Daniel Sjöström

(1)

Unlicensed and licensed

low-power wide area networks

Exploring the candidates for massive IoT

DANIEL SJÖSTRÖM

K T H R O Y A L I N S T I T U T E O F T E C H N O L O G Y

(2)

Unlicensed and licensed

low-power wide area networks

Exploring the candidates for

massive IoT

Daniel Sjöström

2017-09-24

Master’s Thesis

Examiner

Gerald Q. Maguire Jr.

Academic adviser

Anders Västberg

Industrial adviser

Viktor Dahl

KTH Royal Institute of Technology

School of Information and Communication Technology (ICT) Department of Communication Systems

(3)

Abstract

In the Internet of things (IoT), many applications will require low-power and low-cost to achieve long lifetime and scale (respectively). These types of applications are referred to as massive IoT, as opposed to critical IoT, which emphasizes ultra-high reliability and availability and low latency. One type of network catering to massive IoT applications are Low-Power Wide Area Networks (LPWANs), and presently, many LPWAN standards are trying to assert their role in the IoT ecosystem.

This thesis explores LPWANs from both technical and non-technical perspectives to ascertain their use-case versatility and influence on the future telecommunications’ landscape. With respect to spectrum, the studied LPWANs may be categorized as unlicensed LPWAN or licensed LPWAN. The prior category typically refers to proprietary solutions and in this thesis are represented by SigFox and LoRaWAN. The latter group includes EC-GSM-IoT, eMTC, and NB-IoT and can be considered synonymous with cellular LPWAN because they are designed to be integrated into existing cellular infrastructures.

The results indicate that all of the different types of explored LPWANs support applications without strict downlink, payload size, and latency requirements. For use cases without these specific demands (typically sensors, meters, tracking, etc.), it is not a question of whether or not a network fulfills the requirements, but rather how flexible the requirements are. As a result the choice of network will be determined by non-technical aspects and a cost versus functionality trade-off where unlicensed LPWAN is typically cheaper. Hence, both categories of LPWANs offer a unique value proposition; therefore, they can be considered complementary. This notion is reinforced when looking at non-technical aspects such as ecosystem, regulation, network ownership and control, and network coordination, which differ quite significantly. Furthermore, unlicensed LPWANs are likely to be the vanguard of a new type of competitor offering the core service of connectivity.

(4)

(5)

Abstrakt

Inom Internet of Things (IoT) kommer många applikationer att kräva låg effekt och låg kostnad för att uppnå en lång livstid och skala. Dessa typer av applikationer refereras till som massiv IoT, vilket står i motsats till kritisk IoT som kräver ultrahög tillförlitlighet och tillgänglighet och låg fördröjning. En typ av nätverk som ämnar tillgodose kraven av massiv IoT är Low-Power Wide Area Networks (LPWANs), och idag försöker många av dessa hävda sig inom IoT ekosystemet.

Detta examensarbete undersöker LPWANs from ett teknisk och icke-tekniskt perspektiv för att utröna deras mångsidighet och påverkan på det framtida telekomlandskapet. Med avseende på spektrum kan de i detta examensarbete undersökta nätverken kategoriseras som olicensierat LPWAN eller licensierat LPWAN. Den tidigare hänvisar typiskt till proprietära lösningar och representeras i detta arbete av SigFox och LoRaWAN. Den senare kategorin består av EC-GSM-IoT, eMTC, och NB-IoT och kan betraktas som synonymt med mobil LPWAN eftersom de designade för att bli integrerade i existerande mobila nätverk.

Resultaten indikerar att alla nätverk stödjer applikationer utan strikta

krav när det gäller nedlänkens funktionalitet, mängden data per

meddelande, och fördröjning. För applikationer utan dessa specifika krav (typiskt sensorer, mätare, spårning, etc.) är det inte en fråga om huruvida ett nätverk uppfyller kraven eller ej, utan snarare hur flexibla kraven är. Därför kommer valet av nätverk att bestämmas av icke-tekniska aspekter och en avvägning mellan kostnad och funktionalitet vari olicensierat LPWAN är vanligtvis billigare. Därmed erbjuder båda kategorier av nätverk en unik värde proposition och kan därför betraktas som komplementerande. Denna föreställning är förstärkt av att nätverken skiljer sig signifikant när det gäller deras icke-tekniska aspekter såsom ekosystem, reglering, ägandeskap och kontroll, och nätverks koordinering. Dessutom är olicensierade LPWANs troligen är förtruppen av en ny typ av konkurrent som erbjuder den grundläggande servicen av konnektivitet.

(6)

(7)

Acknowledgements

This work was carried out during 2017 at EY (Stockholm) and the school of information and communication technology at KTH.

I would like to thank my supervisor Viktor Dahl, management consultant at EY, for our insightful discussions and for giving me the opportunity to write this thesis at EY. I would also like to express my gratitude to Professor Gerald Q. Maguire at KTH for accepting my thesis proposal, taking on the role of examiner, and especially for his support and understanding. I also owe my thanks to Mats Landstedt, CEO of IoT Sweden, and Magnus Sparrholm, founder of Talkpool, for our valuable discussions.

(8)

(9)

List of Figures

1.1 M2M and IoT paradigms with respect to commercial aspects. This figure appears here with the permission of Nick Hunn [6]. 2

1.2 IoT technologies grouped by range, courtesy of Keysight technologies. . . 3

2.1 LPWA entities, adapted from [19] . . . 8

2.2 SigFox’s network topology, adapted from [49] . . . 16

2.3 SigFox’s coverage in Europe as of 2017-05-25. Blue and purple represents live coverage and country under roll-out respectively [61] . . . 24

2.4 Semtech SX1272 chip, Spreading factor & bandwidth vs. Sensitivity (range), data points from [75] . . . 30

2.5 LoRaWAN network topology, adapted from [74] . . . 31

2.6 LoRaWAN deployments [98] . . . 41

2.7 3GPP IoT standardization, adopted from Rohde & Schwarz [12], [116]. . . 44

2.8 PSM and eDRX battery life optimization, adapted from

Qualcomm [126]. . . 49

3.1 Thesis flowchart . . . 69

5.1 Maximum data rate and coverage profiles in different

environments. . . 77

5.2 SigFox battery lifetime for 2 or 12 messages per day (24 h) for both an ideal and non-ideal battery (2% self-discharge per month) . . . 85

5.3 Technical comparison: Performance is indicated on a scale from low (0) to high (10). The confidence of each ranking is indicated by the percentage associated with the name of each metric. . . 87

(14)

(15)

List of Tables

2.1 Regulatory parameters for the EU 863-870 MHz SRD band and sub bands [33] . . . 11

2.2 Atmel ATA8520E [56] with pricing information from [57]. . . . 20

2.3 Overview of SigFox’s technological characteristics according to SigFox and other sources. . . 21

2.4 The available offerings from SigFox’s official web shop [70]. . 25

2.5 The SigFox partner network [72]. Note not all SigFox entities are registered as a SigFox partner, and some companies may appear in several categories. . . 26

2.6 LoRa modules, all uncited data is from Semtech [87] . . . 36

2.7 Overview of LoRaWAN’s technological characteristics

according to the LoRa Alliance, Semtech and other sources. . . 38

2.8 A selection of LoRaWAN deployments. . . 41

2.9 Technological EC-GSM-IoT characteristics . . . 51

2.10 3GPP LTE UE category evolution, adopted from Nokia [127], Radio-electronics.com [130] and Qualcomm [126]. . . 54

2.11 Commercially available LTE-IoT modules . . . 55

2.12 Technological LTE-IoT characteristics according to various sources . . . 61

2.13 Use case categories and characteristics according to GSMA [22]. . . 64

2.14 Use case categories and characteristics according to Ericsson [20] . . . 65

(16)

(17)

List of Abbreviations

3GPP 3rd Generation Ppartnership Project

ABP Activation By Personalization

API Application Program Interface

CDMA Code Division Multiple Access

CEPT Conference of Postal and Telecommunications Administrations

EAB Extended Access Barring

EC-EGPRS Extended Coverage - Enhanced GPRS

EC-GSM-IoT Extended Coverage GSM-IoT

EDGE Enhanced Data rates for GSM Evolution (Also known as EGPRS)

EGPRS Enhanced GPRS

eMTC Enhanced Machine Type Communication

EPS Evolved Packet System

ERC European Radiocommunications Committee

ETSI European Telecommunications Standards Institute

FCC (United States of America) Federal Communications Commission

FCC e-CFR FCC electronic Code of Federal Regulations

FEC Forward Error Correction

FHSS Frequency Hopping Spread Spectrum

FOTA Firmware Over The Air

GFSK Gaussian Frequency Shift Keying

GPRS General Packet Radio Service

GSM Global System for Mobile Communications

GSMA GSM Association

HARQ Hybrid Automatic Repeat Request

IETF Internet Engineering Task Force

(18)

ITU International Telecommunications Union

IoT Internet of Things

LPWA Low-Power Wide Area

LTE Long Term Evolution (4G)

M2M Machine-2-Machine

MAC Media Access Control

MNO Mobile Network Operator

MTC Machine Type Communication

NB-IoT Narrow Band IoT

OFDM Orthogonal Frequency-Division Multiplexing

OTAA Over-The-Air Activation

PAPR Peak-to-Average Power Ratio

PTS The Swedish Post and Telecom Authority

QoS Quality of Service

SC-OFDM Single Carrier-OFDM

SRD Short Range Device

SLA Service Level Agreement

SIG Special Interest Group

SoC System on Chip

TOA Time On Air

(19)

Chapter 1

Introduction

Machine Type Communication (MTC), Machine-to-Machine (M2M)

communication, and the Internet of things (IoT) are three terms that are part of a grand vision of the near-future. Ericsson calls this vision the Networked Society, in which everything and everyone that benefits from being connected, will be connected [1].

Laya, Ghanbari, and Markendahl claim that the three terms above are used interchangeably, and although they complement each other, they are not equivalent [2]. They offer the following definitions [2]: M2M is the communication between machines and central management applications. MTC has a similar definition; the difference is that at least one party must be a machine, but not all. Within the field of telecommunications, MTC is more or less regarded as the segment of M2M carried over cellular networks . IoT was coined in 1999 by K. Ashton to describe "a world of seamless connected devices that would save us time and money" [3]. Unlike MTC and M2M, the term IoT includes access to the broader Internet.

Minerva, Biru, and Rotondi in contrast with Laya, Ghanbari, and Markendahl argue that there exists no clear all-inclusive definition of IoT that covers the range from small localized systems to larger globally distributed systems [4]. This raises the issue of different types of systems differentiated by communication range as will be highlighted in Section1.1.

The most official definition of the Internet of Things is arguably the international telecommunications union’s (ITU) version:

"A global infrastructure for the information society, enabling advanced services by interconnecting (physical and virtual) things based on existing and evolving interoperable information and communication technologies." [5]

Another point of view regarding the terms IoT and M2M is expressed by Nick Hunn [6]. Hunn suggests that the term IoT was a moniker given to M2M in an attempt to either revive a flagging M2M market by introducing a more exciting name or to democratize M2M beyond its vertical sectors [6]. Hunn claims that the acronym was brought back from obscurity into the mainstream by Ericsson in their annual report of 2009, and their discussion paper titled "Towards 50 Billion Connected Devices", published in 2010

(20)

[6]. Ericsson’s annual report of 2009 mentions 50 billion connected "things" [7], and their discussion paper "Towards 50 Billion Connected Devices" published the following year uses the term "Internet of Things" [8].

Hunn goes on and offers a definition of M2M and IoT with respect to the value of data, number of things, and cost of deployment over time, conveyed in Figure1.1 [6]. That the number of connected things increases because the cost of deployment decreases is self-explanatory. A more intricate proposal is that the value of the data will decrease before it starts to increase [6]. This is explained by the fact that as the cost decreases, sensors will become less application specific, therefore the data per sensor will decrease [6]. Instead, the value of the data will arise from long-term accumulation of data or by combining data from multiple sensors, which both take time to achieve [6].

FIGURE 1.1: M2M and IoT paradigms with respect to commercial aspects. This figure appears here with the

permission of Nick Hunn [6].

Yockelson expresses an opinion [9] that builds on Hunn’s interpretation of IoT. He claims that the winners in IoT are those best able to process the harnessed information to create value for the masses and other organizations [9]. This is in contrast to the M2M paradigm, in which the generated information tends to be tied to a single application, and used for a single purpose.

1.1 Background

Ericsson defines two segments within IoT: Massive IoT and Critical IoT. Massive IoT refers to applications that require large numbers of devices, low cost, low energy, and little data per message [10]. In contrast, Critical IoT requires very low latency, high reliability, and availability [10]. However, these two are considered as the extremes ends of the spectrum, and there are

(21)

many use cases in between them.

A fundamental categorization of different IoT network technologies may be made with respect to range, as seen in Figure1.2 from KeySight. In each of these categories, and to some extent in between them, several technologies label themselves as the sole enabler of IoT. In fact, Aref Meddeb maintains that there are so many IoT standards in development, that fragmentation is a real issue [11]. Among other things, Meddeb claims that without common standardization efforts, seamless interoperability will be difficult to achieve, and therefore obstruct the realization of IoT [11]. In addition, Meddeb proposes that the sheer number of proposals is likely to exacerbate the on-going discussion of what IoT really is, thereby delaying necessary regulatory frameworks [11].

FIGURE 1.2: IoT technologies grouped by range, courtesy of Keysight technologies.

This thesis will explore networks that cater to the long range Massive IoT segment, indicated as Low-Power Wide Area Networks (LPWANs) in Figure 1.2. Thus far unlicensed proprietary LPWANs (LoRa, SigFox, and others in Figure 1.2) have commercial deployments. These proprietary players are arguably the vanguard of the IoT paradigm defined by Hunn, as depicted in Figure1.1. In response, the cellular industry has via the 3rd

Generation Partnership Project (3GPP) scrambled to standardize equivalent licensed cellular LPWANs (Narrowband (NB)-IoT in Figure 1.2). To use more dramatic terms, one could describe the current situation as a prelude to war between the cellular hegemony and new entrants.

Because IoT is already happening, especially regarding massive IoT, the role of 5G within IoT is in the present context increasingly diminishing.

(22)

However, 5G will to a large extent be an evolution of 4G technology. Hence, the two cellular LPWANs based on 4G that are explored in this thesis, NB-IoT, and eMTC, are likely to pave the way for 5G IoT, as mentioned by Rhode & Schwarz [12].

1.2 Problem definition

Mobile network operators (MNOs) and other players will have to decide which LPWAN or combination of LPWANs to include in their connectivity portfolios. However, before doing so, it is necessary to ascertain the value proposition of each LPWAN and whether or not it overlaps with others .

Some MNOs are quite clear about their choice: According to Vodafone’s director of innovation and architectures Matt Beal, NB-IoT will "crush" proprietary LPWANs such as SigFox and LoRa [13]. In Sweden, MNOs have to some extent already started to explore their options. Hans Dahlberg, Vice President of Telia Global IoT Solutions, says that Telia awaits cellular LPWAN connectivity and are unlikely to make use of unlicensed LPWANs, according to Iain Morris at LightReading [14]. Tele2 recently deployed a LoRa network in Gothenburg [15], and Ervins Kampans, CTIO of Tele2 Estonia, claims that Tele2 has positioned itself at the "forefront of NB-IoT commercialization in Europe and worldwide" [16].

1.3 Purpose

The purpose of this thesis is to compare the value propositions of SigFox, LoRaWAN, EC-GSM-IoT, NB-IoT, and eMTC to conclude which ones are needed and which not.

1.4 Goals

The goals of this degree project are:

• Ascertain the need for different LPWANs, and therein the need for licensed or unlicensed LPWANs, by mapping use case requirements to the following categories:

– Not possible with one or more LPWANs or

– Possible with all LPWANs, but favored by one of them.

• Discuss the implications and role of LPWANs in the future telecommunications landscape.

(23)

1.5 Delimitations

The choice of which unlicensed LPWANs to explore was based on commercial relevance and availability. According to Aris Xylouris at Analysys Mason, the two proprietary LPWNAs with the highest number of deployed networks are SigFox and LoRaWAN [17]. Furthermore, SigFox and LoRaWAN each have a different technology, business model, and approach regarding their intellectual property, which in turn makes a comparative assessment all the more interesting.

Concerning technological delimitation. Unlicensed LPWANs are by nature simplistic and are therefore explored to a greater extent than unlicensed LPWANs. This is because licensed LPWANs, which is in this thesis synonymous with cellular LPWANs, is integrated into the already existing complex wide area cellular networks and would require another thesis to explain fully. Hence, for unlicensed LPWANs, the cellular network is only described to the extent that is necessary to understand the enhancements and modifications resulting in cellular LPWAN. A reader interested in gaining a complete understanding of cellular networks is referred to Sauter [18].

1.6 Structure of thesis

Chapter 2 presents background information about each of the compared

LPWANs. Chapter 3 explains the methodology used to perform said

comparison in order to ascertain the "need" for the respective LPWANs. Chapter 4contains novel information gathered by interviews. The analysis based on the information presented in Chapter2and Chapter4is performed in Chapter5. Then, last but not least, conclusions along with future work are presented in Chapter6.

(24)

(25)

Chapter 2

Background

This chapter conveys information about each LPWAN based upon publicly available information. SigFox is presented in Section2.3 and LoRaWAN in Section 2.4. Cellular LPWAN, included in Sections 2.6 and 2.7, is preceded by an introduction to the development of cellular technology in Section2.5. Finally, the results of previous use case analyses are explored in Section2.8.

2.1 LPWANs

Raza, Kulkarni, and Sooriyabandara state that LPWANs are designed to meet the requirements of Massive IoT [19]. As such, LPWANs features wide-area connectivity for delay tolerant, low power, low data rate devices [19]. Furthermore, Raza, Kulkarni, and Sooriyabandara argue that proprietary LPWANs developed by new entrants are gaining much attention since they both complement and supersede legacy wireless technologies [19]. In response, rather than waiting for 5G-IoT, the cellular industry has raced to develop cellular LPWAN techniques that can operate on top of their existing cellular network.

According to Aris Xylouris at Analysys Mason, the number of deployed LPWA networks (unlicensed and licensed) tripled in 2016 compared to 2015, i.e. going from 29 to 85 [17].

The forecasts for number of devices to utilize IoT LPWANs varies quite significantly. In Ericsson’s latest mobility report, published November 2016, Ericsson anticipates a total of 29 billion devices by 2022 including everything ranging from phones, IoT devices, and tablets to PCs, fixed phones, and laptops [20]. Ericsson further claims that 16 billion of these devices will be short-range IoT devices (Bluetooth, Wi-Fi), and 2.1 billion will be wide-area massive- and critical IoT devices [20]. In comparison, 2016 had 400 million wide-area massive- and critical IoT devices [20]. According to the latest forecast by GSM Association (GSMA) (a special interest group (SIG) representing MNOs worldwide), the number of LPWAN devices are predicted to overtake 2G, 3G, and 4G as the leading form of IoT connectivity by 2022 with 1.4 billion connections [21]. In an earlier report in 2016, GSMA, anticipated 5 billion LPWAN devices by 2022 [22]. According to SigFox, the LPWAN market is expected to grow 90% annually and reach $24.5 billion in

(26)

annual revenue 2021 [23].

As for the LPWAN providers themselves, the most prominent proprietary alternatives are: SigFox based in France, founded in 2009; LoRa(WAN); Ingenu headquartered in the US, founded in 2008; and Telensa. However, as explained in Section 1.5, only SigFox and LoRaWAN are explored in this thesis, but it is important to remember that there are many competing technologies. There are quite a few standardized LPWANs, as indicated in Figure2.1. However, in this thesis, only unlicensed LPWANs are considered in depth.

FIGURE2.1: LPWA entities, adapted from [19]

2.2 Regulation and standards

There are several international organizations involved in the regulation and standardization of telecommunications. Pertinent to this particular thesis are those regulations and standards governing the use of spectrum in both North America and Europe. In addition, the International Telecommunications Union (ITU), a United Nation’s specialized agency in ICT technologies, assists and coordinates the development of technical standards. ITU

develops so-called International-Mobile-Telecommunications(IMT)

standards for mobile telecommunications [24]. Furthermore, each IMT consist of a set of ITU-R recommendations [24]. Although these IMT requirements are referred to in the industry as different cellular generations, ITU emphasizes that it does not recognize the terminology of

(27)

generations, and does not take any position on whether or not a specific IMT recommendation is labeled as 3G or 4G, see for example [24]. However, broadly speaking, IMT-2000 is commonly referred to as 3G, IMT-Advanced as 4G, and IMT-2020 as 5G.

2.2.1 Standardization terminology

This thesis utilizes the following definitions regarding standards and intellectual property:

• ITU writes that "Open Standards" are standards made available to the general public and are developed (or approved) and maintained via a collaborative and consensus driven process. "Open Standards" facilitate interoperability and data exchange among different products or services and are intended for widespread adoption. [25]

• PCMag states that proprietary standards are: "Specifications for hardware or software that are controlled by one company." [26]

• The European Commission defines a harmonised standard as: "A harmonised standard is a European standard developed by a recognised European Standards Organisation: CEN, CENELEC, or ETSI. It is created following a request from the European Commission to one of these organisations." [27]

2.2.2 Europe

In Europe, the entities of interest are the European Telecommunications

Standards Institute (ETSI), the European Conference of Postal

and Telecommunications Administrations (CEPT), and Electronic

Communications Committee (ECC). The Swedish Post and Telecom Authority (PTS) explains that CEPT, founded in 1959, is a cooperation between European post and telecom regulatory bodies, and is since 1992 exclusively involved in regulatory issues [28]. PTS also writes that ECC is CEPT’s committee for electronic communications [28]. According to CEPT, ECC was formed as the result of the merger between ECTRA (responsible for general telecommunications matters) and ERC (European Radiocommunications Committee) [29]. In 1988, PTS writes that ETSI was founded with the purpose of taking on CEPT’s standardization efforts, and it is today a cooperation between telecom authorities, operators, manufacturers, service providers, and users [28].

Real Wireless claims in their whitepaper on M2M technologies that LPWAN devices falls under the category of Short Range Devices (SRDs) within the regulatory framework [30]. Similarly, ETSI states that SRDs are "essentially low power radio communication systems", and may include data, audio, video, telemetry, sensors, radar, etc. [31]. Provided that they conform

(28)

to the relevant ETSI standard to ensure coexistence, SRDs can operate license free in Europe in some bands [31]. Furthermore, ETSI states that once compliance has been demonstrated, there is no need for an end user license or any paperwork, and a compliant device may be used freely throughout Europe [31].

In a pamphlet on regulatory matters published by ETSI, it is stated that ETSI’s Technical Committee for EMC and Radio Spectrum Matters are responsible for SRD standardization [31]. The standards governing LPWANs include technology-specific standards (wideband, narrowband, radar, inductive, etc.) and generic standards for different bands of the spectrum [31]. The harmonized and generic standard for communication between 25 MHz to 1000 MHz is ETSI EN 300 220 [31]. EN 300 220 specifies conformance limits and associated certified testing methods [32].

ETSI indicates that the regulatory parameters included in the harmonized ETSI standards are not based on ITU’s recommendations. This is because SDRs are not, as ITU defines it, considered to be a "Radio Service" [31]. The regulatory parameters in the harmonized ETSI standards are instead based on recommendations laid forth by CEPT or ECC [31]. The ECC recommendation (REC) for SRDs is ERC/REC 70-03 [33]. Hence, according to Real Wireless, RRC/REC 70-03 is the recommendation that applies for SRDs / license-exempt devices, including LPWANs [30]. Note that although ERC/REC 70-03 represents the most accepted position regarding SRD frequency allocation throughout the EU, this allocation is not mandated and not all countries have adopted it, hence national implementations differs [33]. It is often claimed that LPWANs operate within the ISM bands. According to ITU, the organization that allocates ISM bands worldwide, ISM applications are in the ITU radio regulations (RR) defined as "Operation of equipment or appliances designed to generate and use locally radio frequency energy for industrial, scientific, medical, domestic or similar purposes, excluding applications in the field of telecommunications." [34]. The ISM bands are defined in ITU RR No. 5.138 and 5.150 [34].

Haim Mazar, the vice-chairman of ITU-R study group 1 (Spectrum

management) and author of Radio Spectrum Management: Policies,

Regulations, and Techniques [35], writes that SRDs are not ISM applications as defined by ITU Radio Regulations [36]. Mazar states SRD bands are different from ISM bands, but that it is a common mistake to confuse the two [35]. However, Mazar writes that SRDs may be deployed in both ISM bands and non-ISM bands [36]. Furthermore, the SRD bands, as defined by the ERC/REC 70-03, are also designated for industrial, scientific, and medical (ISM) applications as defined by the ITU according to ERC/REC 70-03 [33]. Mazar further writes that individual licenses for SRDs are usually not required [36]. This is in line with ETSI’s view on the matter (that no license is required) as was described above.

(29)

According to Real Wireless, ERC/REC 70-03 identifies bands at 169 MHz, 433 MHz, 863 MHz, 860 MHz, and 915 MHz as suitable for LPWANs [30]. Moreover, Real Wireless states that the 863-870 MHz, 870-875.6 MHz, and 915-921 MHz band are the most practical for LPWANs [30]. This is because the 169 MHz band would require large antennas, while the 433 MHz band permits too little radiated power [30]. Real Wireless points out that almost all CEPT countries have adopted the ERC/REC 70-03 recommendation within the 863-870 MHz band [30]. Real Wireless goes on and justifies the popularity of this band by that it provides a good balance between range, building penetration, and the ability to use small antennas [30]. The European 863-870 MHz bands and sub-bands as defined by ERC/REC 70-03 are summarized in Table2.1.

TABLE 2.1: Regulatory parameters for the EU 863-870 MHz SRD band and sub bands [33]

70-03 annexes Band Power Spectrum access & mit-igaiton requirements

Notes

h1.1 863-870 MHz 25 mW e.r.p ≤ 0.1 % duty cycle or LBT (See below)

FHSS

h1.2 863-870 MHz 25 mW e.r.p ≤ 0.1 % duty cycle or LBT + AFA (See below)

Wideband tech-niques other than FHSS

h1.3 863-870 MHz 25 mW e.r.p ≤ 0.1 % duty cycle or LBT + AFA (See below)

Narrow/Wide-band modulation h1.4 868-868.6 MHz 25 mW e.r.p ≤ 1 % duty cycle or LBT

+ AFA

Narrow/wide-band modulation h1.5 868.7-869.2

MHz

25 mW e.r.p ≤ 0.1 % duty cycle or LBT + AFA

Narrow/wide-band modulation h1.6 869.4-869.65

MHz

500 mW e.r.p ≤ 10 % duty cycle or LBT + AFA Narrow/wide-band modulation h1.7 869.7-870 MHz 5 mW or 25 mW e.r.p. No req. for 5 mW. ≤ 1 % duty cycle or LBT + AFA for 25 mW e.r.p.

Narrow/wide-band modulation

In Table2.1, the duty cycle may be increased to ≤ 1% for bands h1.1, h1.2, and h1.3 if the band is limited to 865-868 MHz [33]. The duty cycle is defined as "the ratio, expressed as a percentage, of the maximum transmitter "on" time monitored over one hour, relative to a on hour period" [32]. Hence, ≤ 1% equates to less 36 seconds time on air (TOA) per device per hour. The duty cycle restriction may be lifted if the devices employ adaptive frequency agility (AFA) or listen before talk (LBT). However, considering the scope of this thesis, these methods are irrelevant since neither LoRaWAN [37] or SigFox employs them. All of the bands in Table2.1have been adopted by all of the CEPT member countries (i.e., most of the EU), with the exception of Georgia (none adopted), Sweden (h1.1, h1.2, h1.3 not adopted), Russia (h1.4, h1.6, h1.7 not implemented), and Ukraine (h1.5, h1.6, h1.7 not implemented) [33].

(30)

2.2.3 US

In the US, the Federal Communications Commission (FCC) regulates communication including wireless transmission. According to Mazar, regulation regarding radio frequency devices (without an individual license) is stipulated in the FCC Electronic Code of Federal Regulations (e-CFR) part 15, and ISM regulation in part 18 [36]. Real Wireless writes that for transmission within the 902-928 MHz ISM band, which SigFox and LoRaWAN use in the US (see Sections2.3.1and2.4.2), e-CFR part 15 §15.247 [38] is the relevant provision and is summarized below [30]:

• There are no duty cycle limitations in the US. However, there are limitations on dwell times (time spent transmitting in one frequency/channel).

• Maximum dwell time:

< 250 kHz bandwidth: 0.4 second within a 20 second period and more than 50 different channels.

> 250 kHz bandwidth: 0.4 second within a 10 second period and more than 50 different channels.

• Maximum transmit power:

– At least 50 channels: 1 W.

(31)

2.3 SigFox

This section describes SigFox in terms of technical- and non-technical aspects.

2.3.1 Spectrum and Regulation

SigFox says that it transmits in the 868 MHz ISM band in Europe and the 900 MHz ISM band in North America [39]. Strictly speaking, the 868 MHz band is not defined as an ISM band in Europe, according to ITU RR No. 5.138 and 5.150 [34]. Hence, the correct terminology would be referring to it as an SRD band, as defined by the ERC(REC 70-03) [33]. The 900 MHz band, however, is an ISM band, in the sense that RR No. 5.150 defines 902-928 MHz as an ISM band for North America [34].

SigFox documentation does not specify which sub-bands are used. However, looking at the frequencies specified for a typical SigFox chip gives an indication. In the EU, the Atmel ATA8520E, summarized in Table 2.2 in Section2.3.2, transmits in the frequency ranges 868-868.6 MHz and receives in 869.4-869.65 for the uplink and downlink respectively. In the US, the uplink and downlink frequency ranges are within 902-906 MHz.

The regulatory parameters in Europe for the 868 MHz band, as recommended by CEPT/ECC, and adopted by ETSI [32], are summarized in Table 2.1 in Section 2.2.2. The regulatory parameters in the US for the 900 MHz band are imposed by the FCC in e-CFR §15.247 [38] and are described in Section2.2.3.

The regulatory parameters in Section 2.2.2 translate to a maximum number of allowed messages per day. For example, for the EU uplink band, 868-868.65 MHz band, limited by ≤ 1% duty cycle and 25 mW e.r.p, SigFox claims that the uplink duty cycle restriction restricts the number of messages sent per device to 140 per day [39]. Assuming per day here implies 24 hours, this equates to 6 messages per hour, which is in agreement with Radiocrafts AS’s interpretation [40]. Similarly, the less than 10% duty cycle restriction on the downlink band can be used to calculate the maximum allowed number of downlink messages per day per device. Raza, Kulkarni, and Sooriyabandara set this number to 4 messages per day [19]. There is no equivalent duty cycle restriction in North America. However, North American regulations, as stated in Section 2.2.3) on dwell time lead to a limitation on number of message per day per device according to Singh and Kaur because they claim that a SigFox transmission usually takes around 3 seconds, which is much more than the allowed 0.4 seconds [41].

SigFox writes that the SigFox protocol is designed to comply with ETSI and FCC regulations [39]. Furthermore, SigFox chips may support both FCC and ETSI compliance, such as the Atmel ATA8520 (see Table 2.2), or only compliance with one set of regulations, such as the ETSI compliant

(32)

M2COMM M2C8001 [42].

2.3.2 Technology of SigFox

This section presents the technological aspects of SigFox.

Physical layer

Raza, Kulkarni, and Sooriyabandara state that the physical layer is based on ultra-narrow band (UNB) technology in regional sub-GHz ISM bands [19]. KeySight claims that the number of supported uplink channels are 360 UNB 100 Hz channels [43], and that multiple access is realized through frequency hopping spread spectrum (FHSS) [43]. Raza, Kulkarni, and Sooriyabandara and Laya et al. explains that to increase the likelihood of successful reception, each device transmits three times on different and randomly selected UNB channels [19] [44].

SigFox has designed a communication system capable of transmitting signals over long distances at the expense of data rate. SigFox has achieved this by operating in sub-GHz bands using UNB in combination with robust (simple) modulation techniques.

There are several reasons why UNB may increase the attainable range. One is that UNB receivers listen to an equally narrow portion of the spectrum and therefore capture less noise, which in turn increases the range. Another inherent benefit in UNB is that receivers employ very steep filters, which effectively cancels out sideband noise. Moreover, using FHSS with a small signal bandwidth (100 Hz) and a relatively large listening bandwidth (200 kHz) mitigates frequency-selective fading according to Real Wireless [30]. Furthermore, SigFox claims that, compared to wideband techniques, allocating the energy in a UNB channel increases the signal to noise ratio [45]. Last but not least, a small bandwidth translates to a long symbol period, thereby increasing the energy per symbol, and therefore the range. The downside is that a long symbol period results in a low a data rate. Indeed, such is the fundamental trade-off in any radio system; an increase in data rate (bandwidth) will decrease the energy per bit, and therefore decrease the range.

KeySight notes that the uplink modulation is differential binary phase shift keying (D-BPSK) [43]. This is a robust modulation because there are only two states possible (1 bit per symbol), thereby reducing the possibility of erroneous interpretation by the receiver. Furthermore, since the "resolution" of the signal or symbols are degraded over distance, having only two states increases the range. In BPSK, these two states (1 and a 0), are conveyed by changing (modulating) the phase of the carrier wave. Hence, with respect to a reference signal, a cosine might represent 1, whereas a sine would

(33)

correspond to a 0. However, in D-BPSK it is the change of phase, not the phase itself, that represents the data. Hence, a D-BPSK transmission can operate with respect to only itself, which simplifies implementation and lowers the cost, since the demodulator does not need to have a copy of the reference signal. Also according to KeySight the downlink, which does not have the same power constraint as the end nodes, uses Gaussian frequency shift keying (GFSK) [43].

Raza, Kulkarni, and Sooriyabandara write that the range is 10 km and 50 km in urban and rural environments respectively [19]. According to M. Centenaro et al. the range is 3-10 km and 30-50 km in urban and rural environments respectively [46]. According to KeySight [43] the maximum acceptable path loss (including fading) is in between 146-162 dB.

Regarding the transmission of data, Raza, Kulkarni, and Sooriyabandara and KeySight specify the uplink- and downlink data rate as 100 bps and 600 bps (respectively) [43] [19]. Link Lab explains this asymmetry is because the endpoint’s receiver sensitivity is not as good the more expensive base-station receiver [47]. The number of allowed messages per day was explored in Section 2.3.1. Raza, Kulkarni, and Sooriyabandara write that the size per message is 12 bytes for the downlink and 8 bytes for the uplink [19].

Network

The network topology, shown in Figure 2.2, is a star network as stated by Raza, Kulkarni, and Sooriyabandara [19], in which each gateway realizes a cell. Raza, Kulkarni, and Sooriyabandara note that an end node is not tied to a single gateway. Instead, each end node’s transmission is picked up by all gateways in range [19]. Hence, SigFox writes that there is no need for a handover procedure since that capability already is inherent in the SigFox network [39]. Centenaro et al. write that SigFox claims that each gateway can handle up to a million connected devices [46]. Although bi-directional communication is supported, Raza, Kulkarni, and Sooriyabandara observe that the radio access is asymmetric because only 4 downlink messages per day per device is allowed and downlink communication can only occur if an uplink communication (immediately) precedes it as the end device can open a receive window and listen for a while after it has sent a message [19]. According to Nicolas Lesconnec of SigFox’s developer relations, SigFox supports acknowledgment: When sending a uplink message, a device can request a downlink acknowledgment [48].

(34)

FIGURE2.2: SigFox’s network topology, adapted from [49]

Laya et al. state that SigFox employs no carrier sensing [44]. Similarly, SigFox’s founder and CSO Christopher Fourtet mentions that there is no channel management and hence no means of collision-avoidance [49]. In other words, SigFox is a contention-based protocol, in which end nodes are "free running" as Fourtet puts it [49] and transmit asynchronously (as mentioned by Raza, Kulkarni, and Sooriyabandara [19]). Furthermore, as pointed out by Raza, Kulkarni, and Sooriyabandara and Laya et al., each message is transmitted three times on random channels [19][44]. Moreover, SigFox states that UNB reduces the likelihood of collision compared to wide-band techniques [39]. Allegedly, this is because more UNB signals can simultaneously co-exist and be carried over a given bandwidth than wide-band signals [39]. Finally, SigFox claims that the likelihood of collision is close to zero [39].

SigFox claims that its lightweight protocol has eliminated the need for signaling and reduces the required overhead [45]. Allegedly, this reduced both design cost and power consumption [45]. Furthermore, less overhead means that more data can be carried in each message [45]. SigFox claims that 14 bytes of overhead are required for a payload of 8 bytes [45]. Along the same line of reasoning, Laya et al. note that SigFox has a Media Access Control (MAC) layer equivalent to an unslotted (pure) ALOHA and argues that it is an energy-efficient and cost-efficient scheme [44]. Laya et al. explain that asynchronous transmission does not involve listening to the channel (which would require power) [44]. Laya et al. explain that asynchronous transmission also removes the need for time synchronization in the network; therefore, there is no need for expensive oscillators in SigFox devices [44]. This observation is in agreement with one of the rationales behind SigFox as expressed by Fourtet:

"Network should not ask modems for long disciplining processes. But it should be at the service of modems to compensate for their imperfection, contributing to their low cost" [49].

(35)

Co-existence and resilience

Regarding co-existence with other communication systems, Real Wireless conducted a co-existence analysis that mapped the impact of different interference conditions [30]. The impact of the interference was categorized as (i) easily mitigated, (ii) mitigated against with some effort, and (iii) difficult to mitigate [30]. Interference from a UNB system on another UNB system was categorized as (ii) mitigated against with some effort. Interference from a wide-band technique (DSSS) was deemed to be (iii) difficult to mitigate [30]. However, SigFox claims that UNB allows for coexistence with other communication system using broadband techniques due to its ultra narrow bandwidth [39].

Juan Carlos Zuniga and Benoit Ponsard in a talk at a meeting of the Internet Engineering Task Force (IETF) write that SigFox exhibits anti-jamming capabilities due to the intrinsic ruggedness of UNB combined with the fact that multiple base stations may receive each message (spatial diversity) [50]. Furthermore, according to Grover, Lim, and Yang FHSS is regarded as a general anti-jamming technique [51]. The same authors, however, observe that there are some jamming techniques like channel hopping, pulsed noise, and flow jamming that can affect multiple channels at once [51]. Furthermore, these authors claim that so-called follow-on jamming is particularly effective against systems using FHSS [51]. SigFox contends that non-compliant jamming signals have no impact on SigFox devices [39].

Cloud/backend

SigFox explains that a SigFox solution must connect to the SigFox cloud to retrieve messages and, if necessary, to manage devices and users [52]. Moreover, the SigFox cloud supports services such as device information (signal quality, events, statistics, and more), billing, user-and fleet management, network management suite, service maps, user-and so on [52].

SigFox describes its cloud as reachable via three different types of interfaces: Internet, Application Program Interfaces (APIs), and callbacks [52]. Via internet, i.e. a simple web browser can be used to access a SigFox web-portal that provides access to all SigFox cloud functions [52]. The API is used to facilitate service delivery and to integrate SigFox functionality, such as device registration, in third party platforms [52]. The data itself is pushed automatically to the customer’s application server using callback functions [52].

Access to the SigFox cloud services and tool suite depend on the type of user according to SigFox [52]. These types range from an administrator (who manage users/fleet, contracts, see service maps, etc.), distributor (who

(36)

creates contracts), and operator (whom has access to a radio planning tool and a network monitoring suite) [52].

Security & authentication

Radiocrafts AS explains that each device is programmed with a device ID, porting authorization code (PAC), and an AES-128 encryption key [40]. The device ID and PAC are required to register the device in the SigFox network [40].

Similarly, the RF module company "TD next" writes that each end node is associated with a device ID and a porting authorization code (PAC) and that both of these are required to register the SigFox device in the SigFox cloud [53]. TD next writes that the SigFox ID is stored in the module memory when it is manufactured [53]. Both the device ID and PAC code are sold with the connectivity module used in the end device, and these values are either printed on the packaging or embedded in the device [53]. Furthermore, TD next emphasizes that the device’s SigFox ID is unique and it will not change during its lifespan [53].

TD next further explains that the PAC code corresponds to the device ID, and its purpose is to manage the ownership of SigFox devices [53], specifically it is used to establish ownership of the device, or to port (transfer) ownership of the device. Once it is used, it is discarded and a new PAC code is generated, so that future porting is possible [53].

According to Radiocrafts AS, network authentication is carried out with an AES-128 encrypted [40] code in the header of each transmitted message. This code is inspected by the SigFox cloud before the data is forwarded to the application server [54]. Radiocrafts AS further describes that the encryption key is securely stored in the device and it cannot be accessed, thus this avoids man-in-the-middle and replay mechanisms [40]. The payload data is not encrypted; however, SigFox points on that developers may implement encryption at the application layer [45].

SigFox explains that spoofing is averted by including a sequence counter in each transmission, which is then verified by the SigFox network [54].

(37)

Modules

A chip, or system on chip (SoC), or integrated network transceiver circuit, will typically realize the physical layer and the MAC layer. Vendors generally provide a reference design (which may be tweaked and added to in terms of functionality) for each of their products. This is valuable because building upon an already working reference design enables a designer to rapidly get a new end-product to the market rather than needing to develop the platform from scratch. Additionally, the network transceiver chip may include a microcontroller that implements some upper protocol stack and security. If the chip includes application layer software or supports application development (often via AT commands), then the chip may be regarded as a radio module with additional functionality.

By having the network compensate for the end node’s weaknesses, SigFox’s founder Christophe Fourtet claims that there is no need for signal processing and expensive hardware in the end nodes which enables basic-and low-cost devices [49]. Anthon Charbonnier, Head of Startups Relations & Programs from Sigfox, emphasizes that the end nodes are designed to be energy efficient and that a transmission which typically takes a few seconds draws 20-35 mA at 25 mW (14 dBm) output power [55]. Another point that Charbonnier highlights is that the devices are idle (and hence the radios need not be powered on) more than 99 % of the time [55].

Besides the data-rate and other network imposed properties, common to all SigFox chips, individual properties such as current consumption (and hence operating lifetime) and device cost are important aspects in determining the viability of an application. As an example of SigFox single-chip transceiver, the properties of Atmel ATA8520E is summarized in Table2.2(see next page).

Besides the power consumption of the radio module, the battery lifetime depends on how often the devices transmit as well as upon the battery itself. According to Radicrafts, a device communicating one message per day can last more than ten years using a 3.6V lithium AA-cell battery [58]. Another example by Radiocrafts AS is that a device transmitting ten times a day would last six years using the same battery [58]. SigFox claims that two messages a day equate to a 20 year lifetime using the energy contained in one AA battery [45].

(38)

TABLE 2.2: Atmel ATA8520E [56] with pricing information from [57].

Properties Atmel ATA8520E

Description Integrated SoC that provides connectivity, embedded firmware, protocol handling and, ID/PAC. The SoC consist of a RF front end, digital baseband, and 8-bit micro controller.

Cost per device

+1500 units: US$2.276 +6000 units: US$2.15 +15000 units: US$2.065

Frequencies EU: uplink 868.0 - 868.6 MHz, downlink 869.4 - 869.65 MHz. US: uplink and downlink 902 - 906 MHz.

Output power EU: 13.8 dBm (typical) US: 9.5 dBm (typical) Current consumption (typical)

Supply voltage: 3 V and 3.3V to 5.5V Transmit: 32.7mA (EU) / 16.7mA(US) Receive: 10.4mA (EU) / 10.5mA (US) Off: 5nA

Base stations

There is no information publicly available about SigFox base stations. Moreover, the only manufacturer seems to be SigFox. It is likely that only certified SigFox operators (see Section 2.3.3 can buy SigFox base stations. According to RadioCrafts AS, SigFox base stations are only provided by SigFox [40]

Roaming

As stated bt Juan Carlos Zuniga and Benoit Ponsard (IETF), SigFox authentication is global, and there are therefore no roaming requirements [50]. The user can enjoy seamless mobility across the global SigFox network [39]. However, for devices to roam internationally (assuming coverage is provided), they must support the local RF regulations, such as ETSI and FCC compliance [39].

SigFox Technical Summary

(39)

TABLE 2.3: Overview of SigFox’s technological characteristics according to SigFox and other sources.

Characteristic SigFox

Frequency EU: 868 MHz ISM band [39]

EU uplink: 868.0 - 868.6 MHz (Atmel ATA8520E) [56] EU downlink: 869.4 -869.65 MHz (Atmel ATA8520E) [56]

North America: 900 MHz ISM band [39] US uplink & downlink 902 - 906 MHz. (Atmel ATA8520E) [56]

Multiple access method uplink & downlink: Proprietary FHSS/UNB [43] MAC Asynchronous / Unslotted ALOHA [19] [44] (transmit 3

times and hope for the best)

Overhear/Signaling No Signaling, 14 bytes overhead per uplink message [45]

Modulation GFSK (downlink), DBPSK (uplink) [43] Range 3-10 km (URBAN), 30-50 km (RURAL) [19] [46] Link budget ∼ 146-162 dB [43]

Forward Error Correction (FEC) No [19]

Output power Max. 25 W (14 dBm) [33]

Bandwidth Total/listening bandwidth: 200 kHz [43] Channel bandiwdths: 100 Hz (uplink), 600 Hz (downlink) [43]

Number of UNB channels 360 [19]

Peak data rate 100 bps (uplink), 600 bps (downlink) [19] Payload length 12B (uplink), 8B (downlink) [19] Allowed number of messages (Europe) per day 140 (uplink), 4 (downlink) [39] [19] Network topology Star [19]

Roaming Yes [50]

Handover Not needed, devices are not tied to a single base station [19]

Co-existence and resilience Although sophisticated anti-FHSS jamming techinques exist (e.g. follow-on), it is likely that the inherent narrowness and hence ruggedness of UNB effectively mitigate jamming and co-existence issues (See section

2.3.2) Adaptive data rate No [19] Bidirectional Yes [19]

Authentication & security AES-128 encrypted authentication [40] Unique device ID and PAC code [53]

VPN tunnel from gateway to SigFox cloud[45] No Payload encryption [45]

Over The Air (OTA) updates No [19]

SIM form factor No cards, eUICC or Soft-Sim equivalent. See Section

2.3.2

Localization No [19] Nodes per base station (BS) Variable

1 million [46] (No specified assumptions on msgs / day / device)

Battery lifetime 3.6V AA battery: 1 msg / day: 10 years [58] 10 msg / day: 6 years [58]

2 messages a day using one AA battery with equates to 20 years [45].

Atmel ATA8520E Current consumptiom: Transmit: 32.7mA (EU) / 16.7mA (US) [56] Receive: 10.4mA (EU) / 10.5mA (US) [56] Off: 5nA [56]

(40)

2.3.3 SigFox’s Non-technological Aspects

This section presents non-technological aspects of SigFox’s LPWAN.

Business model

SigFox, launched in 2009, was the first proprietary LPWAN to hit the IoT market, and since then it has been growing rapidly [46]. According to Ingrid Lunden at TechCrunch, SigFox has received investments from Silicon/product manufacturers such as Samsung and Intel, utility giant Air Liquide, and service providers such as Telefonica and Eutelsat [59]. According to Aris Xylouris at Analysys Mason, SigFox increased their number of active or planned networks from 9 in 2015 to 27 in 2016 [17]. Raza, Kulkarni, and Sooriyabandara summarize the offering by SigFox as end-to-end LPWAN connectivity or device-to-cloud connectivity based on patented technology [19].

According to Link Labs, SigFox’s business model revolves around two approaches [47]. The first business model is to deploy the network itself and act as a operator, as they have done in France and the US [47]. The second business model is to allow another operator to exclusively deploy and commercialize their network in a particular area in exchange for royalties [47]. Such operators are referred to as certified SigFox operators. Bob Emmerson at No Jitter (part of UBM Technology) summarizes his view of the SigFox business model with the following [60]

"All a new entrant needs to do, apart from writing the app, is buy compatible devices. They don’t need SIM cards or multi-band air interfaces. SIGFOX issues IDs to the modem manufacturers, who subsequently give them to their customers, and then they use the local service provider to activate them. This allows each customer to manage their devices and subscribe to the messages they send."

Judging from SigFox’s web page, there are 34 SigFox operators (referred to as local service providers in above quote), each in a different country [61]. The majority of these operators focus solely on SigFox connectivity, whereas others use this service offering to complement their existing connectivity portfolios, such as is the case for Arqiva or Cellnex [61]. Some SigFox operators have struck deals with traditional MNOs. For example, a press release by Tele2, revealed their partnership with the SigFox operator Aerea in the Netherlands [62]. Similarly, Telefonica is in negotiation with local Brazilian SigFox operator WND [63].

(41)

Intellectual property rights and compliance

Although SigFox is a proprietary standard (as defined in Section 2.2.1), SigFox says that it gives away the protocol stack (which defines the physical layer up to the transport layer) to modem manufacturers free of charge, i.e., without royalties [64].

As a means of ensuring compliance, SigFox says that "Any device operating on the SigFox network must pass certification delivered by SigFox" [65]. It is further stated that technical testing required for the SigFox ready certificate is accredited by external labs selected by SigFox [65]. Furthermore, according to Atmel, modem (transceivers) integrating basic SigFox connectivity must be certified as SigFox compliant, in order to be used by a SigFox ready product [66]. Atmel also points out that the SigFox certificate does not cover local regulatory compliance, such as that imposed by ETSI or the FCC [66].

Commercial aspects

Information concerning infrastructure cost, such as price per base station, is not publicly available. However, according to Bob Emmerson at No Jitter, the nationwide SigFox network in France, consisting of around 1000 base stations cost US$4 million [60]. Emmerson further claims that this is about 100 times cheaper than an equivalent GSM/CDMA cellular network [60].

SigFox calls their modules starting at US$2.00 ultra-low cost modules and claims that they are up to 20 times cheaper than LTE cellular modules and five times cheaper than the closest competing technologies [23].

SigFox points out that a data subscription is required for a certified SigFox chip to communicate over the network [39]. Furthermore, SigFox explains that the subscription price depends on the subscription level, which defines the maximum number of messages per day per device and the total number of connected devices [39].

Reach

According to SigFox, they are present in 29 countries, in reach of 471 million people [67], and on track to be present in 60 countries by 2018 [68]. The total coverage in Europe is shown in Figure 2.3. In the US, in spite of the regulatory obstacles explain in Section 2.3.1, SigFox writes that it has attained a 20 percent population coverage, and they are rolling out networks in more than a 100 cities [69].

(42)

FIGURE2.3: SigFox’s coverage in Europe as of 2017-05-25. Blue and purple represents live coverage and country under roll-out

respectively [61]

Actors and activity

According to SigFox’s founder Christophe Fourtet, the simplistic nature of the SigFox connectivity implies "that almost every available chip can run SigFox!" [49]. Indeed, the simplistic design, combined with their giveaway of their intellectual property (see Section 2.3.3) to chip manufacturers, has facilitated a thriving hardware ecosystem. So far, large manufacturers such as STMicroelectronics, Atmel, and Texas Instruments are manufacturing SigFox chips [47].

The two reasons behind this approach, according to Anthony

Charbonnier who is responsible for SigFox startup relations, are cost effectiveness and selection diversity [55]. Charbonnier highlights that because of the wide range of different radio modules available, each new product or device has a greater opportunity of finding a chip which fits the intended use [55]. As shown in Table 2.4, there is there is a wide variety of hardware, solutions, and services presently available in the SigFox ecosystem.

Table 2.4 includes a wide variety of chips, meters, sensors, actuators monitors, and other low-end devices along with more niche solutions such as smoke detectors, outdoor parking management systems, beehive scales, street lighting management, or a simple watch with an alarm button for older persons in case they fall [70]. SigFox emphasizes that the development kits can both accelerate and simplify the prototyping stage, although these development kits are not intended for industrial use [55]. Indeed, Arduino and Raspberry Pi, which are extremely popular hardware development platforms, are both compatible with the SigFox network [55].

(43)

TABLE 2.4: The available offerings from SigFox’s official web shop [70].

Type of offering Comment Number Development kits Used for prototyping and testing 44 Transceivers Required for basic SigFox connectivity 4 SoC Basic SigFox connectivity with a

reference design to alleviate design process

7

Modules Integrates everything required for SigFox connectivity and additional functionality

39

Devices Sensors, actuators etc 150 Solutions Includes one or more devices, cliud

connectivity and SigFox subscription

19

Cloud IoT platforms

Provides data services connected to the SigFox cloud

40

In March, 2016, SigFox launched the so-called SigFox partner network [71]. The SigFox partner network is intended to bring together key partners, ranging from technology enablers, operators, and solution enablers to startups [71]. New players who wish to design their product can, in SigFox’s words: "Using this platform, contact the SIGFOX ecosystem partners directly, including chip manufacturers, module makers, device makers, solution providers, and design houses" [71]. The different type of companies that have joined the SigFox partner network are listed in Table2.5.

(44)

TABLE 2.5: The SigFox partner network [72]. Note not all SigFox entities are registered as a SigFox partner, and some

companies may appear in several categories.

Company type Comment Number Antenna designers Provide service in antenna design for SigFox

devices

4

Chip makers Provide either transceiver chips (PHY layer) or System on Chip (SoC)

7

Consulting companies

Provides assistance in IoT projects 7

Design houses Provides custom hardware & electronic designs 31 Device makers Companies that develop different kinds of

devices/solutions compatible with the SigFox network

80

Manufacturers OEM/ODM/EMS 19 System integrators Integrators have experience rolling out connected

solutions for larger corporations and integrating them with their IT system

18

IoT platform providers

Wide range of solutions for data-management, infrastructure (IaaS), hosting, vertical-specific platforms etc

52

Kit makers Chip or module makers that also builds evaluation boards, development boards, development boards, SDKs

18

Module makers Modules include everything needed for

connectivity with the SigFox network and typically additional functionality

17

Solution providers Provides complete end-to-end solutions (usually an app, device and SigFox connectivity)

43

Tech hubs Accelerators or Incubators for startups within SigFox IoT

4

SigFox operators Sigfox Operators roll-out and commercialize Sigfox connectivity around the world

(45)

2.4 LoRaWAN

This section describes LoRaWAN in terms of technical- and non-technical aspects.

2.4.1 Overview

Before diving into the details, there is a need to clarify the following terms that are used throughout in this section: LoRa, LoRaWAN, and the LoRa Alliance. Centenaro et al. define LoRa (Long-Range) a novel physical layer for LPWANs by Semtech Corporation [46]. According to Semtech, LoRa is "agnostic" to higher-layer implementations, thereby allowing LoRa to coexist and interoperate with existing network protocols [73]. On top of LoRa, which is Semtech’s proprietary intellectual property, a special interest group (SIG) named LoRa Alliance, comprised of industrial- and commercial partners are developing an open standard referred to as LoRaWAN for the MAC layer (data link layer) [74]. The LoRa alliance states that "LoRaWANTM

defines the communication protocol and system architecture for the network while the LoRa R _{physical layer enables the long-range communication link." [}₇₄_].

2.4.2 Spectrum and Regulation

According to the LoRa Alliance in 2015, LoRaWAN is compatible with European and North American regulation [74]. However, significant progress has arguably been made; for example, Sk Telecom has deployed LoRaWAN in South Kora as will be mentioned in Section2.4.4. LoRaWAN transmits in Europe within the 863-870 MHz ISM band [37]. However, as explained in Section 2.3.1, it is strictly speaking an SRD band, and not an ISM band. In the US, the LoRaWAN specifications state that it transmits in the 902-928 MHz ISM band as regulated by the FCC [37].

As to which specific sub-bands are employed, the LoRaWAN

specification states that all devices must be capable of transmitting within the 863-870 MHz ISM band within the EU [37]. The specific bands that are mentioned are three defaults which each device must be able to transmit in (within the 868-868.6 MHz band, corresponding to h1.4 in Table2.1). In North America, the LoRa Alliance explains that the segments 902.3 to 914.9 MHz, 903-914.9 MHz, and 923.3-927.5 MHz are used (their particular use is described in Section2.4.3) [74].

The regulatory parameters in Europe for the 868 MHz band, as recommended by CEPT/ECC, and adopted by ETSI [32], were summarized in Table 2.1. It is difficult to define the maximum allowed number of messages per day since this depends on the time on air (TOA), which in turn depends on the data rate, which is variable (see Section 2.4.3). The regulatory parameters in North America for the 900 MHz band are imposed

(46)

by the FCC in e-CFR §15.247 [38] and were described in Section 2.2.3. The LoRaWAN specification states that transmitters shall use duty-cycle limited transmissions exclusively and by default transmit at 25 mW [37]. Worth mentioning, however, is that LoRaWAN uses a frequency hopping scheme (described in Section2.4.3) to circumvent the maximum allowed dwell time in North America.

2.4.3 Technology of LoRaWAN

This section presents the technological aspects of LoRaWAN.

Physical layer

Semtech explains that the LoRa physical layer is based on spread spectrum techniques, such as direct sequence spread spectrum (DSSS) [73]. However, there are certain issues in utilizing DSSS for LPWAN purposes, which Semtech claims to have resolved by using a variation of so-called Chirp spread spectrum (CSS) [73].

In DSSS, Semtech states that the data that is to be transmitted is multiplied (encoded) with a spreading code, i.e., a chip sequence [73]. For multiple DSSS signals to be transmitted simultaneously, the codes must be uncorrelated with each other, i.e., orthogonal [73]. Metaphorically speaking, this is equivalent to a room full of persons talking at the same time, with each conversation in a distinct language, thereby making it easier to distinguish each conversation from the background noise. The spreading sequence itself utilizes a much higher data rate than the original data and therefore occupies a much larger bandwidth [73]. Hence, we say that the underlying data signal is "spread out" in frequency when multiplied with the chip sequence. The receiver multiplies the received message by the inverse of the spreading sequence, thereby de-spreading the signal back to its original bandwidth, and is then able to recover the original data [73].

The strength of spread-spectrum techniques such as DSSS lies in the spreading and subsequent despreading of the signal. In short, the process strengthens the desired signal compared to undesired signals. Semtech maintains that it enables the receiver to successfully recover the data even if the received noise power is larger than the signal power (negative signal-to-noise ratio when expressed in decibels) [73]. According to Semtech, the degree of spreading depends on the length of the spreading sequence and the rate with which it changes, i.e., the so-called "chip rate" [73]. Furthermore, Semtech writes that the "de-spreading" process by the receiver on the signal has an opposite effect on interfering signals, making them look like noise to the receiver and therefore easy to filter out [73].

Daniel Sjöström

Unlicensed and licensed

low-power wide area networks

Exploring the candidates for massive IoT

DANIEL SJÖSTRÖM

Unlicensed and licensed

low-power wide area networks

Exploring the candidates for

massive IoT

Daniel Sjöström

Master’s Thesis

Examiner

Gerald Q. Maguire Jr.

Academic adviser

Anders Västberg

Industrial adviser

Viktor Dahl

Abstract

Abstrakt

Acknowledgements

Contents

List of Figures

List of Tables

List of Abbreviations

Chapter 1

Introduction

1.1

Background

1.2

Problem definition

1.3

Purpose

1.4

Goals

1.5

Delimitations

1.6

Structure of thesis

Chapter 2

Background

2.1

LPWANs

2.2

Regulation and standards

2.2.1

Standardization terminology

2.2.2

Europe

2.2.3

US

2.3

SigFox

2.3.1

Spectrum and Regulation

2.3.2

Technology of SigFox

2.3.3

SigFox’s Non-technological Aspects

2.4

LoRaWAN

2.4.1

Overview

2.4.2

Spectrum and Regulation

2.4.3

Technology of LoRaWAN