2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:

From M2M Communications to Internet of Things

1

Measurements, Performance and Analysis of LoRa

FABIAN, a real-world implementation of LPWAN
Tara Petri , Mathieu Goessens , Lout Nuaymi , Laurent Toutain , Alexander Pelov
Telecom Bretagne

{name.surname}@telecom-bretagne.eu

AbstractUp to recently, two main approaches were [1] is one prominent technology of this type. LoRa is a
used for connecting the "things" in the growing Internet Layer 1 Network Protocol designed to work on sub-1GHz
of Things (IoT) - one based on multi-hop mesh net- spectrum (109MHz, 433MHz, 866MHz, 915MHz). As those
works, using short-range technologies and unlicensed frequencies are globally available, without any licences,
spectrum, and the other based on long-range cellular they are good candidates for the Internet of Things (IoT)
network technologies using corresponding licensed fre-
quency bands. New type of connectivity used in Low-
communication needs.
Power Wide Area networks (LPWAN), challenges these LPWANs based on LoRa technology, are still insuffi-
approaches by using low-rate long-range transmission ciently researched and tested. As far as we are aware, only
technologies in unlicensed sub-GHz frequency bands. a few publications are available on this topic.
In this paper, we do performance testing on one such Introduction and discussion of LoRa LPWAN technol-
star-topology network, based on Semtechs LoRaTM ogy is given by [2]. This work is followed by [3] where
technology, and deployed in the city of Rennes - LoRa a description of an experimental deployment of a LoRa
FABIAN. In order to check the quality of service network is provided and a rough estimation of the number
(QoS) that this network can provide, generally and in of gateways needed to cover a city is done. The perfor-
given conditions, we conducted a set of performance mance of LoRa when transmitting through materials such
measurements. We performed our tests by generating
and then observing the traffic between IoT nodes and
as water and concrete is done in [4]. First short and long-
LoRa IoT stations using our LoRa FABIAN protocol range measurements was conducted by [5], however in a
stack. With our experimental setup, we were able to notably open/semi-rural environment. The measurement
generate traffic very similar to the one that can be was done with few or no obstacles concealing the line of
used by real application such as sensor monitoring. sight (LoS).
This let us extract basic performance metrics, such In this paper we build upon this work, extending it
as packet error rate (PER), but also metrics related to different areas, conditions, radio parameters and LoRa
specifically to the LoRa physical layer, such as the devices.
Received Signal Strength Indicator (RSSI) and Signal We first describe the LoRa FABIAN network setup that
to Noise ratio (SNR), within various conditions. Our was developed and installed by the authors, in coopera-
findings provide insight about the performance of LoRa
networks, but also about evaluation methods for these
tion with other colleagues, students and two companies
type of networks. We gathered measurement data that - Kerlink [6] and TDF [7]. We give details about the
we make freely available together with the tools we overall architecture, equipment and protocols used. Then,
used. after providing the current radio parameters used in our
network, we detailed the measurements done and analyzed
KeywordsIoT communication needs, LPWAN, LoRa the performance observed.
performance, long-range radio, access networks.

II. LoRa FABIAN

I. Introduction LoRa FABIAN [8] is a Network Protocol Stack and ex-
Tens of billions of machines and different type of sensors perimental network setup, deployed in Rennes, France, for
are expected to be deployed and connected in the short IoT needs. Although being mostly designed for LoRa and
term. A large part of these is expected to be covered the associated constraints (most notably, low bandwidth)
by Wide Area Networks (WANs). In addition to cellular it can be reused on top of any Layer 1 technology.
networks, the recently emerging Low-Power Wide Area LoRa FABIAN aims at democratizing access to low-
Networks (LPWAN) are also a very suitable solution for power long-range technologies, by abstracting the net-
this type of coverage. work complexity using common Internet protocols such as
These networks use a physical layer technology that CoAP, DNS, HTTP(s) etc.
trades bitrate for range, providing a wide coverage (up IoT objects are designed to run a local CoAP server
to tens of kilometers) and energy efficiency at cost of and client which offers resources and access to the object,
low datarate (in order of hundreds of bits or kbits per which can be used to control the object and interact with
second, and, sometimes, much less). LoRa from Semtech its environment. This CoAP server is directly accessible

978-1-5090-3254-9/16/$31.00 2016 IEEE

 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:
From M2M Communications to Internet of Things
2

from the Internet, using an HTTP translator. That way, to the right Gateway that the object is associated to, and
it can leverage all web technologies. vice versa.
Currently, the LoRa FABIAN network (as shown in
Fig. 1) consists of three LoRa IoT stations (in different
A. LoRa FABIAN Architecture
locations), one Gateway, one Service node and number of
LoRa FABIAN uses different components that are nec- IoT objects (used mostly for testing purposes).
essary to test and provide coverage for IoT applications.
This include both communicating and gathering data from
IoT objects, and connecting them to the Internet (both B. Specification and radio parameters of LoRa IoT stations
to send data, and receive remote commands). Therefore, and experimental IoT objects
the setup (as shown on Fig. 1) includes: IoT objects, LoRa LoRa IoT stations are running a customized Linux
IoT stations (that can communicate with the objects), and System on top of an ARM platform that uses a Semtech
components that enable two-way communication with the SX1276 chip for providing the LoRa connectivity with a
Internet and its services (the Gateway and the Service 5dBm antenna.
node). Out of the three LoRa stations hosted by LoRa
The experimental IoT objects are composed of an Ar- FABIAN, two are used for the production network. They
duino [9] and a FroggyFactory [10] LoRa Shield running are hosted by Tl Diffusion de France (TDF) [7] on two
a modified version of contiki OS [11], [12]. The LoRa high points (one in 9 avenue Jean Janvier, 35000 Rennes,
Shield handles the low level communication and low level France, elevation: 85m, and the other in 2 Rue du Clos
network access (registration, authorization, etc.), following Courtel, 35510 Cesson-Svign, Rennes, France, elevation:
a mechanism, that was designed by the authors and pro- 160m). The last LoRa station is situated on the roof
posed in [13]. It forwards the data traffic to the Arduino, of Tlcom Bretagne (2 Rue de la Chtaigneraie, 35510
which allows the user to prototype application in a simple Cesson-Svign. Rennes. France., elevation: 55m) used as
environment, by defining CoAP [14], [15] based REST [16] an experimental platform.
resources, that can then be accessed from the Internet. In All those antennas are configured to use the following
the same way, the user can use the Arduino interface to parameters:
send data traffic to the network. The LoRa shield can have Bandwidth: 868.1 to 868.225Mhz
an embedded antenna or an external, soldered-on, antenna Channel size: 125kHz
that provides additional gain. Spreading Factor (SF):
The LoRa IoT station, built by Kerlink [6] is a LoRa
antenna, that has the ability to connect with the outside Sending (downlink), fixed = 7
via Ethernet or 3G. It can listen and send traffic (to/from Receiving, variable, from 7 to 12
IoT objects) using LoRa technology through multiple fre- Coding Rate: 4/7 (for sending, but can receive 4/5
quencies and code-rate at the same time. It redirects this to 4/8)
traffic through a simple UDP tunnel to/from the LoRa Transmitting Power: 14dBm
FABIAN Gateway exporting data and meta-data in a The IoT object, for now, in order to receive traffic
simple JSON [17] text format. from LoRa stations, needs to use the same parameters, in
The Gateway can communicate with many LoRa IoT terms of frequency, coding rate, channel size and SF. For
stations in order to concentrate traffic for a global area transmitting they use a fixed power of 14dBm. The Frog-
like a city or a country. It is in charge of sending bea- gyFactory LoRa shield comes with an embedded antenna,
cons to inform IoT objects about the network availability however, in our setup, some shields were modified to have
and to receive their registration messages, informing IoT an external, soldered-on, antenna to test the performance
objects they may have to change their radio parameters gain.
(frequency, code rate). All the traffic intended for the regis-
tered IoT object (downstream), and all the packets coming
from the objects (upstream) then go through the Gateway. C. LoRa FABIAN Protocol Stacks
The downstream traffic can be stored here, to allow the LoRa FABIAN uses two different kind of messages
IoT objects to be powered off and receive messages once to communicate between nodes, over the LoRa technol-
they awake. Once registered, all the messages coming from ogy. For interoperability reasons, we choose to use IEEE
the IoT objects are redirected using HTTPS tunnels to the 802.15.4 [20] instead of using LoRaWan [21] as Layer-2
Service node. The code for the gateway was written in Java to transport our CoAP-based messages. As of now and for
by the authors and it is based on Californium [18], [19], as the sake of simplicity, the CoAP messages are send in clear
well as other libraries (JPA, H2 database,...) text. Encrypting them using DTLS [22] and distributing
The Service node, where every IoT objects DNS name the needed keys are part of our next steps.
points to, runs a simple translator from HTTP to CoAP a) Signaling messages: are exchanged between gate-
protocol (and vice versa) written in Java. The code was way and objects. Following [13], they are using CoAP to
developed by the authors and is based on the Californium handle network announces (beacons), nodes registration,
proxy. It receives traffic from the Internet and redirects it nodes control messages (e.g. radio parameter changes).
 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:
From M2M Communications to Internet of Things
3

LoRa IoT station

gateway Internet
Service
node
IoT object

Fig. 1. LoRa FABIAN Architecture

Fig. 3. LoRa IoT stations in LoRa FABIAN network in area of

Rennes city, northwest of France

the QoS, such as the Packet Error Rate (PER), but also
some parameters related to the Radio and LoRa layers,
like the Signal to Noise Ratio (SNR), and the Received
Signal Strength Indicator (RSSI).
At the time of performing the tests, all the commu-
HTTPS
nication within the LoRa FABIAN network occurred on
UDP
the 868.1KHz frequency band. The bandwidth that was
CoAP
IP IP used was 125kHz. These parameters, as well as the coding
802.15.4 rate (that was set to 4/5), and the packet payload (25
LoRa
bytes, including MAC frame) remained fixed throughout
the measurements. The parameters that were varied are:
spreading factor (SF), distance, surroundings, antenna size
on IoT object and elevation/location of LoRa IoT stations
(by observing results on differently placed stations).
With this simple setup, we were able to generate traffic
really similar to one that can be used by real application
Fig. 2. LoRa FABIAN protocol stacks (like sensor monitoring), and then, from that, extract
QoS metrics within various conditions. For this round of
measurements we chose to focus on testing only the uplink
b) Data messages: are also exchanged between the (from IoT object to the LoRa IoT station), as this is the
object and the Service node. They offer the user the ability more common case in IoT.
the access the object directly through the use of CoAP to First we did a basic test to get a rough idea about the
provide a REST semantic available via HTTP. range and performance of the three LoRa IoT stations
The protocol stacks implemented on each of the entities (TDF Suburban, TDF CityCenter and TB on Figure 3),
in the LoRa FABIAN architecture is shown on Figure 2. based on the station location but also on the distance, SF,
and the type of antenna used on the IoT object. Following
that, we gathered some statistical data of performance of
III. Testing of the LoRa FABIAN network
LoRa for a certain distance (but with varying the location
These measurements are supposed to give us insight and surroundings). This data allowed us to evaluate the
about the Quality of Service (QoS) that the LoRa FABIAN different parameters (such as SNR and RSSI) in these
(using LoRA technology) network can provide, and give us scenarios, but also provided us with insight into various
statistical data for different configurations of the network. influences on the performance.
Other than that, they can also bring to light some of the
factors that need to be considered when testing these kinds
of networks and provide some statistical data for different A. Range tests
configurations of the network. To perform these measurements we moved on a trajec-
For this, a number of measurements were performed in tory (going up to roughly 3km away from TDF Suburban
order to get basic metrics that can be used to describe and TB stations, and 6km away from the TDF CityCenter
 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:
From M2M Communications to Internet of Things
4

Fig. 7. Received packets on Tlcom Bretagne LoRa station with

on-board antenna

Fig. 4. Received packets on TDF Suburban LoRa station

using SF 10. In order to isolate the influence of the different
node antenna, the measurement was repeated, but with
switching the SFs (Gamma used SF 10 and Beta SF 12)
All three LoRa stations (positioned as shown in Figure
3) were listening, and logging the information (detailed
in Appendix A, most importantly - RSSI and SNR) for
each received packet. In this way we could compare the
reception on each LoRa station. Since two stations (TB
and TDF Suburban) were relatively close to each other,
we could observe the influence of their placement (most
notably, the elevationa).
The results from the three stations, when receiving from
Gamma (configured with SF 12), can be seen in Figures
4, 5 and 6. Each pin represents one packet, as received
on the given station. The pin color represents the RSSI of
the packet. Red is used for packets received between -120
and -106 dBms, yellow for the range -105 to -94 dBms
Fig. 5. Received packets on Tlcom Bretagne (TB) LoRa station and green for the range -93 to -82 dBms and finally blue
for the range between -81 to -69 dBms. (To access all the
data, interactively on the map or to download it, go to
[23]).
By comparing the results on the TB station (elevation
55m) and on the TDF Suburban station (elevation 160m),
we can see that the elevation of the antenna makes a
dramatic difference. At some areas along the road, there is
complete blackout for the TB station (mainly in valleys),
while TDF Suburban stations reception remains active
throughout the whole trajectory. During the entire mea-
surement, in total, 1447 packets were sent. Out of that, TB
station received only 748, while the TDF Suburban station
received 1400. Maybe surprisingly, the station in the city
center of Rennes (TDF CityCenter, elevation 85m) that is
around 6 km away from the farthest point, received 1319
Fig. 6. Received packets on LoRa station in the center of Rennes packets. We can also observe that the RSSI given by the
(TDF CityCenter) TB station, throughout the whole route, rarely goes over
-112 dBm, while on the TDF Suburban, more than half of
the packets are received with RSSI between -92 and -81
one, shown by pins on Fig. 4 and 5), with two IoT nodes dBm. It is interesting to note that the TDF CityCenter
that were periodically sending packets. One node (that we station is able to receive packets with RSSI below -124
call Gamma) had a soldered-on antenna, while the other dBm. Moreover, most of the packets (1088 out of 1319),
one (that we call Beta) had a weaker, on-board, antenna. on this station, are received in range between -124 and -
For the first round Gamma was using SF 12 while Beta was 103 dBm. This shows us that, in some cases, even thought
 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:
From M2M Communications to Internet of Things
5

Fig. 9. Correlation between PER and SNR, 3km from the LoRa IoT
station

Fig. 8. Fixed measurment points, 3km distance from TV tower LoRa

IoT station

the average RSSI is worse, the reception can be better.

Figure 7 shows the result on the TB station, for the
same SF, when using Beta (on-board antenna), rather than Fig. 10. Correlation between PER and RSSI, 3km from the LoRa
Gamma. In this case only 59 out of 1000 packets were IoT station
received, showing that the use of a soldered-on antenna
makes a significant difference, with the success rate jump-
ing from 6% to 62% when adding the antenna. might be partly explained by the elevation profile between
the LoRa station and measurement point. Namely, when
the elevation profile between the measurement point and
B. Fixed point measurements the LoRa station is strictly degrading - the reception is
In the fixed point measurement we sought to take a good. When some "hills" are present that might block the
closer look at the influence of location and environment reception - it is much worse. In B you can find elevation
on the performance of the LoRa network, so we chose 8 profiles between non-urban measurement points and the
points (as shown on map in Fig.8) that were on the same TDF Suburban station.
distance (3km) from the LoRa station we chose to focus It can also be noted that the correlation between RSSI
on - TDF Suburban. and PER is not straightforward. Namely, sometimes even
Half of these points are located in a more urban part of when the RSSI mean is high the PER is going to be worse.
Rennes, while the other half is on the borders or outside. This is more obvious by looking at a graph such as in
The environment varied from suburb-like found in Cesson- Figure 10 showing the overall PER for each RSSI value at
Sevigne, to ones mostly empty (with nothing but roadside a certain location (the annotations on each point denote
restaurants), to a farm in a rural environment, and finally the number of packets received - both successfully and
to points in the city of Rennes. unsuccessfully - for a given RSSI) On the other hand, SNR
On each of these measuring points, 500 packets (each (Fig. 9), since it takes into account also the current noise
having a 25 byte payload) were sent out for SF 7,9 and level, shows the expected property - more packets are lost
10. Gamma (with soldered-on antenna) was using SF 10 to when the SNR is lower.
send 500 packets, meanwhile Beta (with on-board antenna)
was sending 500 packets for both SF7 and SF9. Data availability
This allowed us, to gather some statistical data about For the sake of reproducibility, and in order to allow
the performance of the LoRa network on distance of 3km everybody interested in this topic to reuse, extend and
in different areas. The results are synthesized in Table I. build on top of our research, we published all our acquired
It was expected that in urban areas the results are going data, and all the scripts used for its parsing, plotting
to be worse, and this was proven true by the packet error and analyzing. They can be found at [24] under a liberal
rate (PER) not going under 40% in this areas, however opensource licence.
what is a bit surprising is the dramatic degradation of
results in non-urban (even can be called rural) areas, such
as the Joualt Hubert farm (C on Fig. 8) or Parc du Bois IV. Conclusion and Future work
de la Justice (H on Fig. 8), with the PER worsening by In this paper, we described our experimental LoRa
30% as compared to Golf de Cesson (A on Fig. 8). This setup in the city of Rennes - LoRa FABIAN, and we
 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:
From M2M Communications to Internet of Things
6

TABLE I. Synthesis of measurements

Point SF RSSI(dBm) mean RSSI std. dev SNR(dB) mean SNR std. dev. PER Packets Rx/Tx
SF10 -106.52 4.74 5.17 2.92 44.0% 280/500
Cemetery Est SF7 -117.81 1.42 -7.08 1.12 90.4% 48/500
SF9 -113.68 3.71 -6.08 4.46 51.0% 245/500
SF10 -100.51 3.64 6.47 1.54 3.2% 484/500
La Terasse
SF7 -107.75 3.93 2.86 3.85 47.0% 265/500
SF10 -106.20 5.89 5.10 3.03 53.4% 233/500
Rue de la
SF7 -111.63 3.33 -0.00 3.90 33.8% 331/500
Frebardiere
SF9 -109.74 3.55 -0.72 4.75 35.2% 324/500
SF10 -109.87 4.15 3.71 2.66 44.2% 279/500
Maison Medicale SF7 100.0% 0/500
SF9 -116.06 2.82 -9.33 2.97 87.2% 64/500
SF10 -109.45 4.68 4.02 3.13 41.6% 292/500
Parc Hamelin SF7 -115.70 2.48 -4.93 1.94 57.2% 214/500
SF9 -113.51 3.74 -4.85 3.80 43.0% 285/500
SF10 -107.67 4.33 5.03 2.38 33.6% 332/500
Parc du Bois
SF7 -113.97 2.80 -2.13 3.10 49.2% 254/500
de la Justice
SF9 -110.85 4.72 -3.40 4.56 21.8% 391/500
Golf de Cesson SF10 -106.25 4.82 5.46 2.58 3.0% 485/500
SF10 -106.82 5.55 4.73 2.93 35.4% 323/500
Joualt Hubert farm SF7 -114.01 2.73 -2.51 2.78 42.2% 289/500
SF9 -114.30 3.65 -6.08 3.58 49.2% 254/500

Fig. 12. Elevation Profile between TDF Suburban LoRa station

(left) and point A (see Fig. 8)

Fig. 11. LoRa packet structure

A. Example of packet received by LoRa IoT station

designed, performed and analysed measurements for it. Packets flowing into the LoRa FABIAN network are
LoRa technology offers a excellent outdoor coverage either received and sent by LoRa stations. In this section, we
in urban or rural area. As expected, the antenna location will present some example of LoRa FABIAN data, as they
and especially its elevation plays a major role in the are received by one LoRa station. We will start by a simple
network performances. In the best conditions, the frames raw example and then describe the LoRa layer meta data,
losses is very low (about 3%). One of the goal of our study and the 802.15.4 and CoAP data.
was to define criteria to switch from one spreading factor
to another one to guaranty the best trade-off between NEW PACKET: received packet (size 43,
channel utilization and error rate. It appears that the RSSI modulation 16,BW 2, DR 2, RSSI -89.0)
alone may not be a good metric since measurement do
not exhibit a strong correlation. SNR could be a better JSON up: {"rxpk":[{"tmst":3936040845,
candidate. A next step will be to combine uplink and "time":"2016-01-25T16:40:15.164887Z","chan":8,
downlink traffic to find if some correlation exists between "rfch":0,"freq":868.100000,"stat":1,
measurements. This will be helpful to determine which "modu":"LORA","datr":"SF7BW250","codr":"4/7",
LoRa station will be the best to join a node. "lsnr":8.2,"rssi":-89,"size":43,
"data":"IjYB+rAAAAAAAAgAAEYBqv9MxGidNU49AX
Appendix A RoZXRhLnQuZXUub3Jng25vMg=="}]}
LoRa packet structure
Figure 11 shows the format of the packet in the LoRa Message in HEX. Size: 43
network. The first field is the preamble, which is in charge 223601FAB000000000000800004601AAFF4CC4689D3
of the synchronization of the receiver and the incoming 54E3D0174686574612E742E65752E6F7267836E6F32
packet. The next field is the header which provides in-
formation such as the SF used, the FEC code rate and In this example, we are able to see the meta data in JSON
size of the payload. Then, the payload field containing the format, as made available by the LoRa Station and the
data to be transmitted and finally the last field is a cyclic Semtech engine [25]. Description of each field can be found
redundancy check (CRC) for error detection. on [25]
The structure may vary depending on whether explicit We can also see the payload of the LoRa data, including
header is used, and also on the length of the preamble thus the IEEE 802.15.4 layer and the CoAP one. Table II
(currently, we are using 8 symbols) shows the interpretation.
 2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC) - Workshop:
From M2M Communications to Internet of Things
7

TABLE II. 802.15.4 frame with CoAP payload

[5] M. Aref and A. Sikora, Free space range measurements with
Message in HEX Interpretation Semtech LoRa technology, in Wireless Systems within the
223601 IEEE 802.15.4 Header: Frame control Conferences on Intelligent Data Acquisition and Advanced
field and sequence number Computing Systems: Technology and Applications
FAB0000000000008 IEEE 802.15.4 Destination address (8
(IDAACS-SWS), 2014 2nd International Symposium on,
Bytes addressing mode) IEEE, 2014, pp. 1923.
[6] LoRa IoT Station 868 MHz - Kerlink,
0000 IEEE 802.15.4 Source Address (2 Bytes
http://www.kerlink.fr/fr/produits/lora-iot-station/lora-iot-
addressing mode)
station-868-mhz.
04601AAFF4CC46 [7] Tldiffusion de france, http://www.tdf.fr.
89D354E3D01746 [8] Rseau Fabian LInternet des Objets dans la ville,
86574612E742E65 CoAP Metadata and Payload
752E6F7267836E6
http://www.reseaufabian.fr.
F32 [9] Arduino - ArduinoBoardUno,
https://www.arduino.cc/en/Main/ArduinoBoardUno.
[10] Froggy Factory, http://www.froggyfactory.com/froggy.
[11] A. Dunkels, B. Gronvall, and T. Voigt, Contiki-a lightweight
and flexible operating system for tiny networked sensors, in
Local Computer Networks, 2004. 29th Annual IEEE
International Conference on, IEEE, 2004, pp. 455462.
Fig. 13. Elevation Profile between TDF Suburban LoRa station [12] Wi6labs/lorafabian code repository,
(left) and point B (see Fig. 8) https://github.com/Wi6labs/lorafabian.
[13] A. Pelov, L. Toutain, and Y. Delibie, Constrained Signaling
Over LP-WAN, Internet Engineering Task Force. [Online].
Available: https://tools.ietf.org/html/draft-pelov-core-cosol.
[14] C. Bormann, A. P. Castellani, and Z. Shelby, Coap: An
application protocol for billions of tiny internet nodes, IEEE
Fig. 14. Elevation Profile between TDF Suburban LoRa station Internet Computing, vol. 16, no. 2, p. 62, 2012.
(left) and point C (see Fig. 8) [15] Z. Shelby, K. Hartke, and C. Bormann, The constrained
application protocol (CoAP), RFC 7252 (Proposed Standard),
Internet Engineering Task Force, Jun. 2014. [Online].
Available: http://www.ietf.org/rfc/rfc7252.txt.
[16] R. T. Fielding, Architectural styles and the design of
network-based software architectures, PhD thesis, University
of California, Irvine, 2000.
Fig. 15. Elevation Profile between TDF Suburban LoRa station [17] T. Bray, The javascript object notation (json) data
(left) and point H (see Fig. 8) interchange format, RFC 7159 (Proposed Standard), Internet
Engineering Task Force, Mar. 2014. [Online]. Available:
http://www.ietf.org/rfc/rfc7159.txt.
Appendix B [18] M. Kovatsch, M. Lanter, and Z. Shelby, Californium:
Elevation profiles Scalable cloud services for the internet of things with coap,
in Internet of Things (IOT), 2014 International Conference
Figures 12-15 show the elevation profiles between the on the, IEEE, 2014, pp. 16.
TDF Suburban LoRa station and the non-urban measure- [19] Californium (Cf) CoAP framework,
ment points as calculated by Google Earth [26] software. https://www.eclipse.org/californium.
[20] IEEE 802.15.4 TG, 802.15.4-2011. IEEE Standard for Local
and metropolitan area networksPart 15.4: Low-Rate
Acknowledgment Wireless Personal Area Networks (LR-WPANs), September.
2011, vol. 2011, p. 314, isbn: 9780738166834.
Authors would like to thank the Android app RainTo- [21] N. Sornin, M. Luis, T. Eirich, T. Kramp, and O. Hersent,
day. Lorawan specifications, LoRa Alliance, 2015.
[22] E. Rescorla and N. Modadugu, Datagram transport layer
References security version 1.2, RFC 6347 (Proposed Standard),
Updated by RFCs 7507, 7905, Internet Engineering Task
[1] Lora Alliance web page, https://www.lora-alliance.org/. Force, Jan. 2012. [Online]. Available:
[2] L. Vangelista, A. Zanella, and M. Zorzi, Long-range IoT http://www.ietf.org/rfc/rfc6347.txt.
technologies: The dawn of LoRa, in Future Access [23] Range tests for different stations, node antennas and SFs,
Enablers for Ubiquitous and Intelligent Infrastructures, interactive map, https://drive.google.com/open?id=1_-
Springer, 2015, pp. 5158. 30lffhl8i49GLAaq-KUkrWtE8&usp=sharing.
[3] M. Centenaro, L. Vangelista, A. Zanella, and M. Zorzi, [24] Measurements data and scripts,
Long-range communications in unlicensed bands: The rising https://lorafabian-pimrc2016.github.io/.
stars in the IoT and smart city scenarios, ArXiv preprint [25] Description of JSON metadata on Semtechs LoRa-net code
arXiv:1510.00620, 2015. repository, https://github.com/Lora-
[4] T. Wendt, F. Volk, and E. Mackensen, A benchmark survey net/packet_forwarder/blob/master/PROTOCOL.TXT.
of long range (LoRa) spread-spectrum communication at [26] Google Earth, https://www.google.fr/intl/fr/earth.
2.45 ghz for safety applications, in Wireless and Microwave
Technology Conference (WAMICON), 2015 IEEE 16th
Annual, IEEE, 2015, pp. 14.