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

net/publication/320057356

LoRa and LoRaWAN testbeds: A review

Conference Paper · September 2017

DOI: 10.1109/AFRCON.2017.8095703

CITATIONS READS

30 4,297

3 authors:

Jaco Morné Marais Reza Malekian

University of Pretoria Malmö University
6 PUBLICATIONS   63 CITATIONS    179 PUBLICATIONS   1,886 CITATIONS   

SEE PROFILE SEE PROFILE

Adnan M. Abu-Mahfouz
Council for Scientific and Industrial Research, South Africa
144 PUBLICATIONS   2,210 CITATIONS   

SEE PROFILE

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

Water Distribution System Modelling, State Estimation, and Parameter Design View project

Special Issue (MDPI Sensors) : Sensors and Modern Technologies for Road, Robotic, and Intelligent Vehicle View project

All content following this page was uploaded by Jaco Morné Marais on 08 November 2017.

The user has requested enhancement of the downloaded file.

 LoRa and LoRaWAN Testbeds: a Review
Jaco M. Marais and Reza Malekian Adnan M. Abu-Mahfouz
Electrical, Electronic and Computer Engineering, Meraka Institute, CSIR
University of Pretoria, Pretoria, 0002, South Africa Pretoria, 0184, South Africa

Abstract—The Internet of Things (IoT) is a fast-growing developing their own products, for example NB-IoT, eMTC
movement turning devices into always-connected smart devices and EC-GSM which will use licensed cellular frequencies.
through the use of communication technologies. This facilitates Currently some communication standards, those based on
the creation of smart strategies allowing monitoring and opti-
mization as well as many other new use cases for various sectors. IEEE 802.15.4 such as ZigBee, use a mesh network topology
Low Power Wide Area Networks (LPWANs) have enormous to extend their network coverage, as links between nodes are
potential as they are suited for various IoT applications and each limited to approximately 100 meters [3]. The technologies
LPWAN technology has certain features, capabilities and limi- competing in the LPWAN market however, use a star topology
tations. One of these technologies, namely LoRa/LoRaWAN has as greater receiver sensitivities allow for coverage in the
several promising features and private and public LoRaWANs
are increasing worldwide. Similarly, researchers are also starting kilometres range. This increased coverage comes however at
to study the potential of LoRa and LoRaWANs. This paper the cost of reduced data rates. Using the star topology greatly
examines the work that has already been done and identifies flaws simplifies network management and reduces the required net-
and strengths by performing a comparison of created testbeds. working firmware for end devices [3].
Limitations of LoRaWANs are also identified. All of the LPWAN competitors claim that their approach
Keywords—LoRa, LoRaWAN, LPWAN, Internet of things.
is the optimum solution for this new generation of IoT.
Depending on the use case some connectivity options will be
I. I NTRODUCTION more suited than others but an in-depth analysis is required to
ensure that the correct design choices are made. An evaluation
The IoT bring the promise of a world filled with smart of LoRa and LoRa Wireless Area Network (LoRaWAN) will
cities, constant tracking of objects and everyday life filled with provide insight into the advantages and disadvantages of this
wireless devices to improve every aspect of our existence. The technology.
IoT movement creates the need for new wireless technologies,
capable of supporting the large numbers of devices found in II. L O R A AND L O R AWAN
the IoT space [1]. These low-cost devices can be considered Some of the LPWAN offerings are mainly proprietary but
to form a Low Power Wide Area Network (LPWAN) and the LoRa Alliance develops LoRaWAN as an open standard.
require a multiyear lifespan and are deployed in large amounts. The physical layer (LoRa) was however developed by Semtech
Furthermore, these devices may be located underground, un- which remain the sole LoRa integrated circuit producer [4].
derwater or deep inside buildings [2]. New wireless tech- LoRaWAN is the communication protocol (ALOHA-based)
nologies must therefore have sufficient penetration and range [5] and system architecture for a network using the LoRa
capabilities. Security of these devices and networks are also physical layer [6]. LoRaWAN is not the only communication
a major concern as more and more critical infrastructure are protocol that uses LoRa as the physical layer: Symphony
added to the IoT. Most of these sensors are battery powered. LinkTM and LoRaBlink are other examples. LoRaBlink [7]
Therefore, energy efficiency and low power consumption is adds multi-hop support while Symphony Link offers guar-
also a major concern. anteed Acknowledgements (ACKs), over the air firmware
Previously Machine to Machine (M2M) networks operated updates and many other features. The DASH7 stack can also
on cellular networks in the form of 2G but 2G is now being be configured to use LoRa as its physical layer and can
phased out by some cellular operators to make room for newer potentially run side-by-side with a LoRaWAN stack [8].
technologies (Long-Term Evolution (LTE)) [1]. These new
technologies offer significantly increased Bandwidth (BW) but A. Physical layer (LoRa)
come at a cost of higher power consumption. For the M2M LoRa is a derivative of Chirp Spread Spectrum (CSS)
use case the higher bandwidth capabilities are unnecessary and modulation with integrated Forward Error Correction (FEC)
the increase in power consumption renders these technologies [9]. LoRa uses sub one GHz ISM bands in Europe and
unsuitable. To fill this gap several competitors such as SigFox, North America and its wide band nature allows LoRa to
Random Phase Multiple Access (RPMA), NB-Fi, Weightless, better compensate for a low Signal to Noise Ratio (SNR)
DASH7 and Long Range (LoRa) now compete for market [10]. This allows LoRa to demodulate signals even when
share in the LPWAN domain. The cellular industry is also they are 19.5 dB below the noise floor [11]. CSS allows for
2017
c IEEE. http://ieeexplore.ieee.org.uplib.idm.oclc.org/document/8095703/
a longer communication range than Frequency-Shift Keying copies of a message and replies using the optimum gateway
(FSK) without an increase in power consumption [5]. [5], [6].
Transmitting at higher power levels will increase a LoRa LoRaWAN uses Adaptive Data Rate (ADR) and a mul-
node’s range. Nodes can adjust their output power to meet tichannel multi-modem transceiver in gateways. This allows
regulatory requirements. In Europe +14 dBm is the maximum for simultaneous message capturing on multiple channels [6].
transmit power except in the G3 band (+27 dBm) [5]. Different spreading factors allows multiple different data rates
LoRaWANs deployed in Europe have channel bandwidths on the same channel simultaneously [5]. Nodes with good
of either 125 kHz or 250 kHz and a single FSK transmission link capabilities should use higher data rates to decrease their
channel providing a higher data rate is also available [6]. Data transmit times, allowing other nodes to use larger spreading
rates are region (regulatory restrictions) as well as Spread- factors.
ing Factor (SF) dependent. Increasing the spreading factor LoRaWAN defines three device classes (A, B, C). All end
improves the SNR but results in longer transmission times devices must support Class A but the other classes are optional
[7]. Using a higher bandwidth shortens the transmission times [14], [15]. Class A devices transmit as needed and create two
but reduces the maximum receiver sensitivity [11]. Capacity downlink windows shortly after transmission. Class B devices
calculations performed in [12] revealed that when a single are similar to class A devices but add scheduled receive slots
gateway must serve many devices the majority of them should (scheduling is enabled by beacon broadcast from gateways).
be close to the gateway (SF = 7) as only a few nodes with the Class C devices continuously listen when they are not transmit-
maximum SF can be supported given their long transmission ting. Choosing the optimal class is vital for applications with
times. response time or minimum power consumption requirements.
LoRa uses FEC to allow the recovery from transmission
III. T ESTBED EVALUATION METRICS
errors due to bursts of interference, but the use of FEC
adds some encoding overhead [13]. LoRa’s coding rates are When deciding if a LoRaWAN will be the most suitable
4/(CR + 4) with CR ∈ {1, 2, 3, 4}. A LoRa packet’s header solution the capabilities of this technology should be under-
and its Cyclic Redundancy Check (CRC) will always be stood. These capabilities can be found through simulation or
transmitted using a CR of 4/8 and the payload with its optional through testbeds. Before this can commence a clear definition
CRC at the chosen coding rate [13]. of evaluation metrics is required to accurately compare results.
When LoRa is transmitting with a BW of 125 kHz and a Suggested LoRa and LoRaWAN testbed evaluation metrics
SF of 11 or 12 a low data rate optimization can be enabled. can be viewed in Table I . Furthermore, the following should
This reduces the impact on transmission due to drift in the be kept in mind: the SNR at a node will have an impact of
reference frequency of the oscillator, but does add additional the Bit Error Rate (BER) of the node which in turn impacts
data overhead [13]. the Packet Error Rate (PER) [16]. A chosen SF will have an
LoRa can detect channel activity using Carrier Activity impact on the maximum bit rate, maximum payload size and
Detection (CAD) [7]. This is faster than Received Signal the amount of time a node is required to be off air before
Strength Indicator (RSSI) identification and can differentiate transmitting again [14].
between noise or a desired LoRa signal [10]. In addition to these metrics there are additional metrics that
are implementation or use case specific. These are listed in
B. LoRaWAN Table II.
LoRaWANs in Europe are limited to 10 channels, has
IV. L O R AWAN TESTBEDS
duty cycle restrictions but no channel dwell time limitations.
LoRaWANs in North America have 64 channels, also have A. Outdoor
duty cycle restrictions but no channel dwell time limitations Free space measurements with a BW of 250 kHz took place
[5]. LoRaWAN has 3 common 125 kHz channels for the in Germany and recorded RSSI, SNR and PER for various
868 MHz band namely 868.10, 868.30 and 868.50 MHz distances and payload lengths of 10, 50 and 100 bytes [16].
that devices use to join the network [14]. Once a node has With the SF fixed at 10 the results showed that for a 10 byte
joined the network, the network server can provide additional payload the PER remained zero up till 8 km. The 50 byte
channels to the device. In Europe, the same channels are used payload resulted in invalid packets from 2.3 km with the PER
for uplink and downlink. changing inconsistently as the distance increased. In contrast
The network architecture is a star of stars topology in which with the 50 byte payload the 100 byte payload had a near zero
end nodes connect directly communicate with gateways which PER up until 6.6 km. The chosen bandwidth value of 250 kHz
in turn connect to a central network server [4]. Gateways are differs from the LoRaWAN specifications (LoRaWAN mainly
always on devices and have LoRa capabilities and potentially uses 125 kHz for these channels) [14].
Ethernet or cellular capabilities to connect to the network The water level of troughs is monitored using a 915 MHz
server. LoRaWAN [17]. The troughs are placed on the ground and
In a LoRaWAN transmissions are received by any nearby connectivity with the gateway is negatively affected by objects
gateway(s). This allows mobile nodes to transmit to any in the Fressnel zone. Therefore, distance tests were conducted
gateway without any handover. The network server drops any in which RSSI levels for various hub and node heights and
 TABLE I Connectivity tests was successful up to 2.2 km in one direction
T ESTBED E VALUATION M ETRICS . but only 1.6 km in another due to a hill [6].
Metric Explanation
In Finland LoRa’s long range capabilities was tested using
a mobile node (car and boat trips) with a PCB antenna [18].
Number
of
A LoRaWAN with only one sink will operate differently The received signal strength, GPS coordinates and resultant
than one with multiple sinks. packet loss was recorded for locations up to 30 km. The
gateways
Number Increasing the number of nodes demands higher perfor- node had a SF of 12 and randomly chose between 6 channels
of nodes mance of the LoRaWAN. when transmitting and transmit power was limited to 14 dBm.
The RSSI measured by a node provides an indication Measurements by car showed that even close to the gateway
RSSI of the coverage range of the gateway and how reliable
communication will be.
(≤ 2 km) a 12 % packet loss ratio as experienced. Path loss
The SNR experienced by a node also provides an indication
exponents was also calculated from the resulting data.
SNR Outdoor testing of 6 LoRa nodes in a 1.5 ha area showed
of how reliable communication will be.
Distance Distance measurements can be combined with connectivity a potential lifespan of 2 years on 2 AA batteries [7]. The
from measurements (RSSI, SNR, PER) to determine a gateway’s authors implemented their own Media Access Layer (MAC)
gateway effective range.
protocol (LoRaBlink) to enhance LoRa by adding multi-hop
Payload Ideally network performance was evaluated for multiple
size payload sizes as these impacts transmission times.
capabilities. Evaluations revealed a relatively high packet loss
Transmit Transmitting at higher power levels (dBm) can achieve
of 20 % for their now short range multihop LoRa network.
power greater range but regulatory restrictions apply. Transmissions can be sent on different frequency channels
Network reliability can be defined in several ways: num- and on different spreading factors. Transmissions with dif-
Network
reliability
ber of packets received successfully, the number of valid ferent spreading factors are mostly orthogonal and can still
received packets or examining the PER.
be decoded successfully [7]. This allows gateways to receive
Power
consump-
Power consumption values for the motes can be used to multiple transmissions from end devices simultaneously. An
estimate the network’s lifetime. investigation into LoRa’s CAD found that the detection prob-
tion
abilities are more than 97 %. Furthermore, if two transmissions
sent with the same parameters are received, the strongest
TABLE II of the two will with a strong probability still be recovered
OTHER SITUATIONAL EVALUATION METRICS . [7]. However, false positives increase when different LoRa
Metric Explanation networks are active on different SF/BW combinations [7].
Number of floors or For indoor LoRaWANs this provides an indica- In Padova (Italy) LoRa’s coverage distance for various SF
walls between gate- tion of the penetration capabilities of the network levels was evaluated [3]. The gateway was placed in a worst-
way and node. when combined with e.g. RSSI. case scenario of no antenna gain, antenna elevation and taller
Single or multi-hop LoRaWAN is by default a single hop network but
network multi-hop features can be added (see [7]). buildings surrounding the gateway. Up to 2 km range was
Options are A, B or C. The class used defines achieved (SF = 12) but the researchers felt that 1.2 km should
LoRaWAN class how communication takes place. By default, Lo- be considered as the coverage value to compensate for any
RaWANs are class A.
Are nodes set to use
link budget variations. Calculating the achievable theoretical
When nodes can choose less congested channels, amount of LoRa nodes per person in Padova revealed 2 nodes
a specific channel
the amount of nodes supported increase.
or can they choose. per person when 30 gateways was used (gateway capacity was
set to 15 thousand nodes per gateway) [3].
antenna polarization were recorded. A 100 node network LoRa mobile connectivity tests was conducted on a univer-
was simulated in a lab environment by increasing the data sity campus [19]. A researcher with a PCB antenna LoRa node
transmission rate of 5 nodes, resulting in a 17 % drop of valid navigated his daily routes on campus. Apart from entering an
packets (payload was 26 bytes) [17]. Care should be taken anechoic chamber the LoRa node had 96.7 % connectivity
when comparing their results with others as the regulatory (95 % when moving). A spreading factor of 12 was used
requirements for US and Europe ISM bands differ (no duty and the node randomly chose between six channels when
cycle restrictions in the US band). transmitting [19]. Acknowledgements were not required for
An outdoor evaluation of LoRa and LoRaWAN using 3 uplink messages and no retries were made when transmissions
gateways was performed in Glasgow [6]. Outdoor testing were unsuccessful. The average path loss for different areas on
recorded GPS location, RSSI and reliability of receiving an campus was calculated. One limitation of this test was that one
Acknowledgement (ACK) after transmitting. Reliability test- node was used and the coverage results may differ when using
ing revealed that the gateway’s cellular network connections multiple or different manufacturer’s nodes. The researchers
experience periods of disconnect, due to the cellular network’s point out that using the maximum spreading factor results in
inactivity policies. Continuously pinging the gateways raised long transmission times, leading to limited link availability
the LoRaWAN connectivity rate from 70 % to 95.5 % [6]. In and scalability issues.
2.5 % of attempts the data was received but the LoRa node Line of sight measurements (9.75 km) for a 2.45 GHz LoRa
did not receive an ACK. system resulted in 81.58 % correctly received packets [20].
A range test with 2 LoRa transceivers was also conducted. The SF was set to 12 and when the antennas were misaligned
50 % of packets were still received correctly. The penetration A private LoRa network consisting of 1 gateway and 32
of LoRa through a salty water mixture (0.77 % salt) was also nodes are part of an energy saving program for a building [3].
tested and 94.5 % of packets was correctly received through a Switching to a LoRa network removed the need for repeaters
10-12 cm water layer. When the width of the water layer was and allowed a single gateway on floor 9 to provide coverage
increased all communication promptly stopped. The potential for all 19 floors of the building. This installation demonstrates
of LoRa’s use in safety systems was evaluated in which LoRa LoRa’s penetration abilities as the researchers noted that
provides redundant links via 868 MHz and 2.45 GHz LoRa some nodes could be placed in areas where connectivity had
modules [20]. previously been poor [3].
LoRa’s network coverage in urban environments (Paris) was LoRa’s power consumption as well as penetration capabili-
tested using SFs of 7, 9, 12 and a transmission power of 14 ties were evaluated by varying the spreading factor [20]. Lower
dBm [21]. Coverage was achieved for distances of up to 3.4 SFs are better for single reinforced concrete wall penetration,
km and at 3.4 km a SF of 12 resulted in a ≈ 38 % packet but a SF of 12 enabled transmissions to penetrate 3 walls with
delivery ratio. The closest test point (650 m) revealed that 33 % of packets received.
100 % delivery can be achieved with a SF of 12 and this The RSSI levels for packets being transmitted from an
ratio drops down to ≈ 84 % when a SF of 7 is used. As outdoor end device to an indoor gateway was recorded [21].
acknowledgements and retransmission were disabled, higher The experiment used a transmit power of 2 dBm which
success rates are possible when these are enabled. resulted in a maximum range of 100 m and RSSI levels did
not decrease as the SF was increased. It was noted that lower
B. Indoor RSSI level packets had a large Signal-To-Interference-plus-
An indoor LoRaWAN deployment [14] evaluates LoRa’s Noise Ratio (SINR) [21].
data throughput capabilities when considering the European 1
% duty cycle limitation. In maximum interference conditions C. Simulations
(SF of 12) end devices must wait two and a half minutes As LoRa and LoRaWANs are new technologies a demand
before two consecutive transmissions even when sending an exists for suitable network simulation programs to better
empty payload packet. Using data rate 6 for this empty packet predict network coverage before a network is deployed. A
reduces the required off air time to 2 seconds [14]. The indoor LoRa simulator was created named LoRaSim to attempt to
network also measured RSSI and SNR levels to evaluate address this problem [11]. This simulator can evaluate the
building coverage for a single gateway scenario. ratio of received to transmitted messages for a period and can
Tests show RSSI levels are not influenced by the data evaluate a network’s energy consumption (estimation). It also
rate but that the SNR is however influenced [14]. Repeated has support for single and multiple sink scenarios. The creators
transmission tests at date rate 2 found that nodes close to found that only 120 nodes can be supported by a single sink
the gateway had frame duplications and received packets with (gateway) when a SF of 12 is used with nodes sending 20
bad CRCs. Packet loss versus data rates comparisons shows bytes every 16.7 minutes [11].
big differences between the rates and decreases in error with The collision rate in a LoRaWAN under load was simulated
increases in distance. Therefore, special attention should be for different packet sizes [21]. A collision was defined as
taken when nodes are in basements to ensure they have two packets having overlapping transmission times. Experi-
sufficient coverage [14]. ments performed in [7] show however that LoRa can receive
An indoor evaluation of LoRa and LoRaWAN was per- concurrent transmissions (to a degree). Simulations showed
formed in Glasgow in Scotland [6]. Brief indoor testing results similar to those of pure ALOHA with a maximum
evaluated RSSI by noting the quality of reception in stairwells capacity use of 18 % for a link load of 0.48. However 60
and central corridors, as well as number of floors between the % of packets are dropped resulting in retransmissions for any
node and the gateway. messages requiring acknowledgements [21].
LoRa’s indoor penetration abilities was tested in Prague [2].
LoRa’s reach inside the building was measured by recording D. Comparison of testbeds
a node’s RSSI levels for various locations and floors in the The LoRa and LoRaWAN based testbeds discussed above
building. The impact of placing the gateway on the roof versus are compared in Table III. For the gateway column only true
in the basement was also explored. Ten messages was send gateways were considered, using an end nodes as a single
from each test location and messages had a large payload channel gateway was counted as another node. The table shows
containing info such as identification information, RSSI, SNR that care should be taken when comparing the work presented
and the data rate to name a few. Tests were conducted using a in one paper with the work in another. Directly comparing
IMST iU880A USB adapter LoRa node which does not feature the results are made difficult by some researchers publishing
an external antenna. The output power was reported to be fixed incomplete setup information and the many ways reliability
at 20 dBm but IMST indicates that the maximum output power was measured.
for this product is 18 dBm. An analysis of the recorded RSSI Most of the presented work used only a single gateway
levels showed that placing the gateway on the roof provides and very few nodes. In all the studies the only LoRa network
significantly more coverage than in the basement [2]. present was the one created for the study and no interference
from other LoRa networks were present. Most of the testbeds does however suffer from false positives when multiple SF/BW
evaluated LoRa by only varying one parameter such as dis- combinations are present (multiple networks in an area) [7].
tance, SF or payload size instead of varying multiple aspects. Support for over the air firmware upgrades appear to be
Some researchers consider the validity of received packets limited if not completely unfeasible. Firmware updates are
while others only look at the number of packets received (or restricted by the 1 % duty cycle restriction and class A
lost). Payload size has a substantial impact on performance but nodes are unicast messaging only. The LoRa specifications
many testbeds don’t fully define their payloads. Most testbeds for class B devices leaves setting up multicast groups to the
fixed the transmission power to the maximum value allowed application layer or as part of node personalization [15]. The
by regulations. However, this can be varied and nodes closer Multicast message format requires that no acknowledgement
to the gateway can use less transmission power to improve be requested for multicast messages. Similarly support for over
battery life. the air security updates are lacking, this is a major concern for
security sensitive applications such as critical infrastructure.
V. L IMITATIONS OF L O R AWAN AND FUTURE WORK Semtech announced in a June 2016 press release that native
geolocation has now been added to LoRa. Limited information
Examining LoRaWAN from an Open Systems Interconnec- on this feature is available but the press release states version 2
tion (OSI) model perspective reveals that LoRaWAN is mainly LoRaWAN gateways are required. LoRa supports geolocation
a MAC layer protocol with a set network architecture and the by using timestamps and differential time of arrival algorithms
rest of the stack is left up to the developer [5], [21]. Advanced [1]. Mobile nodes attempting geolocation will receive inac-
features such as a mechanism to auto-fragment payloads that curate results as especially in Europe the 1 % duty cycle
is larger than the maximum size will have to be developed. limitation will restrict the frequency of geolocation attempts.
Gateways can use ADR to control data rates of end devices Geolocation features will increase the amount of supported
but the optimal manner to execute ADR is still unknown [4]. use cases but secure localisation algorithms are required to
Class A devices establish two receiver windows only after preventing fraudulent or malicious nodes from disrupting the
a message is transmitted. By default window two is fixed network.
to use 869.525 MHz and a SF of 12 in Europe [22]. A When devices re-join a network, they revert back to the
potential scenario could arise where a busy gateway is forced default MAC settings [22]. The network server will then
to use the second window to reply to devices, resulting in very have to re-adjust any changes to these settings on every join
long transmit times during which a channel is occupied [4]. procedure. As a result, a major traffic spike for the network
Furthermore, in Europe several of the channels are limited to server will occur if a large amount of devices suddenly have
a maximum of 1% duty cycle also restricting the time on air to re-join/join a network. This could occur when a gateway
for devices. There is a delay window between when a gateway lost power and it served as the only gateway for all nodes in
has received a message and when the first receive window of its coverage area.
the sender node opens. During this time, the gateway could The LoRaWAN protocol allows end devices (Class A)
keep listening to the channel and start receiving a message to decide when to send ACKs for received messages and
from another node [4]. when to receive pending messages [22]. This allows end
LoRa uses different sub 1 GHz bands depending on your devices to ensure they stay within duty cycle limits. However,
region. This results in region-specific configuration problems gateways don’t have an upper bound on when messages will
as end devices must be re-configured or detect their region. be acknowledged or when devices will receive any pending
The LoRaWAN specification suggests methods for end devices message(s). Applications with time critical requirements will
to determine their region. Suggestions include adding GPS need to either carefully design their end node firmware, choose
hardware to end devices, equipping beacons with GPS and a different LoRaWAN class or opt for a different LPWAN
transmitting their coordinates or instructing beacons to trans- technology.
mit a list of join frequencies. Programming beacons to transmit
a list of join frequencies avoids the cost of GPS hardware but VI. C ONCLUSION
requires that beacon nodes be present [15]. There are many challenges and factors to consider when im-
The long-term viability of LoRaWANs remains in question. plementing a LPWAN. Not only are there several technologies
Presently there is only one LoRa manufacturer (Semtech) and and protocols to choose from, comparing these options can be
as the unlicensed bands become more and more congested a difficult. The long-term performance of a network isn’t only
network may no longer provide the performance it originally impacted by the logical and geographical layout of the network
did. Public and private LoRa networks will interfere with one but also by interference from other networks, the ability to
another as especially in urban environments all the end devices perform firmware upgrades and regulatory restrictions (which
will be in close proximity. This is a widespread problem for could change).
all wireless solutions using the unlicensed bands. Usage of LoRa and LoRaWAN has the potential to create viable
protected licenced bands will always have an advantage. This LPWANs but only when this option’s strengths and weakness
problem is alleviated to a degree by LoRa’s configurability of are considered. The major advantage is the variety of state of
SF, BW and transmit power as well as CAD [7]. LoRa’s CAD the art performance improvement mechanisms included which
 TABLE III
T ESTBEDS FOUND IN LITERATURE .

ISM RSSI SNR How was reliability

Gate- Tx BW Distances Payload
Ref. Nodes SF band mea- mea- measured?
ways (dBm) (kHz) covered (km) (bytes)
(MHz) sured? sured?
Conc building [2] 1 1 20 12 125 - Yes - 7 floors large RSSI values
Glasgow CBD [6] 3 2 14 varied - 868 Yes - 2.2 seq. number Record ACKs
LoRaBlink [7] 0 6 17 varied varied - - - 0.342 varied Packet reception rate
Urban coverage [3] 1 1 - varied - - - - 2 - SF needed for coverage
Number of lost packets
Indoor [14] 1 1 14 varied 125 868 Yes Yes 0.5 m to 60 m varied
and packet error
Outdoor [16] 1 1 - 10 250 868 Yes Yes 0.276 to 8.52 10, 50 and 100 PER
Water troughs [17] 0 7 - - - 915 Yes - 0.5 to 2.7 26 % valid packets received
seq. number,
Car and boat [18] 1 1 14 12 125 868 Yes - 15 % packets lost.
status
Campus eval [19] 1 1 14 12 125 868 - - 65 m to 195 m - % received packets
2.45 GHz LoRa 9.75 outdoor,
0 2 3 varied - 2450 - - 21 % valid packets
[20] 30 m indoor
3.4 outdoor, % packets received and
Paris testbed [21] 1 1 2/14 several 125 - Yes - 1, 25, 51
100 m indoor avg. throughput (bytes/s)

enables optimisation for different use cases and contexts. The [10] Semtech, “LoRa Modulation Basics,” pp. 1–26, May 2015.
open nature of the LoRaWAN specifications enables designing [11] M. C. Bor, U. Roedig, T. Voigt, and J. M. Alonso, “Do LoRa Low-
Power Wide-Area Networks Scale?” in Proceedings of the 19th ACM
your application around LoRaWAN’s strengths. At the same International Conference on Modeling, Analysis and Simulation of
time care must be taken before choosing to opt for LoRa and Wireless and Mobile Systems, Malta, Malta, November 2016, pp. 59–67.
LoRaWAN as not all use cases are best supported by this [12] K. Mikhaylov, J. Petäjäjärvi, and T. Hänninen, “Analysis of Capacity and
Scalability of the LoRa Low Power Wide Area Network Technology,”
technology. European Wireless 2016, pp. 119–124, 2016.
Further research into the capabilities, limitations and opti- [13] Semtech, “SX1272/3/6/7/8: LoRa Modem Designer’s Guide,” pp. 1–9,
mal use cases for LoRa and LoRaWAN is required. Creating July 2013, AN1200.13.
[14] P. Neumann, J. Montavont, and T. Noël, “Indoor Deployment of Low-
better simulation tools will provide a better understanding of Power Wide Area Networks (LPWAN): a LoRaWAN case study,” in
the true performance of a potential network. Additionally, 2016 IEEE 12th International Conference on Wireless and Mobile
more comprehensive evaluations using larger testbeds are Computing, Networking and Communications (WiMob), New York, NY,
USA, October 2016, pp. 1–8.
required to better understand the role of LoRa/LoRaWAN in [15] N. Sornin, M. Luis, T. Eirich, T. Kramp, and O. Hersent, “LoRaWAN
the bigger LPWAN space. TM Specification,” pp. 1–70, July 2016, V1.0.2.
[16] M. Aref and A. Sikora, “Free space range measurements with Semtech
R EFERENCES LoRaTM technology,” in 2014 2nd International Symposium on Wireless
[1] ABI Research, “Best Fit Use Cases for LPWANs,” pp. 1–16, August Systems within the Conferences on Intelligent Data Acquisition and Ad-
2016. vanced Computing Systems, IDAACS-SWS 2014, Offenburg, Germany,
[2] L. Gregora, L. Vojtech, and M. Neruda, “Indoor Signal Propagation of September 2014, pp. 19–23.
LoRa Technology,” in 2016 17th International Conference on Mecha- [17] Lukas, W. A. Tanumihardja, and E. Gunawan, “On the application of
tronics - Mechatronika (ME), Prague, Czech Republic, 2016, pp. 1–4. IoT: Monitoring of troughs water level using WSN,” in 2015 IEEE
[3] A. Zanella and M. Zorzi, “Long-Range Communications in Unlicensed Conference on Wireless Sensors, ICWiSE 2015, Melaka, Malaysia,
Bands : The Rising Stars in the IoT and Smart City Scenarios,” IEEE August 2016, pp. 58–62.
Wireless Communications, vol. 23, no. 5, pp. 60–67, October 2016. [18] J. Petäjäjärvi, K. Mikhaylov, A. Roivainen, T. Hänninen, and M. Pet-
[4] D. Bankov, E. Khorov, and A. Lyakhov, “On the Limits of LoRaWAN tissalo, “On the coverage of LPWANs: Range evaluation and channel
Channel Access,” in 2016 International Conference on Engineering and attenuation model for LoRa technology,” in 2015 14th International
Telecommunication, Moscow, Russia, November 2016, pp. 10–14. Conference on ITS Telecommunications, ITST 2015, Copenhagen, Den-
[5] LoRa Alliance Technical Marketing Workgroup , “A technical overview mark, December 2016, pp. 55–59.
of LoRa and LoRaWAN,” November, 2015. [19] J. Petäjäjärvi, K. Mikhaylov, M. Hämäläinen, and J. Iinatti, “Evalu-
[6] A. Wixted, P. Kinnaird, A. Tait, A. Ahmadinia, and N. Strachan, ation of LoRa LPWAN technology for remote health and wellbeing
“Evaluation of LoRa and LoRaWAN for Wireless Sensor Networks,” monitoring,” in International Symposium on Medical Information and
in 2016 IEEE SENSORS, October 2016, pp. 1–3. Communication Technology, ISMICT, Worcester, MA, USA, March
[7] M. Bor, J. Vidler, and U. Roedig, “LoRa for the Internet of Things,” 2016, pp. 1–5.
in Proceedings of the 2016 International Conference on Embedded [20] T. Wendt, F. Volk, and E. Mackensen, “A benchmark survey of Long
Wireless Systems and Networks, Graz, Austria, February 2016, pp. 361– Range (LoRa TM ) Spread-Spectrum-Communication at 2.45 GHz for
366. safety applications,” in Wireless and Microwave Technology Conference
[8] Norair, JP, Haystack Technologies, Inc., “DASH7 Specification-DRAFT (WAMICON), 2015 IEEE 16th Annual, Cocoa Beach, FL, USA, April
16-An Advanced Communication System for Wide-Area Low Power 2015, pp. 1–4.
Wireless Applications and Active RFID,” pp. 1–92, January 2015. [21] A. Augustin, J. Yi, T. Clausen, and W. Townsley, “A Study of LoRa:
[9] B. Reynders and S. Pollin, “Chirp Spread Spectrum as a Modulation Long Range & Low Power Networks for the Internet of Things,”
Technique for Long Range Communication,” in 2016 Symposium on Sensors, vol. 16, no. 9, 2016.
Communications and Vehicular Technologies (SCVT), Mons, Belgium, [22] LoRa Alliance Technical comittee, “LoRaWAN Regional Parameters,”
November 2016, pp. 1–5. July 2016, V1.0.

View publication stats