DEGREE PROJECT IN INFORMATION AND COMMUNICATION TECHNOLOGY,
SECOND CYCLE, 30 CREDITS STOCKHOLM, SWEDEN 2019
Energy Consumption of Low Power Wide Area Network
Node Devices in the Industrial, Scientific and Medical Band
Rúni Eriksen
KTH ROYAL INSTITUTE OF TECHNOLOGY
Abstract
Low-Power Wide-Area Networks, LPWANs, achieve long communication ranges with a low energy consumption by communicating at low bit rates. Most LPWAN devices are battery powered and are required to operate for an extended period of time, which stresses the requirements for energy efficiency. This thesis investigates the energy consumption of LPWAN devices operating in the Industrial, Scientific and Medical, ISM, band and how use cases affect the consumption. Specifically, LoRa/LoRaWAN and Sigfox are examined. Their key characteristics are described and energy consumption is modelled. The models are verified by comparing the model outputs with measured power consumption of LoRa and Sigfox devices. Through the models, design parameters are investigated with regards to consumption, and product lifetime are estimated.
The influence of use cases on energy consumption is explored by measuring the Package Delivery Ratio, PDR, at different ranges using various bit transmission rates.
The results showed that the bitrate, data redundancy and protocol overhead were among parameters which could be used to optimise energy efficiency. It was also shown, that the device lifetimes could be significantly increased by increasing the transmission interval and removing message acknowledgements. Realistically, LoRa devices can have a lifetime of more than 10 years and Sigfox 3 years, using a 2800 mW h battery. The use case tests showed that a 100 % PDR should not be expected at any bitrate, but lower bitrates and messaging redundancy increase the likelihood of a successful package delivery. Hence, there is a tradeoff between low energy consumption and range/reliability. Additionally, it was found that a low node to gateway distance and a high gateway density increase the probability of a successful transaction. Thus, the power consumption is tightly coupled to the network configuration.
Keywords
LoRa, LoRaWAN, Sigfox, IoT, LPWAN, energy consumption, energy
optimisation
Sammanfattning
Low-Power Wide-Area Networks, LPWANs, uppnår långa
kommunikationsräckvidder med låg energiförbrukning genom att kommunicera med låga bithastigheter. De flesta enheter är batteridrivna och måste operera över längre tid, vilket ökar kraven för energieffektivitet. Denna avhandling undersöker energiförbrukningen för LPWAN enheter i det industriella, vetenskapliga och medicinska ISM bandet och hur olika användningsfall påverkar förbrukningen.
Specifikt undersöks LoRa/LoRaWAN och Sigfox. Deras viktigaste egenskaper beskrivs och deras energiförbrukning modelleras. Modellerna verifieras genom att jämföra resultaten från modellerna med uppmätt effektförbrukning av LoRa och Sigfox-enheter. Genom modellerna undersöks även designparametrar med avseende på strömkonsumtion och produktens livslängd uppskattas. Påverkan användningsfall har på energiförbrukning undersöks genom att mäta Package Delivery Ratio, PDR, vid olika avstånd och bitöverföringshastigheter.
Resultaten visade att bitraten, dataredundansen och protokollstorleken var bland parametrar som kunde användas för att optimera energieffektiviteten. Det visades också att enhetens livslängd kunde ökas signifikant genom att öka överföringsintervallet och ta bort meddelandebekräftelser. Realistiskt kan LoRa- enheter ha en livslängd på mer än 10 år och Sigfox 3 år, med ett batteri på 2800 mW h. Resultatet av olika test visade att en 100 % PDR inte bör förväntas vid någon bitrate, men lägre bitrater och redundans för meddelanden ökar sannolikheten för en paketleverans. Det finns därför en avvägning mellan låg energiförbrukning och räckvidd och sannolikheten för en lyckad packetleverans. Dessutom konstaterades att en låg nod till gateway-avstånd och en hög gateway-densitet ökar sannolikheten för att transaktioner lyckas. Således är energiförbrukningen tätt kopplad till nätverkskonfigurationen.
Keywords
LoRa, LoRaWAN, Sigfox, IoT, LPWAN, strömforbrukning,
energioptimering
Acknowledgements
First of all, i would like to thank my industrial supervisor, Jens Barkval, for all the support, feedback and resources which have been provided throughout the span of the project.
I would like to thank everyone from Prevas being available to discuss practices and technical matter. A special thanks to Oliver von Semkov for providing assistance during the implementation of the cloud platform.
I would like to thank my supervisor Bengt Molin and my examiner Carl- Mikael Zetterling from KTH, for always being available to answer questions and provide feedback.
I would like to thank Mats Landstedt and Carina Dahlberg from IoT Sweden, and to thank Stefan Lindgren and Christopher Koontz from Talkpool for providing information necessary to carry out experimental parts of the work.
I would like to thank my family, friends and girlfriend, Rosaria Notarangelo, for being there for me throughout my education. I would not be able to do this without you.
Finally, to my beloved father, who passed away last year, thank you for all that you have given me in life, for believing in me and always being proud of me.
Rúni Eriksen
Author
Rúni Eriksen <runi@kth.se>
Embedded Systems, Control Track
Information and Communication Technology KTH Royal Institute of Technology
Place for Project
Prevas AB Löfströms allé 5
172 66 Sundbyberg, Sweden
Examiner
Carl-Mikael Zetterling
Professor, Head of Department of Electronics and Embedded Systems School of Electrical Engineering and Computer Science
Box Electrum 229, SE-164 40-Kista KTH Royal Institute of Technology
Academic Supervisor
Bengt Molin
Lecturer, Department of Electronics
School of Electrical Engineering and Computer Science Box Electrum 229, SE-164 40-Kista
KTH Royal Institute of Technology
Industrial Supervisor
Jens Barkval
Consultant Manager Prevas AB
Löfströms allé 5, 172 66 Sundbyberg, Sweden
Contents
1 Introduction 1
1.1 Background . . . . 1
1.2 Problem . . . . 2
1.3 Purpose . . . . 2
1.4 Goal . . . . 3
1.4.1 Benefits, Ethics and Sustainability . . . . 3
1.5 Methodology . . . . 3
1.5.1 Background Study and Modelling Phase . . . . 3
1.5.2 Implementation, Test and Analysis Phase . . . . 4
1.6 Stakeholders . . . . 4
1.7 Delimitations . . . . 4
1.8 Outline . . . . 4
2 Background 7 2.1 Low-Power Wide-Area Networks . . . . 7
2.2 Deployment and Network Topologies . . . . 8
2.3 Modulation Techniques . . . . 9
2.3.1 Narrow and Ultra Narrowband Modulation . . . 10
2.3.2 Chirp Spread Spectrum . . . 12
2.4 Link Budget and Range . . . 14
2.5 Protocol Overhead, Security, Coding Rate, Data Rate and Message Frequency . . . 16
2.6 Chapter Summary . . . . 17
3 LPWAN Enabling Technologies 19 3.1 LoRa and LoRaWAN . . . 19
3.2 Sigfox . . . 25
3.3 Chapter Summary . . . 27
4 Power Consumption Modelling 29 4.1 LoRa / LoRaWAN Model . . . 30
4.2 Sigfox Model . . . 32
4.3 Lifetime Estimation . . . 34
4.4 Chapter Summary . . . 34
5 Implementation 37 5.1 Hardware . . . 37
5.1.1 LoRa Hardware Platform . . . 38
5.1.2 SigFox Hardware Platform . . . 39
5.2 Firmware . . . 40
5.2.1 Periodic Transmission . . . 40
5.2.2 Transmission On Demand . . . 41
5.2.3 Command Line Interface . . . 41
5.3 Cloud Computing, Data Storage and Reporting . . . 41
5.4 Chapter Summary . . . 42
6 Experiments 43 6.1 Power Consumption Tests . . . 43
6.2 Range and Reliability Tests . . . 44
6.3 Chapter Summary . . . 45
7 Results 47 7.1 LoRa Power Consumption . . . 47
7.1.1 LoRa State Durations and Power Consumption Levels . . . . 47
7.1.2 LoRa Power Consumption Characteristics . . . 49
7.1.3 LoRa Lifetime Estimate . . . 53
7.2 Sigfox Power Consumption . . . 54
7.2.1 Sigfox State Durations and Power Consumption Levels . . . 54
7.2.2 Sigfox Power Consumption Characteristics . . . 56
7.2.3 Sigfox Lifetime Estimate . . . 57
7.3 Range and Reliability Tests . . . 59
7.3.1 LoRa Range and Reliability . . . 60
7.3.2 Sigfox Range and Reliability . . . 63
7.4 Chapter Summary . . . 66
8 Discussion and Conclusion 67 8.1 Discussion . . . 67
8.1.1 Power Consumption Optimisation . . . 67
8.1.2 Range and Reliability . . . 69
8.1.3 Related Research . . . . 71
8.2 Conclusion . . . 72
8.3 Future Work . . . 72
References 75
List of Figures
2.1 Network Types . . . . 7
2.2 Network topologies . . . . 8
2.3 Modulation Techniques. Top signal is a digital binary, second is a ASK, third is a FSK and fourth is a PSK modulation. . . . 9
2.4 Spread Waveform and Narrowband Waveform . . . 10
2.5 Narrowband signal in time domain and in spectrogram. . . . 11
2.6 Narrowband bins. . . . 11
2.7 Binary vs Quadrature Phase Shift Keying . . . 12
2.8 Chirp signal in time domain and in spectrogram. . . 12
2.9 Chirp Demodulation. Left image shows a chirp signal modulated with discontinuities, the centered image is an inverse chirp and the right shows the extracted information. . . 14
2.10 Link budget. . . . 15
3.1 LoRa Stack . . . 19
3.2 Device classes in LoRa . . . 20
3.3 LoRa packet. . . 24
3.4 Sigfox Stack . . . 25
3.5 Sigfox transmission. . . 26
3.6 Sigfox packet. Blue is uplink packet and red is downlink packet. . . 27
4.1 General state-based energy consumption model. . . 29
4.2 LoRa Class A . . . . 31
4.3 Sigfox transmission. . . 33
5.1 Generic LPWA Network. . . 37
5.2 Generic IoT Sensor Model. . . 38
5.3 Microchip’s SAM R34 Xplained Pro Evaluation kit. . . 39
5.4 Atmel ATA8520-EK4-E Evaluation kit. . . 40
6.1 Power consumption test setup. Device under test, Otii power supply and data logger and host PC. . . 43
7.1 LoRa packet power consumption using different spreading factors. . 47
7.2 LoRa packet time on air using different spreading factors. The coding rate is fixed at 4/8. The CRC and header are enabled. The
settings payload is 13 bytes. . . 49
7.3 LoRa energy per byte using different spreading factors with a fixed payload of 12 bytes and coding rate of 4/8. The CRC and header are enabled. The settings payload is 13 bytes. . . 50
7.4 LoRa energy per byte using different spreading factors. The coding rate is fixed at 4/8. The CRC and header are enabled. The settings payload is 13 bytes. . . 50
7.5 LoRa packet time on air using different coding rates. The spreading factor is fixed at 9. The CRC and header are enabled. The settings payload is 13 bytes. . . . 51
7.6 LoRa energy per byte using different coding rates with fixed payload of 12 bytes. The spreading factor is fixed at 9. The CRC and header are enabled. The settings payload is 13 bytes. . . 52
7.7 LoRa energy per byte using different coding rates. The spreading factor is fixed at 9. The CRC and header are enabled. The settings payload is 13 bytes. . . 52
7.8 Daily energy distributed in states. . . 53
7.9 Lifetimes. . . 54
7.10 Sigfox transaction. . . 55
7.11 Sigfox packet time on air. . . 56
7.12 Sigfox energy per byte. . . 57
7.13 Daily energy distributed in states. . . 58
7.14 Lifetimes. . . 58
7.15 Measurement locations. . . 59
7.16 LoRa package delivery ratio from various locations using minimum and maximum spreading factor. . . 60
7.17 LoRa number of receptions per transmission from various locations using minimum and maximum spreading factor. . . 61
7.18 LoRa received signal strength indication based on distance. . . 62
7.19 LoRa Package delivery ratio based on distance using minimum and
maximum spreading factor. Including all gateways in the ranges. . . 63
7.20 Sigfox package delivery ratio from various locations. . . 64 7.21 Sigfox number of receptions per transmission from various locations. 64 7.22 Sigfox received signal strength indication based on distance. . . 65 7.23 Sigfox package delivery ratio based on distance using minimum and
maximum spreading factor. Including all gateways in the ranges. . . 65 C.1 LoRa package delivery ratio from various locations using all
spreading factor. . . 90 C.2 LoRa number of receptions per transmission from various locations
using all spreading factors. . . 92 C.3 LoRa received signal strength indication based on distance. . . 94 C.4 LoRa Package delivery ratio based on distance using minimum and
maximum spreading factor. Including all gateways in the ranges. . . 96
List of Tables
3.1 Achievable error detection and correction capabilities in LoRa. . . . 22 3.2 Example of data rates in Europe. . . 22 3.3 Downlink data rates in Europe. . . 23 5.1 User-adjustable variables for LoRa and Sigfox. . . 41 7.1 LoRa power consumption and duration in each phase. Using a
supply voltage of 5V. . . 48 7.2 Sigfox power consumption and duration in each phase. Using a
supply voltage of 3.3V. . . 55 A.1 Calculated vs measured time on air with an estimated settings
payload of 13 bytes and 1 byte of useful data. Header and CRC are enabled. CR is set to 4/5. . . 85 A.2 Calculated vs measured time on air with an estimated settings
payload of 13 bytes and 11 bytes of useful data. Header and CRC are enabled. CR is set to 4/5. . . 85 A.3 Calculated vs measured time on air with an estimated settings
payload of 13 bytes and 1 byte of useful data. Header and CRC are enabled. SF is set to 9. . . 86 A.4 Calculated vs measured time on air with an estimated settings
payload of 13 bytes. Header and CRC are enabled. SF is set to 8.
CR is set to 4/5. . . 86
B.1 Measured time on air with an increasing payload. . . 87
Acronyms
AES Advanced Encryption Standard.
ASK Amplitude-Shift Keying.
BPSK Binary Phase Shift Keying.
CRC Cyclic Redundancy Check.
CSS Chirp Spread Spectrum.
FSK Frequency-Shift Keying.
GFSK Gaussian Frequency Shift Keying.
IoT Internet of Things.
ISM Industrial, Scientific and Medical.
LoRaWAN Long Range Wide Area Network.
LPWA Low-Power Wide-Area.
LPWAN Low-Power Wide-Area Network.
MAC Medium Access Control.
MQTT Message Queuing Telemetry Transport.
NFC Near-Field Communication.
PDR Package Delivery Ratio.
PSD Power Spectral Density.
PSK Phase-Shift Keying.
QPSK Quadrature Phase Shift Keying.
RF Radio Frequency.
RFID Radio-Frequency Identification.
RSSI Received Signal Strength Indication.
SNR Signal to Noise Ratio.
UNB Ultra Narrow Band.
1 Introduction
This chapter covers the background of the project. First it gives a short introduction to Internet of Things, IoT, and explains where Low-Power Wide-Area Networks, LPWAN, fall into the IoT networks. Thereafter, the problem, purpose and goal are described. Finally, a brief description of the methodology used in the project, project stakeholder and delimitations will be given.
1.1 Background
The term IoT, being an umbrella term, covers a wide area of applications. While the internet connects desktops, laptops, servers, phones and tablets, IoT envisions to connect ordinary everyday non-internet-connected devices to the internet, such as house and appliances. The IoT enables everyday objects to communicate over the internet, which allows them to be continuously monitored, easily controlled, as well as to correspond with each other. This opens up a great amount of possibilities, that benefit multiple industries as well as improve the well-being of individuals.
The IoT is present in numerous applications such as smart homes, wearables, smart cities, smart grids, industrial internets, smart cars, healthcare, retail, supply chain and smart farming [1].
The rapid development within electronics, networking, RF-technologies and increase in computational power have made internet enabling technologies more affordable, and continue to do so. Cisco has estimated that in year 2000 about 200 million devices were connected, in 2013 about 10 billion devices and forecast that in 2020 around 50 billion devices will be connected [2]. To put this into perspective. Ericsson reported that there are around 8 billion mobile subscriptions in 2018 [3].
Numerous internet enabling technologies exist. Most well-known technologies
include WiFi and Bluetooth, which are characterised by their short-range
and high-throughput, Near-Field Communication, NFC, and Radio-Frequency
Identification, RFID, that are ultra-short-ranged and have a low-throughput,
and the cellular networks 2G, 3G and 4G that are long-ranged and have a high
throughput. In addition to these, there exists another class of networks which
go under the name LPWANs. LPWANs allow devices to communicate over great distances, while maintaining a low power consumption as a trade-off from communicating at lower bit rates.
Energy consumption plays an important role in IoT. Especially for battery powered devices, which are mounted in remote or inaccessible areas where a lifetime of 10+ years is desired. Such systems need to be carefully developed and the design choices have a big influence on the product lifetime. These design choices and trade-offs will be the subjects for investigation in this project.
1.2 Problem
The design choices, as mentioned in the background section 1.1, heavily influence device lifetimes. The choice of internet enabling technology, the hardware overhead, the device firmware and use are among parameters which impact the lifetime of the devices.
Designing a low power consumption device within IoT, requires cross-disciplinary skills within hardware, software and RF. Attention to the consumption needs to be a key component in all design phases. Additionally, the use cases need to be taken into consideration, when designing devices, as they are tightly coupled to the consumption.
This leads to the questions which this thesis seeks to answer:
• How can LPWAN devices’ power consumption be minimised?
• How do different use cases affect the power consumption of LPWAN devices?
1.3 Purpose
The purpose of the thesis is to identify the properties that are related to LPWAN
technologies’ power consumption and how these can be configured to optimise
device lifetimes. Additionally, the purpose is, to identify the influence use cases
have on the consumption.
1.4 Goal
The goal of this work is to gain insight into competing LPWAN technologies, especially with regards to power consumption, which can help IoT developers make well informed decisions when choosing internet enabling technologies.
Additionally, the goal is to develop a model, that allows developers to estimate device lifetime and estimate the needed battery energy capacities for systems.
1.4.1 Benefits, Ethics and Sustainability
Ethic and sustainability wise, the project contributes to a more sustainable future, as it can help limiting the power consumed by IoT devices. There can be ethical concerns with some use cases for IoT devices, such as surveillance and remote controlling, but this is not something that this project contributes to in particular.
1.5 Methodology
The project will make use of multiple methodologies, depending upon the phase.
The project will be divided into a background study and modelling phase and an implementation, test and analysis phase. The prior phase will be of a more deductive character, whereas the second will be inductive.
1.5.1 Background Study and Modelling Phase
The background study and modelling phase includes two parts.
The first part is a literature study, which seeks to give a broad qualitative introduction to LPWAN technologies, including the key design parameters and modulation techniques.
The second part covers an in-depth description of three chosen LPWAN
technologies and three deduced power consumption models.
1.5.2 Implementation, Test and Analysis Phase
The implementation, test and analysis phase consists of three parts.
The first part describes the hardware platforms that will be used for testing, as well as the software implementation.
The second part covers the construction of test cases, which will be carried out in the project.
The third part presents and analyses the results of the conducted experiments.
1.6 Stakeholders
The project is carried out in collaboration with Prevas. Prevas is an industrial- IT and embedded systems consulting firm. The interest of Prevas in the project is to gain insights in how to choose suitable LPWAN technologies for given use cases, especially with concerns to power consumption. Additionally, the interest is, to obtain models that can be used to estimate device lifetimes of LPWAN devices.
1.7 Delimitations
The project is a master thesis project, which has a duration of 20 weeks and a workload of 40 hours per week. Therefore, a comprehensive analysis and experiments can’t be performed for all LPWAN technologies. This thesis will focus mainly on LoRa and Sigfox, because they are currently the most recognised LPWAN technologies and all of them have network coverage in Stockholm, Sweden, which also makes them testable.
It would be impossible to test everything about the two technologies, so the power consumption tests will mainly focus on the uplinks and the range and coverage tests will be conducted only in an urban area.
1.8 Outline
Chapter 2 presents the background and the key characteristics of LPWANs.
Chapter 3 introduces two LPWAN technologies: LoRa and Sigfox.
Chapter 4 investigates how energy consumption can be modelled and proposes one model for LoRa devices and one for Sigfox devices.
Chapter 5 explains an end-to-end implementation of a LoRa and Sigfox system, which will be used for testing.
Chapter 6 presents the power consumption, range and reliability experiments, that are performed in the project.
Chapter 7 displays the results, that have been obtained through the models and experiments.
Chapter 8 discusses the results and concludes the project.
2 Background
This chapter gives a broad introduction to and describes the key parameters of LPWANs.
2.1 Low-Power Wide-Area Networks
There are many types of networks, as described in the background section 1.1.
The use case will usually determine which type of network that applications would operate on. A common division of network types is illustrated in figure 2.1, where range and computational power are key parameters.
P o w er
Range
LPWA
2G 3G 4G 5G
Wi-Fi
Ultrawide- band
BLE
NFC
Bluetooth
ZigBee
LoRa
NB-IoT LTE-M
SigFox
Figure 2.1: Network Types
We see in figure 2.1, that the low-range and low-power applications include NFC, Bluetooth, BLE and ZigBee, low-range high-power include WI-FI and Ultrawideband, high-range high-power include 2G, 3G, 4G and 5G, and finally high-range low-power devices include LoRa, SigFox, NB-IoT and LTE-M.
The following sections will go more in debt with a range of system design
parameters, which characterise LPWANs.
2.2 Deployment and Network Topologies
The technologies may be deployed in the licensed bands, which is the case for LTE-M and NB-IoT, opposed to technologies such as LoRa and Sigfox, that are deployed in the unlicensed Industrial, Scientific and Medical (ISM) radio bands.
There are advantages and disadvantages of operating in the unlicensed bands. The ISM band has a duty cycle regulation of 1% per hour [4]. The duty cycle regulation applies to communication in both directions, from the node to the base station and from the base station to the node, which gives an asymmetrical communication when the gateway has many nodes connected. The duty cycle regulations also make it challenging to perform over-the-air programming of devices. However, advantages of operating in the unlicensed bands include, that the devices don’t have to continuously listen to messages from base stations, as well as adapt complicated protocols, which is required in the licensed band [5].
LPWANs can be configured in several different network topologies as shown in figure 2.2.
Star Topology Hybrid Topology Mesh Topology
Figure 2.2: Network topologies
The hybrid topology is rarely/never used, since it would require devices to always be on and listening to messages, which significantly increases the consumption.
The mesh topology might be used for relaying or repeating information from one
device to a gateway, when the gateway is out of reach, as Ochoa et al. show in
[6]. The most common configuration is the star of stars configuration, where
gateways or base stations are connected to the internet and nodes are connected
to the gateways. The star topology has several advantages in the LPWANs. They
provide deterministic and longer battery lifetimes for the nodes and require less
network equipment [7, 8]. Additionally, the maintenance is decreased as nodes
can run out of power or stop functioning without affecting other nodes.
2.3 Modulation Techniques
Modulation is the process of varying one or more properties of a carrier wave. The carrier wave is a periodic waveform and information is modulated onto the carrier wave by changing either its amplitude, frequency or phase. Figure 2.3 shows an example of a modulated signal using Amplitude-Shift Keying (ASK), Frequency- Shift Keying (FSK) and Phase-Shift Keying (PSK) respectively.
0 1 0 1 0 1 1 0
Figure 2.3: Modulation Techniques. Top signal is a digital binary, second is a ASK, third is a FSK and fourth is a PSK modulation.
The modulation techniques showed in figure 2.3 are used in RF applications.
When the carrier wave is modulated at a slower rate, the receiver sensitivity can be increased [9]. Currently there are two carrier wave trends in the LPWA area.
The carrier waveform is either a narrowband or a spread waveform. Figure 2.4
illustrates the two waveforms.
Frequency [Hz]
Noise Level
Narrowband Waveform
Spread Waveform
P o w er [dB]
Figure 2.4: Spread Waveform and Narrowband Waveform
The narrowband waveform transmits messages in a narrow bandwidth. As the throughput is decreased, the frequency bandwidth can be decreased. This results in a high Power Spectral Density, PSD, in the narrow band, that can punch through the noise floor. Rather than composing a signal that has high PSD, spread spectrum techniques spread the carrier signal across the available frequency bandwidth and acquire the modulated signal by using post-processing techniques.
2.3.1 Narrow and Ultra Narrowband Modulation
Narrowband techniques have regained popularity due to advances made in signal processing and within random access schemes [10].
Narrowband systems only use, as previously mentioned, a narrow bandwidth,
which gives a high PSD. A narrowband signal is illustrated in figure 2.5.
Time (s)
Amplitude
(a) Narrowband signal in time domain.
Time (s)
Frequency (Hz) Power (dB)
(b) Spectrogram of narrowband signal.
Figure 2.5: Narrowband signal in time domain and in spectrogram.
Figure 2.5, shows the narrowband signal in time domain and a spectrogram of the signal. To decode the signal, the receiver needs to listen to the narrow frequency range, that the signal is modulated onto. Using only a part of the allocated bandwidth allows multiple systems to communicate with the same receiver simultaneously, as illustrated in figure 2.6. In narrowband systems it is also possible to communicate on multiple tones, by using more bins, for faster transmission rates.
Frequency [Hz]
Noise Floor Narrowband Waveform Bandwidth
P o w er [dB]
Figure 2.6: Narrowband bins.
The bandwidths range from 3.75 to 15 kHz as used by NB-IoT to as low as 100 Hz, which Sigfox uses [11, 12]. The later is often referred to as Ultra Narrow Band, UNB. The high PSD makes the signal resistant to interference and jamming [10].
Narrowband signals are modulated using PSK as shown in figure 2.7.
11 Q
I
10 00
01
QPSK
1 Q
I 0
BPSK
Figure 2.7: Binary vs Quadrature Phase Shift Keying
For the highest energy per encoded bit Binary Phase Shift Keying, BPSK, is used, where the phase is changed by 180 degrees and Quadrature Phase Shift Keying, QPSK, which changes the phase in steps of 90 degrees, is used as it doubles the data rate, but it also doubles the probability of faults when the signal is to be demodulated. The modulation rate can be increased greatly by modulating the carrier more densely, but it increases the requirements to the receiver.
2.3.2 Chirp Spread Spectrum
Chirp Spread Spectrum, CSS, is a spread spectrum technique, where information is spread out by encoding the information onto a chirped signal. A chirp is a linear frequency sweep. Figure 2.8 shows a chirp signal in the time domain as well as a spectrogram of the chirp.
Amplitude
Time
(a) Chirp in time domain.
Time (s)
Frequency (Hz) Power (dB)
(b) Spectrogram of chirp.
Figure 2.8: Chirp signal in time domain and in spectrogram.
As shown in figure 2.8, the CSS technique utilises the full allocated bandwidth by
spreading out the signal from f min to f max . This makes it resistant to multi-path fading even when operating on very low power as well as resistant to the Doppler effect [4].
The way spread spectrum signals can transmit information, at signal strengths lower than the noise floor can easiest be explained based on the Shannon - Hartley theorem [13]. The Shannon - Hartley theorem determines the maximum rate at which information can be transmitted, within a specified bandwidth wherein noise is present. The Shannon - Hartley theorem is given by:
C = BW · log 2
( 1 + S
N )
(1) Where C is the channel capacity in bit/s, BW is the channel bandwidth in Hz, S is the average received signal power in W atts and N is the average noise or interference power. S/N is the Signal to Noise Ratio ,SNR, which is expressed as a linear power ratio. Rearranging the above equation from log base 2 to the natural log, e, and by noting that ln = log e , the equation can be manipulated to:
C
BW = 1.433 · S
N (2)
In spread spectrum applications, the signal to noise ratio is small. Often it is below the noise floor. Using the assumption that S/N << 1 we can rewrite the equation to:
C
BW ≈ S
N (3)
Or its reciprocal:
BW C ≈ N
S (4)
Here we see that by increasing the bandwidth, a greater throughput can be achieved as well as a resistance to higher noise levels.
Having established how a signal can be received at such low signal strengths, by
multiplying the original signal with a spreading code, we can now examine the
relationship between bit-rates and chip-rates. The chip sequence, which in CSS is
a chirp sequence, occurs at a much higher rate than the data signal. The higher
the chip rate is compared to the bit rate, the higher is the processing gain as given
by the equation below [13]:
G p = 10 · log 10
( R c R b
)
(5)
Where R c is the chip rate and R b is the bit-rate.
The processing gain enables correct recovery of weak data signals, even with negative SNR. Interfering signals are also reduced by the processing gain of the receiver [14]. The interference is spread across the desired information bandwidth and can easily be removed by filtering [13].
In CSS the information is coded onto the carrier wave by making frequency discontinuities as shown in figure 2.9.
Frequency (Hz)
Time (s)
Frequency (Hz)
Time (s)
020406080100120
Time (s)
X =
Figure 2.9: Chirp Demodulation. Left image shows a chirp signal modulated with discontinuities, the centered image is an inverse chirp and the right shows the extracted information.
Figure 2.9 also shows the demodulation process of the CSS signal. Multiplying the received signal with the inverse chirp gives the data.
2.4 Link Budget and Range
The range of wireless communication systems depend on the link budget. The link budget consists of all gains and losses a system experiences. The gains and losses include system gains, losses associated with the antenna, matching networks and losses due to signal propagation [15]. An illustration of the link budget can be seen in figure 2.10, where the gains and losses are sketched.
Mathematically, link budget of a wireless system can be expressed as [15]:
P RX = P T X + G T X − L T X − L F S − L M + G RX − L RX (6)
As a little sidenode, the units dBm and dBi are explained here, to make the subsequent paragraph easier to understand. The unit dBm stands for decibels relative to milliwatts and is the amount of relative power transmitted by an amplifier. The unit dBi stands for decibels relative to an isotropic antenna and describes the gain an antenna puts on a signal.
Where P RX is the receiver power in dBm, P T X is the transmission power in dBm, G T X is the transmitter antenna gain expressed in dBi, L T X represent the losses in the transmission due to short antennas, feed lines and similar, L F S are the propagation losses in the system and M is the fading margin in dB, G RX is the receiver antenna gain expressed in dBi and L RX represent the losses in the receiver.
Tx
Rx
Free Space
Receiver Sensitivity
P o w er (dBm)
-100 14
Tr ansmission P o w er C onnec tor L oss , ec t. A n tenna G ain A n tenna G ain C onnec tor L oss , ec t. R ec eiv ing P o w er
Free Space Loss and Fading
Figure 2.10: Link budget.
In LPWA systems the transmission power is limited, which increases the receiver sensitivity requirements. The needed reception depends on the transmission frequency. The path loss is inversely proportional to the transmission frequency, so low frequency systems can communicate over greater distances [15]. This is also evident from the radio engineer formula, which calculates the path loss based on the distance and wavelength:
L F S = 20 · log 10
( 4πd λ
)
(7)
Where d is the distance and λ is the wavelength.
We see from the radio engineer formula, that by doubling the frequency the path loss doubles. WIFI-5, as an example, has an operating frequency of 5 GHz. Comparing this to LPWAN technologies in the unlicensed band, which in Europe operate on 868 MHz, it can be seen that the losses are significantly decreased. The low operating frequencies contribute to LPWA systems’ the ability to communicate over larger distances.
The LPWA systems operating in the ISM band, have a limited transmission power of 14dBm whereas the transmission power in the cellular bands are 20 to 23dBm.
Most LPWA systems are designed to achieve a link budget of around 150 ± 10dB [16], whereas other wireless systems such as ZigBee and Bluetooth have a link budget of −125 and −90 dBm [17].
2.5 Protocol Overhead, Security, Coding Rate, Data Rate and Message Frequency
Transmitting messages over longer distances while maintaining a low power consumption requires messages to have a high information rate. Every extra bit, that is added to the message payload contributes to an increased transmission time. Adding extra information is, however, necessary to make sure the message arrives in the correct place, is secure and to ensure correctness of received messages [16].
Generally, a transmitted packet will, in addition to the payload, contain a destination, device information and integrity checks. The destination ensures that the packet arrives at the correct location, the device information contains information about the transmitting device, which ensures that it is possible to identify the device from the receiver, and the message integrity check makes sure that the information has not been corrupted during the propagation process [1].
In LPWA systems, the protocol overhead is low compared to the overhead in other
systems, but amongst themselves vary greatly [18]. Minimising the overhead
will improve device lifetime, but might compromise safety. Furthermore, adding
extra information, such as linear error correcting codes, can give the devices an
opportunity to detect and correct errors, instead of relying on re-transmission.
As mentioned in section 2.4, the data transfer rate is usually low in LPWA systems, so the energy per transmitted bit is high. For this reason, it is important to pack as many bits of information into each message as possible to minimise protocol overhead.
2.6 Chapter Summary
In this section LPWANs have been characterised and put into perspective with
regards to other network types. Compared to other network types, LPWANs have
a long range, low power consumption as a compromize for a low throughput. The
deployment strategies have presented, showing that most LPWANs are deployed
in a star network topology, where the edge devices communicate with a gateway,
which forwards the message to the internet. The reason being that these networks
are easily configurable, robust and energy efficient. The modulation techniques
used in LPWANs are simple (binary) and slow, which increases the power per
bit and thereby the range and reliability. In LPWANs two carrier waves are
dominant. The narrowband waveform, where the digital signal is modulated in a
narrowband and a spread waveform, where the signal is modulated onto a spread
wave. Finally, properties of LPWANs such as range, reliability, protocol overhead,
security, coding rate, data rate and message frequency have been presented.
3 LPWAN Enabling Technologies
This section will give an in-depth description of LoRa/LoRaWAN and Sigfox, covering their business models, stacks, protocols, key characteristics and packet structures.
3.1 LoRa and LoRaWAN
LoRa, short for Long Range, is a wireless communication modulation method, which uses a variation of CSS to transmit messages. LoRa also offers the possibility of communicating using Gaussian Frequency Shift Keying GFSK [19].
Long Range Wide Area Network, LoRaWAN, is the protocol, which is used together with LoRa. The physical layer of LoRa is closed and proprietary technology, that belongs to and is maintained by Semtech, while LoRaWAN is an open standard. The LoRaWAN protocol is promoted by LoRa Alliance, which consists of more than 500 member companies [20]. There exist both private and public LoRa networks and everyone is permitted to set up their own LoRa gateway.
The network is a star of stars type networks as specified in section 2.2.
Stack The stack of LoRa is illustrated in figure 3.1.
LoRa MAC
LoRa Modulation Application Layer
Class A
EU 868
EU 433
US 915
AS
430 ...
Class B Class C
Figure 3.1: LoRa Stack
In figure 3.1, we see that the application is the top level of the stack, that includes
possible cloud services, databases and applications. The Medium Access Control, MAC, layer is the high-level LoRaWAN protocol. LoRa modulation is the physical layer, which is modulated onto the regional ISM bands.
Device Classes Three different classes are offered in LoRa as shown in figure 3.2 [19].
Transmit Receive 1 Receive 2
Receive delay 1 Receive delay 2
Class A
Beacon Ping Transmit Receive 1 Receive 2 Beacon
Beacon period
Class B
Transmit Receive 2 Receive 1 Receive 2
Time Class C
Figure 3.2: Device classes in LoRa
All LoRa devices must implement class A, which is the most simple class, whereas the other two are optional.
Class A is expected to be the most used class, as it has the best power saving capabilities. It features an end-device initiated bi-directional communication.
The end-device transmits a message at a random instance of time and the gateway responds after two predefined delays. The messages in both receive windows are identical.
Class B allows periodic receive slots. It allows the device to receive like class A devices, but additionally opens the receive window at periodic intervals, where the gateway will send messages to the device.
Class C always listens to the gateway, so it implements a traditional bi-directional
communication system, that can be initiated from both the end-device as well as
the gateway/user.
Spreading Factors The spreading factor is a key variable in LoRa, which has great influence on both the range, transmission speed and power consumption, as it increases and decreases the time and energy used to transmit each bit. The signals for each spreading factors are orthogonal, meaning that messages can be transmitted simultaneously without causing interference. LoRa uses spreading factors SF , in the range 7 to 12 to control the data rate of the transmitted signals [13]. The symbol period, T s , is given by:
T s = 2 SF
BW (8)
So, the symbol rate, R s , is the reciprocal of the symbol period:
R s = BW
2 SF (9)
The chip rate, R c , which is the number of pulses per second, can be calculated as:
R c = R s · 2 SF = BW
2 SF · 2 SF = BW (10)
The modulation rate or bit rate, R b , is:
R b = SF · BW
2 SF (11)
Coding Rates LoRa uses forward error correction. Although LoRa modulation is proprietary, reverse engineering attempts show that LoRa uses Hamming codes [21, 22], that adds an overhead to the transmitted messages and the nominal bit rate is therefore:
R b = SF · BW
2 SF · CR (12)
Where the coding rate, CR, is given by:
CR = 4
4 + n (13)
Where n is a number in the range 1 and 4. The Hamming codes add error detection
and correction capabilities to the code. By increasing n by one, the code distance
increases by one, which gives the capabilities specified in table 3.1 [23].
Table 3.1: Achievable error detection and correction capabilities in LoRa.
Code Rate Error Detection [Bits] Error Correction [Bits]
4/5 Parity 0
4/6 1 0
4/7 2 1
4/8 3 1
We see in table 3.1, that the minimum coding rate corresponds to a parity check bit, that is capable of detecting all uneven number of bit faults, and the maximum can detect 3 bit errors and correct 1.
Uplink Data Rates The bit rates, which can be achieved, depend on the regions in which the system is deployed. Table 3.2, specifies the bit rates which are achievable in Europe [19].
Table 3.2: Example of data rates in Europe.
Data Rate Spreading Factor Bandwidth [kHz] Coding Rate Bit Rate [bps]
DR0 12 125 4/5 293
DR1 11 125 4/5 537
DR2 10 125 4/5 977
DR3 9 125 4/5 1758
DR4 8 125 4/5 3125
DR5 7 125 4/5 5469
DR6 7 250 4/5 10938
DR7 FSK - 4/5 50000
Table 3.2 shows that the data rate halves every-time the spreading factor is
increased. Equivalently, the energy per bit increases which gives a greater range
and higher energy consumption. LoRa features an adaptive data rate, that
regulates the data rate according to the RSSI, where it seeks to maximise the data
rate. LoRa also offers communication via standard FSKing, at a high bit rate,
which can be used to update firmware over the air [16].
Downlink Data Rates The first receive window RX 1 uses a frequency that is a function of the uplink frequency. Likewise, RX 1 uses a data rate which is a function of the data rate used for the uplink. The relationship between the uplink and RX 1 downlink data rate is region specific. The downlink data rate as a function of uplink data rates in Europe are shown in table 3.3 [24].
Table 3.3: Downlink data rates in Europe.
RX1DROffset Upstream data rate
0 1 2 3 4 5
Downstream data rate in RX1 slot
DR0 DR0 DR0 DR0 DR0 DR0 DR0
DR1 DR1 DR0 DR0 DR0 DR0 DR0
DR2 DR2 DR1 DR0 DR0 DR0 DR0
DR3 DR3 DR2 DR1 DR0 DR0 DR0
DR4 DR4 DR3 DR2 DR1 DR0 DR0
DR5 DR5 DR4 DR3 DR2 DR1 DR0
The second receive window RX 2 uses a fixed frequency and data rate. The data rates can be changed through MAC commands. The default data rate in Europe is DR0 which uses SF 12 and 125kHz [24].
Packet Structure The LoRa packet structure is specified by the LoRa Alliance
[19]. Figure 3.3 shows the structure for uplinks and downlinks. Downlinks have
the same structure except the Payload Cyclic Redundancy Check, CRC, is omitted
in the physical layer.
Preamble
Variable
PHY Header
2 bytes
Header CRC
4 bits
PHY Payload
Variable
Payload CRC
2 bytes PHY Layer
MAC Header
1 byte
MAC Payload
Variable
MIC
4 bytes MAC Layer
Frame Header
7-22 bytes
Frame Port
1 byte
Frame Payload
Variable App Layer
Dev Addr
4 bytes
Frame Ctrl
1 byte
Frame Cnt
2 bytes
Frame Opts
0-15 bytes
Uplink
Downlink
Preamble
Variable
PHY Header
2 bytes
Header CRC
4 bits
PHY Payload
Variable
Figure 3.3: LoRa packet.
The packet has three layers, as shown in figure 3.3. The physical layer, the MAC layer and the application layer.
The physical layer is the layer which is transmitted between nodes and gateways.
It contains a preamble, that is used to synchronise transmitters and receivers. The duration varies across regions, but in EU it is specified to 8 symbols [24]. This is followed by a PHY header and CRC, that are encoded with the highest coding rate of 4/8. The header contains information the payload size and whether the payload has a CRC. The payload contains the MAC frame.
The MAC layer’s destination is the network server. It contains a header, that specifies the protocol version, if the message is an uplink or downlink, if the message should be ack’ed or if a device wishes to join the network. The MAC header and payload are used to calculate the message integrity code (MIC), using a network session key N wk_SKey. Once the MIC has been calculated, the MAC payload is sent to the application.
The application layer frame contains a frame header, a frame port depending on
the application and the frame payload, which is encrypted with an application session key App_SKey using an encryption method based on the Advanced Encryption Standard, AES, 128 algorithm.
The frame header contains the device address, a frame control, that specifies if the previous message was received, which data rate should be used and similar, the frame count is a running number. Finally the frame options contains optional settings to change data rate, transmission power and other parameters.
3.2 Sigfox
Sigfox, launched in 2012, was one of the first technologies on the LPWAN market [25]. Opposed to LoRa, the Sigfox node device standard is open, whereas the gateways and network servers are proprietary and run by Sigfox, so they provide an end-to-end solution.
Sigfox, has to comply to the ISM regulations of a maximum duty cycle of 1% in both directions. To comply to the regulations, Sigfox has limited the communication to 140 uplink messages with a payload of 12 bytes per day and downlink messages of 8 bytes, but only allows 4 downlinks per day [26].
Stack The Sigfox layers are illustrated in figure 3.4.
Sigfox MAC
Sigfox Modulation (UNB) Application Layer
EU 868
EU 433
US 915
AS
430 ...
Figure 3.4: Sigfox Stack
The application layers implement the user requirements, the MAC layer handles
the MAC assembly for uplinks and disassembly for downlinks, it handles
authentication and applies error detection using frame check sequence. The PHY layer applies the modulation and adds or removes the preamble. Sigfox uses Ultra Narrow Band, UNB, together with BPSK to modulate its messages and operates within the same ISM band, from 868.180 to 868.220 MHz. It divides the band into 400 sub bands of 100 Hz, where of 40 are reserved for future use [27]. The bitrate is 100 bps for uplinks and 600 bps for downlinks [26].
Communication Method To increase the probability of successful message delivery, Sigfox uses time and frequency diversification, illustrated in figure 3.5 [28].
Transmit 1
Transmit 2
Transmit 3 Receive
Receive delay
20s 20s
F requenc y [H z]
Time [s]
Figure 3.5: Sigfox transmission.
It transmits data on a random frequencies within the band and then repeats the process twice on two other pseudo random frequencies. The time and frequency diversification hinders packet collision and makes the system more resistant to jamming and noise [10].
Packet Structure The Sigfox packet structure is illustrated in figure 3.6
[29].
Preamble
19 bits
Frm Sync & Header
29 bits
Device ID
32 bits
Payload
0-96 bits
Hashcode
16-40 bits
CRC
16 bits
Preamble
91 bits
Frm Sync & Header
13 bits
ECC
32 bits
Payload
0-64 bits
Hashcode
16 bits
CRC
8 bits
Uplink
Downlink
Figure 3.6: Sigfox packet. Blue is uplink packet and red is downlink packet.
As seen on figure 3.6, the packet structure is relatively simple with a low protocol overhead. The preamble is used to notify the base station and end-device, that a message is incoming. The preamble for uplinks is noticeably shorter than the one for downlinks, because of the link budget imbalance due to the transmission rates.
The frame sync and header contain information about the frame, destination and similar. The device ID, which only exists in uplinks, contains information about the end-device. The error correction code ECC, that is in the downlinks, is likely added because of the link budget imbalance, due to the difference in data modulation rates. The hashcode or authentication code is calculated based on the payload. Finally, the CRC serves as an integrity check.
3.3 Chapter Summary
In this section the two most dominant LPWA technologies, in the unlicensed
ISM band, LoRa/LoRaWAN and Sigfox have been described. LoRa uses a wide
bandwidth chirp to modulate the signal onto and offers a lot of customisability
from device classes, that offer different communication methods, variable data
redundancy, data rates and payload sizes. The device technology is proprietary
and monitored by Semtech, whereas the network is open source and monitored
by the LoRa Alliance, which allows users to set up their own gateways. Sigfox
modulates the signal onto a narrowband carrier wave. While Sigfox offers the user
to vary the payload, it does have a fixed overhead and data rate. Sigfox’s business
model is similar to the cellular operators, they own the network and charge for
usage, whereas the devices are open source, so everyone is free to create their own
Sigfox hardware.
4 Power Consumption Modelling
The energy consumption of IoT sensor nodes can be modelled by identifying the phases that the product operates in and thereafter the power consumed in each phase, as proposed in several publications on sensor networks [30–34].
The model assumes a constant duration and consumption. When the energy consumption in one message transaction is identified, the distribution of power dissipation based on the phases can be determined as well as the product battery lifetime.
Figure 4.1 shows a phase division of a typical IoT sensor node.
P o w er
Time
Wake-up system
Measure
Process data
Wake-up tranciever
Data transmission
Data reception
Sleep preparation
Sleep
Duration < 1% Duration > 99%
T
WUT
mT
procT
WUTT
TrT
RT
SPT
SleepFigure 4.1: General state-based energy consumption model.
In figure 4.1, the period is divided into 8 phases.
The total energy can be be calculated as:
E T ot = E Active + E Sleep (14)
Where E Active is the energy consumed when the system is awake and E Sleep energy consumed when the system is in sleep mode. The energy consumed in sleep mode is calculated as:
E Sleep = P Sleep · t Sleep (15)
Where P Sleep and t Sleep are the power consumption and duration in sleep mode,
respectively. The energy consumed in active mode can be calculated as:
E Active = E W U + E m + E proc + E W U T + E T x + E Rx + E SP (16)
Where the energies from the states in figure 4.1 are represented. The energies are calculated like the energy in sleep mode, by multiplying their power consumption with their duration.
The relevance of each state depends on the application as the duration and power consumption vary greatly. The data processing, for example, can in some cases cover complex operations whereas in others it can be a simple conversion.
However, usually the most important states are the transmission, reception and the sleep state, as the transmission and reception consume most power and the sleep state is the state where the system will be 99% of the time.
Using this general energy model, the technology specific models can now be derived based on their specific characteristics.
4.1 LoRa / LoRaWAN Model
LoRa is an ALOHA type protocol meaning that it can send information without doing any synchronisation with the network. This simplifies the modelling process as only the data transmission and reception are of variable time [35].
Recalling from section 3.1, LoRa has three classes. Class A which is a node-
initiated bi-directional communication, class B which implements A in addition
to a beacon period and class C that also implements A, but where the node
continuously listens to the network when it is not transmitting. Most sensor node
applications fall into class A. Class A is also the most energy optimal class, so this
project will only model class A devices. Figure 4.2 illustrates the communication
flow of LoRa class A devices.
Transmit Receive 1 Receive 2
Receive delay 1
Receive delay 2
Class A
Time
Figure 4.2: LoRa Class A
Transmission The LoRa transmitted messages consist of a preamble and payload:
t packet = t preamble + t payload (17)
The payload size can be adjusted by enabling or disabling parts of the payload as well as by changing the spreading factor, coding rate and effective payload size.
The number of payload symbols can be modelled as [36]:
n payload = 8 +
ceil
( 8 · P L − 4 · SF + 28 + 16 − 20 · H 4 · (SF − 2DE)
)
· (CR + 4), if ≥ 0
0, otherwise
(18) Where DE is data optimization, H is a header and CRC is a cyclic redundancy check, which individually can be enabled or disabled depending on the application needs. SF is the spreading factor and P L is the payload including both settings and the message payload:
P L = P L settings + P L usef ul (19)
The number of preamble symbols can be modelled as:
n preamble = 4.25 + n regional (20)
Where n regional is a regional constant, which is 8 in Europe.
The symbol duration can be calculated as:
t symbol = 2 SF /BW (21)
Where SF is the spreading factor and BW is the bandwidth. Using the symbol duration the packet time on air can finally be determined as:
t packet = (n preamble + n payload ) · t symbol (22)
Finally,
multiplying the packet duration with the transmission power consumption, the energy consumption per transmission can be calculated as shown below:
E packet = t packet · P T X (23)
Reception The downlink messages are constructed similarly to the uplink messages, but the CRC is omitted. As shown in figure 4.2, every transmission is followed by two reception windows. In the first reception window, the downlink data rate is a function of the uplink frequency and data rate as shown in table 3.3.
The message in the second receive window is sent with spreading factor 12 [24].
The messages sent in receive window 1 and receive window 2 are identical, so if the first message is successfully received, the second message must not be decoded.
Essentially, the second receive window is there as an insurance that the message will be received correctly [19].
This leaves the device with three possible message transaction scenarios:
• An unacknowledged transmission, where both receive windows are ignored.
• An acknowledged transmission, where only one window is decoded.
• An acknowledged transmission where both receive windows are decoded.
The least consuming scenario is the unacknowledged transmission and the most consuming is the double acknowledged transmission.
4.2 Sigfox Model
Sigfox also uses an ALOHA type protocol [10]. Messages can be initiated without
any network synchronisation. As described in section 3.2, Sigfox uses time and
frequency diversification when transmitting messages, as shown in figure 4.3, in
order to avoid message collisions [26].
Transmit 1
Transmit 2
Transmit 3 Receive
Receive delay
20s 20s
