Data Download on the Move in Visible Light Communications: Design and Analysis

C585etukansi.kesken.fm Page 1 Tuesday, October 11, 2016 3:13 PM

C 585 OULU 2016 C 585

UNIVERSITY OF OULU P.O. Box 8000 FI-90014 UNIVERSITY OF OULU FINLAND ACTA UNIVERSITATISUNIVERSITATIS OULUENSISOULUENSIS ACTA UNIVERSITATIS OULUENSIS ACTAACTA

TECHNICATECHNICACC Helal Chowdhury Helal Chowdhury Professor Esa Hohtola DATA DOWNLOAD ON University Lecturer Santeri Palviainen THE MOVE IN VISIBLE LIGHT

Postdoctoral research fellow Sanna Taskila COMMUNICATIONS: DESIGN AND ANALYSIS Professor Olli Vuolteenaho

University Lecturer Veli-Matti Ulvinen

Director Sinikka Eskelinen

Professor Jari Juga

University Lecturer Anu Soikkeli

Professor Olli Vuolteenaho UNIVERSITY OF OULU GRADUATE SCHOOL; UNIVERSITY OF OULU, FACULTY OF TECHNOLOGY Publications Editor Kirsti Nurkkala

ISBN 978-952-62-1361-3 (Paperback) ISBN 978-952-62-1362-0 (PDF) ISSN 0355-3213 (Print) ISSN 1796-2226 (Online)

ACTA UNIVERSITATIS OULUENSIS C Technica 585

HELAL CHOWDHURY

DATA DOWNLOAD ON THE MOVE IN VISIBLE LIGHT COMMUNICATIONS: DESIGN AND ANALYSIS

Academic dissertation to be presented with the assent of the Doctoral Training Committee of Technology and Natural Sciences of the University of Oulu for public defence in the OP auditorium (L10), Linnanmaa, on 2 December 2016, at 12 noon

Supervised by Professor Marcos Katz

Reviewed by Professor Dominic O’Brien Professor Thomas D.C. Little

ISBN 978-952-62-1361-3 (Paperback) ISBN 978-952-62-1362-0 (PDF)

ISSN 0355-3213 (Printed) ISSN 1796-2226 (Online)

Cover Design Raimo Ahonen

JUVENES PRINT TAMPERE 2016 Chowdhury, Helal, Data download on the move in visible light communications: Design and analysis. University of Oulu Graduate School; University of Oulu, Faculty of Technology Acta Univ. Oul. C 585, 2016 University of Oulu, P.O. Box 8000, FI-90014 University of Oulu, Finland

Abstract In visible light communication (VLC), light emitting diodes (LEDs) are used as transmitters; the air is the transmission medium and the photodiodes are used for receivers. This is often referred to as light fidelity (Li-Fi). In this thesis, we provide the methodology to evaluate the performance of VLC hotspot networks in the context of data downloading on the move scenarios by using throughput-distance relationship models. In this context, first we study the different properties of optical transceiver elements, noise sources, characterization and modelling of artificial light interference, different link topologies and then we introduce the throughput-distance relationship model. Secondly, the analytically based throughput-distance relationship has been developed for evaluating the performance of VLC hotspot networks in indoor environment in both day and night conditions. Simulation results reveal that background noise has a significant impact on the performance of VLC hotspots. As expected, in both indoor and outdoor environments the VLC hotspot performs better at night than during day. The performance of VLC hotspot networks is also quantified in terms of received file size at different bit error rate requirements and velocities of the mobile user. Thirdly, we study the performance of hybrid (Radio-Optical) WLAN-VLC hotspot and compare its performance with stand-alone VLC-only or WLAN-only hotspot cases. In this case, we also consider the data download on the move scenarios in an indoor environment for a single- user as well as for multi-user cases. In this hybrid WLAN-VLC hotspot, both the WLAN and the VLC are characterized by their throughput and communication range. Simulations have been performed to evaluate the performance of such network for data downloading on the move scenario by taking into account performance metrics such as filesize, average connectivity and system throughput. Simulation results reveal that the considered hybrid WLAN-VLC performs always better than stand-alone VLC-only or WLAN-only hotspot both for a single and multi-user cases. Finally, this thesis analyses the feasibility and potential benefits of using hybrid radio-optical wireless systems. In this respect, cooperative communication using optical relays are also introduced in order to increase the coverage and energy efficiency of the battery operated device. Potential benefits are identified as service connectivity and energy efficiency of battery operated device in an indoor environment. Simulation results reveal that user connectivity and energy efficiency depend on user density, coverage range ratio between single-hop and multi-hop, relay probabilities and mobility of the user.

Keywords: optical hotspots, optical wireless communications, throughput-distance relationship model, visible light communications

Chowdhury, Helal, Tiedon lataus liikkeessä käyttäen näkyvään valoon pohjautuvaa tiedonsiirtoa: suunnittelu ja analyysi. Oulun yliopiston tutkijakoulu; Oulun yliopisto, Teknillinen tiedekunta Acta Univ. Oul. C 585, 2016 Oulun yliopisto, PL 8000, 90014 Oulun yliopisto

Tiivistelmä Näkyvään valoon pohjautuvassa tiedonsiirrossa (VLC) valodiodeja (LED) käytetään lähettiminä, ilma on siirtokanava ja valoilmaisimia käytetään vastaanottimina. Tätä kutsutaan usein nimellä light fidelity (Li-Fi). Tässä työssä tarjoamme menetelmiä VLC ”hotspot” verkkojen suorituskyvyn arviointiin tiedonsiirtonopeus-etäisyysmalleilla skenaarioissa, jossa tietoa ladataan liikkees- sä. Tässä kontekstissa tutkimme ensin optisen lähettimen komponenttien eri ominaisuuksia, kohinan lähteitä, keinovalon häiriömalleja ja tiedonsiirtolinkkien topologioita, jonka jälkeen esittelemme tiedonsiirtonopeuden ja etäisyyden välisen mallin. Toiseksi kehitetyn analyyttisen tiedonsiirto-etäisyys mallia käytetään arvioitaessa VLC hotspot verkkojen suorituskykyä sisäympäristössä sekä päivä että yö olosuhteissa. Simulointien tulokset osoittavat, että taustakohinalla on suuri vaikutus VLC verkkojen suorituskykyyn. Kuten odotettua, sisä- ja ulkotiloissa VLC hotspot toimii paremmin yöllä kuin päivällä. VLC hotspot verkkojen suorituskyky arvioidaan myös vastaanotetun tiedoston koon, eri bittivirhesuhteen vaa- timuksilla ja liikkuvan käyttäjän nopeuden suhteen. Kolmanneksi tutkimme hybridi WLAN-VLC hotspot verkon suorituskykyä ja vertaamme sen suorituskykyä pelkän VLC- tai WLAN hotspot tapauksessa. Käsittelemme myös skenaarioita jossa tiedoston lataus tapahtuu liikkeessä sisätilassa yhden käyttäjän sekä monen käyttäjän tapa- uksissa. Tässä hybridi WLAN-VLC hotspot, sekä erilliset WLAN- ja VLC verkot ovat määritel- ty niiden tiedonsiirtonopeuden ja kantaman perusteella. Näiden verkkojen suorituskykyä arvioitaessa on tehty joukko tietokonesimulointeja verkossa tapahtuvasta tietojen lataamisesta liik- keessä ottamalla huomioon suorituskyvyn mittarit kuten tiedoston koko, keskimääräinen yhtey- den kesto ja saavutettu läpäisy. Simuloinnin tulokset paljastavat, että hybridi WLAN-VLC toimii aina paremmin kuin pelkkä VLC tai WLAN hotspot sekä yhden että monen käyttäjän tapauksessa. Lopuksi työssä analysoidaan ehdotetun järjestelmän toteutettavuus ja mahdolliset edut käy- tettäessä hybridejä radio-optisia langattomia järjestelmiä. Tältä osin esitellään myös kooperatii- viseen viestintään perustuvat optiset releet parantamaan verkon kattavuutta ja energiatehokkuut- ta akkukäyttöisissä laitteissa. Mahdolliset hyödyt tunnistetaan palvelun konnektiivisuudessa ja energiatehokkuudessa akkukäyttöisissä laitteissa sisätiloissa. Simulointien tulokset osoittavat, että käyttäjien konnektiivisuus ja energiatehokkuus riippuvat käyttäjätiheydestä, kantaman ja etäisyyden välisestä suhteesta yhden hypyn ja monen hypyn välillä, releointi todennäköisyydes- tä ja käyttäjien mobiliteetista.

Asiasanat: optinen kuormittajat, optisen langattoman viestinnän, suoritusteho matkan suhde malli, valaistusdatansiirto

To my father & Suha, Basil and Fougia 8 Preface

I would first like to express my sincere gratitude to my thesis advisor Prof. Marcos Katz for his continuous guidance and support in my pursuit of Ph.D studies. I would also like to express my deepest gratitude to Prof. Kaveh Pahlavan who showed me the path of research and make me understand the karma of researcher. In addition, I want to thank Adj. Prof. Sastri Kota for his valuable suggestions and comments regarding the research in satellite communications. I want to thank Prof. Pentti Leppänen, Prof. Matti Latva-aho, Prof. Jari Iinatti for giving me the opportunity to pursue my doctoral studies in the department of communication engineering, University of Oulu, Finland.. I want also to thank the reviewers of this thesis, Prof. Dominico O’Brien from University of Oxford, England, and Professor Thomas Little from Boston University, Boston, USA for their valuable comments that helped to improve the final version of the thesis. It is my pleaser to thank many of my colleagues and co-author at CWC, who have been involved in the related work resulting with this thesis. Specially, I want to thank Prof. Ari Pouttu, Ijaz Ahmed, Kari Kärkkäinen, Babar Shahzad Chaudary, Hamidreza Bagheri, Dr. Janne Lehtomäki, Dr. Juha Pekka Mäkelä, Dr. Jaakko Huusko, Dr. Ani- mesh Yadav, and Dr. Pradeep Kumar. I would like to thank all administrative personnel of CWC. Specially, Jari Sillanpaa, Kirsi Ojutkangas, and Eija Pajunen for computer and official related problems. Finally, I would like to thank my family and friends for their continuous support and encouragement when needed. I would like to dedicate this thesis to my family - my father Aminul Haque Chowdhury, my daughter Suha, my son Basil, and to my wife Fougia Hoque for their patience and endless love.

9 10 List of abbreviations and symbols

4G 4th generation of mobile communication technology 5G 5th generation of mobile communication technology AIr advanced infrared ANSI American national standards institute AoA angle of arrival APD avalanche photodiode AWGN additive white gaussian noise BER bit error rate BS base station CENELEC committee for electrotechnical standardization CIE international commission on illumination CPC compound parabolic concentrator CSK color shift keying COST committee on science and technology dB decibel DC direct current DCO-OFDM DC biased optical orthogonal frequency division frequency DD direct detection DH-PIM dual header pulse interval modulation DMT discrete multi-tone DPIM digital pulse internal modulation DPPM differential pulse position modulation D2D device-to-device DSD dynamic spot-diffusing EMI electro-magnetic interference EMO European mobile observatory EU European Union EPA environmental protection agency FOV ﬁeld of view FSO free-space optical GHz giga Hertz

11 GaN gallium nitride HD high-definition HAN home access networks HetNet heterogeneous networks ICT information and communication technology IEC international electrotechnical commission IEEE institute of electrical and electronics engineers IFFT inverse fast fourier transform IM/DD intensity modulation/direct detection IR infra-red IRC infrared communication IrDA infrared data association ISI inter symbol interference ISM industrial, scientific and medical IoT internet of things JEITA Japan electronics and information technology industries association Li-Fi light fidelity LCD liquid crystal display LD laser diode LED light-emitting diode LOS line-of-sight LSD light shaping diffusers MAC medium access control MAI multiple access interface MIMO multi-input-multi-output MSD multi-spot diffusing MT mobile terminal M2M machine to machine NLOS non line-of-sight NRZ non-return-to-zero OFDM orthogonal frequency division multiplexing OMEGA home gigabit access OLED organic light emitting diode OOK on-off keying OW optical wireless

12 OWC optical wireless communication PAM pulse amplitude modulation PCM pulse coded modulation PHY physical layer PARR peak-to-average power ratio PDF probability density function PLC power line communication PSK phase shift keying PoE power-over-ethernet PoF plastic optical fiber PIN p-type-insulator-n-type PN pseudo-noise PPM pulse position modulation PPP public private partnership PSD power spectral density PTM pulse time modulation PWM pulse width modulation QAM quadrature amplitude modulation QoS quality of service RAT radio access technology RF radio frequency RZ return-to-zero SDR software define radio Si APD silicon avalanche photodiode Si PIN-PD silicon p-type-insulator-n-type photodiode SINR signal-to-interference-plus-noise ratio SISO single-input-single-output SNR signal-to-noise ratio SSL solid-state lighting TIA transimpedance amplifier TOV turn-on voltage TTA telecommunication technology association UV ultra-violet US United States UWB ultra wide band

13 VLAN visible local area network VLC visible light communication VLCA visible light communication association VLCC visible light communication consortium VPPM variable pulse position modulation WDM wavelength division multiplexing WLAN wireless local area network WLED white light emitting diode WPAN wireless personal area network WWRF wireless world research forum

α1 entrance angle

α2 exit angle λ wavelength

λscale scaling factor Ω spatial angle Φ luminous ﬂux

Φmax full angle of LED ϕ angle of irradiance

ϕen energy ﬂux

Ψ 1 half power semi angle 2 θ travelling angle

Ad area of photodetector

A f effective collection area of the PD B receiver bandwidth D distance between transmitter and receiver

Dt travelling distance dhor horizontal separation between transmitter and receiver dvlc total travelling distance in VLC coverage dwlan total travelling distance in WLAN coverage g(Ψ) concentrator gain

H0 channel DC gain

HB average steady background irradiance

Hreﬂec gain from reﬂected path h vertical separation between transmitter and receiver iLED current through the LED

14 imax maximum current iD diode current io saturation current

IB total DC photocurrent

Inatural DC current produce by natural light

Iartiﬁcial DC current produce by artiﬁcial light I luminous intensity I(0) luminous intensity at center

Ihybrid ﬁle size in WLAN-VLC

KB Boltzmann’s constant

Kmax maximum visibility k knee factor kr refractive index

Km maximum visibility

Kscale scaling factor L minimum distance

L0 pathloss at the ﬁrst meter m Lambertian order mopt optimal Lambertian order

N0 Gaussian noise nd diode ideality factor nﬂoor number of ﬂoors N number of users n(t) signal independent shot noise

Pmax maximum power

Pn average power of ambient light

Pr received optical power

Pt transmitted power

Ptotal total received power q electron charge Q(x) Q function r horizontal distance R radius of VLC coverage rsr radius of single-hop coverage

Rmr radius of multi-hop coverage

15 Rb data rate

Ri radiant intensity

Rr photodiode responsivity

R0(ϕ) Lambertian pattern

RF feedback resistance S(r) spatial throughput

Svlc VLC throughput

Swlan WLAN throughput

Tk( f ) absolute temparature

Tsg signal transmission of the ﬁlter tt residence time in VLC and WLAN coverage v velocity of the mobile user V(λ) standard luminosity curve

VT ( f ) thermal voltage vin power ampliﬁer input voltage vout power ampliﬁer output voltage vmax maximum output voltage vLED voltage through LED xi(t) instantaneous optical power

16 Contents

Abstract Tiivistelmä Preface 9 List of abbreviations and symbols 11 Contents 17 1 Introduction 21 1.1 Background ...... 21 1.2 Visible light communications ...... 23 1.3 State-of-the-art and related work ...... 26 1.3.1 Data rate improvement in VLC testbed ...... 27 1.3.2 Hybrid radio-optical wireless communications ...... 29 1.3.3 Motivation ...... 30 1.3.4 Thesis contributions ...... 31 1.3.5 Author’s contributions and thesis outline ...... 32 2 Optical transceiver, noise and interference 35 2.1 Optical transmitters in VLC ...... 35 2.1.1 Generation of white light with LEDs ...... 37 2.1.2 Radiometry and photometry ...... 38 2.1.3 LED I-V characteristics ...... 41 2.2 Wireless optical receiver ...... 43 2.2.1 Elements of a photodetector ...... 44 2.3 Noise sources in wireless optical communications ...... 45 2.3.1 Shot noise ...... 45 2.3.2 Thermal noise...... 46 2.3.3 Measurement-based noise characterisation ...... 46 2.4 Basics building blocks of optical transceiver ...... 47 2.5 Modulations in VLC ...... 49 3 Link characterisation in VLC 53 3.1 Basics link types of VLC ...... 53 3.1.1 Directed LOS link design ...... 56 3.1.2 Non-directed link design ...... 58

17 3.2 Measurement based link design ...... 60 4 Data downloading on VLC coverage 63 4.1 Motivation and related work ...... 63 4.2 Geometry of VLC hotspot coverage...... 65 4.3 Transformation of inclined path to horizontal path ...... 66 4.4 Mathematical framework for a data downloading scenario...... 67 4.5 VLC hotspot design parameters and their relationships...... 70 4.6 Theoretical and empirical throughput-distance models ...... 71 4.6.1 Empirical polynomial based throughput-distance model ...... 72 4.6.2 Throughput vs. distance at daytime...... 72 4.6.3 Throughput vs. distance at nighttime ...... 73 4.6.4 Calculation of average throughput and ﬁle size ...... 73 4.7 Numerical results ...... 74 4.8 Conclusions ...... 76 5 Data downloading in hybrid WLAN-VLC networks 79 5.1 Hybrid WLAN-VLC networks in indoor environment ...... 79 5.1.1 Coverage and data rate of VLC small cell ...... 80 5.1.2 Coverage and data rate of WLAN ...... 81 5.2 Hybrid WLAN-VLC: single-user case ...... 83 5.3 Hybrid WLAN-VLC: multi-user case ...... 86 5.4 Performance evaluation of hybrid WLAN-VLC ...... 88 5.5 Conclusions ...... 94 6 Co-operative relays in VLC and hybrid WLAN-VLC networks 97 6.1 Overview and background ...... 97 6.2 Scenario description...... 99 6.2.1 Relay selection...... 100 6.3 Performance evaluation of co-operative VLC ...... 101 6.3.1 Co-operative hybrid WLAN-VLC networks ...... 103 6.3.2 Energy consumption analysis ...... 105 6.4 Performance evaluation of hybrid WLAN-VLC networks ...... 106 6.5 Conclusions ...... 108 7 Summary and future directions 111 7.1 Summary ...... 111 7.2 Future work ...... 114

18 References 115

19 20 1 Introduction

It is difﬁcult to make predictions, especially about the future.

Attributed to Niels Bohr

1.1 Background

Double-digit annual growth rates of network traffic in all network segments is expected to increase remarkably over the coming years and beyond. Recently, we have also wit- nessed a remarkable increase in the use of smart phones. The wireless world research forum (WWRF) predicts that 7 trillion wireless devices will serve 7 billion people by 2020. This prediction reveals that the number of network-connected wireless devices will reach 1000 times the world’s population by 2020 [1]. The paradigm of communications will also expand from ’human-human’ to ’human-thing’, and ’thing-thing’ (also called machine to machine (M2M)) [2]. The Internet will become the Internet of Things (IoT) [3]. In IoT, every object or thing will be connected virtually and the Internet will be the basic infrastructure for supporting connections of these interconnected objects. Therefore, IoT will bring a massive surge of smart connected devices enabling new services and business across industries. As more and more devices go wireless, the substantial growth of tele-traffic and therefore the need for more spectrum usage is also increased tremendously. Hence, this high volume of tele-traffic is clearly leading to a greater thirst for spectrum to support wireless broadband [4]. Radio frequency (RF) in the range of (1-2) giga Hertz (GHz) proven for best propagation conditions is already congested. Therefore, this spectrum scarcity referred to as the spectrum crunch has to be tackled by appropriate countermeasures in future wireless communication systems [1]. Recently, the green communications campaign has gained momentum in a global consensus to reduce the temperature increase due to carbon dioxide (CO2) emissions.

The European Commission (EC) has proposed to cut CO2 emissions by 40% by 2030. As the volume of data trafﬁc increases, the carbon footprint of using mobile networks is also increased [5]. In 2011, 22% of the carbon footprint represented by communication

21 networks is reported and expected to almost double by 2020 unless underlying network technologies are significantly improved [5–7]. It is envisioned that future high data rate wireless services such as 5thgeneration (5G) will be based on exploiting multiple wireless access technologies [8]. Therefore, 5G enabled terminals are also expected to be multi-standard supporting wireless devices. These multi-standard wireless devices will be more complex and power-hungry signal processing hardware. This will raise a significant threat to using mobile devices for continuous service connectivity due to the requirement of high power consumption of transmission, reception and other processing hardware [9]. In this case, the battery power of mobile devices will drain rapidly and mobile users will be relentlessly search- ing for power outlets rather than network access, which will bind the mobile user to a single location. This is known as the "energy trap" [4, 10]. The world mobile and wireless infrastructure community have taken initiatives to meet the above-mentioned grand challenges. As a result, many important technical, regulatory, economical, and social issues are considered in designing future 5G wireless communication systems [9, 11–14]. The standardisation of 5G networks is expected to be finalised around 2018 and the technology will start to be deployed around 2020 [7]. Massive multi-input-multi-output (MIMO), ultra dense networks, moving networks, device-to-device (D2D), ultra reliable, and massive machine communications are considered to be the key components of 5G wireless networks. It is envisioned that 5G networks will achieve 1000 times the system capacity, 10 times the spectral and energy efficiency. 5G wireless networks are also expected to provide peak data rate of 10 giga bits per second (Gbps) for low mobility and peak data rate of 1 Gbps for high mobility [11, 15]. Lighting is a major source of electric energy consumption. It is estimated that 19% of all electricity is used for lighting [16]. Hence, the development of more energy efficient lighting sources is important. Future light emitting didode (LED) lighting not only achieves significant energy savings, but also carbon footprint reductions [5]. As a result, a significant activity toward the development of solid state sources such as white LEDs (WLEDs) have already been started to replace incandescent and fluorescent lights. In Europe, most countries have already started to ban the use of incandescent and fluorescent lights [17]. The phase-out will deliver considerable savings to the environment and the economy. In this respect, energy efficient solid state based lighting such as LED will bring revolutionary advances in the use of light for illumination. It is also

22 expected that LED lighting will surpass the current lighting options in the near future in terms of both cost and efficiency. A ten dollar 60 Watt LED light is already on the market and the expected price of this LED light will be halved by 2020 [17, 18]. Thus, it is expected that the reduction of cost and the improvement of performance of LED in terms of lumen per Watt will also continue in near future. It is predicted that the lumen per Watt of LEDs will reach 200 around 2025 [17]. In addition to lighting capabilities, WLEDs can also be used to provide wireless communications by modulating the light sources with data, at a rate much faster than the response time of the human eye [19, 20]. Thus, it is foreseen that light transceivers will be built in LED light fixtures. LEDs can easily serve as sources for even very high data rate applications. In this case, multiple wired and wireless backbone technologies such as the broadband power line communication (PLC) (IEEE 1901, ITU-T G.9960/61), 60 GHz millimeter (mm) wave and the recently proposed low-cost plastic optical fibre (POF)-based backbone technologies can be used to build VLC network architecture [21–24]. This will extend the indoor range from a single room (radius 3-10 m) to a large building with hundreds of rooms. As a result, illumination and communication can be achieved with a single platform. The dual use of LEDs for illumination and communication promises a sustainable and energy-efficient approach and has the potential to revolutionise how we use light at present. VLC can be possibly used in a wide range of applications including wireless local area networks (WLAN), wireless personal area networks (WPAN) and vehicular networks among others [25–29]. Wireless communication using the visible light spectrum is known as visible light communication (VLC). Optical wireless communication (OWC) enables wireless connectivity using infra-red (IR), visible or ultraviolet bands. VLC is a subset of OWC [30, 31].

1.2 Visible light communications

In VLC, the visible part of the spectrum is used for communication purposes. The visible light spectrum is unlimited and 10,000 times larger than the range of radio frequencies between 0 Hz to 30 GHz as shown in Figure 1 [32]. In VLC, the WLEDs source are used as transmitters; the air is the transmission medium, and the positive intrinsic-negative (PIN) or more sensitive avalanche photodiode (APD) are used for receivers. Communication network builds upon WLEDs and photodiode is often referred as light ﬁdelity (‘Li-Fi’) [33]. VLC provides access to several hundred tera Hertz (THz)

23 of unlicensed spectrum. One of the key merits of optical wireless systems is the relatively low transceiver complexity and low energy consumption of LEDs [5]. As a result, VLC may facilitate the low energy-per-bit required for data transmission in comparison to RF systems. The visible light spectrum extends from 380 nm to 780 nm in wavelength. The VLC spectrum creates no electromagnetic interference to RF systems and vice versa.

Frequency [Hz] 0 3x10 10 3x10 11 4x10 14 7.9x10 14 3x10 16 3x10 19

Micro- Infra- Ultra- Gamma Radio X-rays waves red violet Rays

X ~ 10,000

Fig. 1. The electromagnetic spectrum.

There are already several standards associated with visible light communications [34–36]. In 2003, the visible light communication consortium (VLCC) established in Japan. In 2007, Japan electronics and information technology industries association (JEITA) established standards and VLCC introduced speciﬁcation standards in 2008 [37]. CP-1221 and CP-1222 are the standards for visible light ID and visible light beacon systems respectively [34, 35]. The visible light ID system can be used for applications such as location-based services and digital signage [17]. Another activity is under the umbrella of the telecommunications technology association (TTA) that supports the standardization of VLC in Korea and worldwide. The VLC working group in TTA started in May 2007. In 2011, VLC was standardized and published as the IEEE 802.15.7 standard [36]. The IEEE 802.15.7 standard promises to be a very attractive candidate as a future high data rate and power-efﬁcient technology in 5G. Re- cently, revision of the IEEE 802.15.7-2011 standard for VLC is prepared and denoted as 802.15.7r1. European committee on science and technology (COST) 1101 research network OPTICWISE group was active and contributing to this standard. OPTICWISE is

24 an associate member of the European 5G public private partnership (PPP). The revision addresses low data rate communication based on imaging sensors [38] In the IEEE 802.15.7 standard, both physical (PHY) and medium access layer (MAC) are defined for short-range optical wireless communications using visible light spectrum. The standard is capable of delivering 5G compatible data rates which is sufficient to support audio and video multimedia services and also considers mobility of the visible link [36]. The concept of physical and logical mobility is discussed in [36]. This standard supports the device discovery mechanism through which short range co-operation can also be possible within homogeneous networks with three different classes of devices: infrastructure, mobile, and vehicle [37]. Recently, the successor to the VLCC another visible light communication association (VLCA) established in May 2014 in Japan to further research, develop, plan and standardise advanced visible light communications systems [39]. There may arise multiple applications of such technology in future. It belongs to the green technologies category when used for lighting purposes, becoming even more environmentally friendly as it supports communication functionalities compared to RF alternatives. VLC can be used both for indoor as well as outdoor communications [40]. The VLC market categorizes both for slow and high data rate applications [41–44]. However, it does not imply that VLC is the universal replacement for short-range based RF technologies such as wireless local area networks (WLAN), Zigbee, ultra wide band (UWB), 6LoWPAN and Bluetooth. The application of optical wireless (OW) systems is limited when considering coverage area and user mobility where RF technologies prove invaluable. Moreover, communication with VLC is challenging due to occlusion, which can occur if severe misalignments and the presence of physical obstructions between sources and detectors are present in indoor as well as outdoor environments. Vis- ible light waves predominantly follow line-of-sight (LOS) propagation. Hence, VLC is more vulnerable than RF in terms of reliable connectivity, especially when mobile users move. However, among the key merits of VLC are that it exhibits several ap- pealing attributes when compared to RF [6]. The comparison between radio and optical wireless communications (VLC is a subset of OWC ) is summarised in Table 1. Co-operation between radio frequency and VLC systems may also be envisioned to build hybrid radio-optical communication systems for higher throughput, better coverage, and higher energy efficiency in future heterogeneous networks (HetNets) solutions [4].

25 Table 1. Comparison between RF and optical wireless based systems.

Characteristics Radio Optical Spectral availability Limited Abundant Spectral regulation Highly regulated No regulation Electromagnetic interference created Possible None Sensitive electromagnetic interference Possible None Dominant noise source Other users Natural and artiﬁcial light Quality of service Best effort Best effort

1.3 State-of-the-art and related work

The first optical wireless communication was developed and named as photophone by Alexander Graham Bell in 1880. In this photophone experiment, he modulated sunlight with voice signal and transmitted it over a distance of around 200 m. A vibrating mirror for sending and a parabolic mirror with a selenium cell at its focal point were used to receive the signal. The photophone did not work very well due to severe interference and the use of ordinary receiver. However, in 1960s the discovery of optical sources such as the laser diode (LD) changed the fortune of OWC. Short-range optical IR based wireless communication is discussed in [45, 46]. The first IR-based local-area networks using diffuse links and their advantages and drawbacks are compared in [45]. An experimental pulse coded modulation (PCM) link operating at 125 kbps and phase shift keying (PSK) operating at 64 kbps were established in this work. Communica- tion range were realized up to 50 m. Performance of an experimental 50 mega bits per second (Mbps) on-off-keyed diffuse IR link is also described in [46]. In July 1990, the IEEE 802.11 standard project was started for the specification of WLAN for different technologies including radio and IR [47]. Due to the lack of interest from potential vendors, IR-based WLAN could not penetrate into the market. However, after the invention of high brightness gallium nitride (GaN) LED by Nakamura in 1993, LED technology is maturing into the low-cost, energy-efficient lighting of the future. LEDs make them also attractive candidates for use in short to medium range data links supporting Mbps rates. Pioneer work to transmit data using LEDs in the visible light spectrum began in 2003 at the Nakagawa laboratory in Keio University, Japan [48–50]. In [48], the basic properties of LED lights, illumination using LED lighting and design of WLED light are discussed. Several research papers on VLC [51–53], which are related to [48–50] have been published also to investigate the performance of VLC. The research work in

26 [51] propose a simple method to equalise the WLED. Simulation results indicate that a simple equaliser can double the maximum data transmission rate compared with the unequalised channel. This paper has shown that the maximum 32 Mbps with non-return zero on-off-keying (NRZ-OOK) at a bit error rate (BER) of 10−6 is achievable. In [52], several challenges and possibilities in VLC are discussed. Complex modulation and parallel communication (optical MIMO) are suggested to increase the data rate in [54, 55]. Provision of uplink and regulatory challenges are also discussed in these papers. Sev- eral hardware demonstrations work on data rate improvement and hybrid radio-optical wireless communication system are discussed in the following two subsections.

1.3.1 Data rate improvement in VLC testbed

Numerous papers have been published on increasing the data rate in VLC systems especially for point-to-point link configuration [56–63]. In most of these works, hardware demonstrations were carried out utilizing different modulation schemes, blue filtering as well as pre and post equalizers to improve the data rates. For example, in [56], an experimental demonstration had been conducted building technology-independent MAC layer and PLC for services and connectivity to any number of devices in any room of a house and apartment. This work was part of the European union (EU) funded home gigabit access (OMEGA) project to investigate optical-wireless communications [64]. Data rate were achieved up to 73 Mbps in their demonstration using VLC. The prototype consists of two parts: digital signal processing and an analogue part. In their experimental setup, digital signal processing was implemented on a Vertex-5 field pro- gramming gate array (FPGA) board where serial input was mapped to a 16-quadrature amplitude modulation (16-QAM) symbol stream. On the analogue front end, the transmitter consisted of driving circuit, trans-conductance amplifier, and commercial Osram (OSTAR E3B) high-power LEDs. On the receiver side, imaging optics, a colour filter, a photodiode, a two stage trans-impedance amplifier, band-pass filter were used. The data rate of 80 Mbps using pre-equalized white LED was demonstrated in [57]. In this work, a pre-equlisation technique had been applied to extend the modulation bandwidth. The equalised bandwidth was 45 MHz. A single Luxeon white LED was used as VLC transmitter. The signal was pre-equalised by a driver and combined with the direct current (DC) signal via a bias-tee. The DC current was set to 200 mA. On the receiver side, a blue filter, concentration lens, PIN type photodetector and low noise transimpedance amplifier were used. In [58], 125 Mbps over 5 m distance in an indoor

27 environment by using on-off keying (OOK) was reported. In the experimental setup, a lamp with six chips that provides luminous flux of about 400 lm was considered. Moreover, a blue filter was applied to suppress the phosphorescent component of the white light. A large area silicon PIN diode with effective area 100 mm2 and polymer lens with 70o field of view (FOV) were used to receive the signal. In [59], OOK modulation and without equalization, 230 Mbps and 125 Mbps data rate were achieved using APD and a PIN diode respectively. A typical commercially available LED module consists of four chips providing luminous flux of 250 lm was used as transmitter. A blue filter with a cut-off wavelength of 500 nm was used in front of the photodiode. The diameter of 3 mm of APD and effective area of PIN photodiode were considered in the receiver side. With a PIN based receiver, 125 Mbps was achieved at 1000 lux. On the other hand, a data rate of 230 Mbps was achieved at 1000 lux due to the enhanced sensitivity of the APD. The authors of [59] further claimed 513 Mbps point-to-point visible light communication link and reported in [60] using discrete multitone modulation (DMT) modulation. Further improvement in data rates was continued by the same group and achieved 803 Mbps and reported in [61]. In this demonstration, wavelength division multiplexing (WDM) VLC link with DMT modulation was used for red-green-blue (RGB) LEDs. On the receiver side, commercially available APD of 3 mm diameter combined with a glass lens of 8 mm focal length was used for detection. The gross transmission rate was about 293.7 Mbps, 223.4 Mbps and 286 Mbps for the red, green and blue channels respectively which led to an aggregate data rate of about 803 Mbps. Data rate records in Gbps for VLC had reported in [62, 63, 65]. In [62], the first Gbps VLC link based on RGB LEDs had demonstrated. In this work, QAM on DMT was used in the WDM link. DMT signals consisted of 128 subcarriers within the baseband bandwidth of 100 MHz. Large area silicon APD of 3 mm diameter combined with a glass lens of 20 mm diameter and 20 mm focal length were used for detection. The detector was followed by a low-impedance amplifier to amplify the signal level up to the operation range. The transmission performance measurements were performed with and without the presence of crosstalk. In the case of without crosstalk, the gross transmission rate were about 444 Mbps, 518 Mbps and 589 Mbps for the red, green and blue channels respectively, which led to an aggregate data rate of about 1.55 Gbps. On the other hand, in the case of with crosstalk the gross transmission rate were about 376 Mbps, 439 Mbps and 430 Mbps for the red, green and blue channel respectively, which led to aggregate data rate of about 1.25 Gbps. 1 Gbps for single channel and

28 2.1 Gbps WDM transmission at usual illumination levels were also reported in [63]. A cellular structure using an angle-diversity conﬁgurations had been implemented offer- ing reasonably wide coverage for indoor communications was reported in [19, 66, 67]. Their work was designed for home access networks (HAN). HAN consists of VLC to deliver data in downlink and infra-red communication (IRC) for uplink to provide Gbps bi-directional communications. Recently, 100 Gbps with wide FOV optical wireless communications has been reported in [68]. In this work, the authors described an indoor optical bi-directional wireless link with an aggregate capacity over 224 Gbps operating at around 3 m range with a FOV of 60o. This type of wide FOV will offer practical room-scale coverage for wireless communication.

1.3.2 Hybrid radio-optical wireless communications

Numerous studies also have discussed the use of VLC and the RF systems as complementary technology [6, 69–75]. The main goal of those studies was to investigate and exploit the heterogeneous nature of such technologies and ﬁnd the beneﬁt from them being used in combination. In [69], the authors developed a seamless communication method between radio and optical communication systems where the hybrid (Wi-Fi)-IR communication system continued to use RF transmission when the communication path of the optical wireless transmission was obstructed and switched back to optical wireless transmission as soon as the optical wireless connection had become stable again. The purpose of using this hybrid radio-optical architecture was to use high speed and secure communication via optical communication whenever it was available. In [70], different types vertical handover (VHO) schemes between radio and optical communications are proposed: immediate VHO (I-VHO), dwell VHO (D-VHO) and a fuzzy logic-based vertical handover for an integrated (Wi-Fi)-IR system. In these schemes, VHO decision-making algorithm determines whether a VHO should be performed and when. The purpose of the work was to trigger an optimal handover decision between radio and optical communication systems for providing a better quality of service (QoS) to the users. In [71], authors had studied and compared the radio-optical based sensor network life time with the RF-only based wireless sensor network. Results showed that the radio-optical based sensor network lifetime lasts at least twice as long as its RF-only counterpart because of low energy consumption by optical transmission in hybrid radio- optical systems.

29 Studies in [72] introduced a hybrid scheme in VLC where spotlighting hotspot cells for high data rate and uniform lighting for moderate data rate and relatively wide coverage. They found that spotlighting VLC had several benefits over uniform lighting implementations in terms of the achievable high data rate. The authors in [73] investigated the criteria for vertical handover in a hybrid WLAN-VLC system which was comprised of broadcast VLC hotspots and Wi-Fi connectivity with a goal to avoid service disconnections and to optimally distribute available resources among users. In [6, 74], authors analysed the feasibility and potential benefits of using hybrid radio-optical wireless systems where co-operative communication using optical relays were also introduced in order to increase the coverage and energy efficiency of battery operated devices. In [75], turbo coded data transmission over a hybrid channel consist- ing of parallel free space optical and radio frequency links were considered. Recently, the demonstration of practical indoor hybrid WLAN-VLC Internet access systems have been proposed in [76]. In their demonstration, VLC is used as a supplementary downlink channel along with conventional Wi-Fi connectivity. The main component of this demonstration were: GNU Radio software-defined radio toolkit with universal software radio peripheral (USRP) developed by Ettus research and its parent company [77]. Osram semiconductor LEDs (LUW CN5M) and a commercial photodiode with transimpedence amplifier (PDA36A) were used as transmitter and receiver respectively. They compared the performance between Wi-Fi and hybrid VLC in terms of Website loading time and average download data rate. They had noticed that as the number of users increased, the hybrid WLAN-VLC performance was better than Wi-Fi only.

1.3.3 Motivation

Recently, research in VLC is gaining momentum in many directions. Most of the research works focus on data rate improvement in the testbed system as discussed in Sec- tion 1.3.1. Optical MIMO and OFDM technologies have also recently been introduced in VLC to improve data rate and coverage [78–80]. Research on hybrid architecture of WLAN-VLC has also been proposed and reported as discussed in Section 1.3.2. How- ever, very little has been published so far about the deployment of VLC wireless infrastructures [4, 10]. Although it is very early to speculate on the infrastructure deployment procedure of VLC, it is important to target a speciﬁc network topology and evaluate the performance in terms of achievable data rates in some typical usage scenarios.

30 As it is mentioned, illumination and communication can be provided within a single platform. Therefore, the deployment of such VLC systems has to be jointly optimised with illumination [81]. However, to get the benefits of both illumination and communication in a single platform, it leads to a lot of challenges in validating the applicability of such a communication system [82]. For example, VLC performance depends on many natural events such as day and night events, where natural as well as artificial light sources produce noise and interference. In addition to that, VLC is challenging due to occlusion, which can occur due to severe misalignments and the presence of physical obstructions between the transmitter and receiver. Therefore, convincing conclusions for deployment can only be drawn by evaluating the performance of such networks. The obtained results and proposed methodologies of this feasibility study will then provide fundamental information to design and deployment related issues of VLC hotspots. In this respect, this thesis will focus on the coverage and data rate design of future VLC and hybrid WLAN-VLC networks considering many important essential design parameters related to VLC and WLAN hotspots in day and night environments. However, many research questions have to be left unanswered in order to keep this work at a manageable size. For example, connection set-up time, vertical handover algorithm development in hybrid WLAN-VLC are not considered in this thesis. In this respect, we sufficiently narrowed the goals for this thesis and focused on designing and evaluating the performance of different VLC and hybrid WLAN-VLC hotspot networks with some valid assumptions. We summarize the thesis contribution in a concise problem formulation in the next section.

1.3.4 Thesis contributions

The main contribution of this thesis is to develop the framework for designing and the comparative performance analysis of VLC and hybrid WLAN-VLC hotspot networks. In this respect, we consider three different types of hotspots network topologies: standalone VLC, hybrid WLAN-VLC, and co-operative relay-assisted hybrid WLAN-VLC hotspot networks. In all cases, data download on the move scenario is considered in performance evaluation both in indoor and outdoor environments. Another contribution of the thesis is to present a mathematical framework to calculate the average throughput on data download on the move scenario. In this mathematical framework, ﬁrst we develop the theoretical throughput-distance relationship for VLC and then this relationship is approximated to two empirical throughput-distance models to calculate the downloaded

31 file size while crossing the VLC or hybrid WLAN-VLC hotspot coverage. In case of WLAN, measurement based empirical throughput-distance relationship is used. In this thesis, we provide two types of mobility models. In the first mobility model, the main assumption is that the single mobile user passes through the coverage of the VLC hotspot through a straight inclined path. In this respect, we introduce a simple but novel rotational technique where equivalent straight horizontal paths can be derived using the entry and exit angles of the mobile user. The purpose of this rotation of the inclined path to horizontal is to simplify the calculation of average throughput and file size. Secondly, the random waypoint mobility model is used in hybrid WLAN-VLC and co-operative hybrid WLAN-VLC scenarios in case of multi-user cases. The obtained results and proposed methodologies provide fundamental information to design and deployment related issues of standalone VLC and hybrid VLC-WLAN hotspot networks.

1.3.5 Author’s contributions and thesis outline

This thesis is based on three journal papers [10, 83, 84] and seven conference papers [4, 6, 9, 74, 85–87]. The author of the thesis has had the main responsibility in developing the original ideas, developing the mathematical framework, analysing the performance and writing all of these papers. Other co-authors provided constructive criticism during the writing process of these papers. The research work has been performed under the supervision of Prof. Marcos Katz who also provided invaluable comments and support throughout the writing of this thesis. In addition, the author of the thesis has published a number of supplementary papers which is related to this thesis [88, 89]. The author of the thesis is also the main author of all these papers. Research work [88, 89] has been carried out under the supervision of Prof. Kaveh Pahlavan. Research work in [10] is performed under the supervision of Adj. Prof. Sastri Kota. This thesis is composed of ﬁve technical chapters. The main contributions of this thesis are addressed in Chapter 4 to 6. In the following, the content of the chapters is outlined brieﬂy. In Chapter 2, the elements of solid state lighting (SSL) optical transceiver devices such as LD, LED, PD and their properties, noise and interference characteristics in VLC networks are discussed. The generation of white light with LEDs, conversion of radio to optical properties and vice versa are addressed in this chapter. There are various noise sources present in the wireless optical links. A brief introduction to all these noise sources is addressed in this chapter. Finally, the characterization and modelling

32 of artiﬁcial light interference, basic and advanced modulation techniques used in VLC are also discussed in this chapter. In Chapter 3, four basic link topologies are described. Classiﬁcation of these link characteristics are done mainly according to the degree of directionality and the presence of uninterrupted LOS between the transmitter and receiver. The basic characteristics of these link topologies are described in detail. Different types of diffuse links such as the multi-spot diffuse link and dynamic multi-spot diffuse links are also discussed along with four basic link topologies. Finally, link design of LOS and non-line-of-sight (NLOS) are also derived and discussed in details in this chapter. In Chapter 4, LOS link and noise characteristics of optical wireless system in an indoor and outdoor environment are discussed. Theoretical and empirical throughput- distance relationship development are presented in details in this chapter. The mathematical frame work for data downloading scenario is also presented in this chapter. Simulation results and discussion are reported in Section 4.7. Finally, conclusions are drawn in Section 4.8. In Chapter 5, we study the performance of the hybrid (radio-optical) WLAN-VLC hotspot and compare its performance with stand-alone VLC-only or WLAN-only hotspot cases. We consider the data download on the move scenario in an indoor environment for the single-user as well as for the multi-user cases. Throughput-distance relationships are developed both for WLAN and VLC to evaluate the performance of an indoor environment by taking into account the radio and optical channel characteristics. Simu- lation results and discussion are reported in Section 5.4. Finally, conclusions are drawn in Section 5.5. In Chapter 6, the scenario description of hybrid WLAN-VLC networks where relay selection, mobility and energy consumption are discussed. Numerical examples based on chosen system model parameters of WLAN-VLC are also presented in this chapter. Finally, in Chapter 7, we summarize and conclude the dissertation and some open problems for future research are addressed.

33 34 2 Optical transceiver, noise and interference

The aim of this chapter is to introduce the fundamental building blocks of VLC transceiver systems. Various types of noise and interference present in VLC receivers and optical channels are also discussed in this chapter. It is predicted that LED will be used in future lighting systems because of its energy efﬁciency and reliability compared to existing lighting solutions [90–92]. In VLC, white and coloured LEDs are used as transmitters; the air is the optical transmission medium, and the appropriate photodiodes or image sensors are used as signal receiving components. Apart from LEDs, other SSL devices such as LD and organic LED (OLED) can be used as sources of data transmission [93– 95]. Generation of white light with LEDs, conversion of radio to optical properties and vice versa are addressed in this chapter. Different noise sources and their impact on the performance on VLC networks are also addressed. Finally, basic and advanced modulation techniques used in VLC are discussed in this chapter.

2.1 Optical transmitters in VLC

Many existing conventional lighting sources such as incandescent, fluorescent lights and future SSL sources LD, LED, OLED can be used as optical sources for transmission of data in VLC. The intensity of these optical sources can be modulated to transmit data [26]. SSL devices such as LD/LED can be flickered faster than incandescent and fluorescent lights. In this respect, SSL based lighting sources are better than other conventional lighting sources in higher data rate applications. For example, using fluorescent light, data can be transmitted up to tens of kilo bits per second. On the other hand, SSL devices such as LD/LED can be flickered fast enough to achieve a data rate up to several Gbps [19, 96, 97]. Moreover, reliability, stability, low power consumption, low cost and easy controllability of SSL sources such as LD/LED put them well ahead of conventional lighting sources and they are envisioned as future transmission source for VLC [98]. LEDs/LD can be fabricated to emit light across a wide range of wavelengths from the visible to the infra-red parts of the electromagnetic spectrum [31]. In VLC, the role of all these optical transmitters is to convert an electrical input signal into the corresponding optical signal [31]. In many cases, LDs are preferable over LEDs because of their high data rate transmission capabilities and higher optical power outputs [31]. Data rate in the order of

35 Table 2. Comparison between LED and LD.

Characteristics LED LD Optical output power Low power High power Optical spectral width (25 − 100) nm (0.01 − 5) nm Modulation bandwidth Tens of kHz to hundreds of MHz Tens of kHz to tens of GHz E/O conversion efﬁciency (10 − 20)% (30 − 70)% Eye safety Considered eye safe Must be rendered eye safe Directionality Beam is broader and spreading Beam is directional Reliability High Moderate Source type Lambertian Point

Gbps can be achieved using high powered LDs as transmission source [99]. However, high powered LDs are potentially dangerous to eye safety especially in indoor environment, which may deliver very high power within a small area on the retina thereby resulting in permanent blindness. There are a few international standard bodies such as international electrotechnical commission (IEC), American national standards institute (ANSI), and committee for electrotechnical standardization international (CENELEC) which provide safety guidelines for both LD and LED emission level in terms of maximum permissible exposer (MPE) values [99]. The MPE is the highest power or energy density in W/cm2 or J/cm2 of a light source that is considered safe and represents negli- gible probability for creating damage to the eye. Moreover, LD is not suitable for using as lighting source for illumination. Therefore, LD is not preferable for wide coverage of hotspot networks as so-called visible local area network (VLAN). In indoor VLAN types of network, wide-angle radiation patterns of white and col- ored LEDs may be used as transmitters. On the other hand, LEDs are the preferred, as a light source for most future indoor as well as outdoor lighting applications. These LEDs are available in both visible and IR wavelengths. IR LED wavelengths range between (830-940) nm. On the other hand, visible LED wavelengths fall into the region between (400-700) nm, where a variety of colors including red, yellow, orange, amber, green, blue are available [100]. However, the closer the wavelength is to the visible part of the spectrum, the safer that wavelength is for the eyes. The comparison between the characteristics of LD and LED is summarized and shown in Table 2. Recently usage of another SSL device such as OLED is gaining momentum in soft lighting and display applications [94]. Great picture quality, brilliant colors, lower power consumption; faster refresh rate, better contrast, and greater brightness of OLED put it well ahead of LCD [94, 95]. As such, they are expected to serve in the next generation of full color displays and ﬂat panel lighting [94]. An OLED is made from

36 carbon-based organic materials that emit light when electricity is applied. It can also be easily be adapted for lighting and data communication in a number of applications [41–43, 90]. However, OLED offers much lower bandwidth than non-organic LEDs. The frequency response of OLED is much slower than conventional LED [94]. The bandwidth of the OLEDs depends also on the other factors such as manufacturing process, material and size of the device. According to the recent development of OLED, a moderate data rate can be achieved using a small photoactive area of OLED. A data rate of 2.5 Mbps is achieved using an OLED with low bandwidth reported in [94]. The Luminous efﬁcacy of typical OLED is (40-60) lm/W. However, new materials research in organic devices has already started to increase the high luminous efﬁcacy up to 120 lm/W, which will really satisfy the future need of illumination and moderate data rate applications for VLC communication systems [101–103]. Aluminium Gallium Nitride (AlGaN) based micro-light emitting diode arrays have also been developed for display panels. The range of each pixel of AlGaN is in the range of (14-84) µm [104]. A high data rate in the order of Gbps can be achieved incorporating parallel communications of Micro-LEDs. Micro-LED design and performance are discussed in detail in [105]. There are also other higher bandwidth LED such as resonant-cavity LED (RCLEDs) and edge-emitting LED. RCLEDs provide data rate up-to 1 Gbps. On the other hand, edge-emitting LED provides data rate over hundred of Mbps. However, the edge-emitting LED has better output intensity than the RCLED [106].

2.1.1 Generation of white light with LEDs

The general purpose of white LEDs is to provide white lighting for illumination, while white light emitting diodes are the most common optical sources proposed for transmission of data in VLC. There are mainly two approaches for generating white light using LEDs.

37 White Light White Light

Red Blue Green Yellow phosphor Blue LED

(a) (b) Fig. 2. White light: (a) Trichromatic based white LED. (b) Phosphor based white LED.

In the ﬁrst approach, three primary colors such as red, green, blue are mixed to produce white light, where the wavelength of red, green, and blue are 625 nm, 525 nm, and 470 nm respectively. In this case, these three color emitters consist of a single package with combining optics [32]. These three primary colors can also produce other arbitrary colors such as yellow, orange, amber by modulating the drive currents in each individual LED of trichromatic LEDs. Hence, this kind of LED enables easy color rendering by adjusting each color independently. The production of white light using trichromatic LEDs is shown in Figure 2(a). Trichromatic LEDs are preferred over phosphorescent LEDs due to their faster rise-time and each color can be modulated independently which contributes to tripling the total throughput [62, 87]. In the second approach, white light is generated by the use of phosphor together with a short-wavelength of blue or ultra violet (UV) light. This second approach is called phosphor-converted white light [80]. In this case, phosphor material used in LEDs is illuminated by blue light. Some of the blue light will be converted to yellow light by the phosphor. The remaining blue light, when mixed with the yellow light, will produce white light. Commercial phosphor-based LEDs have limited bandwidth which is typically (2-3) MHz [107]. This is due to the slow response of the phosphor. However, bandwidth can be increased up to 20 MHz by using blue ﬁlter to suppress the slow phosphorescent components at the receiver. The production of phosphor-converted white light using phosphor material is shown in Figure 2(b).

2.1.2 Radiometry and photometry

Electromagnetic radiation within the frequency range of (3×1011 −3×1016) Hz is iden- tiﬁed as optical radiation [108]. Radiometry involves the measurement of this optical

38 Table 3. Radiometry parameters.

Quantity Terminology Units Power radiant power ﬂux Watt (W) Power per unit area Irradiance W/m2 Power per unit solid angle radiant intensity W/sr Power per area per solid angle radiance W/(sr.m2)

Table 4. Photometry parameters.

Quantity Terminology Units Power luminous ﬂux lm Power per unit area illuminance W/m2 Power per unit solid angle luminous intensity candela (cd) (= lm/sr) Power per area per solid angle luminance cd/m2

radiation. The wavelength of the ultraviolet, the visible and the infra-red fall into this frequency range. However, in radiometry, this measurement does not take into account the specific sensitivity of the human eye, rather it represents the light transmitting capability of a light source for the wide spectrum. Typical radiometric units include Watt, irradiance, radiant intensity and radiance as shown in Table 3. On the other hand, photometry examines only the radiation that humans can see. Thus, photometric parameters take only account into the visible band with different weights for various wavelengths [100]. Typical photometric units include lumen, illuminance, luminous intensity and luminance as shown in Table 4. VLC is an opto-electrical wireless communication system where conversion between radiometry to photometry or vice versa is necessary. In order to convert radiometric values into photometric values, and vice versa, the relative visibility of the light of the particular wavelength should be taken into account, which is represented as the eye sensitivity curve [31]. The eye sensitivity curve is also called the luminous efficiency curve and it is the ratio of any photometric unit to its radiometric equivalent unit. The most common unit in radiometry is the Watt, which measures radiant flux (power), while the most common unit in photometry is the lumen, which measures luminous flux. The relationship between radiometric and photometric can be expressed as ( ) lm Photometricunit[lx] = radiometricunit[Watt] × 683 ×V(λ), (1) W where V(λ) is the standard luminosity curve. Hence, for monochromatic light of 555 nm, radiometric 1 Watt is equivalent to photometric 683 lumens. For light at other

760

800 750 740

730 750 720

710 700 700 Illuminance [lux] 690 650 4 680 2 4 0 2 670 0 −2 −2 660 −4 Y[m] −4 X[m]

Fig. 3. Illuminance distribution. wavelengths, the conversion between Watts and lumens is slightly different, because the human eye responds differently to different wavelengths. Two basic properties of an LED are luminous intensity and transmitted optical power. Luminous intensity is used to express the brightness of LED. On the other hand, transmitted optical power is used to indicates the total energy radiated from an LED. Mathematically, luminous intensity can be expressed as luminous flux per solid angle and is given as [31] dΦ I = , (2) dΩ where Φ is the luminous flux and Ω is the spatial angle. Φ can be calculated from the ϕ energy flux ef as ∫ 780 Φ = Kmax V(λ)ϕef (λ)dλ, (3) 380 where Kmax is the maximum visibility, which is about 683 lm/W at 555 nm wavelength.

The integral of the energy ﬂux ϕef in all directions is the transmitted optical power Pt given as [31] ∫ ∫ Λmax 2π Pt = Km ϕef dθdλ, (4) Λmin 0 where Λmin and Λmax are determined from the photodiode sensitivity curve. For an LED lighting with a Lambertian radiation pattern, the radiation intensity at the receiving

40 plane is given as I(ϕ) = I(0)cosm(ϕ), (5) where ϕ is the angle of irradiance and I(0) is the center of luminous intensity. m is the order of Lambertian emission. The illuminance distribution with a semi-angle 60o is shown in Figure 3. According to the international lighting standard organization, full illumination with a mean level of 350 lux is needed in both ofﬁce and home environments.

2.1.3 LED I-V characteristics

In VLC, the real value of baseband signal is modulated onto the instantaneous power of the optical carrier. In this case, a modulation signal is linearly encoded using a current source driver. However, one of the disadvantageous properties of LEDs is their nonlinear relationship between current and voltage. The Shockley Equation models the diode current iD as a function of the diode voltage VD as

VD n V iD = i0(e dif T − 1), (6) where iD, VD, and i0 are the diode current, diode voltage, and saturation current respectively. ndif is known as the diode ideality factor and VT is the thermal voltage. The relationship between LED I-V characteristics is shown in Figure 4. In every LED, minimum current is required for photon emission to occur. This is known as conduction region and there is a certain threshold value of voltage at which it starts to show the linear relationship between diode voltage and diode current. This threshold value is known as the turn-on voltage (TOV) [80]. However, above of TOV, current and voltage of diode will not follow the linearity in LED I-V characteristics. Therefore, the range between minimum and maximum allowed current where the relationship between current and light output is linear can be considered as the operating region for signal modulation. Currently, the question of combating non-linearity is one of the biggest challenges for VLC systems specially for OFDM signal [109–111]. Therefore, one approach to combat non-linearity is to operate the LED in a small range (between minimum and maximum of TOV) where its output characteristic is linear enough [111].

41 0.012

0.01

0.008

Operating region 0.006

0.004 Forward current [A]

0.002

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Forward voltage [V]

Fig. 4. LED I-V characteristics.

In [80, 109], measured LED I-V is approximated with the 6th degree polynomial where the top and bottom of the curve are saturated in a so-called S-shaped curve. In their work, they had performed analysis of OFDM signal using the model of 6th degree ﬁtted polynomial function of LED I-V curve. In a radio frequency systems, a commonly used model to describe the non-linearity behaviour of power ampliﬁer (PA) is Rapps model. The Rapps model can be described as [112]

V ( in) Vout = k , (7) 1 + ( Vin 2 )(1/2k) Vmax where Vout is the PA output voltage, Vin is the PA input voltage, Vmax is the maximum output voltage and k is called the knee factor that controls the smoothness of the transition from the linear to the saturation region. In [80, 109], Rapps model has been modiﬁed to describe the increase the linearity of I-V characteristics of LED. The modiﬁed LED model behavior is described as follows   ≥ hvLED i f vLED 0 iLED =  (8) 0 i f vLED < 0, where iLED and VLED are the current through the LED and the voltage across the LED respectively. The transfer function hvLED describes the dependence of the emitted optical

0.5

0.45

0.4

0.35

0.3

0.25

LED(A) 0.2

0.15

0.1 k=40 0.05 k=3 k=2 0 0 0.2 0.4 0.6 0.8 1 LED(V) Fig. 5. The non-linear modeling. power on the driving current. It is seen in Figure 5 that the saturation part of LED I-V characteristics is controlled by knee factor k and it is also seen that as the value of k increases, the linearity of I-V characteristics also increases. The modiﬁed Rapps model for LED can be written as

f (v ) V = ( LED ) , out / (9) 1 + ( f (vLED) )2k 1 2k imax where imax is the maximum permissible current.

2.2 Wireless optical receiver

In a wireless optical receiver, a photodiode is used to detect light and convert it into electrical signal. Two main components of photo detector circuits are the photo diode and resistor. There are mainly two types of basic detectors: the PIN and the APD. The PIN photodetector performs better at longer optical wavelengths, while APD performs better at shorter wavelengths [113]. The image sensor used in a cell phone camera or digital camera can also be used as an optical receiver. The image sensor used in these devices is basically an array of photodiodes. Some advantages of an image sensor receiving system over the conventional photo diode receiving system are parallel receptors and robustness against interfering light sources. Image sensors used for digital cameras or video cameras usually have a

43 Fig. 6. Basic elements of optical receiver. frame rate of tens of frames per second. The automotive applications using image sensor for vehicle-to-infrastructure-VLC (V2I-VLC) and vehicle-to-vehicle-VLC (V2V-VLC) are presented in [114].

2.2.1 Elements of a photodetector

The performance of most OW systems is determined by the receiver rather than transmitter. The VLC receiver is composed of optical collection elements such as a concentrator, optical filter, photodiode and amplifier, as shown in Figure 6. Light enters the receiver through the concentrator. The optical concentrator is used to compensate for high spatial attenuation due to the beam divergence from the LEDs to illuminate a large area. By using the appropriate concentrator, the effective collection area can be increased. For the optical collection element, the aim is to maximize both the FOV as well as the collection area. In VLC, both natural daylight and artificial illumination sources act as noise sources, most of the energy from the sun is in the visible and infra-red spectrum [46, 115]. There- fore, it is important to employ both electrical as well as appropriate optical filter to reject unwanted DC noise components in the recovered data signal. Scattered sunlight, which

44 concentrates around zero Hz, can be easily removed by an electrical filter [108, 116]. On the other hand, optical filters are used to reduce the background interference from artificial illumination. The optical filter is usually used to pass only the relatively narrow band radiation of the transmitter, while rejecting the (optically) broadband of unwanted light falling outside the useful band. Photodetectors convert the optical power to a current proportional to it [31]. This current is usually converted to a voltage with the aid of a transimpedance amplifier, and then further amplified before data recovery. The photodiodes with good responsivity to visible light are Si PIN-PD (silicon p-type-insulator-n-type photodiode) and Si APD (silicon avalanche photodiode). The silicon material photodiode operates from (400- 1200) nm. There are also many other photodiodes whose bandwidths are over 200 MHz and is much wider than the VLC LED transmitter [31].

2.3 Noise sources in wireless optical communications

There are various noise sources present in the wireless optical links. Sources of this noise are in the channel as well as generated locally in the receiver. The noise generated in VLC are shot noise, optical excess noise, photodetector dark current noise, photodetector excess noise and thermal noise [117]. However, in this work we only consider shot noise produced by natural and artificial noise sources such as sun, fluorescent and incandescent light sources. The basic definitions of shot and thermal noise are given below:

2.3.1 Shot noise

Shot noise is considered to be the dominant noise sources in wireless optical communications [46]. The origin of shot noise is due to the presence of both ambient light and transmitted signal. Shot noise is modeled as a Poisson distribution with a white power spectral density due to the discrete random nature of energy and charge in the photodiode. Mathematically, it can be expressed as [31, 46]

σ 2 shot = 2qRrPn, (10) where q is the electron charge. Rr and Pn are the photodiode responsivity and the average power of ambient light respectively.

45 2.3.2 Thermal noise

In conducting materials, noise is generated due to the random motion of electrons in resistive and active devices. This random motion of electrons which gives rise to noise voltage is called thermal noise [31]. A large number of free electrons and ions are responsible for generating this noise, which is bounded by molecular forces in the conductor. Hence, there is a continuous transfer of energy between the ions and electrons. This is the source of resistance in the conductor. Thermal noise is also a function of temperature. Thermal noise is generated independently of the received signal and can be modeled as

π2 ( ) σ 2 4kBTk 16 kBTk Γ 2 3 thermal = I2Rb + Cd +Cg I3RB. (11) RF gm

The ﬁrst term represents thermal noise generated in feedback resistor. kB and Tk are the

Boltzmann’s constant and absolute temperature respectively. RF is the feedback resistance. Second term describes the thermal noise from the ﬁeld-effect transistor (FET) channel resistance where Γ is the FET channel noise factor, gm is the FET transin- ductance, Cd is the capacitance of a detector. Cg is the FET gate capacitance, and

I3 = 0.0868. Sum of contributions from the shot and the thermal noises can be written as

σ 2 σ 2 σ 2 total = shot + thermal. (12)

2.3.3 Measurement-based noise characterisation

In [115, 118], the shot noise produced by natural and artificial light sources is characterised as DC current. The characterisation of noise and interference was done through extensive measurements. Measurements had been performed: with and without a filter. In the case of a filter, a long-pass absorption optical filter with a cut-off wavelength 800 nm was used. In those measurements, sunlight and artificial lights had been considered as sources of noise and interference. For example, direct and indirect sunlight used as ambient noise source in the receiver. On the other hand, incandescent lamps with tungsten filament, fluorescent lamps with conventional and electronic ballasts used as artificial noise sources. Irradiance produced by sunlight is steady and has a slow variation of intensity in comparison to the irradiance produced by artificial lights.

46 Table 5. Background current in day and night conditions.

Source Without optical ﬁlter (µA) With optical ﬁlter (µA) Direct sunlight 5100 1000 Indirect sunlight 740 190 Incandescent light 84 56 Fluorescent light 40 2

Shot noise power is directly proportional to that current [115, 118]. This induced current can be easily included in the system models to account for the shot noise produced by the background light. Induced noise current also depends on the effective area and type of photodiode used in the optical receiver. In [115, 118], the experiment was performed by measuring the current induced on a 0.85 cm2 Silicon PIN photodiode. The measured shot noise in day and night conditions is shown in Table 5 [115, 118].

Background current Ib can be expressed as [115, 118]

Ib = Inatural + Iartificial, (13) where Inatural and Iartificial are the DC currents produced by natural and artificial light sources respectively. Background noise generated by different noise sources, are listed in Table 5 [115, 118]. The shot noise is independent of the signal and can be modeled as a white Gaussian N0 given by [119]

N0 = 2qIb. (14)

2.4 Basics building blocks of optical transceiver

The equivalent baseband model of an IM/DD optical wireless link can be summarized by the following ⊗ y(t) = Rr xi(t) h(t) + n(t), (15) ⊗ where is the convolution operation. xi(t), h(t), and n(t) represent the instantaneous optical power, baseband channel impulse response, and signal independent shot noise respectively. In optical wireless communications, the instantaneous optical power xi(t) is proportional to the generated electrical current, hence xi(t) represents the power than

47 x(t) Rh(t) y(t)

n(t) Fig. 7. Equivalent optical wireless baseband model.

the amplitude of the signal. This imposes two constraints such as the xi(t) should be nonnegative and maximum optical transmit power should not exceed eye safety requirement limit. According to the ﬁrst constraint, xi(t) should be nonnegative can be written as

xi(t) ≥ 0. (16)

Secondly, according to the second constraint, the average value of xi(t) must not exceed a speciﬁed maximum power value Pmax, that is given by ∫ 1 T Pmax = lim xi(t)dt. (17) T→∞ 2T −T

The impulse response h(t) is generally used to analyze effects of multipath dispersion in indoor optical wireless channels. The channel impulse response can be modeled as [31]   2t t 0 t ≤ t ≤ 0 t3 sin2(FOV) 0 cos(FOV) h(t) =  (18) 0 elsewhere, where t0 is the minimum delay. The channel DC gain can be expressed as [31]

∫ ∞ H(0) = h(t)dt. (19) −∞

In general, signal-to-noise ratio (SNR) is used to express the quality of performance in communication systems. In such case, the performance of wireless optical link at the bit rate Rb can be related to SNR as [46]

2 2 2 Rr H (0)Pt SNR = . (20) RbN0

48 We assume that the transmitter transmits the signal using an OOK modulation technique. The BER of the OOK can be expressed as [46] √ BER = Q( SNR), (21) and the function Q(x) is deﬁned as [46]

∫ ∞ − 2 1 ( y ) Q(x) = √ e 2 dy. (22) 2π x

2.5 Modulations in VLC

In VLC, the information is carried by the intensity (power) of the light. As a result, the information-carrying signal has to be real valued and strictly positive [80]. There are also other issues such as dimming support and flickering that have to taken into account when designing modulation schemes in VLC [37]. Dimming support is needed for power saving and energy efficiency. The purpose of dimming support is to main- tain communication while the user arbitrarily dims the light source. On the other hand, flickers refer to the fluctuations of the brightness of light which can cause noticeable negative and harmful physiological changes in humans. To avoid flicker, the changes in brightness must fall within the maximum flickering time period. Optimal flicker frequency greater than 200 Hz is generally considered safe. Therefore, designing modulation techniques for VLC dimming support and flickering issues should also be taken into account in such way that both illumination and communication are well supported [120–125]. There are number of methods which can be used to modulate the data over the visible light spectrum. The most common methods are: OOK, pulse width modulation (PWM), pulse position modulation (PPM), variable pulse position modulation (VPPM), color shift keying (CSK), and orthogonal frequency division multiplexing (OFDM) [36, 37, 126–130]. In OOK modulation, LEDs are turned on or off dependent on the data bits being ”1” or ”0”. A bit ”1” is simply represented by an optical pulse that occupies the entire or part of the bit duration while a bit ”0” is represented by the absence of an optical pulse. Both the return-to-zero (RZ) and non-return-to-zero (NRZ) schemes can be applied in VLC. In the NRZ schemes, a pulse with a duration equal to the bit duration is transmitted to

49 represent ”1” while in the RZ schemes the pulse occupies only the partial duration of bit. PWM is a way of encoding analog signal levels into a pulsing signal digitally. In such case, the information is encoded into the duration of pulses. In PWM, more than one bit of data can be conveyed in longer pulse than for OOK. In PPM, information is represented by the position of the pulses within the fixed time frames. In comparison to OOK, PPM modulation technique has a higher signal bandwidth and power efficiency. It has the advantage of containing the same amount of optical energy within each frame. However, in the presence of multipath echoes, PPM does not perform well due to find the correct pulse position. Color-shift keying (CSK) is a visible light communication intensity modulation scheme where the data can be carried by the color itself by red, green, and blue light emitting diodes [128]. Variable PPM (VPPM) is a modulation scheme that is compatible with a dimming control that varies the duty cycle or pulse width to achieve dimming, as opposed to amplitude. VPPM combines 2-PPM with PWM for a dimming control. Bits ”1” and ”0” in VPPM are distinguished by the position of a pulse, whereas the width of the pulse is determined by the dimming ratio [36]. In pulse interval modulation (PIM), information is encoded by inserting empty slots between two pulses. The PIM offers a reduced complexity compared to PPM due to its built-in-symbol synchronization. There are also other modulation schemes based on the PPM and PIM have been suggested and investigated in [31, 131]. The suggested modulation schemes either improve throughput or reduce power requirements by adopting a pattern of complex symbols or by adopting multilevel amplitude. Digital (PPM), digital (DPIM) and dual header PIM (DH-PIM) are examples of these three new modulation methods based on PPM and PIM for optical wireless communication, these new modulation techniques can be used as substitution for PPM and PIM for their better performance in terms of power efficiency and bandwidth efficiency. Theoretical analysis and simulation results show that DPPM, DPIM and (DH-PIM) are more applicable for future optical wireless communication system specially in VLC [31, 131]. IEEE 802.15.7 VLC has defined three PHY and their respective modulations techniques according to the applications it used. PHY I is intended for outdoor use with low data rate applications. This mode uses OOK and variable pulse position modulation (VPPM) modulations with data rates in the tens to hundreds of Kbps. PHY II is intended for indoor use with moderate data rate applications. This mode also uses OOK and VPPM with data rates in the tens of Mbps. In PHY III CSK modulation method is used to transmit data through the light’s color property of a multi-color light source.

50 Table 6. Comparison between different modulation schemes.

Modulation Dimming support Susceptibility to Spectral efﬁciency Flicker LED non-linearity OFDM No Hi 3 − 4 Medium SM No Low 2 Hi VPPM Yes Low <1 Medium MPPM Yes Low <1 Low EPPM Yes Low <1 Low MEPPM Yes Low 2 − 3 Very Low

It provides data rates in the range (12-96) Mbps. Reed Solomon codes can be used for forward error correction in these modulation schemes. Recently, many advanced modulation techniques such as expurgated PPM (EPPM), and orthogonal frequency-division multiplexing (OFDM) have been proposed to increase the data rate to multi-Gbps in VLC. EPPM is suitable for peak power limited communication systems [54, 132]. To increase spectral efficiency, multilevel forms of EPPM have also been proposed in [55]. Both EPPM and multilevel EPPM (MEPPM) are able to support a wide range of optical peak to average power ratios (PAPRs) and transmit high speed data even in highly dimmed scenarios. A comparison of different modulation schemes is shown in Table 6. OFDM is a special version of subcarrier modulation where all the subcarrier frequencies are orthogonal [133]. In this modulation technique, intensity is modulated via the time variant OFDM signal to achieve wireless access. There are mainly two techniques to generate unipolar OFDM: DC-biased optical OFDM (DCO-OFDM), asym- metrically clipped optical OFDM (ACO-OFDM) [134, 135]. In DCO-OFDM, a DC bias is put into the signal where constellation size determines the optimum bias. Util- ising a proper DC operating point, the optical carrier intensity is modulated in bipolar time domain results. On the other hand, ACO-OFDM clips the OFDM signal at the zero level where data is carried in only odd subcarriers. ACO-OFDM is far more power efficient than DC biased OFDM although which are coming at the cost of losing half of the available bandwidth. There is also another technique to generate unipolar OFDM signal named itself unipolar OFDM (U-OFDM [136]. In this technique, OFDM sam- ples are rearranged and sent them in separate positive and negative blocks. It has the same spectral efficiency as ACO-OFDM, but from the power efficiency point of view it is better than ACO-OFDM.

51 52 3 Link characterisation in VLC

In VLC, link topologies can be configured in many ways. However, they are typically classified into four different basic links. Classification of these topologies are done mainly according to the degree of directionality and the presence of uninterrupted LOS between the transmitter and receiver. The basic characteristics of these link topologies are described in detail. Different types of diffuse links such as multi-spot diffuse links, dynamic multi-spot diffuse links are also discussed along with four basic link topologies. Link design of LOS and NLOS are also derived and discussed in this chapter. Finally, measurement-based link design and their performance are discussed.

3.1 Basics link types of VLC

Four different basic link characteristics in an indoor environment are shown in Figure 8. These four link characteristics are classified as: directed-LOS, non-directed-LOS, non- directed-NLOS and tracked [31, 45, 46]. There are mainly two criteria used to classify these link topologies. The first is the degree of directionality and the other is the presence of uninterrupted LOS between transmitter and receiver. In directed directionality, both transmitter and receiver have narrow FOV. On the other hand, in non-directed directionality, both transmitter and receiver have wide FOV. If the LOS path is present in directed directionality, then this link topology is defined as directed-LOS, otherwise link topologies can be classified as directed-NLOS and non-directed-NLOS. Non-directed- NLOS is also known as diffuse link [31, 46]. Directed LOS is typically used in point-to-point communication links in indoor and outdoor environments as shown in Figure 8(a). This type of link topology experiences lower path loss and less impact of ambient light noise. As a result, hundreds of Mbps data rates can be achieved in directed LOS indoor as well as outdoor OW links. How- ever, since this link topology requires strict alignment for LOS between transmitter and receiver, mobility is an issue to work with this link topology. In non-directed-LOS, a wide beam transmitter and wide FOV receivers are used to achieve a broader coverage area as shown in Figure 8(b). Non-directed-LOS can be used in point to multi point communications. In this type of link topology, in addition to the LOS link there are also other multiple signal reflections from the walls and objects of the room. As seen in Figure 8(c), the link without LOS and completely depending on

53 (a) (b)

Tx Tx

Rx Rx

Rx Rx Rx

Fig. 8. Link topologies: a) directed-LOS, b) non-directed-LOS, c) non-directed-NLOS, d) tracked. reflections from the walls and ceiling is known as a non-directed-NLOS or diffuse link. Diffuse links are relatively immune to blockage and pointing errors and permit a great degree of mobility for the receivers in an indoor environment. However, the received signal is corrupted by multipath dispersion. As a result, large number of collected reflections received at the receiver limit to achieve a high data rate due to inter-symbol interference [31, 46]. Multibeam transmitters together with a multiple element narrow FOV angle diversity receiver is also suggested in [137, 138]. This type of link configuration is also known as quasi multi-spot beam. In this link architecture, a combination of point-to- point links with the mobility is achieved through diffuse links. In this case, each diffusing spot would be considered as a source of LOS link to a narrow FOV of receiver. Although diffuse and multi-spot diffuse links depend on the reflections but there are noticeable difference between these two link configurations. In the diffuse link, light emits over a large divergence angle, on the other hand, in multi-spot diffuse link a series of narrow divergence beams directed to the ceiling which results into far smaller

54 Table 7. Comparison of three link models.

Characteristics Point-to-point Diffuse Quasi-diffuse Range Long Moderate Moderate Rate High Low Moderate LOS Yes No No Mobility No Yes Yes Implementation cost Low Moderate High path loss in multi spot diffuse systems than in diffuse systems. The multi-spot diffuse receiver consists of a series of narrow FOV elements directed to the ceiling which results into less ambient noise in the receiver in comparison to wide FOV used in diffuse links which allows large ambient noise. Therefore, multi-spot diffuse links provide better communication links and mobility in comparison to diffuse links at the expense of complex transceiver architecture. A comparison of these three link topologies (point-to- point, diffuse, quasi-diffuse) is shown in Table 7. A special case of the general spot diffusing model is proposed as dynamic spot diffusing (DSD) and reported in [99], where the transmitter consists of one or a small number of spots, which are translated across the ceiling in a closed path. The detector is composed of multi-element imaging receivers. Data is received only whenever the spot is in the FOV of the receiver. In DSD, the channel is time-varying because of the spot motion. In comparison to the conventional spot-diffusing techniques, DSD has far fewer spots and hence has a lower multipath distortion. Tracked directed links are a different link architecture than others as mentioned earlier. In this case, the base station (BS) is mounted on the ceiling and the mobile terminal (MT) is placed at the table height as shown in Figure 8(d). In case of both for uplink and downlink, the beam from the transmitter is concentrated onto the receiver. In this architecture, diffuse link, directed link, position detection, and tracking can be realized with one and the same transceiver hardware. In a tracked system, both diffuse and tracked directed links are used for high-speed communications. In this case, diffuse link is used for connectivity and tracked directed links are used for high-speed communications. The transmitter in tracked directed link architecture is composed of a laser diode array in combination with multiple-beam form- ing optics. On the other hand, the receiver is composed of a wide angle lens, and an array of APDs. Because of the narrow beam of transmitter and narrow FOV of the receiver, high bandwidth becomes available due to minimum multipath signals. More- over, interference from the sun is also reduced because of pointing a directive receiver

55 LED

Half-power angle

φ D

ψ Ψ c

Concentrator

Photodetector Fig. 9. Link conﬁguration in LOS case.

selectively toward the transmitter. A manually tracked directed IR link with 155 Mbps over a distance of 2 m is reported in [139].

3.1.1 Directed LOS link design

In OWC, the radiant intensity Ri for LOS propagation model, as shown in Figure 9 can be expressed as [46]:

Ri = Pt R0(ϕ), (23) where Pt is the transmitted power and R0(ϕ) is the Lambertian pattern. The Lambertian pattern at the incidence angle ϕ can be expressed as:

(m + 1) R (ϕ) = cosm(ϕ), (24) 0 2π where m is the order of the Lambertian radiation pattern and related to the LED semi- angle at half-power Ψ1/2 given by

ln2 m = . (25) ln(cos(Ψ1/2))

It should be noted that R0(ϕ) is a function of two angles: angle of incidence ϕ and half power semi angle Ψ1/2.

56 Data Rate [Mbps] 2.5 110 10 10 20 2 100 30 1.5 20 40 30 20 90 50 40 60 10 1 10 70 80 20 80 60 90 50 30 0.5 50 70 70 40 60 110 40

8090 110

0 30 60 100

100 Y[m] 70 90 30 −0.5 20 50 60 20 80 10 70 −1 50 60 40 40 50 10 −1.5 30 40 30 30 20 −2 20 20

10 10 −2.5 10 −2 −1 0 1 2 X[m]

Fig. 10. Data rate in LOS link.

Without considering reﬂections, DC gain H(0) in LOS channel can be expressed as   (m + 1)Ad cosm(ϕ)T (ψ)g(ψ)cos(ψ) 0 ≤ ψ ≤ Ψ H(0) = 2πD2 s c (26)  0 ψ > Ψc, where D is the distance between LED and photo detector (PD), Ad is the detector area,

Ts(ψ) is the signal transmission of the ﬁlter, g(ψ) is the concentrator gain and Ψc is the concentrator FOV that can be represented as a semi-angle. The relationship among received power Pr, Pt and H(0) can be expressed as

Pr = H(0)Pt . (27)

The concentrator is made of transparent materials such as glass or plastic, which can be used to increase the physical area. The concentrator gain g(ψ) of an ideal non-imaging concentrator with refractive index kr can be expressed as   2  kr 0 ≤ ψ ≤ Ψ ψ 2 Ψ c g( ) = sin ( c) (28)  0 ψ > Ψc.

57 LED

φ d 1 1

α φ D

Firstreflector d ψ Ψ 2 c

Photodetector

Fig. 11. Link conﬁguration in N-LOS case.

The effective collection area of the PD can be represented as [46] { Ad cos(ϕ), |ϕ| < FOV A f = (29) 0, |ϕ| ≥ FOV.

Data rate and coverage for the LOS link case is shown in Figure 10. In this case, we consider a general indoor scenario where 5×5×3 m3 room is chosen for LOS communication. In this model, the placement of the receiver is 0.85 m on top of the ground. The numbers of LEDs in the transmitter are 12. The half-power angle of each transmitter is 85o. The transmit power of each LED is 12 mW. The ambient noise is modelled as N = 10−22. The transmission and concentrator gain of the photodetector are chosen to be 1. FOV and responsivity of the receiver are 50o and 0.85 cm2 respectively. The maximum achievable data rate is 110 Mbps.

3.1.2 Non-directed link design

In non-directed LOS and non-directed NLOS (diffuse) links, optical path loss depends on many factors such as room dimensions, ceiling reflectivity, walls and objects within the room. Moreover, the position and orientation of the transmitter and receiver, win- dow size and place and also affects on the performance. The total received power in non-directed link is composed of direct path and the first order reflected path. It can be

58 Data Rate [Mbps] Data Rate [Mbps] 2.5 120 2.5 0.05 0.05 0.35 0.1 0.05 0.1 20 0.2 0.15 0.2 0.15 0.15 0.1 0.05 0.2 0.25 2 110 2 0.25 0.3 0.3 20 0.25 0.35 0.1

0.2 40 0.15 0.30.35 0.25 0.3 1.5 20 100 1.5 0.35 0.25 0.3 40 0.2 60 0.35 0.05 1 90 1 80 60 0.15 0.3 0.3 0.2 0.1 0.25 0.25 100 0.25 0.15 0.05 0.25 0.2 40 0.05 0.5 60 80 80 0.5

0.15 0.1

120 0.1 0.15

0 70 0 0.2 0.2 0.2 0.1 Y[m] Y[m]

0.25

20 0.2 0.05 −0.5 100 60 −0.5 0.25 0.05 40 0.15 0.25 0.1 0.15 0.3 0.15 80 0.3 −1 60 50 −1 40 60 0.3 0.25 0.2 0.35 0.25 0.2 20 0.35

−1.5 40 −1.5 0.05 0.35 40 0.3 0.1

0.15 0.25 0.3 0.35 −2 20 30 −2 0.10.15 0.25 0.3 0.2 0.1 20 0.2 0.20.25 0.10.15 0.10.15 −2.5 20 −2.5 0.05 0.05 0.05 0.05 −2 −1 0 1 2 −2 −1 0 1 2 X[m] X[m]

Fig. 12. Data rate and coverage in LOS Fig. 13. Data rate and coverage in NLOS link. link. expressed as ( ) ( ) Ptotal = Hlos(0) + Hnlos(0)Pt = Hlos(0) + ∑ Hreflec(0) Pt , (30) #reflec where Hreflec represents the channel DC gain on reflection points and can be expressed as

Hreflec(0) =  (m + 1)Ad ρ dA cosm(ϕ)T (ψ)cos(β)cos(α)g(ψ)cos(ψ) 0 ≤ ψ ≤ Ψ (31) π2 2 2 s c  2 d1 d2  0 ψ > Ψc, where d1 and d2 are the distance between an LED and a reflective point and the distance between a reflective point and a receiver respectively. ρ is the reflectance factor. dA is a reflective area of small region. α is the angle of incidence to a reflective point and β is the angle of irradiance to the receiver. Data rate and coverage for non-directed NLOS link design case is shown in Figure 13. In non-directed cases, we keep both the transmitter and receiver specification the same as in the LOS case. The reflection coefficient of the walls of the room is 0.8. Figure 13 shows the data rate of the diffuse link. In this case, we only consider the data rate is achieved only from the reflected paths of the walls. We keep the same value of the parameters of the transmitter, receiver and room size as we consider in case of LOS and non-directed LOS.

59 Data Pre- LED Modulation LED source equalization driver Channel Optical

Data Post- Demodulation Amplifier Detector sink equalization

Fig. 14. Block diagram of VLC transceiver systems.

3.2 Measurement based link design

In this section, a simple GNU radio based testbed has been implemented to evaluate the performance of the VLC link. GNU radio is a free software development toolkit for runtime signal processing that has the capability of interacting with the external hardware typically used for transmitting RF signals [119]. However, in this work, this hardware namely the universal software radio peripheral (USRP) is interfaced with an optical front end to create a point to point VLC link.The purpose of using USRP in the testbed for providing a versatile and cost efficient platform for the system designing. Moreover, the system designs can be modified and duplicated easily, for example, the used parameters like frequencies, amplification and filtering are simply controllable in real time with the software used in USRP. The systematic block diagram of the VLC system implemented in the testbed is shown in Figure 14. In the testbed, both the transmitter and receiver have a signal processing part and an analogue part. For example, in the transmitter side, the data source, modulation and pre- equalization are considered to be the signal processing part. The characteristics of these signal processing modules are implemented in the USRP device. However, the pre- equalization can also be implemented with external circuits. On the other hand, such as the LED driver, the LED and the transmitter optics are considered as the analogue part. Similarly in the receiver side post-equalization, the demodulation and data sink can be considered as the signal processing part and the detector and amplification belong to the analogue part. The raw data from the data source generated in USRP is used to feed the modulation module. The USRP provides many different modulation schemes to modulate incoming data from the data source. Modulation schemes such as OOK, PPM, pulse-amplitude

60 Fig. 15. Visible light communication testbed implemented in University of Oulu. modulation (PAM), PWM are available for selection in USRP. In this testbed, we use GMSK modulation to modulate the incoming data that is already coded in the GNU Ra- dio [140]. Several measurements were conducted during the implementation to verify the adoption of the GMSK in the VLC testbed. Verification is done by calculating an error vector magnitude (EVM) using Agilent E4446A PSA Series spectrum analyzer equipment and Agilent 89600 VSA 16.0. The result of this verification is discussed in [140]. The experimental setup of the VLC prototype is shown in Figure 15. GNU radio software is used to interact between two N210 software radio peripherals (USRP2). Two types of LEDs, such as phosphor coated LED and trichromatic LEDs (RGB) are used as transmitters in VLC link measurement. A commercial RGB LEDs (LED EN- GIN, 03MC00, 40 W, 81 mm footprint) LEDs, and a blue-phosphor LED (LUXEON) were used in this testbed. On the other hand, a Thorlabs photodetector (PDA36A) is used as receiver. This particular photodiode has a built-in transimpedance amplifier that is set at the default value of 10 dB. Data rate and light intensity are recorded by varying

61 the distance between the transmitter and receiver both for phosphorus and trichromatic LEDs as shown in Figure 16 and Figure 17 respectively. It is obvious as it is seen also in both of these Figures that as the distance between transmitter and receiver increases, the data rate and light intensity also decreases. The maximum data rates realized were 6.25 Mbps and 12.5 Mbps for the blue-phosphor and RGB LEDs, respectively. Theo- retically, in case of RGB LEDs, all three colours could be modulated simultaneously to achieve bitrate up to 37.5 Mbps. Lower data rate achieved for blue phosphor due to slow response rates of the phosphor required to produce white. The blue LED emission maximum at 450 nm, corresponds with a 0.3 A/W responsivity in the photodiode used in this testbed, whereas at longer wavelengths the responsivity is up to 0.9 A/W. However, the obtained results are still signiﬁcantly lower than VLC links with custom software and hardware that have achieved several Gbps.

14 800 RGB LEDs Blue LED 12 700

600 10

500 8 400 6 300 Data rate [Mbps] 4 Light intensity[lux] 200

2 100

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 Distance [m] Distance [m]

Fig. 16. Data rate vs. distance. Fig. 17. Light intensity vs. distance.

62 4 Data downloading on VLC coverage

In this chapter, we provide the methodology to evaluate the performance of VLC hotspot networks in the context of data download on the move by using throughput-distance relationship models. In this regard, the motivation and related work of data download on the move in sparse-based coverage are described in Section 4.1. The concept of sparse- based infostation coverage is introduced in this section. Secondly, the geometry of VLC hotspot coverage is described in Section 4.2, where many optical design parameters such as FOV of the receiver, semi-angle at half-power of the transmitter are considered in modeling. Mobility related parameters such as travelling path, velocity of the user are also taken into account in the analytical model. The transformation of travelling inclined path into horizontal path is introduced in Section 4.3. Detailed descriptions of the mathematical framework and VLC hotspot design parameters and their relationships are given in Section 4.4 and 4.5 respectively. Theoretical throughput vs. distance relationship and its approximation to the so-called empirical throughput-distance relationship model is introduced and derived analytically by taking into account the optical design parameters such as FOV of the receiver, semi-angle at half-power of transmitter, ambient and artificial shot noise and so on in Section 4.6. The purpose of this empirical throughput-distance relationship model is to calculate the average throughput analytically in the receiving plane. Performance analysis of such VLC hotspot networks in the context of data download on the move scenarios is given in Section 4.7. Performance of VLC hotspot networks is quantified in terms of received file size. Simulation results reveal that background noise has a significant impact on the performance of VLC hotspots. As expected, the VLC hotspot performs better at night than in the daytime. Finally, conclusions are drawn in Section 4.8.

4.1 Motivation and related work

Data download on the move is seen as a promising application for the next generation wireless communication systems. An example of such a paradigm of wireless communication systems for delivering information services to a mobile user is infostation [141]. The coverage area of infostation is sparse and, as a result, as long as the mobile user is in the coverage area of the infostation he/she may download information to the

63 y

Entry point Exit point θ Moving-in Moving-out -x x AP Travelling angle

-y

Fig. 18. Data download on the move. mobile terminal storage for later usage. The retrieved downloaded information can be a video ﬁle, a piece of music, a weather report, and so forth. The concept of infostation is not new: there exists of such models for information distribution [88, 89], drive-thru In- ternet [142], and roadside infrastructures [27, 143], all of which are examples of sparse coverage-based wireless networks. The naming varies, but in terms of functionalities and characteristics, they share major commonalities with the infostation. The overall coverage of such wireless communication systems is considered to be sparse [10]. The performance of IEEE802.11-based sparse coverage WLAN hotspot is studied in [142]. In [144, 145], measurement-based performance analysis of IEEE802.11b/g-based systems for static and nomadic users in case of outdoor drive-through scenarios also have been carried out extensively and reported in [88, 89]. However, very few papers addressed the issue of performance analysis of VLC-based systems especially in the data download on the move scenario [86, 146]. Unless, otherwise mentioned further, the concept of infostation will be termed as hotspot throughout this chapter. The data download on the move scenario is shown in Figure 18. In this scenario, the concept of moving-in and moving-out is introduced. In the moving-in case, the received throughput by the mobile user gradually increases from the entry point to the point when the distance between the moving user and access point (AP) is at its minimum. On the other hand, in the moving-out scenario, the received throughput starts to

64 Lighting equipment

LED LED

φ h D Vertical Vertical distance

Ψ ψ c

Horizontal distance yxr ),( Optical receiver

Receiving plane Fig. 19. VLC hotspot coverage.

decreases from the maximum throughput to minimum at the coverage end as shown in Figure 18. The angle created by the entry and exit points at the center of coverage is called the travelling angle. In Figure 18, the segments "ab" and "bc" show the travelling distances at a certain travelling angle of a mobile user in moving-in and moving-out scenarios, respectively. The details of travelling path of the mobile user and its mathematical framework for calculation of average throughput in data download in on the move scenario are discussed in Section 4.3 and 4.4 respectively.

4.2 Geometry of VLC hotspot coverage

In any optical transmitter, the half-power angle is deﬁned as the highest angle that the transmitter can illuminate. On the other hand, FOV is the highest angle within which the receiver can receive signal rays for any optical receiver. In this work, we will consider the worst case alignment link geometry of the optical hotspot [147]. In the worst case alignment, the divergence angle and acceptance angle are the same. The divergence o angle varies between 0 to the maximum half-power angle ϕmax. On the the other hand, the acceptance angle belongs to FOV. These angular representations provide the simple link geometry in terms of hypotenuse and the vertical and horizontal distance between transmitter and receiver as shown in Figure 19. As a result, the relationship between

65 divergence angle and acceptance angle in the worst-case alignment can be written as

( h ) ϕi = ψi = arccos , (32) Di where h and Di are the vertical separation and instantaneous hypotenuse distance between transmitter and receiver. According to (32), (26) can be written as   (m + 1)Ad cos(m+1)(ϕ)T (ψ)g(ψ) 0 ≤ ψ ≤ Ψ H(0) = 2πD2 s c (33)  0 ψ > Ψc.

Equation (33) can be written as a function of horizontal and vertical distances as follows   (m + 1)Ad cos(m+1)(ϕ)T (ψ)g(ψ) 0 ≤ ψ ≤ Ψ H(0) = 2π(r2 + h2) s c (34)  0 ψ > Ψc, where r is a function of (x,y) coordinates in the receiving plane. If the coverage of hotspot is considered as circular, then the maximum horizontal separation between transmitter (WLED) and receiver (photodetector) will represent the radius R of VLC hotspot. The relation among horizontal separation r, D, and vertical separation h can be expressed as √ r = D2 − h2 0 < r ≤ R. (35)

4.3 Transformation of inclined path to horizontal path

The inclined and the transformation of the inclined path to the horizontal path are shown in Figure 20(a) and Figure 20(b), respectively. The mobile user enters the coverage at point A′, known as entering point and exits the coverage at point B′ ˛Aknown as exit point, as shown in Figure 20(a). For the inclined path, the entrance angle is denoted with α1, and the exit angle is denoted with α2. The angle between entrance point and exit point is called the travelling angle, which is denoted as θ. It should be noticed that the maximum range of the travelling angle can be limited to the range of (0o- 180o) instead of (0o-360o) for the calculation of average throughput over the simpliﬁed horizontal path. Moreover, the entrance and exit angle of the inclined path can be transformed as α1 = αs and α2 = αs. After transformation, the average throughput over

66 y y

B' Dt A B L α ' 1 θ A θ α B α α 2 s s x

(a) (b) Fig. 20. Transformation from inclined to horizontal path: a) inclined path b) transformed horizontal path.

the inclined path A′B′ will be the same as the calculation of the average throughput over the horizontal path AB. Depending on the difference between the entrance and exit angle of inclined path, after rotation the travelling angle θ of the horizontal path can be expressed as: { |α − α | |α − α | ≤ 180o θ = 1 2 1 2 (36) o o 360 − |α1 − α2| |α1 − α2| > 180 . It should be noticed that the maximum range of the travelling angle can be limited to the range of (1o-180o) instead of (1o-360o) for calculation of average throughput over the simpliﬁed horizontal path.

4.4 Mathematical framework for a data downloading scenario

The footprint of the coverage of LED source is shown in Figure 21. We assume that the coverage of the hotspot is circular. The entrance and exit points of the inclined path are ′ ′ coordinated as A (xa,ya) and B (xb,yb) respectively. Assuming constant velocity, the path that the mobile user travels can be simply expressed with

(x(t),y(t)) = ((xb − xa)t + xa,(yb − ya)t + ya), (37)

67 D t Horizontal path A B O O O O ( ) (− x, y ) x , y L r R θ (− R 0, ) (R 0, ) ' θ B’ ( − ) xb , yb

A’ (− − ) xa , ya

Fig. 21. Footprint of VLC coverage. where the parameter t varies between zero (corresponding to the entrance point) to one (the exit point). To characterise the throughput as a function of the distance, i.e., the throughput at distance r is S(r). The calculation of instantaneous distance from the center of coverage to the mobile user can be written as √ 2 2 r = [(xb − xa)t + xa] + [(yb − ya)t + ya] . (38)

Now, calculation of average throughput of the mobile user (while it passes through the coverage) can be written as (√ ) ∫1 2 2 Sav(r) = S [(xb − xa)t + xa] + [(yb − ya)t + ya] dt, (39) t=0 √ where S( ·) represents spatial throughput-distance relationship function. This function can be either measurement- or analytical-based. However, the calculation of (39) for arbitrary entrance and exit points is cumbersome. Hence, in order to simplify the calculation of the average throughput, we rotate the inclined path to an equivalent horizontal path as discussed in Section 4.3 and shown in Figure 21. In this case, entrance and exit coordinates points of inclined and horizontal path is related as xa = x, xb = −x, ya = y, yb = y. The entrance and exit points of the transformed horizontal path is coordinated ′ ′ as (due to the symmetry of entrance and exit angle) A (−xa,yb) and B (xa,yb) respec-

68 tively as shown in Figure 21. The calculation of average throughput over the simpliﬁed horizontal path is given below: After rotation of the inclined path to the horizontal path, (38) is simpliﬁed and written as √ r = [x(2t − 1)]2 + y2. (40)

In Figure 21, it is also shown that the minimum distance between mobile user and( / the) θ θ center of coverage is L, which for a particular travelling angle is(L /= R) cos 2 , where R is the coverage radius. The travelling distance Dt = 2Rsin θ 2 . However, it should be noted that the travelling distance is equal before and after rotation (e.g ′ ′ Dt = A B = AB) that is the travelling angle of horizontal path θ is not affected by the rotation, that is θ will remain same as it was for inclined path. Finally, average throughput over the simpliﬁed horizontal path can be expressed as

∫1 (√ ) 2 2 Sav(r) = S [x(2t − 1)] + y dt, (41) t=0 ( / ) Dt θ where x = 2 = Rsin 2 and y = L. It is easily seen that (41) leads to much simpli- ﬁed integration in comparison to (39). By taking velocity v into account,/ the/ dwelling time of the mobile user in VLC hotspot can be found with tdwell = Dt v = 2x v. Now by simply multiplying the dwell time with the average throughput, the size of the transferred information or ﬁle size can be written as

It (θ) = Sav ·tdwell. (42)

Using (40), transferred information or ﬁle size can be written as follows:

∫1 (√ ) 2x I (θ) = S [x(2t − 1)]2 + y2 dt. (43) t v t=0

Before deriving relationship between throughput and distance. Let’s represent (33) as

C H(0) = c , (44) D2 where (m + 1)A C = d cos(m+1)(ϕ)T (ψ)g(ψ). (45) c 2π s

69 In this work, the indoor environment is considered into two classes: daytime and nighttime. Note that the effect of background noise will vary depending on its particular class. For speciﬁc value of BER and N0 and using the relationship of data rate and hypotenuse distance D can be expressed as [4]

2 2 2 RrCc Pt Rb = −1 2 4 . (46) (Q (BER)) N0D

Moreover, the relationship between Rb and distance D can be represented as

−4 Rb = K · D , (47) where K is a scaling factor and is given as

2 2 2 RrCc Pt K = −1 2 . (48) (Q (BER)) N0

The value of scaling factor K depends on FOV, BER, transmit power, noise, detector area, concentration gain of the concentrator and others.

4.5 VLC hotspot design parameters and their relationships

Prior to deploying any wireless networks, a number of scenarios and the relationship of key design parameters are targeted to evaluate performance in terms of capacity and reliability. Some examples of these relationships are SNR vs. throughput, SNR vs. distance, BER vs. data rate and so on. These relationships are used as tools in analytical models for analyzing as well as designing wireless networks before being deployed in practice. Figure 22 shows the FOV changes due to the horizontal separation between transmitter and receiver with step changes of 0.1m. As mentioned earlier in the case of worst case alignment, the irradiance and reception angle will be the same. It is seen in Fig- ure 22 that as the step value increases the alignment between irradiance and reception angle also increases. In this particular case the half-power at semi-angle is chosen as 85o and FOV of the receiver is chosen as 60o. Figure 23 shows the variation of received power with Lambertian order m. It is also seen that for a ﬁxed Lambertian order the narrow semi-angle at half power provides the higher received power. This received power difference is more noticeable in higher

70 60 −38

−40 50

−42 40

−44 30 −46

20 −48 Received power [dBm]

Incidence angle [degree] 10 −50

0 −52 0 0.5 1 1.5 2 2.5 3 0 2 4 6 8 10 12 14 16 18 20 Horizontal distance [m] Lambertian order of LED

Fig. 22. FOV vs. horizontal distance. Fig. 23. Lambertian order vs. power. order Lambertian order. For achieving a higher transmission bandwidth the half-power semi-angle of the LED can be tuned to its optimal value. The optimal Lambertian and transmitter semi-angle at half-power is calculated as [148, 149]

− ( 1 ) mopt = − 1. (49) ln cos(ϕmax)

( ( − )) ϕ ln(2) ϕ o 1 = arccos exp − 0 < 1 < 90 . (50) 2 opt 1 2 opt ln(cos(ϕmax))−1

4.6 Theoretical and empirical throughput-distance models

Signal coverage and data rate are the two important design parameters in any wireless communication system. In this work, the relationship between these two design parameters and their approximated representative models are termed as throughput-distance models. Determination of signal coverage and data rate are inﬂuenced by a variety of factors, most prominently the frequency of operation and the terrain [150]. It may also be inﬂuenced by the choice of several design parameters, e.g., modulation and coding, constellation size, power level, multiple access scheme, and many others. The effect of all these parameters to the throughput may be described as the throughput- distance model. Several throughput-distance models for hotspots have been discussed in [10, 88, 89]. Throughput performance dependent on the distance based data rate of VLC hotspots has investigated in [86]. In this chapter, we have developed a polynomial based empirical throughput-distance model. This empirical throughput-distance model has been approximated from an analytically derived throughput-distance rela-

71 tionship. The purpose of using this empirical throughput-distance model is to simplify the problem. This simpliﬁcation will provide a tractable solution for calculating average throughput and ﬁle size of a mobile user when he/she is downloading data on the move.

4.6.1 Empirical polynomial based throughput-distance model

The second order empirical polynomial model is used to approximate the theoretical throughput-distance model. In this case, we ran a MATLAB basic fitting tool to fit the theoretical throughput-distance to polynomial throughput distance model. This basic fitting tool uses the least square (LS) method. The approximated second order polynomial throughput-distance relationship model can be written as { ar2 + br + c, 0 < r < R S(r) = (51) 0, otherwise where the a and b represent the first and second term coefficients of second order polynomial respectively, and c represents the constant term.

4.6.2 Throughput vs. distance at daytime

In the daytime, direct and indirect sunlight causes interference to the performance of the VLC hotspot. By taking into account direct and indirect sunlight and other lighting noise sources the relation between the data rate and distance is shown in Figure 24. Figure 24 shows the theoretical throughput-distance models in case of with and without using ﬁlter. Both of these theoretical models are then approximated using the polynomial as below 2 − Td−wf = 0.69dhor 3.9dhor + 5.6. (52) 2 − Td−wof = 3.282dhor 18.58dhor + 26.74. (53)

It is noticed that the throughput vs. distance relationship are approximated to polynomial functions. In this case we have assumed that coverage of a VLC hotspot is circular. Any point from the center of coverage is a function of both horizontal as well as vertical distance. However, for further analysis we represent this data rate versus distance relationship into two dimensional coordinates, where dhor can be represented by (x,y) coordinates.

72 35 700 Daytime with filter Nighttime with filter Approximated daytime with filter Approximated nighttime with filter 30 Approximated daytime without filter 600 Approximated nighttime without filter Daytime without filter Nighttime without filter

25 500

20 400

15 300

10 200 Throughput [Mbps] Throughput [Mbps]

5 100

0 0 0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3 Distance [m] Distance [m]

Fig. 24. Polynomial ﬁtting to the theo- Fig. 25. Polynomial ﬁtting to the theoret- retical throughput-distance model at day ical throughput-distance model at nighttime. time.

4.6.3 Throughput vs. distance at nighttime

At nighttime, artificial light sources such as incandescent lamps with tungsten filaments, halogen and mercury lamps, fluorescent lamps with different colours and fluorescent lamps are considered to be the major noise sources. By taking into account with filtered and not filtered noise sources, the throughput-distance model can be represented as shown in Figure 25. The approximated throughput-distance model can be expressed as

2 − Tn−wf = 33.03dhor 187dhor + 269.1. (54)

2 − Tn−wof = 70.62dhor 3.99.7dhor + 575.4. (55)

4.6.4 Calculation of average throughput and ﬁle size

In this section, we will calculate the theoretical average throughput and ﬁle size substi- tuting respective throughput-distance models (52, 53, 54, 55) into (41) and performing the integration, the average throughput for each throughput-distance model can be calculated as ( ) ( ) x2 b (−x + R) S (r) = a + y2 + 2Rx − y2 log + c, (56) av 3 4x (x + R)

73 ( / ) θ θ Equation (56)( can/ be) expressed in terms of the travelling angle by using x = Rsin 2 and y = Rcos θ 2 . For the average throughput we get ( ) 2 θ S (θ) = aR2 1 − sin2( ) + av 3 2 ( ( ) ) ( ( )) θ θ (57) 1 cos2 1 + sin bR + ( 2 ) × log ( 2 ) + c. 2 θ − θ 4sin 2 1 sin 2

Similarly, using the transferred information in terms of ﬁle size can be expressed as ( ) 2aR3 θ I (θ) = sin + t v 2 ( ( ( )) ( ) ) ( ) ( ) θ θ (58) bR2 θ θ 1 + sin 2Rsin 2sin + cos2 log ( 2 ) + 2 c . 2v 2 2 − θ v 1 sin 2

4.7 Numerical results

In this section, we evaluate the performance of VLC hotspots, for indoor as well as for outdoor environments using throughput-distance models. The parameters used in the simulation environment are given in Table 8. Numerical results are provided and the performances between day and nighttime cases are compared. The received ﬁle size by a mobile user serves as a performance metric.

Table 8. Simulation model parameters.

Simulation Parameters Value Transmit power 20 mW Photodiode responsivity 0.27 Semi-angle at half power 60o FOV at the receiver 30o Detector physical area of PD 28 mm2 Coefﬁcient of optical ﬁlter 1.0 Mobile user velocity [0.4 − 1] m/s BER [10−9]

It is seen in Figure 26 that as the travelling angle increases, the data download in terms of file size also increases. When the travelling angle increases, the mobile user dwell more time on the coverage of hotspot and download more data on its digital storage. For example, in all cases (with filter and without filter), at relatively higher

74 5 x 10 4.5 v= 0.5 m/sec without filter 4 v= 1 m/sec without filter v= 0.5 m/sec with filter v= 1 m/sec with filter 3.5

2.5

1.5 Filesize [Kbyte]

0.5

0 0 20 40 60 80 100 120 140 160 180 Travelling angle [deg]

Fig. 26. File size vs. travelling angle in an indoor environment at nighttime.

travelling angles such as at 120o, the mobile user can download more in comparison to a relatively lower travelling angle such as 80o. Figure 26 also shows the improvement in performance whenever the noise and interference cancellation filter is used. For example, at 179o travelling angle with using filter, an almost double file size data download in comparison to without using filter at velocity 0.5m/sec is achieved. Figure 27 shows the received file size of low mobility mobile user in the daytime in an indoor environment. Two different velocities are considered for evaluating the performance of the VLC hotspot. It is clearly seen that the performance decreases drastically in the daytime in comparison to the nighttime. At nighttime, there will be zero solar radiation and only artificial light sources such as fluorescent and tungsten light bulbs will contribute to background noise. At relatively higher travelling angles, for example, between (140o − 180o) the received throughput at v = 0.5m/sec with and without filter will be around between (140-270) and (30-60) respectively. Figure 28 shows the received file size and dwelling time in the coverage of VLC hotspots at velocity v = 0.5 m/sec in day and night conditions. It is seen that the maximum dwelling time at v = 0.5 m/sec is 12 sec. At 9.89 sec, when the travelling angle is at 100.6o, the download file size at nighttime will be 200 Mbytes. However, at the same dwelling time the download file size at daytime will be 6.1 Mbytes.

75 4 x 10 2 v=1 m/sec without filter 1.8 v=0.5 m/sec with filter v=1 m/sec without filter 1.6 v=0.5 m/sec with filter

1.4

1.2

0.8

Filesize [Kbyte] 0.6

0.4

0.2

0 0 20 40 60 80 100 120 140 160 180 Travelling angle [deg]

Fig. 27. Filesize vs. travelling angle in an indoor environment at daytime.

400 20 Dwelling time Received Filesize at night time Received Filesize at day time

X: 110.6 Y: 9.89 200 10 X: 110.6 Y: 140.9 Filesize [Mbyte] Dwelling Time [Sec]

X: 110.6 Y: 6.181 0 0 0 20 40 60 80 100 120 140 160 180 Travelling angle [deg]

Fig. 28. Filesize and Dwelling time vs. travelling angle.

4.8 Conclusions

In this chapter, we studied the performance of optical wireless hotspots (a.k.a. VLC hotspots), where the visible light spectrum is considered in local access points. In con-

76 trast to radio frequency based wireless networks, VLC is more sensitive to many factors such as alignment between transmitter and receiver, FOV of the receiver, semi-angle of transmitter, state of the optical channel and different background noise at in the day and at nighttime. To evaluate the performance of such wireless communication networks, the impact of the above mentioned optical designing factors needs to be considered and modelled. In this chapter, we provide a methodology to evaluate the performance of visible hotspot networks using throughput-distance relationship models. First we derived the theoretical average throughput-distance model and then we approximated this to the empirical throughput-distance model to calculate the file size that can be downloaded while mobile users move through the VLC hotspots. Simulation results show that there is a large impact of background noise on the performance of a VLC hotspot. As expected, the VLC performance in both indoor and outdoor environments is better at night than in the daytime. Performance of VLC hotspot networks are also quantified in terms of received file size. Numerical examples show that in indoor environments a mobile user can download more data at nighttime than at daytime. Finally, we ar- gue that the presented framework can be used as a mathematical tool for evaluating the performance of a VLC hotspot before it is deployed in practice.

77 78 5 Data downloading in hybrid WLAN-VLC networks

In this chapter, we study the performance of hybrid (radio-optical) WLAN-VLC hotspots. In this respect, ﬁrstly we determine the coverage and data rate of VLC and WLAN. VLC coverage and data rate are described in Subsection 5.1.1. Very short-range VLC small cells (limited to 1 m coverage range) are considered for higher data rate regions. Unlike in Chapter 4, the data rate of VLC is considered to be constant in this chapter. The coverage and data rate of WLAN are described in Section 5.1.2. The measurement- based throughput-distance model in indoor environment is considered for WLAN. The scenario of hybrid WLAN-VLC for single-user cases is described in Section 5.2. Multi- user hybrid WLAN-VLC is discussed in Section 5.3. Finally, simulation results are discussed in Section 5.4. Simulation results reveal that the considered hybrid WLAN- VLC always performs better than the stand-alone VLC-only or WLAN-only.

5.1 Hybrid WLAN-VLC networks in indoor environment

In 5G, the possible integration of VLC technology with legacy radio access technologies (RATs) has been envisioned to built the hybrid network in future wireless networks [13]. In hybrid wireless networks, a mix of two or multiple wireless technologies, frequencies, cell sizes and different network architectures are used to optimally respond according to the demand of mobile users as well as multi-vendor, multi-service providers. Al- though the concept of hybrid wireless networks is not new, very few research work has been conducted, especially in the area of hybrid WLAN-VLC networks [4, 73, 76, 86]. Detailed descriptions of research work related to hybrid WLAN-VLC has already been discussed in Section 1.3.4 of Chapter 1. In hybrid WLAN-VLC networks, it is assumed that a mobile device is equipped with both optical (VLC) and radio (WLAN) air interfaces that enable connecting to either radio and optical networks. The decision on whether to be connected to an optical or radio network is purely based on the availability or quality of the channel. In special cases, the user may connect both radio-optical networks concurrently for better coverage and higher throughput, and increased reliability.

79 −27.5

−28

−28.5

−29

−29.5 Received power [dBm] −30 10 5 10 0 5 0 −5 −5 −10 −10 Y[m] X[m]

Fig. 29. Spotlighting coverage.

5.1.1 Coverage and data rate of VLC small cell

In this chapter, a small VLC cell and the comparatively larger coverage of a WLAN cell are considered in an indoor hybrid WLAN-VLC network. In [151], the small cell of VLC known as spotlighting had been studied for illumination and communication. In such a scheme, a number of small cells of VLC was used for providing a higher data rate as shown in Figure 29. Authors in [151] also introduced a hybrid scheme where the small cell VLC was responsible for providing both uniform lighting and communication. In [152], the concept of cell zooming was introduced to improve trafﬁc distribution in multi-user environments and handover at cell edges. In their work, the cell coverage regions were dynamically altered according to the need of required signal-to- interference-plus-noise ratio (SINR) distribution in an indoor environment while main- taining constant illumination. However, unlike in [151, 152], in this work we will consider WLAN for providing wide coverage and VLC small cell for providing higher data rates.

In case of VLC small cells, the required light irradiance Pr of circular lighting LED source can be determined from light ﬁeld radius Rs and its transmit power Pt regardless of the distance parameter and can be expressed as [151]

P P = r . (59) t π 2 Rs

80 9 Night Time 8 Day Time

Throughput [Mbps] 2

0 BER=10^−3 BER=10^−5 BER=10^−6 BER

Fig. 30. Data rate at three different BER requirement in indoor environment.

In (59), note that the received power is inversely proportional to the square of radius

Rs. Therefore, for a constant radius of VLC spotlighting the received power Pr will be proportional to the transmit power Pt with a scaling factor π. However, in this work we model the throughput and coverage of VLC small cell according to the two piece-wise throughput distance model studied in [85], where the ﬁrst piece represents the constant data rate till one meter of coverage range. The constant throughputs of small single LED VLC cell at three different BERs at day and night condition are shown in Figure 30.

5.1.2 Coverage and data rate of WLAN

WLAN is a multi-rate system meaning that when a mobile user moves inside the coverage area of the cell, the received SNR changes according to the distance between mobile user and access point (AP). In such a multi-rate system, SNR is a random variable and therefore the mobile user can operate at one of the multi-rate choices of data rates according to the value of its received SNR. In a single mobile user environment, the achieved average throughput is the average of all data rates at which it operates while moving in the area. In a multiple user environment, the average throughput per user is also a function of the MAC technique employed and the number of users in the area. Therefore, the average of the data rates observed by a user located randomly in

81 the coverage area of multi-rate system can be expressed as [150]

ΣNrate Rav = n=1 PnRn, (60) where Rav is the average spatial data rate, Rn is one of the available multi-rate, and Pn is the probability of occurrence of that data rate. If a terminal is located randomly in the coverage area, the probability of being in each area is given by the ratio of the area for the speciﬁc data rate to the total coverage area as

A P = i , (61) n π 2 Dmax where Ai and Dmax are the speciﬁc data rate area and the maximum coverage range respectively. In general, channel models include several geometry and physical phenomena, such as refraction, diffraction, propagation medium (dry or moist air), the distance between the transmitter and the receiver and so on. The simplest way to represent the relationship between the effects of all these factors vs. distance is by using the path-loss model. An indoor path-loss model in multi-storey buildings can be expressed as [150]

Lp = L0 + nF + 10log(Ltr), (62) where F represents the signal attenuation provided by each floor, L0 is the path loss at the first meter. Ltr is the distance between the transmitter and receiver in meter, n is the number of floors through which the signal passes. In fact, no specific channel model is valid for all environments. Thus, a proper model of coverage for any wireless technology can only be established by making a new measurement and calculations for the transmission in that specific environment. In this work, instead of a pathloss model we have used the throughput-distance relationship model to characterise the channel propagation with respect to distance for the indoor WLAN channel model. The simplest model for the throughput-distance relationship is a linear model. In- deed, the experimental data collected in [150] shows the linear throughput-distance relationship of the IEEE 802.11b in a typical office building. In the linear throughput- distance relationship model, the throughput decreases linearly as distance between the mobile user and AP increases. The coverage and throughput-distance relationship model for WLAN are shown in Figure 31 and Figure 32 respectively. Therefore, in

82 40 6

30 5 (5.4=Higher rate<=−3.34) 20

4 10 (1.338

0 5.4 Mbps 3 Y[m] −10 2 4.8 Mbps

−20 Throughput [Mbps]

2.4 Mbps 1 −30 (0<=Lower rate<1.338)

−40 0 −40 −30 −20 −10 0 10 20 30 40 0 5 10 15 20 25 30 35 X[m] Distance [m] Fig. 31. Indoor WLAN coverage. ( [83] Fig. 32. Indoor WLAN throughput. ( [83] ⃝c 2014 John Wiley& Sons). ⃝c 2014 John Wiley& Sons).

the case of WLAN, the relationship between throughput S(rw) and distance is given by: { amaxrw + bmax, 0 < rw < Rmax S(rw) = (63) 0, otherwise.

The parameters amax and bmax are used to calculate the maximum coverage distance and maximum provided data rate of AP. At rw = 0, the maximum data rate received by mobile user is S(0) = bmax. The parameter amax is related to the maximum range distance of AP which follows as a = − bmax . max Rmax

5.2 Hybrid WLAN-VLC: single-user case

The scenario for data downloading in hybrid WLAN-VLC hotspots is shown in Fig- ure 33. In this scenario, limited width and a relative long corridor in the ofﬁce building is considered as the service area of hybrid WLAN-VLC hotspot networks where adjacent walls of the corridor are assumed to be transparent. Therefore, indirect sunlight is assumed to be present at the photo detector in the daytime. A series of optical LED light sources are aligned in a row in the ceiling of the corridor for illumination as well as for the communication purposes. Coverage of each spotlighting cell can be considered as a VLC hotspot. Therefore, depending on the placement, the number of optical light sources will represent the number of hotspots in the area of corridor. We have considered multiple VLC hotspots within a wider range of WLAN coverage. Data rates of WLAN is channel adaptive. As a result, according to the placement of WLAN access point, different data rates such as higher data rate, a moderate data rate as well as

83 5 d : travelling distance in WLAN cell 4 w d : travelling distance in spotlighting cell 3 v

2 ← WLAN Coverage→ 1 B C d A w 0 d (−4,0) (0,0) (4,0) v −1 Width [m] ↑ Spotlighting Coverage −2

−3 Segmented corridor

−4

−5 −6 −4 −2 0 2 4 6 Length [m]

Fig. 33. An example scenario of data downloading in a hybrid WLAN-VLC hotspot coverage. ( [83] ⃝c 2014 John Wiley& Sons). lower data rate will be provided to the mobile user in the coverage of WLAN [88]. The distance between two consecutive VLC hotspots may vary depending on the placement of LED light sources. It is also assumed that the a mobile user terminal is equipped with both optical and radio transceivers. In the proposed scenario a mobile a user will always have ubiquitous coverage of WLAN meaning that when the mobile user is in or out of the coverage of the VLC hotspot, he/she may have a connection with the WLAN link. However, in this work we have assumed that when the user is in VLC spotlighting coverage he will be connected with the VLC link. In this case, we assume the transmitter and receiver are aligned to each other in the VLC hotspot by using an electronic tracking system. We have also assumed that initial and ﬁnal connectivity of the mobile user during data downloading will be started and ended at the VLC coverage, as shown in Figure 33. Seamless vertical handover is considered between the WLAN and VLC hotspot. The amount of information download to the user digital storage device depends also on many optical as well as radio design parameters, such as FOV of LED and PD, physical area of detector, dwelling time in the coverage of WLAN or VLC hotspot, mobile user velocity and environment (e.g. day and nighttime).

84 B ( ) xb , yb

r ( A ) C x a , y a (xc , yc ) θ α α (− R 0, ) 1 2 (R 0, ) ( 0,0 )

Fig. 34. Circular coverage area of radius R for a VLC hotspot. ( [83] ⃝c 2014 John Wiley& Sons).

The mobility in VLC for single-user cases is shown in Figure 34, where the mobile user enters the coverage at point A, with coordinates (xa,ya) known as the entering point and exits the coverage at point B, with coordinates (xb,yb) known as the exit point, as shown in Figure 34. For such an inclined path, the entrance angle is denoted by α1, and the exit angle is denoted by α2. The angle between entrance point and exit point is called the travelling angle which is denoted by θ. Assuming constant velocity, the path that the user travels can be simply expressed as

(x(t),y(t)) = ((xb − xa)tp + xa,(yb − ya)tp + ya), (64) where the parameters tp varies from zero (corresponding to the entrance point) to one (the exit point). Let us characterize the throughput as a function of the distance, i.e., the throughput at distance r is S(r). The instantaneous distance from the center of coverage to the mobile user can be calculated as √ 2 2 r = [(xb − xa)tp + xa] + [(yb − ya)tp + ya] . (65)

For single-user cases, a special case of a random walk mobility model with constant speed and limited movement direction was adopted to predict the mobile user move- ments throughout the hybrid WLAN-VLC hotspot coverage. Two assumptions have been made in dealing with the mobility of the user in WLAN and VLC hotspots coverage as shown in Figure 34. The ﬁrst assumption is that the travelling path within the

85 coverage of VLC hotspot or in the WLAN will be a straight, inclined or horizontal path. Secondly, we have assumed that the mobile user will only turn at the transition point between the VLC hotspot and WLAN. This transition point constitutes the exit point of the present hotspot and the entry point of the next VLC hotspot. For example, if A and B are the entry and exit points of the VLC hotspot then C is the entry point of next VLC hotspot. In that case, the the travelling distance in VLC hotspot will be |A−B|2 and the travelling distance in WLAN hotspot will be |B − C|2, where C is the transition points as shown in Figure 34. | · |2 is a Euclidean norm. It should be noted that the generation of random entry and exit point is done using the concept of travelling angles which is assumed to be uniformly distributed [88]. The size of the transferred information from the VLC hotspot during the dwelling time at velocity v and for travelling distance dvlc and with average throughput Svlc is

Ivlc = (Svlc.dvlc)/v. Similarly, the size of the transferred information from WLAN for travelling distance dwlan and with average throughput Swlan is Iwlan = (Swlan.dwlan)/v. However, the calculation of average throughput in hybrid coverage of WLAN-VLC will be as follows: ( ) Svlcdˆvlc Swlandˆwlan Shybrid = + , (66) dˆvlc + dˆwlan dˆvlc + dˆwlan where dˆvlc and dˆwlan are the total distance traveled by the mobile user in the VLC and WLAN coverage, respectively. Finally, the received ﬁle size in hybrid coverage of WLAN-VLC can be expressed as

Ihybrid = Shybrid tt, (67) where tt is the sum of residence time spent by a mobile user in the VLC spot cell and the WLAN coverage.

5.3 Hybrid WLAN-VLC: multi-user case

The scenario of data download in hybrid WLAN-VLC coverage for the multi-user case is shown in Figure 35. In this case, ten mobile users are shown for ﬁve mobility slots in the hybrid coverage of WLAN and VLC. Mobility slots represent the varying conditions, such as position, and direction at certain time steps. The inner small circles with radius 1 m represents the VLC coverage. On the other hand, the large circle represents the coverage of WLAN with radius 10 m. The mobile users are uniformly distributed

86 10

8 Spotlights 6

2 Mobile moves for 5 mobility slots

Y[m] −2

−4

−6 ← WLAN coverage → −8

−10 −10 −5 0 5 10 X[m] Fig. 35. Random walk mobility for five mobility slots in multi-user case. ( [83] ⃝c 2014 John Wiley& Sons). around the region of both WLAN and VLC. When the mobile users are inside the coverage of the VLC hotspot then they will communicate with the optical VLC interface otherwise they will communicate with radio WLAN interface of the mobile device. For a single-user case, we have considered constant velocity and unidirectional movement of the mobile user as discussed in Section 5.2. However, in practice, a mobility model should attempt to mimic the movement of a mobile user as realistically as possible. The most commonly used mobility model in the literature is the random waypoint model where users move randomly: the destination, speed and direction are all chosen randomly in the course of its movement. In random mobility, a velocity vector v = (v,θdir) is associated itself with two parameters, the first parameter is the magnitude of velocity v of the mobile user itself and the second parameters θdir is the direction of angle. The direction of angle is a random variable. The positions of the mobile user are defined by two dimensional coordinates (x,y). At every ∆t time step, both the velocity and the position of the mobile user are updated. Updated velocity and its direction can be expressed as [4]

x(t + △t) = x(t) + v(t) · cos(θdir(t)). (68)

y(t + △t) = y(t) + v(t) · sin(θdir(t)), (69)

87 Table 9. Simulation model parameters. ( [83] ⃝c 2014 John Wiley& Sons).

Simulation Parameters Value Transmit power 20 mW Photodiode responsivity 0.27 Semi-angle at half power 60o FOV at the receiver 30o Detector physical area of PD 28 mm2 Mobile user velocity [0.4 − 1] m/s No. of spotlights [4,6,8,9,18,27] BER [10−3,10−5,10−6] Rmax 32 m R 1 m

where v is the velocity assumed to be uniformly distributed where it varies between minimum velocity vmin and maximum velocity vmax. A mobile node at each moment has ﬁve different possible directions to move with each direction. In this work, we have considered that each mobile has a higher probability in moving in the same direction as the previous move. The assigned values of probabilities to each direction of angle are as follows: p0 = 0.7, p1 = 0.1, p2 = 0.05, so that p0 + 2 · p2 + 2 · p2 = 0.05.

5.4 Performance evaluation of hybrid WLAN-VLC

In this section, we evaluate the performance of the hybrid WLAN-VLC communication system for indoor environment at day and nighttime for the single-user as well as multi- user cases. Numerical results are achieved by using analytically and measurement- based throughput-distance relationship models for the VLC and WLAN, respectively. Monte-Carlo simulation results are also carried out for 1000 iterations for random mobility for the mobile user along with other system design parameters, as given in Table 9. The received file size of moving user serves as a performance metric for the single- user case and average connectivity and system throughput performance metrics for the multi-user case throughout the whole study of this work. Figure 36 shows the received file size in hybrid WLAN-VLC and VLC-only coverage for the indoor environment at nighttime at different BER requirements for the case. In this particular simulation environment, eight spotlights are considered. It is seen that the received file size decreases as the BER requirement gets higher both for the hybrid WLAN-VLC and VLC-only cases. It should be noted that the performance difference between the hybrid WLAN-VLC and VLC-only is not so significant at all BER require-

88 9 Hybrid 8 VLC

2 Average Filesize [Mbyte]

0 BER=10^−(3) BER=10^−(5) BER=10^−(6) Night Time

Fig. 36. Received Average Filesize for single-user case, [ v = 1m/sec, No. of spotlights = 8 ]. ( [83] ⃝c 2014 John Wiley& Sons). ments. This is due to the arrangement of VLC spotlights and mobility model for the single-user case (see Figure 33). The difference between two adjacent spotlights is 2 m. In such a case, the residence time in WLAN coverage will be much less in comparison to VLC spotlights coverage. As a result, the contribution of data downloaded file size from the WLAN will be much less in hybrid WLAN-VLC. Then, it can be said that in the case of a narrow corridor (for example 2 m width), VLC-only network can provide reasonable downloaded file size in comparison to the hybrid WLAN-VLC, if VLC spotlight hotspots are placed closely enough. Figure 37 shows the file size in hybrid WLAN-VLC and VLC-only coverage for the indoor environment during day at different BER requirements for a single-user case. It is clearly seen that the performance of hybrid WLAN-VLC and VLC-only decrease drastically at all BER levels in comparison to the indoor environment performance at night. This performance decrement happens in the daytime because of the impact of indirect sunlight on the photodetector. Figure 38 shows the three sets of CDF plots of received information file size at night for three different sets of spotlights deployed in a corridor or big hall for the single-user case. In this simulation, we mainly focus on studying the impact of the number of spotlights deployment in a certain length of corridor or room for hybrid WLAN-VLC

89 2.5 Hybrid VLC

1.5

0.5 Average Filesize [Mbyte]

0 BER=10^−(3) BER=10^−(5) BER=10^−(6) Day Time

Fig. 37. Received Average Filesize for single-user case, [v = 1m/sec, No. of spotlights = 8 ]. ( [83] ⃝c 2014 John Wiley& Sons).

0.9 # of spotlight=4 → 0.8

0.7

0.6

0.5 CDF 0.4 ↑ # of spotlight=8 0.3

0.2 ↑ # of spotlight=6

0.1

0 2 3 4 5 6 7 8 9 10 11 Filesize [Mbyte]

Fig. 38. CDF vs.Filesize at nighttime [ BER=10−3, v = 1m/sec ]. ( [83] ⃝c 2014 John Wiley& Sons).

90 1

0.9

0.8

0.7

0.6 ↑ # of spotlight=6

0.5 CDF 0.4

0.3

0.2 ↑ # of spotlight=4 ↑ # of spotlight=8 0.1

0 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 Filesize [Mbyte]

Fig. 39. CDF vs. Filesize at daytime [ BER=10−3, v = 1m/sec ]. ( [83] ⃝c 2014 John Wiley& Sons). networks. The purpose of this is to evaluate the performance of hybrid WLAN-VLC networks in terms of received file size while the mobile user moves through a corridor. It is obvious, as reflected in the simulation results, that when the number of spotlights increases, the received file size also increases. This is due to the fact that when the number of spotlights increase, the high data rate space of VLC and residence time of a mobile user in VLC are also increased in the combined coverage of WLAN-VLC. As a result, received file size is also increased. It is easily seen that with 50% probability the amount of information will be less than or equal to 4 Mbyte, 6 Mbyte and 8 Mbyte when the number of VLC spotlight cells will be 4, 6 , and 8 respectively Figure 39 shows the three sets of CDF plots of received information file size in the daytime for three different sets of spotlights deployed in a corridor. At daytime condition, the received file size will be considerably smaller than at night. With 50% probability, the amount of information will be less than or equal to 1 Mbyte, 1.6 Mbyte and 2 Mbyte when the number of VLC spotlight cells is 4, 6, and 8, respectively. It is also seen that at a certain probability (e.g., 50%) the downloaded received file size doubles as the number of VLC spotlights double. Figure 40 shows the CDF plots of received file size in case of the hybrid WLAN- VLC and WLAN-only at different velocities. The main purpose of this simulation is to

91 1

0.9

0.8

0.7

0.6

0.5 CDF 0.4

0.3

0.2 WO−WLAN:v=0.75 m/sec WO−WLAN: v=0.5 m/sec 0.1 W−WLAN: v=0.5 m/sec W−WLAN: v=0.75 m/sec 0 6 8 10 12 14 16 18 20 22 Filesize [Mbyte] Fig. 40. CDF vs. Filesize at different velocities at nighttime [ # of spotlights = 8,BER=10−3 ]. ( [83] ⃝c 2014 John Wiley& Sons). assess the impact of velocity on the performance of hybrid WLAN-VLC only. It is seen that the performance of both hybrid WLAN-VLC and WLAN-only decreases rapidly when velocity increases. For example, in the case of hybrid WLAN-VLC, it is seen that with 50% probability the amount of information will be less than or equal to 16.1 Mbyte and 10.7 Mbyte at v = 0.5 m/sec and v = 0.75 m/sec, respectively. However, at the same velocity, the performance difference is not much between the hybrid WLAN- VLC and WLAN-only. It is also seen in the simulation results that with 50% probability the amount of information will be less than or equal to 16.1 Mbyte, 15.4 Mbyte for hybrid WLAN-VLC and WLAN-only, respectively at v = 0.5 m/sec. This is due to the placement of number of VLC hotspots. Figure 41 shows the CDF plots of received ﬁle size in case of hybrid VLC-WLAN and WLAN only at daytime for different velocities. The same conclusion can be made as it is done for Figure 40. In both cases (Figure 40, and Figure 41), it can be concluded that when the velocity of the mobile user increases, the residence time either in hybrid WLAN-VLC or WLAN also decreases. As a result, data downloaded ﬁle size will be less when velocity increases. Figure 42 shows the performance of hybrid VLC-WLAN in terms of average connectivity for multuser and random walk mobility cases. Simulation has been performed for three different number of spotlight sets 9, 18 and 27 and the coverage range of the

92 1

0.9

0.8

0.7

0.6

0.5 CDF 0.4

0.3

0.2 W−WLAN:v=0.5 m/sec W−WLAN:v=0.75 m/sec 0.1 WO−WLAN:v=0.5 m/sec WO−WLAN: v=0.75 m/sec 0 1.5 2 2.5 3 3.5 4 4.5 5 5.5 Filesize [Mbyte] Fig. 41. CDF vs. Filesize at different velocities at day time [ # of spotlights = 8, BER=10−3 ]. ( [83] ⃝c 2014 John Wiley& Sons).

80 VLC 70 WLAN

20 Average Connectivity (%) 10

0 # of spotlights=27 # of spotlights=18 # of spotlights=9 # of Users N=10

Fig. 42. Average connectivity in multi-user case. ( [83] ⃝c 2014 John Wiley& Sons).

93 1.4 VLC WLAN 1.2 Hybrid

0.8

0.6

0.4

System Throughput [Gbps] 0.2

0 BER=10^−3 BER=10^−5 BER=10^−6 Number of Spotlights= 27

Fig. 43. System Throughput in multi-user case. ( [83] ⃝c 2014 John Wiley& Sons).

WLAN is 10 m. It can be easily be seen that as the number of spotlights increases, the average connectivity in VLC coverage also increases. However, the average connectivity in WLAN always outperforms average connectivity in VLC. This is due to the large coverage area covered by the WLAN than the coverage area covered by number of spotlights. Figure 43 shows the system throughput in VLC, WLAN and hybrid WLAN-VLC coverage at different BER requirements. Simulation has been performed for a number of users N = 10 and number of spotlights = 27. It shows that even with limited connection, the system throughput with VLC connectivity is better than WLAN connectivity. It can also be easily seen that the system throughput in hybrid VLC-WLAN coverage is far better than in individual VLC or WLAN coverage. System throughput in hybrid VLC-WLAN coverage is the summation of throughput in VLC and WLAN. This is due to the utilisation of both connectivity of VLC and WLAN by mobile users.

5.5 Conclusions

We have studied the performance of hybrid WLAN-VLC hotspot network in a data download on the move scenario. The performance of the hybrid WLAN-VLC hotspot

94 network is characterised by taking into account the throughput of radio and optical technology. Apart from the throughput parameter other key parameters such as velocity of the mobile user, travelling angle and FOV of the LED source are also taken into account. Finally, simulations have been performed to evaluate the performance of such network for the data downloading on the move scenario by taking into account performance metrics such as information ﬁle size. Simulation results helped to understand that there is a signiﬁcant impact on the performance due to the velocity of the mobile user. Moreover, simulation results reveal that as the number of spotlight cells placement in the corridor increases, the performance of the hybrid WLAN-VLC hotspot system increases. Hence, it is worthwhile to further investigate data download on the move issues in the hybrid WLAN-VLC hotspot scenario by taking into account detailed physical layer models, different network loads and various application scenarios.

95 96 6 Co-operative relays in VLC and hybrid WLAN-VLC networks

In this chapter, we evaluate the performance of relay-assisted VLC and relay-assisted hybrid WLAN-VLC hotspot networks. In relay-assisted VLC networks, we have developed an optical relay based multi-hop communication simulation environment. In this respect, an overview and background of multi-hop relaying techniques are introduced in Section 6.1. The scenario of co-operative relays in multi-hop VLC communication is described in Section 6.2. The motivation for this work is to use co-operative relays to reduce the on-demand recovery of connection. A relay selection algorithm is also introduced in this work and described in Section 6.2.1. Numerical results for relay-assisted VLC networks are shown in Section 6.3. The simulation environment of relay-assisted VLC hotspot networks has been further extended to relay-assisted hybrid WLAN-VLC hotspot networks. The purpose of this extension is to exploit the beneﬁts of each technology such as reliable connectivity of WLAN and higher throughput and improved energy efﬁciency of VLC. The scenario of co-operative hybrid WLAN-VLC networks is described in Section 6.3.1. Energy consumption analysis is described in Section 6.3.2. Numerical results for relay-assisted hybrid WLAN-VLC networks are given in Section 6.4. Finally, conclusions are drawn in Section 6.5.

6.1 Overview and background

In a multi-hop relaying technique, if a MT is unable to connect to the AP directly then it can connect with the AP through relay station (RS). The first benefit of multi- hop relaying comes from the reduction in the overall path loss between the AP and the MT. Another benefit of multi-hop relaying is the path diversity gain that can be achieved by selecting the most favorable, multi-hop path in the shadowed environment [153, 154]. The performance of such multi-hop relay based networks depends on many factors, such as the number of RSs in the cell, transmission range of MT, mobility of the relaying node, traffic volume in the network and others. The targeted scenario of multi- hop communication is shown in Figure 44. The description of the scenario is given in Section 6.2. Relay schemes are not considered in the current IEEE 802.15.7 standard, but this standard supports device discovery mechanisms, through which short-range co-

97 LED Source

LOS

Cooperative communication

NLOS

Single-hop

LOS region Multihop region

Idle MT

Fig. 44. Multi-hop communication. operation can also be possible within homogeneous networks [36]. In this regard, many research works have proposed the future adoption of relay schemes in LED identiﬁca- tion (LED-ID) systems [155, 156]. The LED-ID system typically uses the LOS channel to achieve high data rates and bright illumination. In the LED-ID system, the visible spectrum is used for downlink and the IR band is used for uplink communication. In [155], three scenarios have been presented: i) Direct communication, ii) LED-ID co- operative communication and iii) LED-ID multi-hop communications. The motivation for that work is to use co-operative relays and multi-hop communications to reduce latency and on-demand recovery. In [156], ﬁrst a scenario on the shadow region problem is presented and then a co-operative optical relay for LED-ID systems is proposed to extend the coverage range of LED-ID networks. In both [155, 156], co-operative MAC protocols were proposed using co-operative relay schemes for downlink communications in LED-ID systems. In [119], two novel MAC protocols and a novel scheme are studied and designed for multi-hop multi-access in VLC communications. In [6, 146], the main goal of the work is to expand the limited coverage of VLC. In this respect, the connectivity performance of MTs has been performed using varying number of simulation parameters. In [156], the authors develop a VLC based multi-hop audio data transmission system prototype using RS. The demonstration of their prototype has been shown to transmit a high quality of audio signal at a long distance. In their

98 proposed multi-hop transmission system, audio data is encoded using the Sony/Philips Digital Interface Format (S/PDF) standard. The multi-hop system consists of two RSs. At each relay, the received digital audio signal is improved and ampliﬁed before being transmitted to the next RS. At the receiver, the encoded audio signal is decoded, ampliﬁed and converted into an audio signal.

6.2 Scenario description

The scenario of co-operative relays in multi-hop VLC communications is shown in Figure 45. In this ﬁgure, the coverage area of a VLC source is shown in the segmented area of a typical conference room. The inner circle represents the coverage area of VLC which is denoted as AI, known as the single-hop region (SR). On the other hand, the outer circle represents the coverage area of multi-hop region (MR) which is denoted as

Ao. The coverage range of the inner and outer circles are rsr and Rmr respectively. The mobile terminals are uniformly distributed around the region of both inner and outer circles. We assume that mobile terminals in the inner circle will have LOS communications to the AP and transmit directly to AP. The mobile terminals in the outer circle will connect via relays situated in the inner circle. However, all the MTs that are outside of the inner region will not be able to connect unless any relay presents an active transmitter transmission range. It should be noted that each MT will be considered as an active transmitter when it is in transmission mode, otherwise MTs will act as relays.

For simplicity, we consider the coverage range dmin of MT/relay to be circular. At any instance, MT may act as a relay and during that time it may receive a packet from a particular active transmitter resides in AI. The details of varying conditions of MT in terms of position and operation mode are given below: As seen from Figure 45, there are ten MTs in the coverage of both SR and MR. The position of each MT is shown for three mobility slots. A mobility slot represents the varying conditions such as position, direction and operation mode of an MT at a certain time step. Active transmitters in the SR region are marked as green circles. On the other hand, active transmitters in the MR are marked as red circle. It is also seen that relays are marked as ’+’ sign. Connected with active transmitters and disconnected relays are represented as red + and with blue + respectively. Inactive transmitters are marked as circle (o) with black colour. The operation mode of MTs is also seen in Figure 45. It should be noted that each MT can change its operation mode from relay (+) to active node (o) or vice versa.

99 4 A O 3 R mr 2 A I 1 ← Relays 0 ← MTs in A

Y[m] I r −1 sr

MTs in A → −2 O

−3

−4 −4 −2 0 2 4 X[m]

Fig. 45. VLC hotspot (spotlighting cell) coverage. ( [74] ⃝c 2013 IEEE).

6.2.1 Relay selection

In this work, the scheduling process is mainly controlled by two parameters: relay probability p and transmission range dmin of MTs. In the relay selection scheme as shown in Figure 46, we will follow the following procedure: Let S denotes the set of

MTs in AI as relay candidates and the distance from each relay to the AP is li where i ∈ S. For an active transmitter, j in AO, li j(t) denotes the distance between relay i and active transmitter j at time ti. In this respect, we choose the relay node as follows:

argmin = min{λscale li + (1 − λscale)li j}, i ∈ S (70) i where λscale ∈ [0,1] is a scaling factor. The relationship between the number N of MTs, radius of the multi-hop region Rmr and radius of transmission range dmin of MT/relay can be expressed as [157] · 2 µ N dmin = 2 . (71) Rmr It can be seen that as the radius of multi-hop region increases, the value of node density decreases [157, 158].

100 YES Is VLC AP within the range

NO Is there any relays within the range

YES

NO Is any relay closer to the VLC AP than the mobile itself YES

Transmit to the Transmit directly to Stay idle best relay the VLC AP

Fig. 46. Relay selection algorithm.

6.3 Performance evaluation of co-operative VLC

In this section, numerical results are given on the performance of user connectivity. In this respect, Monte-Carlo simulation results are obtained for 100,000 iterations by using the system design parameters as given in Table 10 and other required parameters described in Section 6.2. Figure 47 shows the CDF vs. average percentage of connectivity of MTs with the AP. In this simulation, we investigate the impact of numbers of MTs on connectivity by considering the single-hop coverage range (2 m), multi-hop coverage range (4 m), and relay probability (p = 0.2). The velocity v is uniformly distributed. It is seen that as the number of MTs increases, the connectivity performance of MTs also increases. This is due to the fact that when the number of MTs increases, the user density also increases, hence the probability of getting a higher number of relays also increases within

Table 10. Simulation model parameters.

Simulation parameters Value No. of nodes [5,10,15] Transmission range of MT 2 m Relay probability [0.2,0.4,0.6] Single hop cell radius 2 m Multi-hop cell radius [4,6,8] m Velocity [0.1,0.2,0.9] m/s

101 1

0.8

WR 0.6 CDF 0.4 N=10 (wor) N=15 (wor) WOR N=20 (wor) 0.2 N=10 (wr) N=15 (wr) N=20 (wr) 0 0 20 40 60 80 Average Connectivity (%)

Fig. 47. CDF vs. average percentage of connectivity, [ Rmr = 4 m, p = 0.2, dmin = 2 m]. ( [74] ⃝c 2013 IEEE).

the transmission range of dmin of MT in AO. As a result, with relay (WR) case, the average percentage of connectivity will increase. On the other hand, in the case of without relay (WOR), an active transmitter in MR will not be able to connect with relays, which results in no user connectivity, hence the performance of average percentage of connectivity will decrease. For example, with 50% of probability for N = 15, the average number of MTs are connected is less than or equal to 37 with relay case compared to less than or equal to 18 for without relay. Figure 48 shows the CDF vs. average percentage of user connectivity for WR and WOR for different coverage ranges of MR. Relay probability and velocity of MTs are the same as considered for the simulation results for Figure 47. In this work, we have considered two-hop communications. According to our assumptions, only those transmitters that are in MR (AO) will communicate to AP via with only those relays those who are in SR (AI). So, when the coverage range of MR increases, the distance between the active transmitter and relays goes beyond the transmission range of the active transmitter and it will unable to connect to the AP via the relay. Hence, the more the coverage range of MR increases, the more the number of active transmitters in AI will be unable to connect to the AP. This phenomenon is true for both WR and WOR. For

102 1

0.8 WR & WOR R=8 m

WR & WOR R=6 m 0.6 CDF 0.4 WR R= 4 m WR R= 6 m WR R= 8 m 0.2 WOR R= 4 m

WR & WOR R=4 m WOR R= 6 m WOP R = 8 m 0 0 20 40 60 80 Average Connectivity (%)

Fig. 48. CDF vs. average percentage of connectivity, [ N = 10, p = 0.2, rsr = 2 m, dmin = 2 m]. ( [74] ⃝c 2013 IEEE). example, in the case of WR, there is a 50% probability the average number of MTs connected will be equal to or or less than 37% when Rmr = 4 m. With the same probability the average percentage of MTs will be connected will be equal to less than 10% and 5% for Rmr = 6 m and for Rmr = 8 m respectively. Figure 49 shows the CDF vs. average percentage of connectivity of users with the AP. In this simulation, three different constants velocities are chosen unlike the other simulations results (Figure 47, and 48), where the considered velocity for each user was chosen as randomly distributed. Again, these simulations demonstrate that the performance of user connectivity increases with WR in comparison to WOR. However, it is seen that the considered velocities in the range [0.1 − 0.9] m/s almost have the same impact on the performance of user connectivity both for WR and WOR. There is no signiﬁcant difference among the considered velocities.

6.3.1 Co-operative hybrid WLAN-VLC networks

The scenario of multi-hop communication (Section 6.2) has been extended to hybrid WLAN-VLC systems where the coverage of WLAN is assumed to be provided. It is assumed that MT is equipped with both radio and optical wireless interfaces. The

103 1

0.8 With Relay 0.6 Without Relay CDF 0.4 v= 0.1 m/sec v= 0.2 m/sec v= 0.9 m/sec 0.2 v= 0.1 m/sec v= 0.2 m/sec v= 0.9 m/sec 0 0 20 40 60 80 Average Connectivity (%)

Fig. 49. CDF vs. average percentage of connectivity [ N = 10, rsr = 2 m, dmin = 2m]. ( [74] ⃝c 2013 IEEE).

coverage range of VLC is assumed to be 2 m. The transmission range of the VLC interface is assumed to be 1 m. Furthermore, we have assumed that VLC will be the primary system, meaning the MT will communicate with the AP either through single- hop or multi-hop communication using an optical wireless interface, or then it will communicate with the WLAN radio interface. If the MT in the multi-hop region and does not ﬁnd any relay to communicate with, the AP will then communicates with the WLAN. Active transmitters (e.g., that are in AI) that are connected directly are considered as VLC users. On the other hand, the active transmitters that are in multi-hop region that connected via a relay are considered as multi-hop VLC users, and those who will connect via the WALN interface are considered as WLAN users. Both VLC and multi-hop VLC users can also be considered as optical interface users. The deﬁnition of other concepts such as single-hop and multi-hop regions, and the relay selection algorithm are considered to be same as described in Section 6.2.

104 4

WLAN Coverage WLAN Coverage 3 5 1 2 6

1 8

Y[m] 3 4 7 −1 VLC Coverage

−2 2

−3 WLAN Coverage Multihop Coverage −4 −5 −4 −3 −2 −1 0 1 2 3 4 5 X[m]

Fig. 50. Layout of hybrid coverages.

6.3.2 Energy consumption analysis

Energy consumption during the transmission of a data packet by MT in the case of a single-hop can be expressed as [8]

Lp ESh = TtxPtx = Ptx , (72) Rr where Ttx is the transmission time, Ptx is the transmitted power, Lp is packet length and

Rr is the transmission rate. It is not straight forward to compare energy consumption for the transmission of data of different wireless systems. However, energy consumption for transmission of data for IEEE 802.11g and optical link have been compared and are shown in Table 11 [159]. In the multi-hop case, the total energy consumption is the sum of energy consumption at the MT and the relay, which can be expressed as

ETh = EMT−RS + ERS−AP, (73) where EMT−RS and ERS−AP are the energy consumption by MT and relay respectively. In this work, the data rate of both relay and active transmitter for VLC interface are considered as same. In that case, energy consumption for multi-hop can be represented

105 Table 11. Energy consumption.

Standard Power consumption (W) IEEE 802.11 (g) 1.25 Optical Link 0.3 as [160, 161] Lp ETh = (Rn + 1)Ptx , (74) rvl where Rn represents the number of hops. Energy consumption in the hybrid system where energy consumption has been calculated according to the technology speciﬁc transmission power and data rate which can be represented as

T T T Pw Pv Pvl EH = Nw ∑ + Nsh ∑ + Nmh ∑ , (75) m=1 rw m=1 rv m=1 rvl where Nw, Nsh, Nmh represent the number of MTs in WLAN, single-hop and multi-hop coverage respectively. Similarly, Pw, Pv, Pvl are the transmission powers of WLAN and VLC interface respectively. T is the number of mobility slots. In all cases, the length of transmitted packet Lp is assumed to be 1.

6.4 Performance evaluation of hybrid WLAN-VLC networks

In this section, the connectivity and energy consumption performance of hybrid WLAN- VLC will be evaluated using system design parameters as given in Table 12.

Table 12. Simulation model parameters.

Simulation Parameters Value No. of Nodes [5,10,15,20, 25] Relay probability [0.2,0.4,0.6] Single hop cell radius [2 m] Multi-hop cell radius [4,5,6,7,8,9,10] m Velocity [0.5, 0.7,0.9] m VLC data rate 10 Mbps WLAN data rate 5 Mbps

Figure 51 shows the three sets of plots of average connectivity in hybrid WLAN- VLC coverage. In this particular simulation, results of 20 MTs are involved with relay probability p = 0.2. In this ﬁgure, the blue plot shows user connectivity via WLAN.

106 100

40 VLC Multihop VLC WLAN 20 Average Connectivity (%)

0 2 2.5 3 3.5 4 4.5 5 Coverage Ratio Fig. 51. Coverage range ratio vs. average connectivity. ( [6] ⃝c 2013 IEEE).

The red plot shows single-hop connectivity via VLC. Finally, the green plot shows user connectivity in a two-hop case. It is seen that as the coverage range ratio increases a large number of active transmitters will be connected via WLAN. This is because when the coverage ratio increases, a large number of users will be positioned at the outer area of the boundary. Such a large number of active transmitters can not connect to the AP via relay due to the short transmission range dmin of MT. As a result, MTs will connect to the AP via WLAN. Figure 52 shows the simulation results of average percentage connectivity vs. number of MTs. In this simulation, we investigate the impact of numbers of MTs on connectivity performance by considering single-hop coverage range (2 m), multi-hop coverage range (4 m), and relay probability p = 0.2. It is seen that as the number of MTs increases, the connectivity performance of MTs is also increased for multi-hop VLC and WLAN interfaces. This is due to the fact that when the number of MTs increases, the node density also increases. Hence, the probability of getting a higher number of relays also increases within the transmission range dmin of MT in Ao. However, many of the MTs in Ao will be disconnected because of not ﬁnding the relays; in such cases the MTs will be connected via WLAN interface. As a result, when the number of users

107 60 VLC Multihop VLC 50 WLAN

Average Connectivity (%) 10

0 5 10 15 20 25 Number of Users (N) Fig. 52. Number of users vs. average connectivity. ( [6] ⃝c 2013 IEEE). increases, the co-operative connectivity as well as connectivity using complementary technology WLAN will increase to extend the short coverage range of VLC. Figure 53 shows the CDF vs. total energy consumption for transmission of active transmitters for ten mobility slots. The red plot shows the energy consumption for transmission when all MTs use radio (WLAN). On the other hand, the blue plot shows the energy consumption for transmission when both radio (WLAN) and optical (VLC) technologies are used for transmission. It is seen that when the active transmitters uses hybrid system for transmission, then the total energy consumption is less than when they use only radio (WLAN) system. For example, at 50% probability total average energy consumption by hybrid system is less than 23 µJ in comparison to less than 36 µJ for the WLAN system. Almost 1.58 times less energy is consumed by hybrid users (VLC, VLC multi-hop, WLAN) users in comparison to only WLAN interface users.

6.5 Conclusions

In this chapter, we have evaluated the performance of relay-assisted VLC and relay assisted hybrid WLAN-VLC hotspot networks. In relay-assisted VLC networks, we developed an optical relay based multi-hop communication simulation environment. Re- lay selection algorithm is also introduced in this work. The motivation of this work is

108 1 Hybrid 0.9 WLAN

0.8

0.7

0.6

0.5 CDF 0.4

0.3

0.2

0.1

0 15 20 25 30 35 40 45 Total Energy Consumption (µJ)

Fig. 53. CDF vs. Average aggregated energy consumption. ( [6] ⃝c 2013 IEEE). to use co-operative relays to reduce on-demand recovery. In this respect, the percentage of connectivity is chosen as the performance metric to evaluate the performance of such co-operative relay-assisted networks. The simulation environment of the relay- assisted VLC hotspot network is further extended to the relay-assisted hybrid WLAN- VLC hotspot network. The purpose of this extension is to exploit the benefits of each technology such as the reliable connectivity of WLAN and the higher throughput and improved energy efficiency of VLC. Simulations have been performed in order to evaluate the connectivity and energy efficiency performance of such homogeneous and hybrid networks. Simulation results reveal that user connectivity and energy efficiency depend on user density, coverage range ratio between single-hop and multi-hop, relay probabilities and mobility of the user. Finally, it can be concluded that hybrid radio- optical wireless systems have a positive impact on the performance of user connectivity and energy consumption of the mobile device with the expense of providing complexity in the communication architecture.

109 110 7 Summary and future directions

The imagination must feed on the past in creating the future.

Bulletin of the Atomic Scientists (Vol.15, No.2)

In this chapter, we will summarise and conclude the research work presented in this thesis. The main ﬁndings are highlighted and conclusions are drawn. Finally, future work related to VLC and hybrid WLAN-VLC are presented.

7.1 Summary

The purpose of this thesis is to study the performance of stand-alone VLC and hybrid WLAN-VLC based networks. Performance analysis of such networks is needed before they deployed in practice. In this regard, first we created three scenarios namely a) stand-alone VLC hotspot b) hybrid WLAN-VLC hotspot, c) co-operative optical relay- based networks. Secondly, we studied the link design parameters for both VLC and WLAN. Thirdly, we developed the mathematical framework to study the performance of such networks in data download on the move scenarios. In this regard, first we highlighted the three grand challenges such as spectrum scarcity, reduction of (CO2) emissions for green communication, and energy efficiency of mobile devices in Chapter 1. VLC may play an important role in mitigating some of these challenges, being a key networks for 5G technology and beyond. An extensive literature review, especially on data rate improvement in hardware testbed and hybrid WLAN-VLC networks, was also carried out in Section 1.3.1 and Section 1.3.2 respectively in Chapter 1. The fundamentals of VLC transceiver systems, such as different types of LEDs, LED I-V characteristics and characteristics of photodetectors were studied in Chapter 2. The presence of different types of noise and interference in the optical channel were also discussed in this chapter. Finally, different modulation techniques in VLC were discussed in Chapter 2. In practice, many of the design parameters such as FOV of the receiver, divergence angle of LED, degree of directionality between transmitter and receiver as well as ambient noise were considered in VLC link designing. We described these four basic link topologies. These four link characteristics were classified as: directed-LOS, non-

111 directed-LOS, non-directed-NLOS and tracked. The coverage and data rate of such link designs were also provided in Chapter 3. Many other advanced link topologies such as a multi-spot diffuse link, DSD were also discussed in Chapter 3. In this thesis, non- directed LOS with tracking link topology was assumed exist between transmitter and receiver. In Chapter 4, we investigated the first scenario (data download on the move in VLC coverage). In this respect, first we introduced the data download on the move scenario with related concepts such as moving-in, moving-out, travelling angle in Section 4.1. A mathematical framework to calculate the average throughput and file size were also introduced in Section 4.3. Theoretical and polynomial-based throughput-distance relationship models were also developed for the day and night conditions. Transformation of straight inclined travelling paths to equivalent horizontal travelling paths using rotational technique was used in this chapter. It was showed that this rotation (inclined path to horizontal path) simplifies the calculation of average throughput. Measurement based shot noise was also included in the mathematical framework. Analytical solutions for calculating average throughput and file size for single-user case were provided. Simulations were also performed to evaluate the performance of such hotspot network in the context of data download on move scenario. The simulation results revealed that there was a large impact of background noise on the performance of VLC hotspots. As expected, in both indoor and outdoor environments, VLC hotspot performed better at night than during the day. The performance of VLC hotspot networks was also quantified in terms of received file size at different bit error rate requirements and velocities of the mobile user. In Chapter 5, we studied the performance of second scenario (data downloading in hybrid WLAN-VLC networks). We considered data download on the move scenario in an indoor environment for single-user as well as for multi-user cases. In this hybrid WLAN-VLC hotspot, both the WLAN and the VLC were characterized by their throughput and communication range. Short-range VLC small cells were considered for high data rate region. A constant data rate was assumed for the VLC small cell, in this respect three sets of data rates at three different BER requirements were provided in Section 5.1.1. The coverage range of the VLC small cell was considered to be one meter. On the other hand, an integrated complementary WLAN hotspot was used for providing reliable connectivity and moderate data rate support. In this respect, measurement-based coverage range and data rate were provided for WLAN as discussed in Section 5.1.2. Descriptions of environments for both single as well as

112 multi-users for two different mobility models were created and discussed in Sections 5.2 and 5.3 respectively. The impact of the velocity of mobile user(s) and the number of VLC small cells present in the considered service area were also considered on the performance of hybrid WLAN-VLC small cells. Simulations had been performed to evaluate the performance of such network for data downloading on the move scenario by taking into account performance metrics such as file size, average connectivity and system throughput. Simulation results revealed that the considered hybrid WLAN-VLC performs always better than standalone VLC-only or WLAN-only hotspot both for single and multi-user cases. Finally, in Chapter 6, we studied the performance of a third scenario (co-operative relays in VLC and hybrid WLAN-VLC networks). In relay-assisted VLC networks, we developed an optical relay based multi-hop communication simulation environment. A relay selection algorithm was also introduced in this work as discussed in Section 6.2.1. The motivation of this work was to use co-operative relays to setup the communication between AP and transmitter if direct communication is not available. The percentage of connectivity was chosen as a performance metric to evaluate the performance of such co-operative relay-assisted networks. The simulation environment of the relay-assisted VLC hotspot network was further extended to the relay-assisted hybrid WLAN-VLC hotspot network as discussed in Section 6.4. The purpose of this extension was to exploit the benefits of each technology, such as the reliable connectivity of WLAN and the higher throughput and improved energy efficiency of VLC. Simulations had been performed in order to evaluate the connectivity and energy efficiency performance of such homogeneous and hybrid networks. Simulation results revealed that user connectivity and energy efficiency depend on user density, coverage range ratio between single-hop and multi-hop, relay probabilities and mobility of the user. Finally, it can be concluded that hybrid radio-optical wireless systems had a positive impact on the performance of user connectivity and energy consumption of the mobile device. The framework of this work may be applied in future deployment of VLC communication system in indoor and and outdoor environments. It is also noticed that the performance of such a network degrades rapidly in the presence of natural as well as artificial noise. Therefore, suitable filtering methods, for example, optical and electrical filtering method can be used to mitigate these problems. Thesis also dealt with the hybrid communication systems and provide also the benefits of using it.

113 7.2 Future work

This thesis essentially focuses on the performance analysis of future VLC and hybrid WLAN-VLC network systems. In doing so, we had few assumptions to simplify our analytical models. For example, connection set-up time, link discovery mechanism, fast link recovery algorithm, switching algorithm between VLC and WLAN or vice versa were not taken into account in the analytical model. Existing mathematical framework of data downloading on move scenario can be extended by considering any of these assumptions. This thesis deals also with data download on move scenario in multi-user environments. However, multi-user interference is not consider in this thesis. So, existing simulation environment can be further extended by considering multiple access interference and resource allocation mechanism to avoid the multi-user interference. Cell zooming technique studied in [152], can be one of the possible solutions to mitigate the interference produced in multi-user VLC communication environment. We have also shown that hybrid WLAN-VLC performs better than the standalone VLC and WLAN in terms of connectivity and throughput. In this respect, development of both theoretical and practical analysis of PHY and MAC designs for hybrid WLAN- VLC will be interesting topic to explore. Therefore, future objective will be to design and analyze suitable unified PHY techniques including channel coding and modulation techniques for energy and spectrally efficient hybrid WLAN-VLC systems. Moreover, in this thesis seamless handover has been assumed to select the best available VLC or WLAN wireless networks. In practice, a suitable handover technique is needed to switch between these heterogeneous networks. However, all-radio vertical handover (VHO) approaches cannot be applied to the hybrid radio and LOS optical circumstances, since the channels have different properties. Hence, characteristics of radio and optical channel models have to be revisited and modified. Feasibility study of integrating VLC in 5G HetNet solution in emerging software defined networks (SDN) will be interesting topic to explore as well.

114 References

1. Hanzo L, Haas H, Imre S, O’Brien D, Rupp M & Gyongyosi L (2012) Wireless myths, realities, and futures: From 3G/4G to optical and quantum wireless. Proceedings of the IEEE 100(Special Centennial Issue): 1853–1888. 2. Fettweis G & Alamouti S (2014) 5G: Personal mobile internet beyond what cellular did to telephony. IEEE Communications Magazine 52(2): 140–145. 3. Tarkoma S & Ailisto H (2013) The internet of things program: the Finnish perspective. Communications Magazine, IEEE 51(3): 10–11. 4. Chowdhury H & Katz M (2013) Data download on move in indoor hybrid (radio-optical) wlan-vlc hotspot coverages. In: Vehicular Technology Conference (VTC Spring), 2013 IEEE 77th, pp. 1–5. 5. Kavehrad M (2010) Sustainable energy-efﬁcient wireless applications using light. IEEE Communications Magazine 48(12): 66–73. 6. Chowdhury H, Ashraf I & Katz M (2013) Energy-efﬁcient connectivity in hybrid radio- optical wireless systems. In: Wireless Communication Systems (ISWCS 2013), Proceed- ings of the Tenth International Symposium on, pp. 1–5. 7. Olsson M, Cavdar C, Frenger P, Tombaz S, Sabella D & Jantti R (2013) 5GrEEn: Towards green 5G mobile networks. In: 2013 IEEE 9th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob), pp. 212–216. 8. Radwan A, Albano M, Rodriguez J & Verikoukis C (2012) Analysis of energy saving using cooperation use-case: WiFi and WiMedia. In: Future Network Mobile Summit (Fu- tureNetw), 2012, pp. 1–10. 9. Chowdhury H, Bagheri H, Ashraf I, ul Hassan ST & Katz M (2013) Techno-economic analysis of visible light communications. In: Wireless Communication Systems (ISWCS 2013), Proceedings of the Tenth International Symposium on, pp. 1–5. 10. Chowdhury H, Lehtomäki J, Mäkelä JP & Kota S (2010) Data downloading on the sparse coverage-based wireless networks. Journal of Electrical and Computer Engineering 2010. 11. Osseiran A, Boccardi F, Braun V, Kusume K, Marsch P, Maternia M, Queseth O, Schell- mann M, Schotten H, Taoka H, Tullberg H, Uusitalo MA, Timus B & Fallgren M (2014) Scenarios for 5G mobile and wireless communications: the vision of the METIS project. IEEE Communications Magazine 52(5): 26–35. 12. Mogensen P, Pajukoski K, Tiirola E, Lahetkangas E, Vihriala J, Vesterinen S, Laitila M, Be- rardinelli G, Da Costa G, Garcia L, Tavares F & Cattoni A (2013) 5G small cell optimized radio design. In: Globecom Workshops (GC Wkshps), 2013 IEEE, pp. 111–116. 13. Wang CX, Haider F, Gao X, You XH, Yang Y, Yuan D, Aggoune HM, Haas H, Fletcher S & Hepsaydir E (2014) Cellular architecture and key technologies for 5G wireless communication networks. IEEE Communications Magazine 52(2): 122–130. 14. Gohil A, Modi H & Patel SK (2013) 5G technology of mobile communication: A survey. In: Intelligent Systems and Signal Processing (ISSP), 2013 International Conference on, pp. 288–292. 15. Boccardi F, Heath RW, Lozano A, Marzetta TL & Popovski P (2014) Five disruptive technology directions for 5G. IEEE Communications Magazine 52(2): 74–80.

115 16. Wu Z & Little T (2010) Network solutions for the ,LOS problem of new indoor free space optical system. In: Communication Systems Networks and Digital Signal Processing (CSNDSP), 2010 7th International Symposium on, pp. 582–587. 17. Haruyama S (2013) Visible light communication using sustainable LED lights. In: ITU Kaleidoscope: Building Sustainable Communities (K-2013), 2013 Proceedings of, pp. 1–6. 18. Evans DD (2008) High brightness matrix LED assembly challenges and solutions. In: 2008 58th Electronic Components and Technology Conference, pp. 1704–1708. 19. Minh HL, Ghassemlooy Z, O’Brien D & Faulkner G (2010) Indoor gigabit optical wireless communications: Challenges and possibilities. In: 2010 12th International Conference on Transparent Optical Networks, pp. 1–6. 20. Langer KD, Grubor J, Bouchet O, Tabach ME, Walewski JW, Randel S, Franke M, Nerreter S, O’Brien DC, Faulkner GE, Neokosmidis I, Ntogari G & Wolf M (2008) Optical wireless communications for broadband access in home area networks. In: 2008 10th Anniversary International Conference on Transparent Optical Networks, volume 4, pp. 149–154. 21. Ma H, Lampe L & Hranilovic S (2013) Integration of indoor visible light and power line communication systems. In: Power Line Communications and Its Applications (ISPLC), 2013 17th IEEE International Symposium on, pp. 291–296. 22. Komine T & Nakagawa M (2003) Integrated system of white LED visible-light communication and power-line communication. IEEE Transactions on Consumer Electronics 49(1): 71–79. 23. Ndjiongue AR, Ferreira HC, Ouahada K & Vinckz AJH (2014) Low-complexity SOCPBFSK-OOK interface between PLC and VLC channels for low data rate transmission applications. In: Power Line Communications and its Applications (ISPLC), 2014 18th IEEE International Symposium on, pp. 226–231. 24. Jovicic A, Li J & Richardson T (2013) Visible light communication: opportunities, challenges and the path to market. IEEE Communications Magazine 51(12): 26–32. 25. Heatley DJT, Wisely DR, Neild I & Cochrane P (1998) Optical wireless: the story so far. IEEE Communications Magazine 36(12): 72–74, 79–82. 26. Elgala H, Mesleh R & Haas H (2011) Indoor optical wireless communication: potential and state-of-the-art. IEEE Communications Magazine 49(9): 56–62. 27. Kumar N, Alves LN & Aguiar RL (2009) Design and analysis of the basic parameters for trafﬁc information transmission using VLC. In: Wireless Communication, Vehicular Tech- nology, Information Theory and Aerospace Electronic Systems Technology, 2009. Wireless VITAE 2009. 1st International Conference on, pp. 798–802. 28. Higgins MD, Leeson MS & Green RJ (2012) An analysis of intra-vehicle optical wireless communications from a passenger perspective. In: 2012 14th International Conference on Transparent Optical Networks (ICTON), pp. 1–4. 29. Elgala H, Mesleh R & Haas H (2009) Practical considerations for indoor wireless optical system implementation using OFDM. In: Telecommunications, 2009. ConTEL 2009. 10th International Conference on, pp. 25–29. 30. URI: http://irdajp.info/speciﬁcations.html . 31. Ghassemlooy Z, Popoola W & Rajbhandari S (2012) Optical Wireless Communications: System and Channel Modelling with MATLAB. CRC Press, Inc., Boca Raton, FL, USA, 1st edition. 32. Elgala H, Mesleh R & Haas H (2011) Indoor optical wireless communication: potential and state-of-the-art. IEEE Communications Magazine 49(9): 56–62.

116 33. O’Brien DC, Faulkner GE, Jim K, Zyambo EB, Edwards DJ, Whitehead M, Stavrinou P, Parry G, Bellon J, Sibley MJ, Lalithambika VA, Joyner VM, Samsudin RJ, Holburn DM & Mears RJ (2003) High-speed integrated transceivers for optical wireless. IEEE Communications Magazine 41(3): 58–62. 34. CP-1221 (2007) Japan electronics industry technology association. In: Japanese Visible Light Communications System. 35. CP-1222 (2007) Japan electronics industry technology association. In: Japanese Visible Light ID System. 36. IEEE1 (2011) IEEE standard for local and metropolitan area networks–part 15.7: Short- range wireless optical communication using visible light. IEEE Std 802.15.7-2011 pp. 1 –309. 37. Rajagopal S, Roberts RD & Lim SK (2012) IEEE 802.15.7 visible light communication: modulation schemes and dimming support. IEEE Communications Magazine 50(3): 72– 82. 38. Jungnickel V, Uysal M, Seraﬁmovski N, Baykas T, O’Brien D, Ciaramella E, Ghassemlooy Z, Green R, Haas H, Haigh PA, Jimenez VPG, Miramirkhani F, Wolf M & Zvanovec S (2015) A european view on the next generation optical wireless communication standard. In: Standards for Communications and Networking (CSCN), 2015 IEEE Conference on, pp. 106–111. 39. URI: http://vlca.net . 40. Minh HL, O’Brien D, Faulkner G, Zeng L, Lee K, Jung D & Oh Y (2008) High-speed visible light communications using multiple-resonant equalization. IEEE Photonics Tech- nology Letters 20(14): 1243–1245. 41. Mangold S (2012) Visible light communications for entertainment networking. In: 2012 IEEE Photonics Society Summer Topical Meeting Series, pp. 100–101. 42. Schmid S, Corbellini G, Mangold S & Gross TR (2012) An led-to-led visible light communication system with software-based synchronization. In: 2012 IEEE Globecom Work- shops, pp. 1264–1268. 43. Giustiniano D, Tippenhauer NO & Mangold S (2012) Low-complexity visible light networking with LED-to-LED communication. In: Wireless Days (WD), 2012 IFIP, pp. 1–8. 44. Uysal M & Nouri H (2014) Optical wireless communications x2014; an emerging technology. In: 2014 16th International Conference on Transparent Optical Networks (ICTON), pp. 1–7. 45. Gfeller FR & Bapst U (1979) Wireless in-house data communication via diffuse infrared radiation. Proceedings of the IEEE 67(11): 1474–1486. 46. Kahn JM & Barry JR (1997) Wireless infrared communications. Proceedings of the IEEE 85(2): 265–298. 47. Valadas RT, Tavares AR, d Duarte AM, Moreira AC & Lomba CT (1998) The infrared physical layer of the IEEE 802.11 standard for wireless local area networks. IEEE Commu- nications Magazine 36(12): 107–112. 48. Komine T & Nakagawa M (2003) Integrated system of white LED visible-light communication and power-line communication. IEEE Transactions on Consumer Electronics 49(1): 71–79. 49. Komine T & Nakagawa M (2002) Integrated system of white LED visible-light communication and power-line communication. In: Personal, Indoor and Mobile Radio Commu- nications, 2002. The 13th IEEE International Symposium on, volume 4, pp. 1762–1766

117 vol.4. 50. Komine T, Haruyama S & Nakagawa M (2006) Performance evaluation of narrowband OFDM on integrated system of power line communication and visible light wireless communication. In: 2006 1st International Symposium on Wireless Pervasive Computing, pp. 71–79. 51. Zeng L, O’Brien D, Le-Minh H, Lee K, Jung D & Oh Y (2008) Improvement of date rate by using equalization in an indoor visible light communication system. In: Circuits and Systems for Communications, 2008. ICCSC 2008. 4th IEEE International Conference on, pp. 678–682. 52. Zeng L, Minh HL, O’Brien D, Faulkner G, Lee K, Jung D & Oh Y (2008) Equalisation for high-speed visible light communications using white-LEDs. In: Communication Systems, Networks and Digital Signal Processing, 2008. CNSDSP 2008. 6th International Sympo- sium on, pp. 170–173. 53. Zeng L, O’Brien DC, Minh HL, Faulkner GE, Lee K, Jung D, Oh Y & Won ET (2009) High data rate multiple input multiple output (MIMO) optical wireless communications using white led lighting. IEEE Journal on Selected Areas in Communications 27(9): 1654– 1662. 54. Noshad M & Brandt-Pearce M (2012) Expurgated PPM using symmetric balanced incomplete block designs. IEEE Communications Letters 16(7): 968–971. 55. Noshad M & Brandt-Pearce M (2012) Multilevel pulse-position modulation based on balanced incomplete block designs. In: Global Communications Conference (GLOBECOM), 2012 IEEE, pp. 2930–2935. 56. Bouchet O, Porcon P, Wolf M, Grobe L, Walewski JW, Nerreter S, Langer KD, Fernández L, Vucic J, Kamalakis T, Ntogari G & Gueutier E (2010) Visible-light communication system enabling 73 mb/s data streaming. In: 2010 IEEE Globecom Workshops, pp. 1042–1046. 57. Minh HL, O’Brien D, Faulkner G, Zeng L, Lee K, Jung D & Oh Y (2008) 80 Mbit/s visible light communications using pre-equalized white LED. In: 2008 34th European Conference on Optical Communication, pp. 1–2. 58. Vucic J, Kottke C, Nerreter S, Habel K, Buttner A, Langer KD & Walewski JW (2009) 125 Mbit/s over 5 m wireless distance by use of OOK-modulated phosphorescent white LEDs. In: 2009 35th European Conference on Optical Communication, pp. 1–2. 59. Vucic J, Kottke C, Nerreter S, Habel K, Büttner A, Langer KD & Walewski JW (2010) 230 Mbit/s via a wireless visible-light link based on OOK modulation of phosphorescent white LEDs. In: Optical Fiber Communication (OFC), collocated National Fiber Optic Engineers Conference, 2010 Conference on (OFC/NFOEC), pp. 1–3. 60. Vucic J, Kottke C, Nerreter S, Langer KD & Walewski JW (2010) 513 Mbit/s Visible Light communications Link Based on DMT-Modulation of a White LED. Journal of Lightwave Technology 28(24): 3512–3518. 61. Vucic J, Kottke C, Habel K & Langer KD (2011) 803 Mbit/s visible light WDM link based on DMT modulation of a single RGB LED luminary. In: Optical Fiber Communication Conference and Exposition (OFC/NFOEC), 2011 and the National Fiber Optic Engineers Conference, pp. 1–3. 62. Kottke C, Hilt J, Habel K, Vucic J & Langer KD (2012) 1.25 Gbit/s visible light WDM link based on DMT modulation of a single RGB LED luminary. In: 2012 38th European Conference and Exhibition on Optical Communications, pp. 1–3.

118 63. Cossu G, Corsini R & Ciaramella E (2014) High-speed bi-directional optical wireless system in non-directed line-of-sight conﬁguration. Journal of Lightwave Technology 32(10): 2035–2040. 64. Javaudin JP, Bellec M, Varoutas D & Suraci V (2008) OMEGA ICT project: Towards con- vergent Gigabit home networks. In: 2008 IEEE 19th International Symposium on Personal, Indoor and Mobile Radio Communications, pp. 1–5. 65. Minh HL, O’Brien D, Faulkner G, Bouchet O, Wolf M, Grobe L & Li J (2010) A 1.25- Gb/s Indoor Cellular Optical Wireless Communications Demonstrator. IEEE Photonics Technology Letters 22(21): 1598–1600. 66. Minh HL, O’Brien D & Faulkner G (2010) A gigabit/s indoor optical wireless system for Home Access Networks. In: Communication Systems Networks and Digital Signal Pro- cessing (CSNDSP), 2010 7th International Symposium on, pp. 532–536. 67. Minh HL, O’Brien D, Faulkner G, Bouchet O, Wolf M, Grobe L & Li J (2010) A 1.25- Gb/s Indoor Cellular Optical Wireless Communications Demonstrator. IEEE Photonics Technology Letters 22(21): 1598–1600. 68. Gomez A, Shi K, Quintana C, Sato M, Faulkner G, Thomsen BC & OŠBrien D (2015) Be- yond 100-gb/s indoor wide ﬁeld-of-view optical wireless communications. IEEE Photonics Technology Letters 27(4): 367–370. 69. Sakurai Y,Nishimaki K, Toguchi S & Sakane M (2003) A study of seamless communication method with the adequate switching between optical and rf wireless lan. In: Consumer Electronics, 2003. ICCE. 2003 IEEE International Conference on, pp. 264–265. 70. Hou J & O’Brien DC (2006) Vertical handover-decision-making algorithm using fuzzy logic for the integrated radio-and-ow system. IEEE Transactions on Wireless Communi- cations 5(1): 176–185. 71. Sivathasan S & O’Brien DC (2008) RF/FSO Wireless Sensor Networks: A Performance Study. In: IEEE GLOBECOM 2008 - 2008 IEEE Global Telecommunications Conference, pp. 1–5. 72. Borogovac T, Rahaim M & Carruthers JB (2010) Spotlighting for visible light communications and illumination. In: 2010 IEEE Globecom Workshops, pp. 1077–1081. 73. Vegni AM & Little TDC (2012) Handover in VLC systems with cooperating mobile devices. In: Computing, Networking and Communications (ICNC), 2012 International Conference on, pp. 126–130. 74. Chowdhury H & Katz M (2013) Cooperative multihop connectivity performance in visible light communications. In: Wireless Days (WD), 2013 IFIP, pp. 1–4. 75. Tapse H, Borah DK & Perez-Ramirez J (2011) Hybrid optical/RF channel performance analysis for turbo codes. IEEE Transactions on Communications 59(5): 1389–1399. 76. Shao S, Khreishah A, Rahaim MB, Elgala H, Ayyash M, Little TDC & Wu J (2014) An indoor hybrid WiFi-VLC internet access system. In: 2014 IEEE 11th International Confer- ence on Mobile Ad Hoc and Sensor Systems, pp. 569–574. 77. URI: http://www.ettus.com . 78. Azhar AH, Tran TA & O’Brien D (2010) Demonstration of high-speed data transmission using MIMO-OFDM visible light communications. In: 2010 IEEE Globecom Workshops, pp. 1052–1056. 79. Fath T & Haas H (2013) Performance comparison of MIMO techniques for optical wireless communications in indoor environments. IEEE Transactions on Communications 61(2): 733–742.

119 80. Elgala H (2010) A Study on the Impact of Nonlinear Characteristics of LEDs on Optical OFDM. Ph.D thesis. Jacobs University. 81. Sevincer A, Bhattarai A, Bilgi M, Yuksel M & Pala N (2013) Lightnets: Smart lighting and mobile optical wireless networks x2014; a survey. IEEE Communications Surveys Tutorials 15(4): 1620–1641. 82. Gancarz J, Elgala H & Little TDC (2013) Impact of lighting requirements on VLC systems. IEEE Communications Magazine 51(12): 34–41. 83. Chowdhury H & Katz M (2014) Cooperative data download on the move in indoor hybrid (radio-optical) wlan-vlc hotspot coverage. Transactions on Emerging Telecommunications Tecnologies, John Wiley Sons 25(7): 666–677. 84. Chowdhury H, Pouttu A & Katz M (2014) Performance analysis of visible light communications-based hotspots in indoor and outdoor environments. Wireless Personal Communications pp. 1–14. 85. Chowdhury H, Chaudary B, Saud MS, Bashir N & Katz M (2015) Data downloading on the coverage of visible light communication. In: European Wireless 2015; 21th European Wireless Conference; Proceedings of, pp. 1–6. 86. Chowdhury H & Katz M (2013) Data download on move in indoor hybrid (radio-optical) WLAN-VLC hotspot coverages. In: Vehicular Technology Conference (VTC Spring), 2013 IEEE 77th, pp. 1–5. 87. McBride A, Derveni B, Proko J, Kamsula P, Chowdhury H, Hanninen T, Mäkelä JP & Katz M (2014) Transitioning to hybrid radio/optical networks: Development of a ﬂexible visible light communication testbed. In: 5G for Ubiquitous Connectivity (5GU), 2014 1st International Conference on, pp. 222–228. 88. Chowdhury H, Makea JP & Pahlavan K (2006) Trafﬁc statistics on the coverage of WLAN- BASED music shower. In: 2006 IEEE 17th International Symposium on Personal, Indoor and Mobile Radio Communications, pp. 1–5. 89. Chowdhury H, Makela JP & Pahlavan K (2006) On the random crossing of an infostation coverage. In: IEEE Vehicular Technology Conference, pp. 1–5. 90. Armstrong J, Sekercioglu YA & Neild A (2013) Visible light positioning: a roadmap for international standardization. IEEE Communications Magazine 51(12): 68–73. 91. Jovicic A, Li J & Richardson T (2013) Visible light communication: opportunities, challenges and the path to market. IEEE Communications Magazine 51(12): 26–32. 92. Gancarz J, Elgala H & Little TDC (2013) Impact of lighting requirements on VLC systems. IEEE Communications Magazine 51(12): 34–41. 93. Xu T, Jiang B, Wen D & Shi Y (2011) Design of a wireless laser communication system based on PPM technique. In: Electrical and Control Engineering (ICECE), 2011 Interna- tional Conference on, pp. 4427–4430. 94. Chun H, Chiang CJ & O’Brien DC (2012) Visible light communication using OLEDs: Il- lumination and channel modeling. In: Optical Wireless Communications (IWOW), 2012 International Workshop on, pp. 1–3. 95. Park J (2010) Speedup of dynamic response of organic light-emitting diodes. Journal of Lightwave Technology 28(19): 2873–2880. 96. Smulders P (2013) The road to 100 Gb/s wireless and beyond: basic issues and key directions. IEEE Communications Magazine 51(12): 86–91. 97. Tsiatmas A, Baggen CPMJ, Willems FMJ, Linnartz JPMG & Bergmans JWM (2014) An illumination perspective on visible light communications. IEEE Communications Magazine

120 52(7): 64–71. 98. Inan B, Lee SCJ, Randel S, Neokosmidis I, Koonen AMJ & Walewski JW (2009) Impact of led nonlinearity on discrete multitone modulation. IEEE/OSA Journal of Optical Com- munications and Networking 1(5): 439–451. 99. Shlomi A & John B (2012) Advanced Optical Wireless Communication Systems. Cam- bridge University Press. 100. Muthu S, Schuurmans FJP & Pashley MD (2002) Red, green, and blue LEDs for white light illumination. IEEE Journal of Selected Topics in Quantum Electronics 8(2): 333–338. 101. Haigh PA, Ghassemlooy Z, Rajbhandari S & Papakonstantinou I (2013) Visible light communications using organic light emitting diodes. IEEE Communications Magazine 51(8): 148–154. 102. Haigh PA, Ghassemlooy Z, Minh HL, Rajbhandari S, Arca F, Tedde SF, Hayden O & Pa- pakonstantinou I (2012) Exploiting equalization techniques for improving data rates in organic optoelectronic devices for visible light communications. Journal of Lightwave Tech- nology 30(19): 3081–3088. 103. Punke M, Valouch S, Kettlitz SW, Gerken M & Lemmer U (2008) Optical data link employ- ing organic light-emitting diodes and organic photodiodes as optoelectronic components. Journal of Lightwave Technology 26(7): 816–823. 104. Karunatilaka D, Zafar F, Kalavally V & Parthiban R (2015) LED based indoor visible light communications: State of the art. IEEE Communications Surveys Tutorials 17(3): 1649– 1678. 105. McKendry JJD, Green RP, Kelly AE, Gong Z, Guilhabert B, Massoubre D, Gu E & Dawson MD (2010) High-speed visible light communications using individual pixels in a micro light-emitting diode array. IEEE Photonics Technology Letters 22(18): 1346–1348. 106. Delbeke D, Bockstaele R, Bienstman P, Baets R & Benisty H (2002) High-efﬁciency semiconductor resonant-cavity light-emitting diodes: a review. IEEE Journal of Selected Topics in Quantum Electronics 8(2): 189–206. 107. Khalid AM, Cossu G, Corsini R, Choudhury P & Ciaramella E (2012) 1-gb/s transmission over a phosphorescent white LED by using rate-adaptive discrete multitone modulation. IEEE Photonics Journal 4(5): 1465–1473. 108. Cui K (2012). University of Calfornia Revirside, USA. 109. Elgala H, Mesleh R & Haas H (2010) An LED model for intensity-modulated optical communication systems. IEEE Photonics Technology Letters 22(11): 835–837. 110. Elgala H, Mesleh R & Haas H (2009) A study of LED nonlinearity effects on optical wireless transmission using OFDM. In: 2009 IFIP International Conference on Wireless and Optical Communications Networks, pp. 1–5. 111. Stefan I, Elgala H & Haas H (2012) Study of dimming and LED nonlinearity for ACO- OFDM based VLC systems. In: 2012 IEEE Wireless Communications and Networking Conference (WCNC), pp. 990–994. 112. Elgala H, Mesleh R & Haas H (2010) An LED model for intensity-modulated optical communication systems. IEEE Photonics Technology Letters 22(11): 835–837. 113. Vijay A & Green RJ (2012) Radio on visible light x2014; a new approach for indoor mobile telecommunications. In: Telecommunications Forum (TELFOR), 2012 20th, pp. 572–575. 114. Yamazato T (2015) Image sensor based visible light communication for V2X. In: 2015 IEEE Summer Topicals Meeting Series (SUM), pp. 165–166.

121 115. Moreira AJC, Valadas RT & de Oliveira Duarte AM (1995) Characterisation and modelling of artiﬁcial light interference in optical wireless communication systems. In: Personal, Indoor and Mobile Radio Communications, 1995. PIMRC’95. Wireless: Merging onto the Information Superhighway., Sixth IEEE International Symposium on, volume 1, pp. 326– 331 vol.1. 116. Cui K, Chen G, Xu Z & Roberts RD (2010) Line-of-sight visible light communication system design and demonstration. In: Communication Systems Networks and Digital Signal Processing (CSNDSP), 2010 7th International Symposium on, pp. 621–625. 117. Kaiyun C (2012) Physical layer characteristics and techniques for visible light communications. Ph.D thesis, UC Riverside . 118. Moreira AJC, Valadas RT & de Oliveira Duarte AM (1996) Performance of infrared transmission systems under ambient light interference. IEE Proceedings - Optoelectronics 143(6): 339–346. 119. Wu Z (2012). Ph.D. thesis,Boston University, USA. 120. Ntogari G, Kamalakis T, Walewski J & Sphicopoulos T (2011) Combining illumination dimming based on pulse-width modulation with visible-light communications based on discrete multitone. IEEE/OSA Journal of Optical Communications and Networking 3(1): 56–65. 121. Oh M (2013) A ﬂicker mitigation modulation scheme for visible light communications. In: Advanced Communication Technology (ICACT), 2013 15th International Conference on, pp. 933–936. 122. Feng Z, Papageorgiou G, Gao Q, Atya AF, Krishnamurthy SV & Chen G (2014) Perfor- mance of visible light communications with dimming controls. In: 2014 IEEE Wireless Communications and Networking Conference (WCNC), pp. 1756–1761. 123. Kim HS, Kim DR, Yang SH, Son YH & Han SK (2012) Transmission performance variation by dimming control in carrier allocation based visible light communication. In: Opto- Electronics and Communications Conference (OECC), 2012 17th, pp. 743–744. 124. Wang Z, Yu C, Zhong WD, Chen J & Chen W (2012) Performance of variable M-QAM OFDM visible light communication system with dimming control. In: Opto-Electronics and Communications Conference (OECC), 2012 17th, pp. 741–742. 125. Lee K & Park H (2011) Modulations for visible light communications with dimming control. IEEE Photonics Technology Letters 23(16): 1136–1138. 126. Cho E, Choi JH, Park C, Kang M, Shin S, Ghassemlooy Z & Lee CG (2011) NRZ-OOK sig- naling with LED dimming for visible light communication link. In: Networks and Optical Communications (NOC), 2011 16th European Conference on, pp. 32–35. 127. Ali AY, Zhang Z & Zong B (2014) Pulse position and shape modulation for visible light communication system. In: Electromagnetics in Advanced Applications (ICEAA), 2014 International Conference on, pp. 546–549. 128. Monteiro E & Hranilovic S (2014) Design and implementation of color-shift keying for visible light communications. Journal of Lightwave Technology 32(10): 2053–2060. 129. Singh R, OŠFarrell T & David JPR (2014) An enhanced color shift keying modulation scheme for high-speed wireless visible light communications. Journal of Lightwave Tech- nology 32(14): 2582–2592. 130. h Choi J, b Cho E, Kang TG & Lee CG (2010) Pulse width modulation based signal format for visible light communications. In: OECC 2010 Technical Digest, pp. 276–277.

122 131. Pergoloni S, Biagi M, Colonnese S, Cusani R & Scarano G (2014) Merging color shift keying and complementary pulse position modulation for visible light illumination and communication. In: Euro Med Telco Conference (EMTC), 2014, pp. 1–6. 132. Mesleh R, Elgala H & Haas H (2011) On the performance of different OFDM based optical wireless communication systems. IEEE/OSA Journal of Optical Communications and Networking 3(8): 620–628. 133. Armstrong J (2009) OFDM for optical communications. Journal of Lightwave Technology 27(3): 189–204. 134. Elgala H, Mesleh R & Haas H (2009) A study of LED nonlinearity effects on optical wireless transmission using OFDM. In: 2009 IFIP International Conference on Wireless and Optical Communications Networks, pp. 1–5. 135. Armstrong J & Lowery AJ (2006) Power efficient optical OFDM. Electronics Letters 42(6): 370–372. 136. Tsonev D, Sinanovic S & Haas H (2012) Novel unipolar orthogonal frequency division multiplexing (U-OFDM) for optical wireless. In: Vehicular Technology Conference (VTC Spring), 2012 IEEE 75th, pp. 1–5. 137. Yun G & Kavehrad M (1993) Indoor infrared wireless communications using spot diffusing and fly-eye receivers. Canadian Journal of Electrical and Computer Engineering 18(4): 151–157. 138. Yun G & Kavehrad M (1992) Spot-diffusing and fly-eye receivers for indoor infrared wireless communications. In: Wireless Communications, 1992. Conference Proceedings., 1992 IEEE International Conference on Selected Topics in, pp. 262–265. 139. Jungnickel V, Pohl V, Nonnig S & von Helmolt C (2002) A physical model of the wireless infrared communication channel. IEEE Journal on Selected Areas in Communications 20(3): 631–640. 140. Kamsula P (2015) Design and implementation of a bi-directional visible light communication testbed. M.Sc thesis, University of Oulu . 141. Goodman DJ, Borras J, Mandayam NB & Yates RD (1997) INFOSTATIONS: a new system model for data and messaging services. In: Vehicular Technology Conference, 1997, IEEE 47th, volume 2, pp. 969–973 vol.2. 142. Ott J & Kutscher D (2005) A disconnection-tolerant transport for drive-thru internet environments. In: Proceedings IEEE 24th Annual Joint Conference of the IEEE Computer and Communications Societies., volume 3, pp. 1849–1862 vol. 3. 143. Sichitiu ML & Kihl M (2008) Inter-vehicle communication systems: a survey. IEEE Com- munications Surveys Tutorials 10(2): 88–105. 144. Ott J & Kutscher D (2004) The "drive-thru" architecture: WLAN-based internet access on the road. In: Vehicular Technology Conference, 2004. VTC 2004-Spring. 2004 IEEE 59th, volume 5, pp. 2615–2622 Vol.5. 145. Ott J & Kutscher D (2004) Drive-thru internet: IEEE 802.11b for "automobile" users. In: INFOCOM 2004. Twenty-third AnnualJoint Conference of the IEEE Computer and Com- munications Societies, volume 1, p. 373. 146. Chowdhury H & Katz M (2013) Cooperative multihop connectivity performance in visible light communications. In: Wireless Days (WD), 2013 IFIP, pp. 1–4. 147. O’Brien D & Katz M (2004) Short-range optical wireless communications. Wireless World Research Forum (WWRF) .

123 148. Wu D, Ghassemlooy Z, Minh HL, Rajbhandari S & Khalighi MA (2012) Optimisation of lambertian order for indoor non-directed optical wireless communication. In: Communica- tions in China Workshops (ICCC), 2012 1st IEEE International Conference on, pp. 43–48. 149. Wu D, Ghassemlooy Z, d Zhong W & Chen C (2014) Cellular indoor OWC systems with an optimal lambertian order and a handover algorithm. In: Telecommunications (IST), 2014 7th International Symposium on, pp. 777–782. 150. Pahlavan K (2011) Principles of Wireless Networks: A Uniﬁed Approach. John Wiley & Sons, Inc., New York, NY, USA, 2nd edition. 151. Borogovac T, Rahaim M & Carruthers JB (2010) Spotlighting for visible light communications and illumination. In: 2010 IEEE Globecom Workshops, pp. 1077–1081. 152. Rahaim M & Little TDC (2013) SINR analysis and cell zooming with constant illumination for indoor VLC networks. In: Optical Wireless Communications (IWOW), 2013 2nd International Workshop on, pp. 20–24. 153. Chowdhury H, Iinatti J & Pirinen P (2008) Broadband services on move in the coverage of relay-based network. In: 2008 IEEE 10th International Symposium on Spread Spectrum Techniques and Applications, pp. 227–231. 154. Chowdhury H, Lehtomaki J, Iinatti J & Pirinen P (2009) Data downloading in relay assisted mobile vehicles. In: Wireless Communication, Vehicular Technology, Information Theory and Aerospace Electronic Systems Technology, 2009. Wireless VITAE 2009. 1st International Conference on, pp. 361–365. 155. Le NT, Choi S & Jang YM (2011) Cooperative mac protocol for led-id systems. In: ICTC 2011, pp. 144–150. 156. Park IH, Kim YH, Cha J, Lee K, Jang YM & Kim JY (2010) Scalable optical relay for LED-ID systems. In: 2010 International Conference on Information and Communication Technology Convergence (ICTC), pp. 415–420. 157. Sivathasan S & O’Brien DC (2008) RF/FSO wireless sensor networks: A performance study. In: IEEE GLOBECOM 2008 - 2008 IEEE Global Telecommunications Conference, pp. 1–5. 158. Sivathasan S (2009) RF/FSO and LEACH wireless sensor networks: A case study compar- ing network performance. In: 2009 IFIP International Conference on Wireless and Optical Communications Networks, pp. 1–4. 159. URI: http://www.ict-omega.eu . 160. Fitzek F, Pedersen M, Perrucci GP & Larsen T (2008) Energy and link measurements for mobile phones using ieee802.11b/g. In: Modeling and Optimization in Mobile, Ad Hoc, and Wireless Networks and Workshops, 2008. WiOPT 2008. 6th International Symposium on, pp. 36–36. 161. Fitzek FHP, Heide J, Pedersen MV, Ertli G & Katz M (2011) Multi-hop versus overlay networks: A realistic comparison based on energy requirements and latency. In: Vehicular Technology Conference (VTC Spring), 2011 IEEE 73rd, pp. 1–5.

124 C585etukansi.kesken.fm Page 2 Tuesday, October 11, 2016 3:13 PM

ACTA UNIVERSITATIS OULUENSIS SERIES C TECHNICA

570. Postila, Heini (2016) Peat extraction runoff water purification in treatment wetlands constructed on drained peatlands in a cold climate 571. Happonen, Tuomas (2016) Reliability studies on printed conductors on flexible substrates under cyclic bending 572. Soderi, Simone (2016) Evaluation of industrial wireless communications systems’ security 573. Harjula, Erkki (2016) Energy-efficient peer-to-peer networking for constrained- capacity mobile environments 574. Tolonen, Arto (2016) Product portfolio management over horizontal and vertical portfolios 575. Suliman, Isameldin Mohammed (2016) Performance analysis of cognitive radio networks and radio resource allocation 576. Karjalainen, Satu Maaria (2016) Identification of processes leading to long-term wastewater purification in northern treatment wetlands 577. Ohenoja, Markku (2016) Computational methods for exploiting image-based data in paper web profile control 578. Väliheikki, Ari (2016) Resistance of catalytic materials towards chemical impurities : the effect of sulphur and biomaterial-based compounds on the performance of DOC and SCR catalysts 579. Kinnunen, Tuomo (2016) Product management perspectives on stakeholder and business opportunity analyses in the front-end of product creation 580. Heiderscheidt, Elisangela (2016) Evaluation and optimisation of chemical treatment for non-point source pollution control : purification of peat extraction runoff water 581. Su, Xiang (2016) Lightweight data and knowledge exchange for pervasive environments 582. Kaijalainen, Antti (2016) Effect of microstructure on the mechanical properties and bendability of direct-quenched ultrahigh-strength steels 583. Lanz, Brigitte (2016) Compact current pulse-pumped GaAs–AlGaAs laser diode structures for generating high peak-power (1–50 watt) picosecond-range single optical pulses 584. Kähäri, Hanna (2016) A room-temperature fabrication method for microwave dielectric Li2MoO4 ceramics and their applicability for antennas

Book orders: Granum: Virtual book store http://granum.uta.fi/granum/ C585etukansi.kesken.fm Page 1 Tuesday, October 11, 2016 3:13 PM

C 585 OULU 2016 C 585

UNIVERSITY OF OULU P.O. Box 8000 FI-90014 UNIVERSITY OF OULU FINLAND ACTA UNIVERSITATISUNIVERSITATIS OULUENSISOULUENSIS ACTA UNIVERSITATIS OULUENSIS ACTAACTA

TECHNICATECHNICACC Helal Chowdhury Helal Chowdhury Professor Esa Hohtola DATA DOWNLOAD ON University Lecturer Santeri Palviainen THE MOVE IN VISIBLE LIGHT

Postdoctoral research fellow Sanna Taskila COMMUNICATIONS: DESIGN AND ANALYSIS Professor Olli Vuolteenaho

University Lecturer Veli-Matti Ulvinen

Director Sinikka Eskelinen

Professor Jari Juga

University Lecturer Anu Soikkeli

Professor Olli Vuolteenaho UNIVERSITY OF OULU GRADUATE SCHOOL; UNIVERSITY OF OULU, FACULTY OF TECHNOLOGY Publications Editor Kirsti Nurkkala

ISBN 978-952-62-1361-3 (Paperback) ISBN 978-952-62-1362-0 (PDF) ISSN 0355-3213 (Print) ISSN 1796-2226 (Online)