Research Article
International Journal of Distributed
Sensor Networks
2017, Vol. 13(3)
Ó The Author(s) 2017
DOI: 10.1177/1550147717699412
journals.sagepub.com/home/ijdsn
Performance of a low-pow er wide-area
network based on LoRa technology:
Doppler robustness, scalability, and
coverage
Juha Peta
¨
ja
¨
ja
¨
rvi
1
, Konstantin Mikhaylov
1
, Marko Pettissalo
2
,
Janne Janhunen
1
and Jari Iinatti
1
Abstract
The article provides an analysis and reports experimental validation of the various performance metrics of the LoRa
low-power wide-area network technology. The LoRa modulation is based on chirp spread spectrum, which enables use
of low-quality oscillators in the end device, and to make the synchronization faster and more reliable. Moreover, LoRa
technology provides over 150 dB link budget, providing good coverage. Therefore, LoRa seems to be quite a promising
option for implementing communication in many diverse Internet of Things applications. In this article, we first briefly
overview the specifics of the LoRa technology and analyze the scalability of the LoRa wide-area network. Then, we intro-
duce setups of the performance measurements. The results show that using the transmit power of 14 dBm and the high-
est spreading factor of 12, more than 60% of the packets are received from the distance of 30 km on water. With the
same configuration, we measured the performance of LoRa communication in mobile scenarios. The presented results
reveal that at around 40 km/h, the communication performance gets worse, because duration of the LoRa-modulated
symbol exceeds coherence time. However, it is expected that communication link is more reliable when lower spreading
factors are used.
Keywords
Internet of Things, low-power wide-area network, LoRa, mobility, performance, experiment
Date received: 6 September 2016; accepted: 18 February 2017
Academic Editor: Seong-eun Yoo
Introduction
The low-power wide-area networks (LPWANs) repre-
sent a new trend in the evolution of telecommunication
designed to enable broad range of Internet of Things
(IoT) applications. In contrast to the existing and per-
spective communication technologies (e.g. fourth gener-
ation (4G) or fifth generation (5G)), the high data rate
for each device is not considered to be the most impor-
tant design factor for LPWANs. Instead, the data rates
in LPWANs are intentionally kept low and are traded
for long communication ranges. The other critical
design metric for LPWANs is the energy efficiency,
since many of end devices are expected to be powered
by a battery or even with energy harvesting. Finally,
the high network capacity and low hardware complex-
ity of an end device are also important to keep the cost
of a device low. Although these features inevitably limit
1
Centre for Wireless Communications, University of Oulu, Oulu, Finland
2
Nokia, Oulu, Finland
Corresponding author:
Juha Peta
¨
ja
¨
ja
¨
rvi, Centre for Wireless Communications, University of
Oulu, Erkki Koiso-Kanttilan Katu 3, 90570 Oulu, Finland.
Email: juha.petajajarvi@oulu.fi
Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 4.0 License
(http://www.creativecommons.org/licenses/by/4.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (http://www.uk.sagepub.com/aboutus/
openaccess.htm).
the range of LPWAN applications (by excluding, for
example, the ones requiring data-hungry media stream-
ing), the number of applications which will benefit from
using these technologies is really tremendous. To give
few practical examples, the LPWANs suit quite well the
requirements of, for example, smart metering applica-
tions (e.g. gas, water, electricity, and garbage), civil
engineering and infrastructure monitoring (e.g. tunnels,
bridges, and buildings), and environment conditions
(pollution and climate). Besides, LPWANs can be
employed for tracking of vehicles (e.g. cars, bicycles,
and motorcycles), and monitoring peoples’ well-being.
Figure 1 depicts few potential IoT example use cases
for LPWANs.
Although the LPWANs have much in common with
the traditional wireless sensor networks (WSNs), there
are few critical differences, especially when this comes to
the requirements for networks and end devices. The first
and the major difference is that unlike the traditional
WSNs which usually employ mesh topology, the state-
of-the-art LPWAN technologies require setting up the
gateways (referred to as concentrators or base stations,
depending on the terminology) to serve an end device.
The end devices communicate directly to one or more
gateways as shown in Figure 1. Depending on the tech-
nology, the coverage area of a single gateway may range
from hundreds of meters to tens of kilometers and may
include thousands or even millions of end devices.
Over the last few years, LPWAN technologies have
drawn a lot of attention in the media due to the large
investments from private sector. Today, several com-
peting technology providers are actively trying to gain
ground of the global markets. For example, Sigfox
1
acts
as both a service and a technology provider for
LPWAN and covers multiple countries of the central
Europe and many countries are under roll out such as
USA, Australia and Finland. The second major player
is the LoRa Alliance,
2
which was officially established
in 2015, and which stands after the LoRa technology.
Also, the start of deployment of the first Weightless net-
works based on the technology handled by the
Weightless special interest group
3
has been recently
announced.
4
In addition to these LPWAN technologies,
the traditional telecom industry is also driving toward
IoT. The long-term evolution for machine-to-machine
(LTE-M) and the narrowband IoT (NB-IoT) have
recently been shaped in the Release 13 standard by the
3rd Generation Partnership Project (3GPP).
5
This will
bring further optimizations for device cost, battery life-
time, and coverage. Namely, reduced transmit power in
addition to power spectral density improvement and
allowing, for example, higher error rate or longer acqui-
sition time are expected to enable enhanced coverage
and energy efficiency. Also part of Release 13, Global
System for Mobile Communications (GSM)–Enhanced
Data rates for GSM Evolution (EDGE) radio access
network will be standardized as an Extended Coverage
GSM solution
6
that supports over 10 km range. In gen-
eral, future cellular IoT can be seen having benefits
from large number of vendors and operators.
Moreover, there are numerous technologies featuring
similar characteristics such as WAVIoT, Nwave,
Telensa, Cyan’s Cynet, Accellus, SilverSpring’s
Starfish, and Ingenu/On-Ramp.
7
Figure 1. In the future IoT applications, infrastructure, people, trash bins, bicycles, cars, etc. are possibly monitored with LPWAN
technologies, such as LoRa.
2 International Journal of Distributed Sensor Networks
In this article, we focus on the performance of the
LPWAN and, namely, the LoRa radio technology. The
major contribution of this article includes three aspects.
First, we analyze the performance of LoRa modulation
for mobile use case when the end device’s radio signal is
affected by Doppler. Second, we characterize the capac-
ity of a LoRa wide-area network (LoRaWAN) for sev-
eral illustrative use-case scenarios. Third, we report
results of the field trial measurements conducted using
commercial LoRa end devices. The presented results
provide an insight on the real-life performance of the
technology.
Since the LPWAN concept in general and the LoRa
technology in particular are quite new, they have not
got much of attention from the academic community
yet. Overview of nine different technologies has been
done in Sanchez-Iborra and Cano,
8
while an overview
focusing specifically on LoRa and Sigfox is introduced
in Nolan et al.
9
In Reynders et al.,
10
the performance of
the LoRa-like spread spectrum and Sigfox-like ultra
narrowband (UNB) technologies are compared under
interference. The simulation results revealed that for
the applications which require higher throughput at rel-
atively short range, the spread spectrum approach suits
better than UNB. For the applications requiring longer
range, the UNB systems outperform the spread spec-
trum ones. The quality of service of the LoRa was mea-
sured with fixed 3-km range in Petric
´
et al.
11
Augustin
et al.
12
focused on the LoRa technology by introducing
the network architecture, physical (PHY), and medium
access control (MAC) layers. The LoRa outdoor cover-
age was addressed in Peta
¨
ja
¨
ja
¨
rvi et al.
13
We used the
lowest bandwidth (i.e. 125 kHz) and the maximum
spreading factor (SF) possible (i.e. 12) in our experi-
ments. With these settings, we observed the communi-
cation ranges of over 15 km on ground and almost
30 km on water with 14 dBm transmit power. We mea-
sured indoor coverage at the University of Oulu pre-
mises with the same settings in Peta
¨
ja
¨
ja
¨
rvi et al.
14
The
results showed that the entire campus area can be cov-
ered with an average success delivery ratio of almost
97%. Even from a cellar, almost 95% of the packets
were delivered successfully. We have also studied the
scalability and capacity of the LoRa technology in
Mikhaylov et al.
15
In Wendt et al.,
16
the performance
and indoor through-obstacle penetration of the LoRa-
like modulation in the 2.4 GHz frequency band are
studied. Although the results of this study require adap-
tation (since LoRaWAN operates exclusively in sub-
GHz bands), some of the witnessed effects and made
conclusions can be valid also for the lower frequency
bands. This article introduces novel findings which are
tied together with results from our previous works
introduced in Peta
¨
ja
¨
ja
¨
rvi et al.
13
and Mikhaylov et al.
15
in order to get a more comprehensive picture of the
LoRa technology and its application possibilities.
This article is organized as follows. In section ‘‘The
LoRa technology,’’ we provide a brief introduction and
overview of the LoRa technology. In section ‘‘Analysis
of the LoRaWAN performance,’’ we give the theoreti-
cal analysis of the performance of LoRa modulation
and LoRaWAN performance, namely, the robustness
against Doppler effect, throughput, and network
capacity. In section ‘‘Experimental measurements and
results,’’ we present the results of the real-life measure-
ments illustrating the practical performance of the tech-
nology. Finally, in section ‘‘Discussion,’’ we conclude
the article with discussion about the obtained results.
The LoRa technology
Technically, the LoRaWAN specification
17
includes
three major components, namely, the PHY layer, the
link layer, and the network architecture.
PHY layer
The communication between an end device and a gate-
way is handled in the different sub-GHz frequency
bands depending on the local frequency regulations. In
this article, we address specifically the operation in the
EU industrial, scientific, and medical (ISM) 868 MHz
band. For this band, the LoRaWAN specification
enables eight PHY options. Six of them are based on
LoRa modulation with SF between 7 and 12 and with
bandwidth of 125 kHz. One option is based on
250 kHz bandwidth and SF of 7 LoRa modulation,
and the eight option is Gaussian frequency-shift keying
(GFSK) with 50 kbps data rate.
The LoRa modulation is based on chirp spread
spectrum (CSS) scheme that uses wideband linear
frequency-modulated pulses whose frequency decreases
or increases over a specific amount of time based on
the encoded information.
18
The use of high bandwidth-
time product makes the radio signals resistant against
in-band and out-of-band interferences, while the use of
sufficiently broadband chirps enables to improve
robustness against multipath fading.
19
This results in
the maximum link budget of about 157 dB, which
enables to achieve long communication ranges or to
reduce the transmit power, thus saving the energy of
the end devices. The used modulation scheme is also
expected to help mitigating the Doppler effect.
Furthermore, LoRa modulation includes a cyclic error-
correcting scheme, which improves the communication
robustness by adding redundancy.
19
To improve the
spectral efficiency and increase the network capacity,
LoRa modulation features six different data rates
resulting from orthogonal SF codes. This enables mul-
tiple access method on the same channel
19
without
degrading the communication performance.
Peta¨ja¨ja¨rvi et al. 3
Link layer
The MAC protocol in 1.0 version of the LoRaWAN
specification
17
defines that end devices access the
medium for transmitting their packets in a pure
ALOHA fashion. The MAC layer also defines three
options for scheduling the receive window slots for
downlink communication, which are named as classes
A, B, and C. The end device must have a support for
class A, but support for classes B and C is optional. As
shown in Figure 2, two receive windows are opened
after each uplink transmission in class A. In addition
to the two receive slots after each uplink frame, in class
B, an extra receive window is open at scheduled times.
To have a support for class B, gateway periodically
transmits beacon packets for providing the time refer-
ence and maintaining the synchronization. Class C
devices stay in receive mode unless they are transmit-
ting. In this article, we consider explicitly the end
devices of class A.
Network architecture
The LoRaWANs typically employ a star-of-stars topol-
ogy where the gateways relay data messages between
the end devices and the network server as shown in
Figure 1. The important feature of the LoRa technol-
ogy, named adaptive data rate (ADR), resides in the
network server. The ADR allows adapting and opti-
mizing the data rate for the static end devices. Mobile
end devices should use fixed data rate since mobility
can cause significant temporal variations for the radio
channel characteristics.
17
However, in many mobile
applications, the end devices are actually static most of
the time that makes possible for them to request the
network server to optimize data rate. For example, an
end device mounted to a trash bin’s lid is moved when
trashes are put there or it is being emptied, but remains
static rest of the time. Another important component
of the network server is a mechanism used to filter out
the redundant packets. Since the technology does not
employ any handover method, a single packet trans-
mitted by an end device may get received by several
gateways, each of which will forward such a packet to
the server. Although this technical solution inevitably
introduces redundancy in respect to the backbone com-
munication, it enables to eliminate the handover-related
signaling, thus bringing some energy savings. The net-
work server is also responsible for security, diagnostics,
and acknowledgements.
20
Analysis of the LoRaWAN performance
In this section, we discuss how robust the LoRa tech-
nology is against Doppler effect. Also, the throughput
and the network capacity of the LoRaWAN are ana-
lyzed and discussed.
Doppler effect
It is well known that when a source of a wave is moving
relative to an observer, the observer receives a fre-
quency which differs from the one radiated. The differ-
ence depends on the source’s movement direction and
velocity. When this comes to wireless communications,
this effect may hamper the correct reception of the
signal.
Let’s assume that a chirp signal is transmitted from a
moving end device, which is given by
21
s(t )=
A(t) cos v
0
+ v
D
ðÞt +
mt
2
2
hi
, T=2\t\T=2
0, elsewhere
(
ð1Þ
where A is the amplitude of the signal, v
0
is the angular
carrier frequency, v
D
is the angular frequency shift
caused by Doppler effect, t is time, m is the chirp rate,
and T is the duration of the chirp. This CSS signal is
called up-chirp when frequency linearly increases (m
. 0) and down-chirp when frequency decreases
(m \ 0).
The frequency shift due to the Doppler effect causes
the autocorrelation peak on the receiver to shift in time.
The time shift can be calculated as v
D
/m.
21
If the chirp
rate is large, the time shift is so small that it can be
neglected. This makes CSS to perform well in the pres-
ence of Doppler effect. However, the LoRa technology
provides long-range communication link at the cost of
data rate. This inevitably has an impact on the chirp
Figure 2. Communication phases of a class A LoRaWAN device.
4 International Journal of Distributed Sensor Networks
rate. With a low chirp rate, the time shift is increased,
which makes receiving packets correctly more difficult.
Another approach to analyze performance of the
LoRa technology is to compare coherence time (T
c
)
and symbol time (T
s
). Coherence time is inversely pro-
portional to Doppler shift as
T
c
=
2p
v
D
ð2Þ
If T
s
. T
c
, fast fading occurs due to the Doppler
effect, which leads into signal distortion.
22
Symbol time
in LoRa modulation can be calculated as
19
T
s
=
2
SF
BW
ð3Þ
where SF is the spreading factor and BW is the band-
width. As can be noticed from equation (3), T
s
doubles
when SF is increased by one, given that the bandwidth
does not change.
In order to see when fast fading occurs, coherence
time with center frequency of 868 MHz at different velo-
cities is shown in Figure 3 along with periods of LoRa-
modulated symbols with different SFs and bandwidths.
When the velocity is under 38 km/h, the T
c
is larger than
T
s
with shown SFs. At 38 and 76 km/h, T
c
and T
s
curves with SF = 12 and SF = 11 cross, respectively.
Therefore, the LoRa technology might experience
packet losses at relatively low velocities with these SFs.
Lower SFs can tolerate higher speeds.
End device data rate
According to Semtech,
23
the duration of a LoRaWAN
frame is composed of a preamble and the actual packet
payload and is given by
T
LoRa
= T
preamble
+ T
packet
=
1
R
s
n
preamble
+ SW + max ceil
8PL 4SF + 28 + 16CRC 20IH
4(SF 2 DE)
(CR + 4), 0
ð4Þ
for LoRa modulation and by
T
GFSK
= T
preamble
+ T
packet
=
8
R
GFSK
L
preamble
+ SW + PL + 2CRC
ð5Þ
for GFSK, where R
s
is the symbol rate, n
preamble
=
12.15 is the number of preamble symbols for a LoRa-
modulated packet, L
preamble
= 5 bytes in GFSK, SW
is the length of synchronization word (SW = 1 byte
for LoRa and 3 bytes for GFSK), PL is the number of
payload bytes, SF is the spreading factor, CRC speci-
fies the presence of payload cyclic redundancy check.
CRC = 1 when enabled and zero otherwise. IH indi-
cates the operation mode. IH = 0 in explicit mode and
IH = 1 in implicit mode. For LoRa modulation, DE
stands for data rate optimization which introduces
overhead to increase robustness to reference frequency
variations over the timescale of the LoRa frame (man-
datory for SFs exceeding 10 at 125 kHz bandwidth).
DE = 1 when the optimization is enabled and DE =0
otherwise. CR is the coding rate and ranges from 1 to
4. R
GFSK
is GFSK-modulated data rate (50 kbps). max
denotes the function returning the maximum of the two
arguments in the brackets separated by the comma,
and ceil designates the function mapping a real number
argument to the smallest following integer.
Based on the frame formats defined in Semtech,
23
the length of the PHY layer payload in bytes is given
by
PL = MHDR + MAC
payload
+ MIC
= MHDR + FHDR
ADDR
+ FHDR
FCTRL
+ FHDR
CNT
+ FHDR
OPTS
+ F
port
+ FRM
payload
+ MIC
= 12 + FHDR
OPTS
+ F
port
+ FRM
payload
ð6Þ
where MHDR = 1 is the length of MAC header;
FHDR
ADDR
= 4 is the length of the frame header
(FHDR) address field; FHDR
CTRL
= 1 and
FHDR
CNT
= 2 are the lengths of the FHDR’s frame
control and frame counter fields, respectively;
FHDR
OPTS
is the optional FHDR field’s length;
F
port
= 1 is the port identifier for an application spe-
cific; and MIC = 4 is the message integrity code.
Figure 3. Comparison of the coherence time and symbol
times for LoRa signals with different spreading factors.
Peta¨ja¨ja¨rvi et al. 5