Vous êtes sur la page 1sur 5

2018 IEEE Middle East and North Africa Communications Conference (MENACOMM)

MAC layer-based evaluation of IoT technologies:


LoRa, SigFox and NB-IoT
H. Mroue1,4 , A. Nasser3 , S. Hamrioui1 , B. Parrein2 , E. Motta-Cruz1 and G. Rouyer4
1 IETR UMR CNRS 6164, Université de Nantes, Polytech Nantes, France
2 LS2N UMR CNRS 6004, Université de Nantes, Polytech Nantes, France
3 American University of Culture and Education, Beirut, Lebanon
4 SPIE City Networks, France

Email: hussein.mroue@univ-nantes.fr

Abstract—In this paper, we explore functionalities of MAC [7]. This paper deals also with a simulation of a dense network
layer for three Low Power Wide Area Networks (LPWAN) for each of these technologies. The contribution of this paper is
technologies (LoRa, Sigfox and NB-IoT) dedicated for Smart the comparative study of these technologies according to their
Cities applications and Internet of Things (IoT). These networks
are dedicated to long-range (up to dozens of kilometers) and low functionalities: carrier frequency, packet duration, number of
rate communication to ensure a good autonomy up to 10 years. channels and spectrum access.
The technical differences at the level of MAC layer between Sigfox The paper is organized as follows. Related works are
and LoRa and NB-IoT are explained and evaluated. We apply discussed in section II. MAC Layer for LoRa, Sigfox and
our simulator to evaluate the performances of these technologies NB-IoT are reviewed in section III. Probability calculations of
in terms of Packet Error Rate.
Index Terms—Low Power Wide Area Networks (LPWAN), Packet Error Rate are explained in section IV. In this section,
Internet of Things (IoT), Smart City, Medium Access Control we describe the calculation of PER in different cases. Next the
Layer (MAC Layer), Packet Error Rate (PER), LoRa, Sigfox, results of simulations are presented in section V and finally
NB-IoT. the conclusion is given in section VI.

I. I NTRODUCTION II. R ELATED W ORK


Today, we live the Internet of Things (IoT) time, where Iot Recent research works about the LPWAN technologies
networks need to be connected to disseminate the huge amount extended to evaluate the performance of LPWAN [8]–[17].
of information [1]. IHS Markit [2] announces 75 billions of Most of these works describe a single technology or compare
connected objects to come and spread in all sectors for 2025. only 2 technologies based on physical layer or make an
Such objects will be devices, automata, sensors surrounding experimental study which is not quite the case described.
our houses and public places. They can transmit many types In [8], several structures of LoRaWAN are illustrated and
of information such as temperature, humidity, public lighting, modeled with many scenarios taking into account the physical
home automation, parking spot (occupied or vacant) and e- layer and focusing on the problem of network scalability. In
health [3]. However, the connection of these objects to the [9], the performance of pure Aloha used by LoRa and Sigfox
Internet requires a compatible and scalable wireless infras- is investigated and the theoretical throughput expressions of
tructure capable of absorbing the exponential increase in the both technologies are derived. In [10], the capabilities and
number of devices and managing such amount of information. limitations of LoRaWAN are investigated, when a strategy is
This process results in new technical challenges that need to be adopted in order to identify in which use cases a LoRaWAN
investigated. Low Power Wide Area Networks (LPWAN) are technology works, and in which use cases it does not work.
often used when other wireless networks, such as Bluetooth- [11] presented a comparative and description study of the
BLE, Wi-Fi and ZigBee, are not suitable for long-range perfor- technical differences between LoRa and NB-IoT in terms of
mance. LPWAN is a solution of low-energy wireless compared physical features, network architecture and MAC protocol.
to M2M cellular networks in terms of energy consumption and In [12], a technical simulation study of an Ultra Narrow
hardware complexity. The international standardization bodies, Band (NB) and Chirp Spreading Spectrum (CSS) protocol for
like 3GPP, as quickly pushed research towards infrastructures long range communication and a measurement study of their
with wireless access network taking into account the technical interference sensitivity, and then a network scalability study
specificities of the connected objects such as GPRS and LTE- to see how they cope with both inter and intra-cell collisions
M [4]. But recently, a multitude of new radios technology have and interference, but the numerous network parameters are not
emerged. The precursors of this LPWAN technology are today exposed enough thus lacking of transparency. In [13], some
LoRa [5] and Sigfox [6]. interesting insights on the decoding performance in LPWAN
In this paper, we describe the features of the Medium Access with packet collisions but the model is based on physical
Control (MAC) layer for the existing solutions of LPWAN layer. Several experimental studies of LPWAN technologies
technologies, LoRa, Sigfox and Narrow Band IoT (NB-IoT) are being introduced into [14]–[17]. In this article, the main

978-1-5386-1254-5/18/$31.00 ©2018 IEEE


2018 IEEE Middle East and North Africa Communications Conference (MENACOMM)

contributions of this work are : (a) a technical study of MAC one random channel among a set of available one (in our
layer for 3 LPWAN technologies, (b) a probabilistic model is simulation : 200 channels) [21]. The packet duration is up
proposed, based on the use of calculation probabilities to find to 2 ms [22].
the best technology.
C. Narrow band IoT
III. M EDIUM ACCESS C ONTROL L AYER According to [18], NB-IoT reuses the LTE design ex-
This section presents an analysis on the MAC layer of the tensively. For uplink, NB-IoT use single-carrier frequency-
LPWAN technologies and how collision domain is handled. division multiple-access (SC-FDMA). An NB-IoT carrier uses
Note that, LoRaWAN and Sigfox start from the unslotted twelve 15 kHz sub-carriers for a total of 180 kHz. There
ALOHA MAC protocol [12], which implies that essentially are two types of transmission for NB-IoT, multi-tone and
all devices can access the channel whenever they want. But, single-tone. Multi-tone transmission is based on SC-FDMA
the uplink of NB-IoT uses Single-Carrier Frequency-Division with the same 15 kHz sub-carrier spacing, 0.5 ms slot and 1
Multiple-Access (SC-FDMA) [18], in SC-FDMA, multiple ms subframe as LTE. Single-tone transmission supports two
access among users is made possible by assigning different numerologies, 15 kHz and 3.75 kHz. The 15 kHz numerology
users different sets of non-overlapping sub-carriers. This sec- is identical to LTE and thus achieves the best coexistence
tion compares the 3 technologies in terms of MAC layer performance with LTE in the uplink. The 3.75 kHz single-
specificities. Table I illustrates numerical values were obtained tone numerology uses 2 ms slot duration. Like the downlink,
from [5] for LoRa, [19] for Sigfox and [18] for NB-IoT. an uplink NB-IoT carrier uses a total system bandwidth of 180
kHz. Table III illustrates the different types of NB-IoT.
Table I
LPWAN PROTOCOLS Table III
NB-I OT SPECIFICATIONS
LoRa Sigfox NB-IoT
Band 868/915 MHz 868/915 MHz LTE Band Sub-carrier No of No of Transmission
PHY CSS UNB NB spacing tones SC-FDMA symbols time
Spreading Factor 27 − 212 — — 15 KHz 12 14 1 ms
Bandwidth 500-125 KHz 1 KHz 180 KHz 15 KHz 6 28 2 ms
Data rate (Kbps) 27-0.37 0.1 250-226.7 15 KHz 3 56 4 ms
Range (Km) 22 63 35 15 KHz 1 112 8 ms
3.75 KHz 1 112 32 ms
A. LoRaWAN
In Europe, LoraWAN has 3 ”125 kHz” channels in the ISM IV. PROBABILITY CALCULATIONS OF PACKET E RROR RATE
band. The default channels are 868.10, 868.30 and 868.50 (PER)
MHz, LoRaWAN uses the duty-cycled limited transmissions This section describes how to set up the Packet Error Rate
of 1%. The LoRaWAN provides dynamic data rate, which (PER) calculations, utilizing symbol error rate. PER is the
is obtained according to the use of different spreading factor probability of packet error (PE) :
(SF). It is possible for the data rate to shift up to 9.375 Kbps. P ER = P (P E) (1)
LoRaWAN specifies data rates for LoRa modulation between
0.3 kbps and 22 kbps based on the SF. In our simulation, the Let N be the number of symbols that collide. N is a random
choice of SF is chosen randomly. The relation between SF, bit variable with probability mass function :
rate and packet duration is described in [20]:
fN (m); m = 0,1,2,...M (2)
Table II Where m is the number of symbols in the packet. By the
L O R AWAN PROTOCOL principal of total probability [25] :
Spreading Factor bit rate (b/s) packet duration (ms) M
X
12 293 682 P ER = P (P E) = P (P E | m)fN (m) (3)
11 547 365
m=0
10 976 204
9 1757 113 The conditional probability of not having a packet error
8 3125 64 is the probability that all the symbols are correct. So the
7 5468 36
6 9375 21
conditional probability of a packet error is the complement
rule of probability [25] :
B. Sigfox P (P E | m) = 1 − (1 − p)m (4)
The MAC layer in Sigfox sends 3 messages using a random Where p is the symbol error rate. Therefore, the PER formula
frequency within a macro channel of 200 kHz in the ISM becomes :
band. The default channel is 868.130 MHz, Sigfox uses M
100Hz ultra-narrowband BPSK modulation. Sigfox uses also X
P ER = [1 − (1 − p)m ]fN (m) (5)
multiple channels. For every transmission, the device chooses m=0

2
2018 IEEE Middle East and North Africa Communications Conference (MENACOMM)

This probability mass function of N depends on the duration


of the packet and the duty cycle. For Sigfox, there is a unique
mass function because using a constant packet duration of
2 ms. For LoRa and NB-IoT, there are many cases of mass
function because of the use of different packet duration. Two
cases will be studied in the following sections.
Figure 2. Packet Collision with Packets longer than Collide Packets
A. Packet shorter than collide packets
If the collide packets are of fixed duration we can see
Figure. 1 illustrates the case of where the packet is shorter
that the probability mass function will take a similar form
than collide packet. The figure shows two collide packets.
to the previous case. The only difference is that the maximum
When the packet occurs relative to the collide packets is
number of symbol collisions is F not M. We can use this form
a random process. This figure shows three possibilities. In
for both cases, since we can replace M by F if appropriate.
possibility 1, the packet collides completely with one of the
collide packets. In possibility 2, the packet does not collide

with collide packets. And in possibility 3, the packet partially c1 if m=0

collides with the collide packets. We can relate these cases to 
c if m = 1, 2, ..., F − 1
2
the value of colliding symbols. In possibility 1, the number fN (m) =
c3 if m=F
of colliding symbols is M. In possibility 2, the number of 

0 if m = F + 1, F + 2, ..., M − 1

colliding symbols is zero. And finally, in possibility 3, the
number of colliding symbols is less than M. (8)
Then the packet error rate is given by :

F
X −1
P ER = c2 [1 − (1 − p)m ] + c3 [1 − (1 − p)F ] (9)
m=1

V. PACKET E RROR R ATE E VALUATION OF L O R AWAN,


S IGFOX AND NB-I OT
Figure 1. Packet Collision with Packets shorter than Collide Packets
In this section, a comparison among the three technologies
LoRa, Sigfox and NB-IoT is presented. PER standing as the
If the collide packets are of fixed duration we can see that comparison criteria, PER is calculated according to the number
the probability mass function will take a special form : of collisions or fails during transmission.

 A. System Model
c1
 if m=0
The system model used for measurement and comparison is
fN (m) = c2 if m = 1, 2, ..., M − 1 (6) based on three homogeneous networks consisting LoRaWAN,

c3 if m=M Sigfox and NB-IoT links, as the data in sensor networks

generally flows in uplink way. Our system model focuses only
There is some probability of no collision. There is also some on the random uplink transmission in a specified time interval.
probability of a full collision. The interesting situation in this The simulation also takes into consideration the spectrum use
case is that the probability of m collisions for m between one around one gateway.
and M-1 is constant. At first, for each technology, we consider a discrete time
The packet error rate for this case is given by : of 60 seconds slotted into slots with equal duration, and each
transmission occupies one time slot.
M −1
X Secondly, we start to calculate the collisions. For LoRa,
P ER = c2 [1 − (1 − p)m ] + c3 [1 − (1 − p)M ] (7) the collision occurs when two packets using the same SF at
m=1
the same time. For Sigfox and NB-IoT, the collision is when
two packets arrive at gateway at the same time on the same
B. Packet longer than collide packets
channel.
Figure. 2 illustrates the case whene the sent packet is longer After that, we calculate PER. For LoRa and NB-IoT, the
than collide one. This figure shows two collide packets of PER is when the received packet does not look like the sent
equal duration. Here three cases can be distinguished. Since packets. For Sigfox, the PER is when the 3 packets are badly
the packet is longer than an interference pulse there is a received. In addition, for LoRa, two types of simulations were
maximum number of symbol collisions in the packet. Let us realized. The first one is when devices send packets with a
call this number F. randomly selected Spreading Factor. The second one is when

3
2018 IEEE Middle East and North Africa Communications Conference (MENACOMM)

devices send packets with a unique Spreading Factor. For, NB- results obtained for LoRa with several SF are logical with
IoT, we simulate both cases, multi-tone and single tone. A respect to equation (3) and Table III, where using a higher SF,
MATLAB simulation is used in this study. the number of symbols in the packet is larger than that in the
case of lower SF and the duration packet for a higher SF is
B. Simulations Results bigger compared to the duration packet of lower SF.

100

90

80
Packet Error Rate (%)

70

60
100
50

40
80
30 SF 12

Packet Error Rate (%)


SF 11
20 SF 10
SF 9 60
10 SF 8
SF 7
0
0 2000 4000 6000 8000 10000 40
Number of messages / minute
12-tones
Figure 3. LoRa PER simulation using a unique Spreading Factor with 104 6-tones
20
devices transmitting within 1 minute 3-tones
1-tone
Single-tone

100 0
0 1 2 3 4 5
90 Number of messages / minute ×104
80
Figure 5. NB-IoT PER simulation with 50.103 devices transmitting within
1 minute
Packet Error Rate (%)

70

60

50

40

30 SF 12
SF 11
20 SF 10
SF 9
10 SF 8
SF 7
0 Figure.5 shows the PER of all types of NB-IoT using Multi-
0 200 400 600 800 1000 tones and Single-Tone as function of number of messages
Number of messages / minute transmitting within 1 minute. Figure.5 shows that the PER
decreases as number of tunes increases, PER of 12-tones is
Figure 4. LoRa PER simulation using a unique Spreading Factor with 103
devices transmitting within 1 minute better than the PER of 6-tones, 3-tones, 1-tone and single-
tone. Single-tone maintains a 100% PER when using 6.103
Figure. 3 shows the PER of LoRa using a unique SF as devices, this PER is achieved when using 20.103 devices for
function of number of messages transmitting within 1 minute. 1-tone and 40.103 devices for 3-tones. In the other hand, for
Figure. 3 shows that the use of lower SF is able to support 12-tones, PER is only 38% for 50.103 devices.
more devices with lower PER than using a larger SF.
Figure. 4 represents a ”zoom” of Figure. 3, the number of
devices corresponding to 100% PER grows with the SF order, The results obtained for NB-IoT are compatible with equa-
for example, for SF 10, the 100% PER is achieved for a tion (3), where using 12-tones, the number of symbols in the
number of devices of 800, whereas this probability is achieved packet and the packet duration are smaller than all of other
for a number of devices of 250 when SF 12 is adopted. The tones.

4
2018 IEEE Middle East and North Africa Communications Conference (MENACOMM)

100 R EFERENCES
[1] A.S. John, ”Research Directions for the Internet of Things”, Internet
of Things Journal, vol. 1, 2014, pp: 3-9.
80 [2] IoT platforms: enabling the Internet of Things, March 2016. [Online]
Available: https://www.ihs.com/industry/telecommunications.html
Packet Error Rate (%)

[3] S. Hamrioui and P. Lorenz, ”Efficient wireless mobile networks com-


munications applied to e-health,” 2017 IEEE International Conference
60 on Communications (ICC), Paris, 2017, pp. 1-6.
[4] Nokia, “LTE-M – Optimizing LTE for the Internet of Things,” p.
August, 2015.
[5] LoRa Alliance Technology, [Online] Available: https://www.lora-
40 alliance.org/what-is-lora
[6] SigFox, 9 2016. [Online] Available: https://www.sigfox.com/
[7] ”Standardization of NB-IOT completed”. [Online] Available:
http://www.3gpp.org/news-events/3gpp-news/1785-nb-iot-complete
20 [8] K. Mikhaylov, .. Juha Petaejaejaervi and T. Haenninen, ”Analysis of
LoRa Random SF
Sigfox Capacity and Scalability of the LoRa Low Power Wide Area Network
NB-IoT 12 tunes Technology,” European Wireless 2016; 22th European Wireless Con-
ference, Oulu, Finland, 2016, pp. 1-6.
0
0 0.5 1 1.5 2 2.5 3 [9] C. Goursaud and Y. Mo, ”Random unslotted time-frequency ALOHA:
Theory and application to IoT UNB networks,” 2016 23rd International
Number of messages / minute ×104 Conference on Telecommunications (ICT), Thessaloniki, 2016, pp. 1-5.
[10] F. Adelantado, X. Vilajosana, P. Tuset-Peiro, B. Martinez, J. Melia-
Figure 6. PER simulation with 20.103 devices transmitting within 1 minute Segui and T. Watteyne, ”Understanding the Limits of LoRaWAN,” in
for LoRa, Sigfox and Nb-IoT 12-tones IEEE Communications Magazine, vol. 55, no. 9, pp. 34-40, 2017.
[11] R. Sinha, Y. Wei and S. Hwang. ”A survey on LPWA technology:
LoRa and NB-IoT”, ICT Exp., vol. 3, no. 1, pp. 14–21, 2017, doi:
Figure. 6 shows the PER of LoRa, Sigfox and Nb-IoT 12- 10.1016/j.icte.2017.03.004
tones as function of number of messages transmitting within [12] B. Reynders, W. Meert and S. Pollin, ”Range and coexistence analysis
of long range unlicensed communication,” 2016 23rd International
1 minute. Figure. 6 shows that the PER of NB-IoT is larger Conference on Telecommunications (ICT), Thessaloniki, 2016, pp. 1-6.
better than the PER of Sigfox and LoRa. For LoRa, the 100% [13] H. Lieske, G. Kilian, M. Breiling, S. Rauh, J. Robert and A. Heuberger,
PER is achieved for a number of devices of 30.103 . This ”Decoding Performance in Low-Power Wide Area Networks With
Packet Collisions,” in IEEE Transactions on Wireless Communications,
is the real case of LoRa when choosing a different SF to vol. 15, no. 12, pp. 8195-8208, Dec. 2016.
transmit data using Adaptive Data Rate (ADR). For Sigfox, [14] P. Neumann, J. Montavont and T. Noël, ”Indoor deployment of low-
the 100% PER is achieved for 14.103 devices which is a power wide area networks (LPWAN): A LoRaWAN case study,”
2016 IEEE 12th International Conference on Wireless and Mobile
huge number of devices compared to LoRa using a unique Computing, Networking and Communications (WiMob), New York,
SF when the same parameters of transmission are adopted, NY, 2016, pp. 1-8.
but when using a random SF, LoRa uses a huge number [15] M. Centenaro, L. Vangelista, A. Zanella and M. Zorzi, ”Long-range
communications in unlicensed bands: the rising stars in the IoT and
of devices compared to Sigfox. For the considered values smart city scenarios,” in IEEE Wireless Communications, vol. 23, no.
of the simulations parameters, the robustness behind NB-IoT 5, pp. 60-67, October 2016.
is due to the larger bandwidth according to a smaller time [16] A. Cenedese, A. Zanella, L. Vangelista and M. Zorzi, ”Padova Smart
City: An urban Internet of Things experimentation,” Proceeding of
slot duration. On the other hand, LoRa present a moderate IEEE International Symposium on a World of Wireless, Mobile and
performance compared to NB-IoT and Sigfox, which has a Multimedia Networks 2014, Sydney, NSW, 2014, pp. 1-6.
very sensitive PER in terms of messages / minute due to the [17] S. Latre, P. Leroux, T. Coenen, B. Braem, P. Ballon and P. Demeester,
”City of things: An integrated and multi-technology testbed for IoT
their bandwidth and time slot duration. smart city experiments,” 2016 IEEE International Smart Cities Confer-
ence (ISC2), Trento, 2016, pp. 1-8.
VI. C ONCLUSION [18] R. Ratasuk, N. Mangalvedhe, Y. Zhang, M. Robert and J. P. Koskinen,
”Overview of narrowband IoT in LTE Rel-13,” 2016 IEEE Conference
In this paper, we described the features at the MAC layer for on Standards for Communications and Networking (CSCN), Berlin,
LoRa, Sigfox and NB-IoT. In addition, we made an evaluation 2016, pp. 1-7.
of the MAC Layer for the three IoT technologies : LoRa, [19] ”Sigfox Radio Technology Keypoints”, [Online] Available:
https://www.sigfox.com/en/sigfox-iot-radio-technology
Sigfox and NB-IoT in terms of PER. To conclude, NB-IoT is [20] LoRa™ Modulation Basics, AN1200.22. [Online] Available:
more robust than LoRa and Sigfox in terms of PER, due to the http://www.semtech.com/images/datasheet/an1200.22.pdf
adopted large bandwidth and to the small time slot duration [21] LE51-868 S RF Module User Guide, Telit, Apr. 2014, rev. 1.
[22] Sigfox Technology Overview.[Online] Available:
relative to LoRa and Sigfox. As future work, we plane to https://www.sigfox.com/en/sigfox-iot-technology-overview
explore the functionalities of MAC layer in order to distinguish [23] J. Petäjäjärvi, K. Mikhaylov, M. Hämäläinen and J. Iinatti, ”Evaluation
the different sources of interference to increase the Quality of of LoRa LPWAN technology for remote health and wellbeing moni-
toring,” 2016 10th International Symposium on Medical Information
service(QoS). and Communication Technology (ISMICT), Worcester, MA, 2016, pp.
1-5.
VII. ACKNOWLEDGEMENT [24] Y. P. E. Wang et al., ”A Primer on 3GPP Narrowband Internet of
Things,” in IEEE Communications Magazine, vol. 55, no. 3, pp. 117-
The authors would like to thank the company SPIE City 123, March 2017.
Networks and the Télécom & Réseaux Chair of Polytech [25] A. Papoulis, ”Probability, Random Variables, and Stochastic Pro-
cesses,” Third Edition, McGraw Hill, 1991.
Nantes for their funding support.

Vous aimerez peut-être aussi