermediate case pro-duced approximately 13 dB higher median interference compared to local licensed case.It means that LBT-based network experienced more rapid increase compared to non-LBT based network. This result already gives the hint of the expected uncontrollableinterference that occurs in LBT-based network, but we still need to see the interferencelevel on high traffic load as well as its impact to throughput.

In both LBT-based cases, we can see that majority of the user experienced highlevel of interference, hence the skewed CDF plot. The lower percentile of LBT-basedcases stays equal or less than non-LBT local licensed case. This is the interference thatwas experienced by users located near to the AP they were connected to, which cause theinterference became less. In this percentile level, we can also see the effect of geographicalseparation. There were more users in intermediate case that experienced low level ofinterference, compared to the unlicensed case. The reason is a user in non-overlappingdeployment will have a better access to AP as all APs inside a building belongs to thesame network.

In contrast with the changes in interference, the median SINR in current traffic loadshows a similar trend as previous load, as intermediate case still has the highest medianSINR with 46.89 dB. Compared to first and third case, users in intermediate case receivedquite higher SINR due to lower interference.

SINR calculation between LTE and WiFi is different. In LTE, resource is allocatedby assigning certain resource blocks to certain user, which is not the case in WiFi. Thisfundamental differences makes the SINR calculation differs since the SINR per resourceblock that each user gets are different between user in LTE system and WiFi system.

24

Page 34: Quantitative Analysis on the Feasibility and Benefits of ...

High traffic load

(a) (b)

Figure 5.6: Downlink interference and SINR CDF on high traffic load

Moving to the testing under heavy traffic, the LBT-based unlicensed and interme-diate case still has higher median interference compared to local licensed. The medianinterference on unlicensed and intermediate case are -96.77 dB and -98.24 dB, whereasthe value is -108 in local licensed case.

However, the local licensed case experienced the highest increase in median inter-ference as the traffic increased from medium load. The difference is 6.9 dB, comparedto around 4.7 dB in LBT-based scenarios. It seems that the interference in LBT-basedsystem has nearly reach its upper limit, as higher interference than current value willcause the Clear Channel Assesment in CSMA system to define the channel as busy. Thebusy status will prevent the APs to transmit any signal. The CS threshold is -85 dBmand we can see the upper percentile is about to reach that level.

An LBT-based system may work better than LTE to prevent any further increasein interference, as the CS threshold detection is network-agnostic. It does not differenti-ate the source of interference, whether it is inter-network or intra-network interference.However, this comes with a cost. The prevention to transmit signal means that thethroughput can drop as more packets are being queued by the system while waiting thechannel to be free. The effect of this situation will be analyzed further in chapter 5.2,talking about bitrate and throughput.

5.4 Impact on User Bitrate and Throughput

Link level performance affects the maximum achievable bitrate in the user. Further-more, different traffic load affects the offered traffic to the scheduler which will ultimatelyaffects the experienced throughput on the user side, which will be analyzed in this section.

25

Page 35: Quantitative Analysis on the Feasibility and Benefits of ...

Low traffic load

(a) (b)

Figure 5.7: Downlink bitrate and throughput on low traffic load

We start by providing the downlink bitrate and throughput in low traffic load, asdepicted in figure 5.7. The figure shows that both LBT-based cases has a lot more medianachievable bitrate at around 71 Mbps, whereas the local licensed only has roughly halfof it. On lower percentile case, we can see that the intermediate case produced a slightlybetter bitrate. The bitrate is mostly influenced by the SINR, thus we can see that thetrend of bitrate CDF is more or less the same with SINR CDF on figure 5.4.

However, we observed an interesting phenomenon that bitrate of WiFi networkis peaked around 71 Mbps. Consequently, it also limits the maximum throughput atthe same value. Thus, 71 Mbps is the performance wall of the setup that was used insimulation.

Since the throughput on local licensed case are more or less the same with its bitratethroughout the percentile range, we can conclude that nearly all of the offered traffic gotserved in this traffic load. Users enjoyed optimal use of the available channel capacity.Different situation happens in the LBT-based cases, especially in the lower percentile.The throughput is much less than the bitrate.

In unlicensed case, the problem worsened due to hidden node problem as the de-ployment of the two networks overlaps each other. There are two different networks in thesame area. However if a node is located near AP from another network, it still can onlybe connected to AP on the same network, meaning that it will suffer stronger channelcontention that come from AP near it. It caused the throughput to drop.

Aside from the throughput drop in LBT-based cases compared to the bitrate, thethroughput is actually still better than the non-LBT case. It shows that LBT-basedsystem handle network traffic in low load condition better. With unnecessary signalingoverhead, WiFi network can use its resources optimally to provide connection to theusers. We also believed that this advantage was not caused by the higher SINR, asfigures 5.4, 5.5, and 5.6 consistently shows that WiFi network always have much betterSINR compared to its LTE counterpart.

26

Page 36: Quantitative Analysis on the Feasibility and Benefits of ...

Medium traffic load

(a) (b)

Figure 5.8: Downlink bitrate and throughput on medium traffic load

Moving on to the testing in medium traffic load, the bitrate graph in figure 5.8shows quite similar trend with the bitrate graph in low load. Intermediate case still holdsthe highest bitrate, followed by the unlicensed and local licensed case. Median bitratein unlicensed case, intermediate case, and the local licensed scenario are now 58.9, 64.2,and 33.56 Mbps respectively. Compared to bitrate in low traffic case, the values havedecreased by 11.74, 6.44, and 1.96 Mbps. Local licensed has the least drop in bitrate,meanwhile the other two shows quite significant decrease.

From the throughput graph, we can see notable difference compared to result in lowtraffic. On medium load, the median throughput of both LBT-based cased have droppedbelow the local licensed’s. The throughput have dropped by 54.16, 57.8, and 3.32 Mbpsin unlicensed, intermediate, and local licensed case respectively.

The big drop in LBT-based cases’ throughput caused their throughput to be lowerthan the local licensed’s, even though the bitrate are still higher due to good link levelperformance. It proves that the channel access probability in LBT-based network becomeworse as the traffic increase, which consequently reduce the throughput at user’s side.The contention-based scheduling mechanism is suboptimal in higher load case. Multipleterminals send offered traffic on the same time, but a large percentage of them is droppedor blocked.

27

Page 37: Quantitative Analysis on the Feasibility and Benefits of ...

High traffic load

(a) (b)

Figure 5.9: Downlink bitrate and throughput on high traffic load

The bitrate graph on figure 5.9 shows that both LBT-based cases produced highermedian bitrate than local licensed, following the same trend as the previous traffic load.The intermediate case now has the least drop in median bitrate with 4.8 Mbps difference.The median bitrate has decreased by 16.7 Mbps in unlicensed case and by 13.7 Mbpsin local licensed. The median throughput in both LBT-based cases have also droppedmuch further. We can see that throughput of these cases stays near zero, except for smallportion of users in 95 percentile region that still experienced acceptable throughput.

This graph proves further the analysis we have on medium load, which the higherbitrate on unlicensed case is still not enough to maintain a good throughput on hightraffic load.

5.5 Discussion on the Feasibility of Local Licensing

We have seen the definition and example of local licensing in chapter 2.3. We havealso validated the concept by simulation and comparison with WiFi based system. Thissection will discuss more about local licensing based on the simulation result and itsfuture potential. First, we recall the term local operator or micro operator as a facilityowner that have their own network inside their building.

Motivation

High quality indoor coverage has been an expected feature that should exist invirtually all public places, as well as enterprise building and vertical industries. Thequality of the connection inside can add value to the venue itself, which further attractsmore guest or tenants. Availability of good connection will attract people to the venueand it is of course good for the bussiness.

From MNO’s point-of-view, however, the better user experience sometimes does notjustify the cost of providing indoor coverage. With the emergence of small cell productsand Distributed Antenna System (DAS), deploying in-building network is now easier and

28

Page 38: Quantitative Analysis on the Feasibility and Benefits of ...

more possible than before. Still, if the infrastructure can be shared, then it will be awin-win solution for all parties, including the users, the venue owner, and the MNOs [32].

Such shared infrastructure is possible by deploying neutral host that facilitate allother MNOs in a building or venue [33] [15]. Multiple MNOs are connected to a neutralhost infrastructure that are aggregates the signal and relay it inside a building to providecoverage as shown in figure 5.10. The MNO and the neutral host owner (or the localoperator) can reach a joint agreement on how they can provide service using this model.

Figure 5.10: Neutral host model

If the neutral host and local operator model can be realized, the next question iswhich spectrum licensing regime would be the best choice for these kind of deployment?Unlicensed spectrum can be one choice, but the need to serve a large base of subscriberwith high QoS requirement make a licensed spectrum a better choice [34]. Local licensedspectrum is very attractive in this case, as it perfectly fits the local operator’s need tosecure a spectrum for its premise. As the spectrum is exclusive to the local operator,it can plan the deployment of indoor wireless network according to the needs, whileminimizing external uncontrollable interference to achieve best possible QoS for the user.In theory, each local operator should be able to obtain a wider band compared to dividedband if each incumbent MNOs deploy their own services with their own spectrum, thusproviding a better service to the users.

Another motivation of having an exclusive spectrum is to deploy a wireless networkfor critical use cases in specific verticals. In example, many predicts that billions of deviceswill be connected to the cloud in the future, and these includes connected healthcaredevices in hospitals. Having a very-reliable connection is an unevitable requirement andthese systems sometimes cannot afford to be interfered by external interference.

In comparison with WiFi

License-exempt band has and advantages of having no entry barrier for new en-trants. However it is also its biggest weakness, which is the aforementioned uncontrollableinterference.

29

Page 39: Quantitative Analysis on the Feasibility and Benefits of ...

Such assumption has been proven in this thesis. We have seen from the simulationresult that performance of LTE network in local licensed scenario is much better thanWiFi to cope with higher traffic load. I admit that the simulation used quite old WiFistandard of 802.11n, but so does the LTE. Newest technology such as 802.11ac might bebetter to cope with such traffic demand, but the limitation in LBT-based WiFi will stillremain as the carrier sensing threshold is fixed and mandatory as defined in the 802.11standard [5].

In short term, we can still rely on WiFi in 5GHz U-NII band. Currently the 5GHzband is underutilized, meanwhile the 2.4GHz ISM band is overly crowded in premisessuch as office or apartment. The interference in ISM band is also much worse becauseit is also used by other technology, such as bluetooth, cordless phone, microwave oven,or Zigbee devices. Also, adjacent channels are overlapping each other. We have to useonly channel 1, 6, and 11 if we want non-overlapping channels. In contrast, U-NII bandprovides more channel (23 channel compared to 13 channel in ISM band) and thosechannel are not overlapping each other. Figure 5.11 shows how the U-NII band dwarfsISM band in terms of available non-overlapping channel.

Figure 5.11: Comparison of available non-overlapping channelsin 2.4GHz ISM band and 5GHz U-NII band. Source: [35]

Given the aforementioned reason, major migration to 5GHz WiFi is a practical andtemporary solution to congestion in 2.4 GHz band. Yet, there is prediction that evenWiFi in 5GHz may not be enough to cope with future traffic demand [15]. Bianchi in [11]provides theoritical analysis on 802.11’s maximum saturation throughput. Kang in [12]went further to simulation of dense 5G WiFi deployment in LoS, shopping mall, and officeenvironment. Office environment has the highest wall density. He found that internal wallcan actually increase the overall system capacity due to worse propagation characteristic,which in turn reduce the interference and increase the spatial reuse. Using bandwith of480 MHz, which is the entire available bandwith in 5GHZ band, the throughput reachmaximum of approximately 5 Mbps/m2 in office environment. Citing the european officestandard by van Meel [36], we obtain average space per employee of 24m2. That willgive throughput of 120 Mbps per user. Of course this is a very rough estimation and theconditions on each figure might differ, but it should give a quick glance on the limita-tion of WiFi on user throughput even on 480 MHz bandwidth. Inside a more crowdedenvironment with less internal wall, the throughput per user will drop significantly.

From Spectrum Regulation Perspective

We have discussed the benefits of local licensing in indoor wireless network deploy-ment from technical point-of-view. However, local spectrum licensing does not exist as ofthe time of writing. It is still in its feasibility study and yet to be discussed officially bythe spectrum regulator. Defining a new spectrum licensing method is not an easy taskthat have to overcome many challenges as we will discuss on this section.

30

Page 40: Quantitative Analysis on the Feasibility and Benefits of ...

One of the challenges are spectrum availability to be used as local licensed spectrum.It is known that current RF spectrum is already fully utilized in various application. InSweden itself, PTS (Swedish Post and Telecom Authority) mentioned that there is almostno unused or unplanned frequency band in Sweden [37]. Therefore, we are left with twooptions. The first option is to free up currently utilized RF spectrum, which will be reallychallenging as the incumbent users will need to migrate to a new spectrum, and possiblyrequired to change the equipment they already using. One example is the migration fromanalog television to digital television. One of the motivation behind it was to clear upfrequency bands to be used in another purposes, as digital television use less broadcastspectrum than its analog counterpart. However the migration is a complicated processthat has to be prepared very carefully by the national regulator. TV-subscribers alsohave to change their receiver from analog to digital one. In some part of the world, suchas Russia, China, Indonesia, etc, the transition is still ongoing.

The second option is to reuse currently utilized frequency band in such a way thatthe new usage will not interfere with incumbent user. 3.5 GHz CBRS band in the USis on good example for this case. FCC did not halt the previous utilization of 3550-3700 MHz band completely. Instead, they created a Spectrum Access System (SAS) tocontrol the access to this frequency band, protecting the incumbent user from interferencecoming from the new users [7]. The benefits will be no spectrum clear-up procedure andmore efficient usage of radio spectrum. The challenge will be setting up a new system tomanage the usage of the spectrum in realtime and on-demand manner.

Another challenge is to define a complete specification of the new licensing methoditself. There are many things to be considered and standardized. From technical side, wehave the technical specification, such as the frequency band, transmit power, deploymenttype, etc. From non-technical side, socio-economic impact and the harmonization betweeninternational and national regulator also have to be studied. The process of managingand possibly introducing a new spectrum assignment can be seen in one example fromPTS in figure 5.12.

Figure 5.12: Process of managing spectrum. Source: [37]

Defining a new spectrum licensing method needs a good cooperation from all stake-

31

Page 41: Quantitative Analysis on the Feasibility and Benefits of ...

holders, including researchers, international and national regulator, as well as wirelessequipment vendor. Nevertheless, given the needs of smarter and more efficient spec-trum authorization method to cater future requirement of high speed, reliable in buildingconnectivity, all stakeholders must cooperate together to realize local licensing as futurecandidate for in-building connectivity.

32

Page 42: Quantitative Analysis on the Feasibility and Benefits of ...

Chapter 6

Conclusion and Future Work

6.1 Conclusion

We first recall the problem statement which questions whether local licensing isworthwhile in terms of QoS improvement compared to unlicensed spectrum in indoor usecase. Specifically, we also wanted to know the effect of LBT and deployment restriction.To answer those question, a simulation has been carried out to model indoor wirelessnetwork deployment in three different scenarios based on the usage of LBT and deploy-ment type. The networks were simulated in low, medium, and high traffic load to see thechanges in QoS as the traffic increase.

The usage of LBT was the most significant reason for the result above. We can seethat the median throughput in local licensed and intermediate case (both using non-LBTMAC protocol) does not have a rapid drop like unlicensed case as the traffic increase,which we prove that contention-based MAC protocol is suboptimal in high load.

Unlike the first and third case, the intermediate case did not have an inter-networkco-existence mechanism. Therefore, it cannot handle traffic load more than 42.26 Mbpsoffered traffic in this simulation model.

We found out that local licensing was better in overall than unlicensed spectrumin higher traffic load. The reason behind this is the limitation in LBT-based systemto utilize a spectrum efficiently in a certain geographical area. The throughput of locallicensing was less prone to increase of traffic load compared to unlicensed case. Capacitywise, the local licensed case has 180% higher capacity in 5 Mbps QoS, and 128% higherin 10 Mbps QoS. Therefore we come to a conclusion that local licensing have a promisingresult if it can be paired with a suitable spectrum management system. It should also ableto serve critical use case as the users experience the lowest interference level comparedto other cases.

6.2 Future work

These are the future work to study the concept of local licensing as this thesisdid not completely cover all aspect about it. First, the simulation model in this thesiswas not completely accurate because it only modelled single storey building with noinner walls. More accurate simulation model using multi-storey building with differentbuilding interior is needed. Second, a simulation using newer wireless standard (e.g. 5Gand 802.11ac) is also needed as the industry is moving towards newer standard for future

33

Page 43: Quantitative Analysis on the Feasibility and Benefits of ...

deployment of wireless network.Apart from technical simulation, we also need further socio-economic analysis of

local licensing. Security issues also need to be studied as MNO’s traffic will pass througha 3rd party if the neutral host model is used. Thorough analysis on bussiness model andsuitable spectrum management scheme is also needed.

34

Page 44: Quantitative Analysis on the Feasibility and Benefits of ...

References

[1] Harri Holma, Antti Toskala, and Jussi Reunanen. LTE Small Cell Optimization:3GPP Evolution to Release 13. January 2016. pp. 154.

[2] Cisco. Cisco Visual Networking Index: Forecast and Methodology, 2015 to 2020.Technical report, June 2016.

[3] Aleksei Ponomarenko-Timofeev, Alexander Pyattaev, Sergey Andreev, YevgeniKoucheryavy, Markus Mueck, and Ingolf Karls. Highly dynamic spectrum manage-ment within Licensed Shared Access regulatory framework. IEEE CommunicationsMagazine, 54(3):100–109, 2016.

[4] Markus Mueck, Wei Jiang, Guolin Sun, Hanwen Cao, Eryk Dutkiewicz, and Seung-won Choi. Novel spectrum usage paradigms for 5G. November 2014.

[5] Du Ho Kang. Interference Coordination for Low-cost Indoor Wireless Systems inShared Spectrum. PhD thesis, KTH Royal Institute of Technology, May 2014.

[6] Spectrum sharing committee public files. http://www.wirelessinnovation.org/

ssc-public-files. Accessed: 2017-02-27.

[7] Mark Lowenstein. Shared spectrum market opportunity for mobile network opera-tors. Technical report, October 2016.

[8] RSPG Opinion on Licensed Shared Access. Technical Report RSPG13-538, ECRSPG, November 2013.

[9] Alexander M Wyglinski, Maziar Nekovee, and Thomas Hou. Cognitive Radio Com-munications and Networks: Principles and Practice. Academic Press, 2009. pp.113.

[10] M. G. Kibria, G. P. Villardi, K. Nguyen, W. S. Liao, K. Ishizu, and F. Kojima.Shared spectrum access communications: A neutral host micro operator approach.IEEE Journal on Selected Areas in Communications, PP(99):1–1, 2017.

[11] G. Bianchi. Performance analysis of the IEEE 802.11 distributed coordination func-tion. IEEE Journal on Selected Areas in Communications, 18(3):535–547, March2000.

[12] D. H. Kang, K. W. Sung, and J. Zander. Attainable user throughput by densewi-fi deployment at 5 ghz. In 2013 IEEE 24th Annual International Symposium onPersonal, Indoor, and Mobile Radio Communications (PIMRC), pages 3418–3422,Sept 2013.

35

Page 45: Quantitative Analysis on the Feasibility and Benefits of ...

[13] James C Chen and Jeffrey M Gilbert. Measured performance of 5 GHz 802.11awireless LAN systems. Atheros communications, 8:27, 2001.

[14] Adrian Foster. GSR 2008 Discussion Paper: Spectrum Sharing. March 2008.

[15] D. H. Kang, K. W. Sung, and J. Zander. High capacity indoor and hotspot wirelesssystems in shared spectrum: a techno-economic analysis. IEEE CommunicationsMagazine, 51(12):102–109, December 2013.

[16] ITU. ITU Radio Regulation, Volume 1. 2004.

[17] IEEE Standard Letter Designations for Radar-Frequency Bands. IEEE Std 521-1976,pages 1–8, Nov 1976.

[18] Martin Cave, Chris Doyle, and William Webb. Essentials of Modern Spectrum Man-agement. Cambridge University Press, New York, NY, USA, 1st edition, 2007.

[19] Paul R Milgrom, Jonathan Levin, and Assaf Eilat. The case for unlicensed spectrum.2011.

[20] Gerald R Faulhaber. The question of spectrum: Technology, management, andregime change. J. on Telecomm. & High Tech. L., 4:123, 2005.

[21] Total wi-fi device shipments to surpass ten billion thismonth. http://www.wi-fi.org/news-events/newsroom/

total-wi-fi-device-shipments-to-surpass-ten-billion-this-month. Ac-cessed: 2017-02-24.

[22] Promoting the shared use of Europe’s radio spectrum. https://ec.europa.eu/

digital-single-market/en/promoting-shared-use-europes-radio-spectrum.Accessed: 2017-03-08.

[23] LTE License Assisted Access. Technical report, Ericsson, January 2015.

[24] Durga Malladi. Best use of unlicensed spectrum. Technical report, Qualcomm Tech-nologies, February 2016.

[25] White paper: Demystifying the Unlicensed LTE Conundrum and UnderstandingLTE-WiFi Coexistence. Technical report, Aricent, 2016.

[26] Guowang Miao, Jens Zander, Ki Won Sung, and Slimane Ben Slimane. Fundamentalsof Mobile Data Networks. Cambridge University Press, February 2016.

[27] Anne Hakansson. Portal of research methods and methodologies for research projectsand degree projects. In Proceedings of the International Conference on Frontiersin Education: Computer Science and Computer Engineering (FECS), page 1. TheSteering Committee of The World Congress in Computer Science, Computer Engi-neering and Applied Computing (WorldComp), 2013.

[28] ITU-R. 1411-7: Propagation data and prediction methods for the planning of short-range outdoor radio communication systems and radio local area networks in thefrequency range 300mhz to 100 ghz. ITU-R Recommendation, September 2013.

36

Page 46: Quantitative Analysis on the Feasibility and Benefits of ...

[29] ITU-R. 1238-8: Propagation data and prediction methods for the planning of short-range outdoor radio communication systems and radio local area networks in thefrequency range 300mhz to 100 ghz. ITU-R Recommendation, July 2015.

[30] IEEE Standard for Information technology–Telecommunications and information ex-change between systems Local and metropolitan area networks–Specific requirementsPart 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY)Specifications. IEEE Std 802.11-2012 (Revision of IEEE Std 802.11-2007), March2012.

[31] A. M. Voicu, L. Simic, and M. Petrova. Inter-technology coexistence in a spectrumcommons: A case study of wi-fi and lte in the 5-ghz unlicensed band. IEEE Journalon Selected Areas in Communications, 34(11):3062–3077, Nov 2016.

[32] Phillip Tracy. Moving in-building forward with shared infrastructure and neutralhost, November 2016.

[33] M. G. Kibria, G. P. Villardi, K. Nguyen, W. S. Liao, K. Ishizu, and F. Kojima.Shared spectrum access communications: A neutral host micro operator approach.IEEE Journal on Selected Areas in Communications, PP(99):1–1, 2017.

[34] Petri Ahokangas, Sara Moqaddamerad, Marja Matinmikko, Alhussein Abouzeid,Irina Atkova, Julius Francis Gomes, and Marika Iivari. Future micro operatorsbusiness models in 5g. The Business & Management Review, 7(5):143, 2016.

[35] Designing a dual-band wireless network. http://www.metageek.com/training/

/design-dual-band-wifi.html. Accessed: 2017-08-01.

[36] Juriaan Van Meel. The European office: office design and national context. 010Publishers, 2000.

[37] Post och Telestyrelsen. Swedish Spectrum Strategy. April 2014.

37

Page 47: Quantitative Analysis on the Feasibility and Benefits of ...

TRITA -ICT-EX-2017-186

www.kth.se