See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/329191085

Low-Power Wide Area Network Technologies for Internet-of-Things: A

Comparative Review (IEEE Internet of Things Journal)

Article  in  IEEE Internet of Things Journal · November 2018

DOI: 10.1109/JIOT.2018.2883728

CITATIONS READS

148 2,768

8 authors, including:

Augustine Ikpehai Bamidele Adebisi

Sheffield Hallam University Manchester Metropolitan University
44 PUBLICATIONS   827 CITATIONS    197 PUBLICATIONS   2,659 CITATIONS   

SEE PROFILE SEE PROFILE

Khaled M. Rabie Kelvin Anoh

Manchester Metropolitan University University of Bolton
200 PUBLICATIONS   2,393 CITATIONS    74 PUBLICATIONS   962 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Advances in Power Line Communication and its Applications View project

CSNDSP 2018: Special Session on Green Communications and Networks View project

All content following this page was uploaded by Kelvin Anoh on 30 November 2018.

The user has requested enhancement of the downloaded file.

Low-Power Wide Area Network Technologies for
Internet-of-Things: A Comparative Review
Augustine Ikpehai, Member, IEEE, Bamidele Adebisi, Senior Member, Khaled M. Rabie, Member, IEEE,
Kelvin Anoh, Member, IEEE, Ruth E. Ande, Graduate Student Member, IEEE, Mohammad Hammoudeh,
Member, IEEE, Haris Gacanin, Senior Member, IEEE and Uche M. Mbanaso, Member, IEEE

Abstract—The rapid growth of Internet-of-Things (IoT) Although IoT connectivity will be dominated by short-
in the current decade has led to the the development of range technologies for many years, the work [3] predicts
a multitude of new access technologies targeted at low- that by 2025, 25% of wireless industrial IoT connections
power, wide area networks (LP-WANs). However, this has also
created another challenge pertaining to technology selection. will be provided with LP-WAN technologies including
This paper reviews the performance of LP-WAN technologies LoRa1 , Sigfox, LTE-M as well as NB-IoT. This signifies
for IoT, including design choices and their implications. We the continued importance of LP-WAN as the IoT landscape
consider Sigfox, LoRaWAN, WavIoT, random phase multiple evolves.
access (RPMA), narrow band IoT (NB-IoT) as well as LTE-M Machine-to-machine (M2M) communication in IoT rep-
and assess their performance in terms of signal propagation,
coverage and energy conservation. The comparative analyses resents a very large market, growing rapidly at compound
presented in this paper are based on available data sheets annual growth rate of over 20% [5]. Nevertheless, the mar-
and simulation results. A sensitivity analysis is also conducted ket is fragmented and awash with many access technologies
to evaluate network performance in response to variations in and vertical solutions that sometimes do not interoperate.
system design parameters. Results show that each of RPMA, The result is that despite the potentials of IoT, organisations
NB-IoT and LTE-M incurs at least 9 dB additional path loss
relative to Sigfox and LoRaWAN. This study further reveals and end users are faced with overwhelming choices of
that with a 10% improvement in receiver sensitivity, NB-IoT access technologies. Therefore, it becomes challenging to
882 MHz and LoRaWAN can increase coverage by up to decide where or how to begin their IoT road map as part
398% and 142% respectively, without adverse effects on the of their wider digital transformation journey. Insufficient
energy requirements. Finally, extreme weather conditions can comparative studies of LPWAN technologies has been
significantly reduce the active network life of LP-WANs. In
particular, the results indicate that operating an IoT device identified as one of the major barriers to potential IoT
◦
in a temperature of -20 C can shorten its life by about half; users [6]. This study does not involve detailed physical or
53% (WavIoT, LoRaWAN, Sigfox, NB-IoT, RPMA) and 48% ◦
media access control (MAC) layer specifications, the results
in LTE-M compared with environmental temperature of 40 C. nevertheless provide indicative performance of the LP-WAN
technologies set-up under the same operating conditions.
Index Terms—Access technologies, low-power wide area net- Until now, most of the studies independently evaluated
work (LP-WAN), Internet-of-things (IoT), LoRaWAN, Sigfox, LP-WAN technologies in different operating environments.
sensitivity analysis, energy conservation, link budget, WavIoT, The main reason is that many of the leading technologies are
random phase multiple access (RPMA), LTE-M, narrow band
IoT (NB-IoT).
based on proprietary protocols whose detailed specifications
are not freely available in the public domain. For example,
LoRaWAN and Sigfox are the top two LP-WAN technolo-
I. I NTRODUCTION gies in terms of installed base. The underlying chirp spread
y 2025, up to 75 billion devices would be connected spectrum (CSS) modulation technique in the physical layer
B in Internet-of-things (IoT), with potential economic
impact of around $11.1 trillion a year [1], [2]. The key
of LoRa is proprietary and owned by Semtech Corporation.
On the other hand, Sigfox deploys and operates the net-
underpinning of IoT is the large number of interconnected work but freely provides the protocol specification to chips
devices that exchange information and enable services. manufacturers. Thus, LoRa is a closed chipset but open
network; (even private networks are possible) and Sigfox
A. Ikpehai, B. Adebisi, K.M Rabie, K. Anoh, R. E. Ande
and M. Hammoudeh are with the Faculty of Science &
is a closed network but open chipset. The implication is
Engineering, Manchester Metropolitan University, Manchester, M1 that in the former, Semtech controls the production, support
5GD, UK. (e-mails: [email protected]; [email protected]; and price of LoRa chipsets, while in the latter, Sigfox
[email protected]; [email protected]; [email protected];
[email protected]).
controls the provision, access and price of the network
Haris Gacanin is a research director with Nokia, Copernicuslaan 50, 2018 resources. Other technologies are different variants of either
Antwerp, Belgium. (e-mail: [email protected]) proprietary physical or MAC layer. Hence, no mainstream
Uche M. Mbanaso is with Centre for Cyberspace Studies,
Nasarawa University, Keffi, Nasarawa State, Nigeria. (e-mail:
LP-WAN protocol stack is currently fully open.
[email protected]) However, given the scope of IoT, it is important for
solution providers, system designers and users to have a
This work was supported in part by “CityVerve: IoTs and Smart
Cities Demonstrator” project funded by Innovate UK under Grant
wide perspective of LP-WAN options without being locked-
102561 and in part by “Triangulum” (part of H2020 Smart Cities and 1 In this paper, LoRa denotes the physical layer modulation only, while
Communities programme) funded by the European Commission under
Grant 646578-Triangulum-H2020-2014-2015/H2020-SCC-2014. LoRaWAN refers to the wide area network protocol suite that adopts LoRa
at physical layer and LoRa MAC at media access control sub-layer [4].
 2

in. Motivated by that gap, the objective of this paper is to environments with higher levels of interference may exhibit
assess the performance of the widely available LP-WAN different performance due to difference in the interference
technology specifications. The system parameters used are sources and patterns. This will not only provide additional
mostly drawn from the datasheets of the respective device insights but also help to analyse the results in the context
manufacturers. of individual field environment.
The contribution of this work is two-fold. First, we For cellular LP-WANs, [22] investigated the capacity and
identify an approach to optimise the LP-WAN by exploiting coverage of LTE-M and NB-IoT in a rural area in Denmark.
the link budget and design parameters without increasing the The paper reported that cellular technologies such as LTE-M
energy requirements of the system. The second contribu- support maximum coupling loss of about 156 dB, however,
tion resides in identifying the multi-variate dependence of in deep indoor applications can experience up to additional
energy consumption on field, application and technology- 30 dB penetration loss. An interesting aspect would be a
specific characteristics, each of which can be exploited to comparison with non-cellular technologies.
different extents. These results provide useful insights which Apart from the various analytical models that have been
are extremely valuable when deciding the trade-off between reported [23]–[25], other experimental studies have been
network performance, design complexity and cost. conducted to individually evaluate different aspects of LP-
WAN including coverage [26], [27], energy consumption
II. R ELATED W ORK [28], capacity [22], [29], [30] and scalability [23], [24],
LP-WAN technologies are increasingly being deployed [30]. The main commonality in these and many other
as last mile connectivity to compliment traditional technolo- existing studies is that they mostly focused on one or
gies and replace them in many use-cases. However, majority two technologies. Comparing results from different studies
of existing studies that compare the LP-WAN technologies may not provide a balanced view as each experiment or
are either literature reviews [7]–[11] or brief experiments simulation model is conducted with a set of assumptions or
with two or three technologies and are often focused on conditions which varies across studies and authors.
a single aspect of the systems. The studies neither include According to [31], until recently, there were very few
rigorous analysis of multiple performance metrics nor inves- studies involving more than three technologies. Although,
tigate how the technologies perform relative to each other in [31] compared the performance of GPRS and NB-IoT with
different environments, using identical system parameters. Sigfox and LoRa, the dearth of publications in this area
In some evaluation cases, the technologies were tested in suggests that a considerable amount of follow-on work still
one environment and are therefore limited in scope. needs to be carried out. The objective of this paper is
In many of the studies where practical systems were therefore to contribute in that regard.
designed, comparisons were not made with competing LP- Recent reviews of LP-WANS are summarised in Table
WAN solutions. In fact, in some of the experiments, single I according to their areas of focus. Many of them de-
technology was deployed for particular use-case which also scribe single technologies while energy consumption was
limits the assessment outcomes to type-specific scenarios not considered in most. However, given its importance in
[12], [13]. In some other works, the comparisons were IoT generally and LP-WANs in particular, it is necessary
restricted to physical and associated MAC layers only to consider the energy performance of existing LP-WAN
while some evaluation studies focused on either cellular- technologies as part of a complete system review.
only or non-cellular LP-WAN technologies [5], [14], [15].
However, as some of the LP-WAN technologies are still III. L INK B UDGET AND I MPLICATIONS IN LP-WAN S
at their infancy, many practical issues are being resolved. LP-WANs in IoT access networks share similar imple-
Thus, a cross-cutting comparison will not only identify mentation methodologies with traditional wireless networks
suitability for specific applications but also ascertain among but with some differences in system design. According to
the LP-WAN specifications, the extent of design diversity Shannon-Hartley theorem, the signal-to-noise ratio (SNR)
in relation to the network performance. of each bit can be written as [38]
LoRaWAN is currently one of the most deployed LP-  2η − 1
Eb ε  C
WAN access networks for IoT [16] and also one of the = 2 ε −1 = (1)
widely reported LP-WAN technologies in literature. Signal N C η
propagation is crucial in IoT [17] as it affects network where C is the capacity (in bits/s), ε is the bandwidth and
performance in terms of coverage, reliability and data rate. η is the spectral efficiency (in bits/s/Hz). For a bit duration
Path loss and coverage have been reported in many LP- of (1/C) seconds, if S denotes the average signal power
WAN studies including LoRa [18] and NB-IoT [19] while of each bit and N the noise power in Watts, the average
S
[20] proposed an indoor channel attenuation model for energy per bit is Eb = C (in J) and the total noise power
LoRa, following a series of measurement campaigns. A few is N · ε Watts. The implication of (1) is that, ENb varies
other empirical studies such as [21] compared the perfor- with η and the modulation scheme in LP-WAN technologies
mance of LoRa’s CSS modulation scheme with frequency- must ensure that η is maximised. For a wireless signal to
shift keying (FSK). The paper found that coverage and be correctly decoded at the receiver, it must meet energy
network reliability are both affected by the payload size. detection threshold. The receiver sensitivity (σ) can be
However, the experiments were limited to the university expressed as [39]
environment and the results did not provide insights into po-
tential behaviours in other environments. More challenging Eb
σ= + 10 · log10 C + (φ − 174 dBm) (2)
terrains such as dense urban areas and other metropolitan N
 3

Table I: Summary of Related Work

Ref. Specification Focus area Contribution/ Methodology Energy? Limitations/Gaps
[7], LoRa & Sigfox, Basic literature review & grouping Little analytic
[32] RPMA or into licensed and unlicensed No rigour. No
WavIoT bands. Description of advantages performance
of LP-WANS over legacy comparison
long range technologies. or evaluation.
[8] LoRaWAN Description of Yes Only LoRaWAN
LoRaWAN studied. Other
in terms of capacity, cellular and
security, battery life. non-cellular
LP-WANs excluded.
[33] LTE-M, NB-IoT, Basic overview Description of each Insufficient depth of
LoRa, Sigfox, & discussion technology and review. No
RPMA. of LP-WAN measurements performance
technologies. of reliability. No comparison
[5], [9], Sigfox, LoRa, Very basic comparison or evaluation.
[11] NB-IoT, 3G/4G of technologies & their differences.
[27] 5G Indoor coverage No Only 1 site &
evaluation based 1 technology
on ray tracing. considered.
[34] 5G, LTE-M, Description of various Comparative
EC-GSM cellular , LP-WANs No performance
enablers & applications. with non-cellular
LP-WANs was
not considered.
Capacity, scalability, No Battery life and
[16], [30] LoRaWAN Comprehensive densification, use-cases its dependencies
review or tutorial and reliability. Used were not covered.
simulation to illustrate the LoRaWAN not
limits of LoRaWAN: compared with
scalability & reliability. other LP-WANs.
Comprehensive review of NB-IoT not
[35] NB-IoT NB-IoT including security, Yes compared with
applications & energy consumption other technologies
[13] Deployment of LoRa No Only LoRa
to monitor a river. considered. One
Measurements taken. use-case studied.
[17] Indoor signal No Only LoRa
propagation characteristics studied.
using RSSI as measure Performance
of performance. relative to
[20] LoRa/LoRaWAN Measurement-based No other
attenuation model: coverage technologies
and reliability for land and sea were
[36] Indoor evaluation: SNR, Yes not reported.
packet loss, throughput
RSSI, energy consumption.
[37] Experimentation, Effects of interference No Single
evaluation, on scalability. Evaluation technology
system design & of LoRa in terms of coverage, considered.
[21] implementation. reliability using CSS & No
FSK in terrestrial deployments.
[28] Experimental evaluation of energy Yes Only NB-IoT
consumption of NB-IoT devices. studied
[14] NB-IoT NB-IoT firmware developed & No Performance
deployed on devices. Cloud relative to other
platform, application server, user other technologies
application also deployed. were not reported.
[18], Simulation of LoRa with multiple Yes
[23] SF and rates. Distribution of SF
and discrete power settings
LoRa to reduce packet rerror rate (PER). Only LoRa
[29] Detailed analysis of LoRa and No investigated.
LoRaWAN, simulation of Performance
throughput & collision rate relative to other
and measurements technologies
of coverage & reliability. not reported.
[19] NB-IoT Simulation or Coverage of NB-IoT No Only NB-IoT
analytical based on in-band investigated.
solution for deployment in LTE.
[22] LTE-M, LP-WANs Coverage, reliability, energy Yes Non-cellular
consumption in indoor LP-WANs
NB-IoT and outdoor cases. not covered
[24] Sigfox Derivation of expression of No Only Sigfox
packet error rate for UNB investigated.
LP-WAN such as Sigfox.
[25] LoRa, NB-IoT Description of MAC of No Energy &
Sigfox LoRa, NB-IoT & Sigfox. its dependencies
Comparison of PER. were excluded.
 4

Table II: Link Budget of LP-WAN Technologies used in the simulation [4], [31], [47]–[52]
Technology NB-Fi (WavIoT) LoRaWAN U-NB(Sigfox) RPMA (Ingenu) LTE-M NB-IoT
868MHz 868MHz 868MHz 2.4GHz 1.8-2.7GHz 700-2100 MHz
Downlink DL DL DL DL DL(2.6GHz) DL(882MHz,1840MHz)
Tx Power, dBm 30 21 24 21 40 35
Tx cable loss,dB -3 -3 -6 -3 -3 -3
Bandwidth (kHz) 0.1 125 0.1/0.6 1000 1400 180
Tx antenna gain, dBi 0 9 9 9 10 16
σ, dBm -147 -137 -129 -133 -129 -141
Rx env. noise 0 0 0 0 0 0
Rx antenna gain 0 0 0 8 0 0
Total (link budget) 174 164 156 168 176 189
Obstacle loss (fixed), dB 20 20 20 30 30 30
Obstacle loss (variable), dB 35.8 28.1 30.7 29.8 27.1 29.4

Table III: Additional Simulation Parameters

The noise figure (φ) is typically around 2 to 6 dB for LP-
WAN radio front-end [39], [40]. Parameter Value
Number of gateways 1
In many access networks such as Sigfox and LoRaWAN, Baseline payload 12 bytes
each message sent by the end node can be received by Radio electronics energy,Eelec 50 nJ/bit
multiple gateways. According to [41]–[43], the potential On-Air time Sigfox-EU:6.24s
throughput available per coverage unit can be expressed as LoRa (varies with SF)
Background noise, dBm -130 dBm
Battery model E91-AA alkaline battery
R(λ, θ) = λ · log2 (1 + θ)Pr(SNR ≥ θ), (3)
where λ is the gateway density, Pr(.) is the probability
average values obtained from measurements in urban areas
operator, θ is the threshold SNR and R is the potential
in Europe through 20 cm concrete walls [48].
throughput in bps/Hz/km2 . However, (3) assumes fixed
The parameters in Tables II and III are used to simulate
data rate and does not provide rate adaptation such that
the transmission from gateway to end device based on the
the bit rate of a gateway is adjusted upward or downward
scenario described in the next section. Sigfox and LoRa
according to the SNR instead of outage when the receiver
are two leading LP-WAN technologies. However, Sigfox
SNR < θ. On the other hand, the average area spectral
supports a maximum payload of 12 bytes while LoRa can
efficiency (ASE) also in bps/Hz/km2 considers SNR of
support up to 242 bytes [36], [53], [54]. In order to compare
individual gateway and avoids outage at low SNR by
the technologies, we adopt Sigfox payload as the baseline in
reducing the data rate instead, this is more amenable to
Table III. Energy consumed by the radio electronic circuit
IoT scenarios. The ASE can be expressed as [44]
in low-power devices such as wireless sensors has been
reported to be about 50 nJ/bit [55], [56]. The choice of
E (λ)=λ · E [log2 (1 + SNR)], SN R ≥ θ. (4)
Alkaline battery is informed by the fact that they are low-
Equation (4) supports bit rate adaptation according to the cost, widely available and are commonly used in low-power
SNR instead of outage. According to [41]–[44], ASE de- applications such as wireless sensor networks. In Europe,
pends on the number of gateway per km2 and path loss. The at 100 kbps, it takes about 2.08 s [57] to send a 12-byte
value of σ can also be reduced to extend the transmission Sigfox frame and since the endpoints do not synchronise
range, however, the performance penalty is reduced bit rate with Sigfox base station, each message is sent three times
which is acceptable in many IoT deployment scenarios. to improve chances of detection [5], [58].
This systematic trade-off underpins the long range operation
of LP-WAN technologies. The value of σ in LP-WANs is IV. N ETWORK M ODEL AND M ETHODOLOGY
typically lower than -128 dBm (-129 dBm to -155 dBm) This paper considers an access network in which the
and ranges from -90 dBm to -110 dBm [45] in traditional gateway antenna is at 25 m height on a tower and the IoT
wireless systems such as Wi-Fi [46]. The implication is end point comprises an antenna located at 8 m (based on
that in terms of absolute signal power in Watts, LP-WAN the average height of 2-storey building in UK [21], [59]).
receivers can detect wireless signals that are 1000 times This set-up applies to many IoT and smart city scenarios
weaker than Wi-Fi and other technologies with -100 dBm including monitoring of multi-storey building packing space
sensitivity. LP-WAN technologies exploits this capability in availability in smart cities, monitoring of structural health
the various specifications. of buildings and bridges, detection of noise map in urban
Table II presents the technical specifications of widely- areas, smart grid, industrial IoT, etc. The downlink is
available LP-WAN technologies from datasheets [4], [47], considered and we discuss the network performance in
[48], measured obstacle losses in European urban envi- different terrains including open/rural areas, suburban, small
ronments [48] as well as other parameters reported in and large cities. In each case, the indicative performance of
literature [31], [49]–[52]. It is generally desirable that the network is analysed following simulation of each access
the main lobe of the transmitted radiation pattern reaches technology in different environments. In this study, EU-868
the receiver to ensure that maximum signal strength and LoRaWAN/Sigfox, WavIoT, 2.4 GHz RPMA-Ingenu, 2.6
coverage are delivered. The field condition can be worsened GHz LTE-M, NB-IoT (882 MHz/1840MHz) specifications
by obstruction losses which will further reduce the link and other parameters outlined in Table II are employed in
budget. The fixed obstacle losses in Table II are the reported
 5

the simulation using MATLAB. The 3GPP recommends in- or infringing the LoS causes attenuation. For the given
band deployment for LTE-M using existing LTE installed scenario, the Fresnel clearance is illustrated in Fig. 1.
base. In this paper, we adopt the 2.6 GHz for LTE-M as it is The Fresnel zone radius at a point P along the LoS is
widely used in Europe and beyond in existing LTE systems. generally defined as [62]
s s
Although NB-IoT supports three modes of deployment, n · cd1 d2 d1 d2
Fzr = = 17 .32 (5)
namely; standalone, in-band LTE and guard band in LTE, it f ·D f ·D
can also coexists with GSM, LTE and UMTS [60]. We adopt
882 MHz and 1840 MHz for NB-IoT which are equivalent where n = 1 denotes the first Fresnel zone, c is the speed of
to bands 5 and 3 downlink frequencies respectively, in 3GPP light, f is the frequency in GHz and D= d1 + d2 is the total
NB-IoT (release 13) [49]. These choices are informed by distance in km. In reality, at least 60% Fresnel clearance is
the fact that many mobile operators already own blocks of required.
frequency in 1800 MHz band and 800 MHz-900 MHz used By applying (5), Fig. 2 presents the minimum clearance
for various services such as GSM and LTE. This clearly requirements for 868 MHz, 2.4 GHz and 2.6 GHz LP-WAN
provides an upgrade path to NB-IoT for such operators technologies as the obstacle is moved to different points
through frequency re-farming. along the transmission path in a 50 km network. It can be
In IoT access network design and implementation, the seen that the required Fresnel clearance increases as the
three most critical questions are: obstacle moves towards the midpoint such that the largest
clearance is required if the obstacle is positioned at the
i) what is the maximum distance the link can cover
middle. Fig. 2 further shows that when the obstacle is moved
without compromising quality-of-service (QoS) or
from 1km to 5 km then 25 km, with LTE-M, the minimum
SNR,
clearance increases from 6.38 m to 13.67 m then 22.79
ii) what field parameters can be explored to reduce
m, respectively. In the Sigfox/LoRaWAN networks with
cost of implementation and
the obstacles at the same positions, the minimum Fresnel
iii) what endogenous variables can be tweaked to
clearance are respectively 11.04 m, 23.66 m and 39.44 m.
conserve energy [61] and optimise network per-
The key observation here is that the closer the obstruction
formance.
to the gateway transmitter, the more impact it has on the
To address these questions, the methodology adopted is access network.
summarised as follows.
1) Use of system-specific attributes of WavIoT, Sigfox,
B. Path Loss
LoRaWAN, RPMA and LTE-M to generate propa-
gation characteristics including Fresnel clearance and Path loss estimation is a crucial part of planning and
path loss. design of the access networks. In wireless communication,
2) Application of equal transmit power across the tech- every time the distance is doubled, only one-fourth of the
nologies, followed by estimation of the corresponding signal power is received and this has some far-reaching
coverage at each power level in rural and urban implications for connectivity in IoT as it degrades the SNR
environments. at the end receiver. The path loss discussed in this paper
3) For each technology, the path loss information previ- relates only to outdoor applications. The Free Space Path
ously calculated is extracted and used to generate the Loss (FSPL) model is based on clear LoS and does not take
received signal strength indicator (RSSI) as well as into account terrestrial objects such as hills, trees, buildings,
SNR at different distances in the field. SNR is bench- etc. In ideal case, the path loss experienced by signals can
marked against 20 dB widely used in conventional be expressed as
wireless systems.
4) As coverage is more challenging in urban areas, a FSPL(dB ) = 20 · log10 l + 20 · log10 f − 32 .44 (6)
sensitivity analysis is conducted for urban deployment
scenario with a view to exploiting system parameters where l is the distance (in km) and f is the frequency (in
for network performance optimisation. Hz). However, the assumption of free space between the
5) Investigation of the key factors affecting energy con- transmitter and receiver in FSPL is invalid in many real life
sumption in end devices. deployments, as RF paths are typically laden with additional
6) For specific LP-WAN system configurations, estimate factors which further degrade the path loss. Such factors
the battery life of the radio transceiver based on 1% include [63] cable losses, antenna gain, LoS condition,
maximum duty cycle allowed in Europe. receiver sensitivity, etc. Particularly in IoT, a wireless signal
While there are no simple or direct answers to the questions will typically encounter different obstacles on its path and
posed, the following sections illustrate these issues and each has an additive effect on the total path loss. Common
provides insights to different aspects of the questions with examples of obstacles and their attenuation factors are
some rationales. summarised in Table IV.
According to Table IV, if Sigfox or LoRaWAN signal
with transmit power of 14 dBm encounters a non-solid
V. S IGNAL P ROPAGATION
metal door on its path, the signal power is reduced to 8
A. Fresnel Clearance dBm (approximately 25% of its original value) when it
Ideally, radio waves should travel in a straight line from exits the door. In urban areas with high vehicular density
the source to destination. An obstruction that is close to and high-rise buildings, signal can suffer from reflection
 6

Fresnel zone
Tx

Fresnel radius
Rx

hg
obstacle

ht

d1 d2
P

Figure 1: LoS condition and Fresnel clearance

40
IoT implementation engineer must therefore ensure that the
received power is often greater than σ.
35
868MHz (LoRa/Sigfox Europe) 1) LP-WAN and Signal Propagation in Different Ter-
2.4GHz (RPMA))
2.6GHz (LTE-M) rains: From (7), it is clear that adequate knowledge of the
30
terrain is required in the network design. Some models de-
Fresnel clearance (m)

rived from empirical studies have been reported to provide

25
more realistic path loss. In particular, the Hata-Okumura
model, which is based on various Okumura correction
20
functions is widely used for predicting path loss in radio
frequency (RF) planning in the wireless communication
15
industry. The key strength of this model is that it caters for
10
different field scenarios such as large cities, small suburban
and rural environments [65]–[68] which has influenced its
5
wide acceptance. In its general form, the model can be
0 5 10 15 20 25 30 35 40 45 50 expressed as
Distance of obstacle from Tx (km)

Figure 2: Effects of obstacle position on Fresnel clearance

PLHata =69.55 + 26.16 · log10 f − 13.82 · log10 hg −
Table IV: Examples of materials and signal attenuation at 2.4 GHz a(ht ) + [44.9 − 6.55 · log10 hg ] log10 d − ct
[63], [64] (8)
Material/object Attenuation Frequency where f is the frequency (in MHz), hg is the gateway
Dry wall 3dB 2.4GHz antenna height (in m) and d is the distance (in km) from
Glass wall with metal frame 6dB 2.4GHz transmitting to receiving antenna. The terrain parameter
Office window 3dB 2.4GHz
Metal door (non-solid) 6dB 2.4GHz a(ht ) is a function of IoT terminal height, defined for urban
Wood 3dB 2.4GHz areas as
Solid concrete (e.g. wall, bridge) 8-15dB 2.4GHz/1.3GHz
Aluminum siding 20.4dB 815MHz
Concrete floor 10dB 1.3GHz 2
a(ht )(urban) = 3 .20 [log10 (11 .75ht )] − 4 .97 (9)
6-inch diameter metal pole 3dB 1.3GHz
for suburban, open and rural areas, a(ht ) is defined as

from metallic surfaces, diffraction or even scattering off a(ht )(sub−rural) = [1.1 · log10 f − 0.7)] ht −
the lampposts and moving vehicles which contribute to (10)
[1.56 · log10 f − 0.8]
the overall losses. If multiple copies of the signal arrive
the receiver, they can destructively interfere and severely where ht is the height of the remote terminal (IoT endpoint)
degrade reception. The path loss can be expressed as antenna above the ground. Also, the parameter ct is the
correction factor defined as
σ = Pt + Gain net − PL (7)

where σ is the receiver sensitivity, P L is total the path 
 4.78 · [log10 f ]2 − 18.33 · log10 f + 40.94 rural/open
 h  i 2
loss in dB and Gainnet is the net gain. The Gnet includes ct = f
2 log10 28 + 5.4 suburban
all gains resulting from transmit and receive antenna as 

0 urban

well as all losses related to the radio and antenna hardware
(11)
including filters, cables, attenuators, antennas and obstacle
For the various access technologies, the path loss in different
losses, etc. The σ in (7) is equivalent to the minimum
environments is presented in Figs. 3 and 4.
signal power that will be detected by the receiver. The
 7

rural
If that is considered against the fact that attenuation is
160
more pronounced in the > 2 GHz region and many ISM
140 radio equipment (e.g. Wi-Fi, ZigBee) also share the 2.4
GHz band, then the system designers and implementation
120
868MHz Europe (LoRa/Sigfox) engineers need to consider the delicate balance between
2.4GHz (RPMA)
100 2.6GHz (LTE-M) network reliability, cost and efficiency in the choice of radio
80
technology. Although more bandwidth is available in 2.4
0 5 10 15 20 25 30 35 40 45 50 GHz and 2.6 GHz compared to 868 MHz, however Figs. 3
suburban and 4 show that at every distance, the 2.x GHz technologies
180
(RPMA, LTE-M) incur at least 9 dB additional path loss
Path Loss (dB)

160 which suggests that they are less able to overcome the
140 effects of obstacles.
2) Non LoS (NLOS) Propagation : For IoT applications
120
in urban environments lined with high rise buildings and tall
100 street furniture, NLOS links arise when the gateways and
0 5 10 15 20 25 30 35 40 45 50
Distance (km) IoT end points are located on two parallel or intersecting
streets. With a rectangular street grid, signals experience
Figure 3: Path loss of access technologies in rural and suburban
environments
reflection off the walls as well as diffraction at the building
corners [69] and other vertical structures such as lamp posts
and bus stop signage, located near the street junction or
small city parallel street. In such cases, the arriving signals undergo
200

180
multiple reflections from the walls on the main and side
streets as well as scattering around the vertical objects. The
160
1-turn path loss, relative to free space can be expressed as
140 868 MHz (LoRa/Sigfox Europe)
2.4 GHz (RPMA))
[70]
120 2.6 GHz (LTE-M)
100 cos2 ϕ x1 x2
0 5 10 15 20 25 30 35 40 45 50 P L1−T urn = P LLOS . . (12)
S 2 x1 + x2
large city
200 where x1 , x2 are the distances from the intersection to
180 gateway and end device respectively, S is the scattering
Path Loss (dB)

160 parameter and ϕ is the viewing angle from the gateway to

140 end device which depends on the heights of the antennas.
120
The P LLOS term in (12) is the ITU-LOS path loss model
defined as
100
0 5 10 15 20 25 30 35 40 45 50
Distance (km)   2  2.5
10δdB/10 γ2 d
d ≤ dc

Figure 4: Path loss of access technologies in urban environments 8π.hg .ht dc
P LLOS =  2
2  4 (13)
10δdB/10 8π.hγ .h
 d
dc d > dc
g t

By applying (8)-(11) using the parameters in Tables II Where δ is the environmental parameter, d is the total
and III, Figs. 3 and 4 respectively, illustrates the path loss in distance x1 +x2 and dc is the cross-over distance. In general,
urban and rural environments for the scenario presented in P LLOS will vary across different environments, depending
Fig. 1. These results generally agree with the literature, for on building heights, antenna height and the amount of
instance Fig. 3a closely aligns with the coupling loss of 156 waveguide provided by the street canyon. By employing a
dB reported in [22] for LTE-M. It can be seen in Fig. 3 that virtual source at the centre of the intersection and capturing
at a distance of 3 km between the transmitter and receiver, local street variables such as main street width, side street
RPMA and LTE-M each incurs additional path loss of about width as well as distance from gateway antenna to the main
9 dB and 11 dB in rural and large cities respectively, relative street wall, it has been reported that the NLOS path loss can
to Sigfox/LoRaWAN. At 10 km, the additional path loss be written as [71], [72]
experienced by RPMA and LTE-M are 8.2 dB and 9.2 dB in
rural areas as well as 12 dB and 13 dB respectively in large   q 2
cities (compared with Sigfox/LoRaWAN). These results are 10.log10 1 2π 4πx1 x2
. d ≤ dc

δ w .x γ
within the range of those in [6], [27] which reported that P LN LOS =  q t t 2
10.log10 1 2π 4πx1 x22
LP-WAN could achieve up to 20 dB improvement over

δ wt .xt . γ.dc d > dc
cellular systems. It is also noted that RPMA and LTE- (14)
M are more susceptible to path loss than LoRaWAN and where wt , xt are the main street width and distance from
Sigfox. The amount of loss also depends on the nature gateway to the wall respectively. While (13) includes the
of the obstacle (Table III), the results in this subsection waveguide and 1-turn corner-turning effects of the path,
however show that as the frequency increases, the ability (14) captures local street characteristics in addition. We
to overcome obstacles drastically reduces. The 868 MHz have chosen wt to cater for shared surface street (used by
technologies require significantly taller masts or towers. pedestrians and vehicles in UK, minimum 4.8m [73]) and
 8

240 240

220 220

200 200
LoRa/Sigfox/WavIoT
RPMA(Ingenu)
180 180
LTE-M
Path loss (dB)

Path loss (dB)

160 160

140 140

120 120 LoRa/Sigfox/WavIoT

RPMA(Ingenu)
LTE-M
100 100

80 80

60 60
0 200 400 600 800 1000 1200 1400 1600 1800 2000 0 200 400 600 800 1000 1200 1400 1600 1800 2000
Distance from Rx to center of intersection (m) Distance from Rx to center of intersection (m)
(a) Gateway height 8m, end device height 0.5m (b) Gateway height 8m, end device height 2m
Figure 5: Non LOS path loss in urban environment with S = f −0.024 [69], δ = 7, wt = 7, xt = 2

assume a regular shaped junction such that main and side that the selected access technology is capable of providing
streets are of equal widths. adequate coverage for the area of interest. While coverage
Fig. 5 illustrates the effects of dc and street char- is relatively less constrained in rural and open areas, it
acteristics on NLOS path loss in urban street canyons is more challenging in large urban environments such as
for low and medium altitude IoT applications. The path metropolitan areas with heavy presence of high-rise modern
loss significantly increases when d > dc . In other words, buildings, road side furniture such as street lights (lamp
reflection dominates the path loss when the receiving end posts) and other man-made obstacles that reduce signal
device is close to the street junction, up to the cross over strength. By substituting (8)-(11) into (7), the maximum
distance dc . This is because the rays from the main street range can be expressed as
bend and arrive on the side street at an incidence angle (Pt +Gainnet −σ−y1 +y2 +kt )
close to 90◦ which results in very little or no refraction. R = 10 bh
(15)
Thereafter, the signals are reflected multiple times as they
where y1 = 69.55 + 26.16 · log10 f , y2 = 13.82 · log10 hg ,
propagate further down the side street. Hence, for end
kt= a(ht ) + ct and bh = 44.9 − 6.55 · log10 hg .
devices located far from the junction, (beyond dc ), the path
loss is dominated by diffraction. By increasing the height
Fig. 6 illustrates the maximum range achievable in ru-
of the end device from 0.5 m to 2 m, Fig. 5 shows that the
ral and suburban areas using the main access technology
path loss immediately after dc in LoRa (146 m), RPMA
specifications outlined in Table II. In particular, this figure
(403 m) and LTE-M (436 m) can be reduced by 77.24 dB,
shows that the coverage of each technology is dependent on
86.07 dB and 86.78 dB respectively.
the transmit power. In practice, coverage will vary among
sites depending on the field characteristics. For example,
VI. N ETWORK P ERFORMANCE A NALYSIS
the presence of multiple obstacles such as trees or walls
In this section, we compare performance of the various along the propagation path will also add to the attenuation
LP-WAN technologies using the system parameters in each of signal power. It is seen in Fig. 6 that in rural areas,
specification according to its data sheet and the scenario in using equal transmit power in the specifications considered,
the previous section. A sensitivity analysis is also conducted NB-IoT offers a significantly wider coverage (57.79 km
for each specification in urban area. The outcomes provide at 40 dBm) than LTE-M (8.57 km at 40 dBm) in the
useful insights for attaining the trade-off between network cellular domain while LoRaWAN offers the longest range
optimisation and design complexity on one hand and cost (29.67 km) among the non-cellular LP-WAN technologies.
on the other. We also examine the energy consumption and Overall, NB-IoT provides the widest coverage out of the
its dependencies in different use-cases. The battery life and six specifications considered. Although, Table II indicates
energy consumption discussed in this subsection are with that LTE-M and RPMA provide higher link budget than
respect to the radio transceiver and includes transmit and LoRaWAN, the low obstacle loss in LoRaWAN means that
receive consumption only. Other device components such as given the same transmit power and operating condition, it
micro-controller unit (MCU), wait windows, sleep currents, is more resilient to obstructions which improves the signal
and other latent consumption in the system are not included. power reaching the end device from the gateway. On the
other hand, even with lower transmit power of NB-IoT than
A. Coverage LTE-M, the former relies on its better sensitivity and higher
One of the key performance metrics in wireless systems antenna gain to significantly extend the network coverage.
is the communication range. Given the scale of IoT in terms Similarly, the range in urban environment is illustrated
of applications, one of the challenges is how to ensure in Fig. 7. The figure shows coverage of 2.89 km (NB-IoT),
2.68 km (LoRa), 1.74 km(WavIoT) in large urban areas,
 9

60 15
WaveIoT WaveIoT
LoRa LoRa
Sigfox Sigfox
RPMA(Ingenu) RPMA(Ingenu)
50
LTE-M LTE-M
NB-IoT 1840MHz NB-IoT 1840MHz

40 10
Maximum range (km)

Maximum range (km)

30

20 5

10

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Transmit Power (dBm) Transmit Power (dBm)
(a) Range in open and rural areas (b) Suburban areas
Figure 6: Maximum range in rural and suburban areas

6 3
WaveIoT WaveIoT
LoRa LoRa
Sigfox Sigfox
RPMA(Ingenu) RPMA(Ingenu)
5 2.5
LTE-M LTE-M
NB-IoT 1840MHz NB-IoT 1840MHz

4 2
Maximum range (km)

Maximum range (km)

3 1.5

2 1

1 0.5

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Transmit Power (dBm) Transmit Power (dBm)
(a) Range in small cities (b) Range in large cities and metropolises
Figure 7: Maximum range of access technologies in urban environments

meaning that NB-IoT end device can receive signal farther city of Padova in Italy reported a network coverage of about
away from the gateway than LTE-M while LoRa receives 1.2 km. This closely aligns with our result for small cities in
the farthest signal among the non-cellular technologies. Fig. 7a (1.4 km) based on LoRaWAN system specification
Thus, the sub-GHz technologies are particularly beneficial with gateway transmit power of 21dBm employed in this
in rural areas where masts are less likely to be spread out. work. Figs. 6 and 7 show that the 868MHz technologies for
They also have some advantages in cities, because the sub- instance do not exhibit the same coverage performance even
GHz technologies have lower penetration losses across brick when operated at the same frequency and equal transmit
walls, vegetation, concrete, metals, etc, which is suitable for power. This is mainly due to the differences in the systems
underground applications such as standpipe monitoring. As design and configurations.
we move from small to large cities, the performance of For each technology type, the equivalent isotropically
RPMA further diverges from Sigfox. This is mainly due radiated power (EIRP) and SNR are also calculated. The
to their lower resilience to obstacle losses in large urban EIRP is the output power radiated from the tip of the
areas. Fig. 7 shows that RPMA and LTE-M are poor at gateway antenna. Using the PL results in Fig. 3, the EIRP
travelling over long distances, which requires the towers and nominal SNR for suburban environment at different
to be closer together to deliver reliable coverage. Hence, link distances are summarised in Table V (for the downlink
2.x GHz technologies may be less attractive in some rural only).
deployments and deep indoor applications. Given the wide In many wireless systems such as Wi-Fi, the target SNR
range of frequency and different modes of deployment sup- is usually 20 dB or higher to ensure reliable communication
ported by NB-IoT, it can be used in > GHz band or sub-GHz [63]. However, as seen in Table V, at link distance of 20 km,
to cater for IoT connectivity in different terrains. In [32], Sigfox and LoRaWAN yield SNR of 9.7 dB while RPMA
the field experiments using LoRa deployment to cover the offered 9.4 dB. These low SNR values are indications that
the connectivity between IoT end point and gateway may be
 10

140
140 Antenna gains
Transmit power Antenna gains
Transmit power
Receiver sensitivity 120
120 Tx antenna height Receiver sensitivity
Tx antenna height
100 100
Percentage change in range (%)

Percentage change in range (%)

80 80

60 60

40 40

20 20

0
0
-20
-20
-40
-40
-60
-60
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
Percentage change in parameter (%) Percentage change in parameter (%)
(a) LoRaWAN urban area (b) Sigfox urban area
Figure 8: Maximum range of LoRaWAN and Sigfox in urban environments

Table V: RSSI and SNR for suburban environments at different

distances for various access technologies Figs. 8 and 9, respectively illustrates the coverage gain
in large cities and metropolitan areas. In each case, the
Spec. WAVIoT LoRaWAN Sigfox RPMA LTE-M
EIRP 27 27 27 27 47
outcome is compared among four access technologies and
P L10 136.5 136.5 136.5 144.9 145.5 it can be seen that the network responds to the changes
P L20 147.3 147.3 147.3 155.6 156.3 in different degrees. In Fig. 8, LoRaWAN and Sigfox
RSSI10 -109.5 -109.5 -109.5 -117.9 -98.5 exhibit similar performance and trend until 3% change
RSSI20 -120.3 -120.3 -120.3 -120.6 -109.3
SN R10 20.5 20.5 20.5 12.1 31.5 in system parameter. Beyond that point, LoRaWAN out-
SN R20 9.7 9.7 9.7 9.4 20.7 performs Sigfox and this is more pronounced from 8-
10% where LoRaWAN provides more coverage gain. LP-
WAN transceivers are engineered to be very sensitive. As
impaired with high frame error rate which will ultimately a result, the end points can detect signals that are smaller
result in sub-optimal link performance. In order to provide than transmit power by many order of magnitude. Fig. 8
remedies for this condition, the following options can be however shows that by increasing the receiver sensitivity
explored by 10%, LoRaWAN and Sigfox can extend coverage by
• Improve receiver sensitivity 141.7% and 129.57%, respectively, meaning a comparative
• Increase the transmit power gain of 12.13% in favour of LoRaWAN. Furthermore, Fig.
• Use higher gain antenna 9 illustrates the additional coverage gain when similar
• Where possible, remove obstacles or other sources of parameters are varied in RPMA and LTE-M. It is observed
losses that a 10 % improvement in receiver sensitivity can increase
• Increase the antenna height coverage by 135.56% in RPMA and 19% in LTE-M. It is
• Increase gateway density by deploying more base further noted that increase in transmit power alone yields
stations which will reduce the distance from gateway about 17% coverage improvement in the best case (Fig. 9a).
to IoT end point. Similarly in Fig. 10, it is observed that with a 10%
Some of these factors are investigated as part of the sensi- improvement in NB-IoT 882 MHz receiver sensitivity, cov-
tivity analysis in the next section. erage can increase by about 398% while a similar change
in transmit power and antenna gain yields 130% and 122%
B. Sensitivity Analysis and Network Optimisation respectively. In the case of NB-IoT 1840 MHz, a 10%
improvement in sensitivity yields about 191% coverage gain
We observed from Fig. 7 that coverage limitation is while a similar change in transmit power and antenna gain
more imminent in urban areas, therefore the objective of enhance coverage by about 34% and 30% respectively.
this subsection is to investigate ways to optimise network Based on the specifications and parameters employed in
performance by conducting a sensitivity analysis. Equations this paper, we can see in Figs. 8-10 that among the sys-
(11) and (15) as well as their dependence on field param- tem parameters employed, receiver sensitivity and transmit
eters suggest that some variables can be tuned to optimise power are two of the key parameters for optimising LP-
performance of the communication network. As the current WAN coverage. This approach affords the opportunity to
LP-WAN specifications employ different values of system identify the most sensitive parameters. With such knowl-
parameters, percentage changes are employed instead of ab- edge, system designers and engineers can explore further
solute values. In each case, a variable is changed by a fixed ways of tweaking the identified parameters in a manner that
percentage and the resultant improvement or degradation in is cost-effective without compromising network reliability,
the network coverage is estimated. Thus, the transmit power, power requirement or other key performance indicators of
antenna gain, receiver sensitivity, height of gateway antenna the access networks.
are varied within ±10% in succession.
 11

140 20
Antenna gains Antenna gains
Transmit power Transmit power
120 Receiver sensitivity 10 Receiver sensitivity
Tx antenna height Tx antenna height
100 0
Percentage change in range (%)

Percentage change in range (%)

80 -10

60 -20

40 -30

20 -40

0 -50

-20 -60

-40 -70

-60 -80
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
Percentage change in parameter (%) Percentage change in parameter (%)
(a) RPMA (2.4GHz ISM) (b) LTE-M (2.6GHz)
Figure 9: Coverage optimisation with 2.4GHz and 2.6GHz LTE in urban environments

400 200
Antenna gains Antenna gains
Transmit power Transmit power
350 Receiver sensitivity Receiver sensitivity
Tx antenna height Tx antenna height
150
300
Percentage change in range (%)

Percentage change in range (%)

250 100

200
50
150

100 0

50
-50
0

-50 -100
-10 -8 -6 -4 -2 0 2 4 6 8 10 -10 -8 -6 -4 -2 0 2 4 6 8 10
Percentage change in parameter (%) Percentage change in parameter (%)
(a) NB-IoT 882 MHz (b) NB-IoT 1840 MHz
Figure 10: Range optimisation with NB-IoT at 882 MHz and 1840 MHz, bands 5 and 3 downlink frequency respectively, 3GPP NB-IoT (rel. 13)
[49]

While increase in antenna gain, transmit power and  −1

antenna height can improve the received signal level, they βσ
τ (∞, σ, | Ψ |) = ln2. | Ψ | (17)
also increase the chances of higher interference. Therefore φN0
the pros and cons of each optimisation approach needs to Equations (16) and (17) further imply that given a fixed
be carefully considered. receiver sensitivity in each LP-WAN specification, the trans-
mission delay depends on the data rate and message length.
C. Transmission Delay Sending 50 byte payload in LoRaWAN in Europe in some
The transmission delay of a message of length | Ψ |bits cases can take approximately 2.8s using the minimum data
can be determined from Shanon-Hartley theorem as rate (SF12) and 0.154 s using the maximum data rate (SF7).
This means that the end devices have to wait 4 min 37 s
  after each transmission in the former and 15.25 s in the
|Ψ | βσ
τ (ε, σ, | Ψ |) = · log2 1 + ε−1 . (16) latter. However, Fig. 11 indicates that if the payload size
ε φN0 is reduced by 20%, the time-on-air (ToA) can decrease by
where σ(dBm) = Pt − α is the receiver sensitivity (Pt is 328 ms and 61.44 ms for SF12 and SF7, respectively.
the transmit power, α is the coupling loss), β is a correction ToA for LoRaWAN is significantly affected by both
factor to account for difference between Shanon limit and payload size and SF. However, as seen in Fig. 11, the impact
realistic systems and N0 is the thermal noise power spectral of SF is higher than that of the payload, which implies that
density (PSD) which is about -174 dBm/Hz. The minimum for a given payload, the ToA increases more rapidly with the
time required to transmit message A is achieved when ε −→ SF value. The reason is that SF 7 and SF 12 offer the highest
∞ . This is supported by the fact that larger bandwidth and lowest data rates respectively. Therefore, increasing the
causes messages to be transmitted faster such that [74] SF is equivalent to reducing the network throughput and if
 12

3 80
WaveIoT
LoRa
Sigfox
70 RPMA(Ingenu)
2.5
SF7 NB-IoT 837MHz
SF8 NB-IoT 1747MHz

Transmit energy per message ( J)

SF9 LTE-M
60
SF10
2 SF11
Time on air (sec)

SF12
50
1.5

40

1
30

0.5
20

0
5 10 15 20 25 30 35 40 45 50 10
0 5 10 15 20 25 30 35 40 45 50
Payload size (Byte) Transmission distance (km)
Figure 11: Time on Air of LoRaWAN messages vs payload size with (a) Comparing transmit energy for WavIoT, LoRa, Sigfox, RPMA, LTE-M
coding rate 4/5, preamble symbols 8, bandwidth 125kHz, duty cycle and NB-IoT
0.1%
14
WaveIoT
LoRa
13.5 Sigfox
RPMA(Ingenu)
Transmitter Receiver NB-IoT 837MHz
d
'Íë :Gá @; 'Ëë :G; k-bit
Transmit energy per message ( J) NB-IoT 1747MHz
k-bit 13
message message
Electronics Amplifier Electronics
12.5
'ØßØÖ Û G ó=IL Û G Û @á 'ØßØÖ Û G

12
Figure 12: The first order radio energy dissipation model
11.5

all other parameters remain unchanged, the penalty for that 11

is higher ToA which will also impact the end-to-end latency
experienced by the software (application). 10.5

10
D. Energy Consumption 0 5 10 15 20 25 30 35 40 45 50
Transmission distance (km)
To investigate the energy consumption of the IoT end
(b) Enlarged Fig. 11a showing WavIoT, LoRa, Sigfox, RPMA and
devices, this paper adopts the first order radio energy dissi- NB-IoT
pation model introduced in [55] and subsequently applied
Figure 13: Transmit energy consumed per message using gateway
in other studies such as [56], [75]. In estimating the battery height of 12 m and end device height of 0.5 m, data rate 100 bps and
life of the transceivers, we adopt the E91-AA alkaline payload of 12 bytes
battery model and apply its operational characteristics [76].
According to [55], the transmitter consumes energy to
operate the radio electronics and the power amplifier while between the gateway and end device. If the transmission dis-
the receiver consumes energy to run the radio electronics tance is less than a certain threshold dc , the free space model
as shown in Fig. 12. (d2 power loss) is used, otherwise the multipath model (d4
The energy required to transmit a k-bit message from power loss) applies [55]. Therefore, power control can be
the IoT end device to the gateway over a distance d can be employed to invert this loss by setting the power amplifier
Tx
defined as value of Eamp at the transmitter in a way that ensures
that adequate signal power arrive at the receiver. Thus, the
E T x (k, d) = Eelec
Tx Tx
(k) + Eamp (k, d) (18) transmitter energy consumption can be written as
Tx
where Eelec is the electrical energy consumed by the elec- (
Tx
tronic circuit of the transmitter radio and depends on signal k(Eelec + F riis d2 ) d < dc
E T x (k, d) = Tx 4
(20)
processing techniques such as modulation, spreading and k(Eelec + T wo−ray d ) d ≥ dc
Tx
coding, while Eamp is the communication-related energy
√ hg ht
consumed by the amplifier which depends on the environ- where the threshold distance is define as dc = 4π L ,
mental factors including the distance from transmitter to γ
L ≥ 1 is the system loss factor and γ is the wavelength
receiver. At the receiver, since the radio only needs energy
of the signal. Applying the parameters in Table II, if
to operate the electronics to detect the signal, the energy
the gateway is 12 m high and the end device is 0.5 m
consumed is
(low altitude applications such as smart agriculture and
smart cities) with L=1 (assume no system loss), the dc for
E Rx (k) = Eelec
Rx
(k) = k · Eelec (19)
WavIoT/Sigfox/LoRa, RPMA and LTE-M are 218.18 m,
Both the free space and multipath-fading channels are 603.26 m and 653.54 m respectively.
considered in the energy model, depending on the distance
 13

16 16

14 14

WaveIoT
Estimated battery life (Years)

Estimated battery life (Years)

12 12
Sigfox WaveIoT
LoRa Sigfox
RPMA(Ingenu) LoRa
10 NB-IoT 1747MHz 10 RPMA(Ingenu)
LTE-M NB-IoT 1747MHz
LTE-M

8 8

6 6

4 4

2 2
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
Distance (km) Distance (km)
◦ ◦
(a) Device battery operating at 40 C (b) Device battery operating at 20 C

12 8

11
7
10
Estimated battery life (Years)

Estimated battery life (Years)

9 6
WaveIoT WaveIoT
Sigfox Sigfox
8 LoRa LoRa
RPMA(Ingenu) 5 RPMA(Ingenu)
NB-IoT 1747MHz NB-IoT 1747MHz
7 LTE-M LTE-M

4
6

5 3

4
2
3

2 1
0 5 10 15 20 25 30 35 40 45 50 0 5 10 15 20 25 30 35 40 45 50
Distance (km) Distance (km)
◦ ◦
(c) Device battery operating at 0 C (d) Device battery operating at -20 C
Figure 14: Estimated battery life in LP-WAN transceiver operating at 1% duty cycle under different weather conditions, using gateway height of
12 m and end device height of 0.5 m, data rate 100 bps and payload of 12 bytes.

Fig. 13 compares the transmit energy consumed per devices are active for 14.9 years and 8.8 years◦ respectively.
message for the LP-WAN technologies considered in this However, when the temperature drops to -20 C, the figure
study. It can be seen that overall, LTE-M consumes the indicates that the battery life reduces to about 7.32 years
highest amount of energy for every message sent by the for LoRaWAN, Sigfox and WavIoT as well as 7 years
IoT end device, followed by RPMA. For distances below 2 and 4.2 years for RPMA and LTE-M respectively. These
km, the technologies consume approximately equal amount results generally suggest that even with the same system
of energy. However, beyond 2 km, the performance gap configurations and battery capacity when IoT devices are
increases rapidly with distance. deployed, their active lifetimes will vary across regions
Here we apply a duty cycle of 1% (corresponding to according to the local temperature and seasonal variations.
140 messages per day) to the devices to conform with the These results can potentially offer useful insights into the
regulatory requirements in Europe and analyse the energy planning, designing and management of the IoT access
consumption of the devices. Fig. 14 presents the estimated networks, especially in use-cases such as haulage/logistics,
battery life in the devices when deployed with the LP- smart agriculture and many smart city scenarios in which
WAN technologies and operated at different temperatures. the sensors may be unavoidably exposed to extreme weather
This figure shows clearly that the operating temperature is a conditions.
key determinant of battery life. In particular, extremely cold In Fig. 15, we present the variation of daily transmit
temperature can significantly shorten the active lifespan of energy consumption with payload size for different cover-
the end device due to rapid degradation of its battery’s ca- age requirements using LoRaWAN. Although the effect of
pacity. Nevertheless, based on the configurations employed distance is generally more pronounced, the figure indicates
in this section, LoRa appears to be the most promising to that energy required is significantly affected by message size
support long battery life while LTE-M provides the shortest as well. For example, for a coverage distance of 50 km,
battery life across the temperatures investigated. For exam- reducing the message size by 20% can result in transmit
ple, with ◦a coverage requirement of 25 km, Fig. 14 shows energy savings of about 47µJ per message and this can
that at 40 C, LoRaWAN, Sigfox, NB-IoT and WavIoT can extend the device life. The availability of SF in LoRaWAN
each support up to 15.65 years while RPMA and LTE-M provides a way to further trade-off between system variables
 14

250
10 km
20 km
30 km
40 km
200 50 km
Transmit energy per message ( J)

N OMENCLATURE
α coupling loss
150
β correction factorbetween Shanon limit and realistic
systems
100 δ environmental parameter
F riis dissipation variable, free space
T wo−ray dissipation variable, multi-path fading
50 η receiver sensitivity
γ wavelength
λ gateway density
0
0 50 100 150 200 250 φ noise figure
Payload (Bytes) σ receiver sensitivity
Figure 15: Effects of payload size on daily energy consumption for θ threshold SNR
different coverage requirements in LoRaWAN using gateway height
12 m, end device height 0.5 m, duty cycle of 1%, and data rate of 25
ε bandwidth
kbps. ϕ the viewing angle from gateway to end device
antenna
Ψ message length
and energy consumption, thereby offering additional degree a(ht ) terrain parameter
of flexibility in IoT access network design. C channel capacity
ct terrain correction factor
VII. C ONCLUSION dc cross-over distance
Tx
Current IoT access technologies are optimised for dif- Eamp electrical energy consumed by transmitting radio
ferent end-use applications. This paper showed that each amplifier
of RPMA and LTE-M incur at least 9 dB additional path Eb energy per bit
loss compared with Sigfox and LoRAWAN. That means Eelec electrical energy consumed by radio electronic cir-
they are more suitable for short range applications. In real cuit
deployments, the alternatives include the use of higher gain Gnet net gain
antennas or reduction of coverage per gateway both of hg gateway antenna height
which impact the deployment cost. N0 thermal noise power
To illustrate the consequences of link budget on system Pt transmit power
performance, parameters were adopted from the LP-WAN wt width of main street
datasheets. The results consistently indicated that receiver xt distance: gateway antenna to wall on main street
sensitivity is the most significant variable in LP-WAN
optimisation, followed by transmit power. For example,
it was shown that with a 10% improvement in receiver R EFERENCES
sensitivity, LoRaWAN can increase network coverage by [1] Statistica, “Internet of things (iot) connected devices installed base
about 142% and Sigfox by 130% whereas with similar worldwide from 2015 to 2025 (in billions),” 2015, accessed on 26
receiver sensitivity in NB-IoT 882 MHz and LTE-M, the May. 2018. [Online]. Available: https://www.statista.com/statistics/
471264/iot-number-of-connected-devices-worldwide/
coverage gains were about 398% and 136% respectively. [2] M. James, C. Michael, B. Peter, W. Jonathan, D. Richard, B. Jacques,
Across the technologies, the active device life is affected and A. Dan, “Unlocking the potential of the Internet of things,”
by payload size, distance and environmental conditions such 2015, accessed on 26 May. 2018. [Online]. Available: https://www.
mckinsey.com/business-functions/digital-mckinsey/our-insights/
as temperature. In addition to the applied load, it was the-internet-of-things-the-value-of-digitizing-the-physical-world
observed that extreme weather conditions can significantly [3] Drivesnccontrol, “1 in 4 wireless IIoT connections will use LPWA
reduce active device life of IoT objects in the LP-WAN. by 2025,” 2017. [Online]. Available: http://m.drivesncontrols.com/
◦ news/fullstory.php/aid/5553/1_in_4_wireless_IIoT_connections_
Compared with environmental temperature of 40 C, the will_use_LPWA_by_2025.html26May.2018)
results◦indicate operating the IoT devices at a temperature [4] T. M. WG, “A technical overview of LoRa and LoRaWAN,”
of -20 C, can reduce the device life by about half; 53% 2015. [Online]. Available: https://docs.wixstatic.com/ugd/eccc1a_
ed71ea1cd969417493c74e4a13c55685.pdf(accessedon29Jan.2018)
(WavIoT, LoRaWAN, Sigfox, RPMA) and 48% in LTE- [5] K. E. Nolan, W. Guibene, and M. Y. Kelly, “An evaluation of low
M. Indoor signal propagation is remarkably different from power wide area network technologies for the Internet of things,” in
outdoor due to the plethora of objects contributing to the Proc. IEEE Int. Wireless Commun. Mob. Comp., 2016, pp. 439–444.
[6] U. Raza, P. Kulkarni, and M. Sooriyabandara, “Low power wide area
interference level as well as attenuation from different networks: An overview,” IEEE Commun. Survs. Tuts., vol. 19, no. 2,
objects some of which are outlined in Table IV. Thus, our pp. 855–873, 2017.
future work in this area includes the investigation of the [7] N. Poursafar, M. E. E. Alahi, and S. Mukhopadhyay, “Long-range
wireless technologies for IoT applications: A review,” in Proc. IEEE
effects of indoor losses on IoT network performance using Intl. Conf. Sensing Technol., Dec 2017, pp. 1–6.
the LP-WAN technologies discussed in this paper and any [8] J. de Carvalho Silva, J. J. P. C. Rodrigues, A. M. Alberti, P. Solic,
future ones available. and A. L. L. Aquino, “LoRaWAN- A low power WAN protocol for
Internet of Things: A review and opportunities,” in Proc. IEEE Intl.
Multidisc. Conf. Computer and Energy Sci.(SpliTech), July 2017, pp.
1–6.
 15

[9] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer, “A comparative study of smart city scenarios,” IEEE Wireless Commun., vol. 23, no. 5, pp.
LPWAN technologies for large-scale IoT deployment,” ScienceDirect 60–67, 2016.
ICT Express, vol. 12, no. 05, 2018. [33] H. Wang and A. O. Fapojuwo, “A survey of enabling technologies
[10] M. Aazam, S. Zeadally, and K. A. Harras, “Offloading in fog of low power and long range machine-to-machine communications,”
computing for IoT: Review, enabling technologies, and research IEEE Communications Surveys & Tutorials, vol. 19, no. 4, pp. 2621–
opportunities,” ScienceDirect Future Generation Computer Systems, 2639, 2017.
vol. 04, no. 057, 2018. [34] G. A. Akpakwu, B. J. Silva, G. P. Hancke, and A. M. Abu-Mahfouz,
[11] S. Al-Sarawi, M. Anbar, K. Alieyan, and M. Alzubaidi, “Internet of “A survey on 5G networks for the Internet of things: Communication
Things (IoT) communication protocols: Review,” in Proc. IEEE Intl. technologies and challenges,” IEEE Access, vol. 6, pp. 3619–3647,
Conf. Information Technol., May 2017, pp. 685–690. 2018.
[12] K. Zheng, S. Zhao, Z. Yang, X. Xiong, and W. Xiang, “Design and [35] M. Chen, Y. Miao, Y. Hao, and K. Hwang, “Narrow band Internet
implementation of LPWA-based air quality monitoring system,” IEEE of things,” IEEE Access, vol. 5, pp. 20 557–20 577, 2017.
Access, vol. 4, pp. 3238–3245, 2016. [36] P. Neumann, J. Montavont, and T. Noël, “Indoor deployment of low-
[13] W. Guibene, J. Nowack, N. Chalikias, K. Fitzgibbon, M. Kelly, and power wide area networks (lpwan): A LoRaWAN case study,” in
D. Prendergast, “Evaluation of LPWAN technologies for smart cities: Proc. IEEE Conf. Wireless and Mobile Computing, Networking and
River monitoring use-case,” in Proc. IEEE Wireless Commun. and Communications (WiMob), 2016, pp. 1–8.
Network. Conf. Workshops (WCNCW), 2017, pp. 1–5. [37] K. Mikhaylov, J. Petäjäjärvi, and J. Janhunen, “On LoRaWAN
[14] J. Chen, K. Hu, Q. Wang, Y. Sun, Z. Shi, and S. He, “Narrowband scalability: Empirical evaluation of susceptibility to inter-network
Internet of things: Implementations and applications,” IEEE Internet interference,” in European Conf. Networks and Communications
of Things Journal, vol. 4, no. 6, pp. 2309–2314, 2017. (EuCNC). IEEE, 2017, pp. 1–6.
[15] W. Guibene, K. E. Nolan, and M. Y. Kelly, “Survey on clean [38] Mathuranathan, “Channel capacity & shannon theorem -
slate cellular-IoT standard proposals,” in Proc. IEEE Int. Conf. demystified,” 2008, accessed on 07 Aug. 2018. [Online]. Available:
Comput. and Inform. Technol; Ubiquitous Comput. and Commun.; https://www.gaussianwaves.com/2008/04/channel-capacity/
Dependable, Autonomic and Secure Comput; Pervasive Intell. and [39] T. Myers, “Back to basics: The Shannon-Hartley theorem,” 2016,
Comput (CIT/IUCC/DASC/PICOM), 2015, pp. 1596–1599. accessed on 17 Dec. 2017. [Online]. Available: https://www.ingenu.
[16] F. Adelantado, X. Vilajosana, P. Tuset-Peiro, B. Martinez, J. Melia- com/2016/07/back-to-basics-the-{S}hannon-{H}artley-theorem/
Segui, and T. Watteyne, “Understanding the limits of LoRaWAN,” [40] Semtech, “LoRa modulation basics,” Application Note, 2015.
IEEE Commun. Mag., vol. 55, no. 9, pp. 34–40, 2017. [41] A. AlAmmouri, J. G. Andrews, and F. Baccelli, “SINR and through-
[17] L. Gregora, L. Vojtech, and M. Neruda, “Indoor signal propagation put of dense cellular networks with stretched exponential path loss,”
of LoRa technology,” in Proc. Intl. Conf. Mechatronics-Mechatronika IEEE Trans. Wireless Commun., vol. 17, no. 2, pp. 1147–1160, 2018.
(ME), 2016, pp. 1–4. [42] X. Zhang and J. G. Andrews, “Downlink cellular network analysis
[18] B. Reynders, W. Meert, and S. Pollin, “Power and spreading factor with multi-slope path loss models,” IEEE Trans. Commun., vol. 63,
control in low power wide area networks,” in Proc. Intl. Conf. no. 5, pp. 1881–1894, 2015.
Commun. (ICC), 2017, pp. 1–6. [43] M. Di Renzo, W. Lu, and P. Guan, “The intensity matching approach:
[19] N. Mangalvedhe, R. Ratasuk, and A. Ghosh, “NB-IoT deployment A tractable stochastic geometry approximation to system-level anal-
study for low power wide area cellular IoT,” in Proc. IEEE Annual ysis of cellular networks,” IEEE Trans. Wireless Commun., vol. 15,
Intl. Symp. Pers., Indoor, Mob. Radio Commun., 2016, pp. 1–6. no. 9, pp. 5963–5983, 2016.
[20] J. Petajajarvi, K. Mikhaylov, A. Roivainen, T. Hanninen, and M. Pet- [44] A. Al Ammouri, J. G. Andrews, and F. Baccelli, “A unified
tissalo, “On the coverage of LPWANs: range evaluation and channel asymptotic analysis of area spectral efficiency in ultradense cellular
attenuation model for LoRa technology,” in Proc. IEEE Intl. Conf. networks,” IEEE Trans. Inform. Theory, 2018.
ITS Telecommunications, 2015, pp. 55–59. [45] K.-H. Phung, H. Tran, Q. Nguyen, T. T. Huong, and T.-L. Nguyen,
[21] M. Aref and A. Sikora, “Free space range measurements with “Analysis and assessment of LoRaWAN,” in Proc. IEEE Intl Conf.
semtech LoRa technology,” in Proc. IEEE Conf. Intel. Data Acqu. Recent Adv. Sig. Process., Telecomm. & Comput., 2018, pp. 241–246.
and Adv. Comput. Syst: Technol. and Appl. Wireless Syst., 2014, pp. [46] Link-Labs, “A comprehensive look at low power, wide area
19–23. networks for Internet of things engineers and decision makers,”
[22] M. Lauridsen, I. Z. Kovács, P. Mogensen, M. Sorensen, and S. Holst, 2016, accessed on 14 Dec. 2017. [Online]. Available: https:
“Coverage and capacity analysis of LTE-M and NB-IoT in a rural //www.link-labs.com/symphony
area,” in Proc. IEEE Veh. Tech Conf., 2016, pp. 1–5. [47] Ingenu, “An educational guide a white paper by ingenu how RPMA
[23] O. Georgiou and U. Raza, “Low power wide area network analysis: works,” 2015, accessed on 14 Dec. 2017. [Online]. Available:
Can LoRa scale?” IEEE Wireless Commun. Lett., vol. 6, no. 2, pp. http://www.ingenu.com
162–165, 2017. [48] WavIOT, “Comparison of LPWAN technologies,” 2017, accessed on
[24] Y. Mo, C. Goursaud, and J.-M. Gorce, “Theoretical analysis of UNB- 23 Dec. 2017. [Online]. Available: https://waviot.com/wp.pdf
based IoT networks with path loss and random spectrum access,” in [49] J.Schlienz and D. Raddino, “Narrowband Internet
Proc. IEEE Int. Symp. Personal, Indoor, and Mobile Radio Commun., of Things whitepaper, Munich, Germany, Rohde &
2016, pp. 1–6. Schwarz,” 2016, accessed on 03 Aug. 2018. [On-
[25] H. Mroue, A. Nasser, S. Hamrioui, B. Parrein, E. Motta-Cruz, and line]. Available: https://cdn.rohde-schwarz.com/pws/dl_downloads/
G. Rouyer, “MAC layer-based evaluation of IoT technologies: LoRa, dl_application/application_notes/1ma266/1MA266_0e_NB_IoT.pdf
SigFox and NB-IoT,” in Proc. IEEE Middle East and North Africa [50] A. Ratilainen, “NB-IoT presentation for IETF LPWAN,” 2016, ac-
Commun. Conf (MENACOMM, 2018, pp. 1–5. cessed on 03 Aug. 2018. [Online]. Available: https://datatracker.ietf.
[26] M. D’Aloia, A. Longo, R. Ruggero, S. Stanisci, D. Amendolare, org/meeting/97/materials/slides-97-lpwan-30-nb-iot-presentation-00
M. Vessia, F. Lomastro, and M. Rizzi, “An innovative LPWA network [51] R. Ratasuk, J. Tan, N. Mangalvedhe, M. H. Ng, and A. Ghosh,
scheme to increase system reliability in remote monitoring.” “Analysis of NB-IoT deployment in LTE guard-band,” in Proc. IEEE
[27] S. Gangakhedkar, O. Bulakci, and J. Eichinger, “Addressing deep Veh. Technol. Conf. (VTC Spring), 2017, pp. 1–5.
indoor coverage in narrowband-5G,” in Proc. IEEE Veh. Tech. Conf, [52] S.-M. Oh and J. Shin, “An efficient small data transmission scheme
2017, pp. 1–5. in the 3GPP NB-IoT system,” IEEE Commun. Letters, vol. 21, no. 3,
[28] C. Y. Yeoh, A. bin Man, Q. M. Ashraf, and A. K. Samingan, pp. 660–663, 2017.
“Experimental assessment of battery lifetime for commercial off-the- [53] D. Hernandez, G. Peralta, L. Manero, R. Gomez, J. Bilbao, and
shelf NB-IoT module,” in Proc. IEEE Intl. Conf. Advanced Commun. C. Zubia, “Energy and coverage study of LPWAN schemes for
Techn.,, 2018, pp. 1–1. industry 4.0,” in Proc. IEEE Intl. Workshop of Electron., Con-
[29] A. Augustin, J. Yi, T. Clausen, and W. M. Townsley, “A study of trol, Measurement, Signals and their Application to Mechatronics
lora: Long range & low power networks for the Internet of things,” (ECMSM), 2017, pp. 1–6.
Sensors, vol. 16, no. 9, p. 1466, 2016. [54] G. G. Ribeiro, L. F. de Lima, L. Oliveira, J. J. Rodrigues, C. N.
[30] K. Mikhaylov, J. Petaejaejaervi, and T. Haenninen, “Analysis of Marins, and G. A. Marcondes, “An outdoor localization system based
capacity and scalability of the LoRa low power wide area network on Sigfox,” in Proc. IEEE Conf. Veh. Technol., 2018, pp. 1–5.
technology,” in Proc. VDE European Wireless Conf., 2016, pp. 1–6. [55] W. B. Heinzelman, A. P. Chandrakasan, and H. Balakrishnan, “An
[31] M. Lauridsen, H. Nguyen, B. Vejlgaard, I. Z. Kovács, P. Mogensen, application-specific protocol architecture for wireless microsensor
and M. Sorensen, “Coverage comparison of GPRS, NB-IoT, lora, networks,” IEEE Trans. wireless commun., vol. 1, no. 4, pp. 660–
and Sigfox in a 7800 km2 area,” in Proc. IEEE Veh. Technol. Conf., 670, 2002.
2017, pp. 1–5. [56] Y. Zhuang, J. Pan, and G. Wu, “Energy-optimal grid-based clus-
[32] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, “Long-range tering in wireless microsensor networks,” in Proc. IEEE Intl. Conf.
communications in unlicensed bands: The rising stars in the IoT and Distributed Computing Systems Workshops, 2009, pp. 96–102.
 16

[57] Sigfox, “Sigfox technical overview,” 2017, accessed on 7 Mar. 325, 1980.
2018. [Online]. Available: https://www.element14.com/community/ [68] Y. Okumura, “Field strength and its variability in VHF and UHF
docs/DOC-87914/l/sigfox-technical-overview land-mobile radio-services,” Review of the Electrical Commun. Lab.,
[58] B. Vejlgaard, M. Lauridsen, H. Nguyen, I. Z. Kovács, P. Mogensen, vol. 16, 1968.
and M. Sorensen, “Interference impact on coverage and capacity for [69] J. S. Lu, H. L. Bertoni, K. A. Remley, W. F. Young, and J. Ladbury,
low power wide area IoT networks,” in Proc. IEEE Wireless Commun. “Site-specific models of the received power for radio communication
and Networking Conf. (WCNC), 2017, pp. 1–6. in urban street canyons,” IEEE Trans. Antenn. and Propagation,
[59] Randallthorp, “West Carbourne building heights,” 2014, accessed on vol. 62, no. 4, pp. 2192–2200, 2014.
10 Jun. 2018. [Online]. Available: https://www.scambs.gov.uk/sites/ [70] J. Lee, M.-D. Kim, H. K. Chung, and J. Kim, “NLOS path loss
default/files/documents/85S-77J%20BuildingHeights.pdf model for low-height antenna links in high-rise urban street grid
[60] A. D. Zayas and P. Merino, “The 3GPP NB-IoT system architecture environments,” Intl. Journal of Antenn. and Propagat., vol. 2015,
for the Internet of Things,” in Proc IEEE Intl. Conf. Commun. 2015.
Workshops (ICC Workshops), 2017, pp. 277–282. [71] T. Mangel, O. Klemp, and H. Hartenstein, “A validated 5.9 GHz
[61] A. Ikpehai, B. Adebisi, and A. Kelvin, “Effects of traffic character- non-line-of-sight path-loss and fading model for inter-vehicle com-
istics on energy consumption of iot end devices in smart city,” in munication,” in Proc. IEEE Conf. Telecommunications (ITST), 2011,
Proc. IEEE Global Infor. Infra. and Networking Symp. pp. 75–80.
[62] H. Mazar, “LOS radio links, clearance above tall buildings,” in [72] H. M. El-Sallabi, “Fast path loss prediction by using virtual source
Convention of Electrical and Electronics Engineers in Israel, Pro- technique for urban microcells,” in Proc. IEEE Veh. Technol. Conf.
ceedings. IEEE, 1991, pp. 145–148. Proceedings, vol. 3, 2000, pp. 2183–2187.
[63] Cisco, “Radio frequency fundamentals,” 2014, accessed on 2 Jan. [73] D. Cottingham, “How wide are roads?” 2018, accessed on
2018. [Online]. Available: https://www.cisco.com/c/en/us/td/docs/ 19 Aug. 2018. [Online]. Available: https://mocktheorytest.com/
solutions/Enterprise resources/how-wide-are-roads/
[64] T. S. Rappaport et al., Wireless communications: principles and
[74] W. Yang, M. Wang, J. Zhang, J. Zou, M. Hua, T. Xia, and X. You,
practice. prentice hall PTR New Jersey, 1996, vol. 2.
“Narrowband wireless access for low-power massive Internet of
[65] V. Abhayawardhana, I. Wassell, D. Crosby, M. Sellars, and
Things: A bandwidth perspective,” IEEE Wireless Commun., vol. 24,
M. Brown, “Comparison of empirical propagation path loss models
no. 3, pp. 138–145, 2017.
for fixed wireless access systems,” in Proc. IEEE Veh. Tech. Conf,
vol. 1, 2005, pp. 73–77. [75] Y. Zhuang, J. Pan, and L. Cai, “Minimizing energy consumption with
[66] W. Debus, “RF path loss & transmission distance calculations,” probabilistic distance models in wireless sensor networks,” in Proc.
2006. [Online]. Available: https://pdfs.semanticscholar.org/f126/ IEEE INFOCOM, 2010, pp. 1–9.
eb84e340cf74d63ab26782c9ad30cd111114.pdf(accessedon16Jan. [76] J. K. Young, “A practical guide to battery technologies
2018) for wireless sensor networking,” 2008, accessed on 18 Jun.
[67] M. Hata, “Empirical formula for propagation loss in land mobile 2018. [Online]. Available: https://www.sensorsmag.com/components/
radio services,” IEEE Trans. on Veh. Tech, vol. 29, no. 3, pp. 317– a-practical-guide-to-battery-technologies-for-wireless-sensor-networking

View publication stats