You are on page 1of 6

This full text paper was peer-reviewed at the direction of IEEE Instrumentation and Measurement Society prior to the

acceptance and publication.

An Experimental LoRa Performance Evaluation in


Tree Farm
.........................
Daeun Yim Jiwon Chung Yulim Cho Hyunji Song
Computer Science and Computer Science and Mathematics Computer Science and
Engineering Engineering Dongguk University Engineering
Sogang University Sogang University Incheon, South Korea Dongguk University
Seoul, South Korea Seoul, South Korea ikisee@gmail.com Seoul, South Korea
deyim33@gmail.com chloe326o.o@gmail.com hysong6059@gmail

Daehan Jin Sojeong Kim Sungwook Ko Anthony Smith


Computer Science and Computer Science and Computer Science and Computer and Information
Engineering Engineering Engineering Technology
Dongguk University Dongguk University Dongguk University Purdue University
Seoul, South Korea Sung-Nam, South Korea Seoul, South Korea West Lafayette, USA
hanguk46@gmail.com thwjddl94@naver.com kso0611@naver.com ahsmith@purdue.edu

Austin Riegsecker
Computer and Information
Technology
Purdue University
West Lafayette, USA
ariegsec@purdue.edu

Abstract— Deployment of internet of things (IoT) devices, I. INTRODUCTION


wireless sensors, and sensor networks, in agriculture can be a great
help in monitoring environment and growing crops; and having a
There is obvious benefit to agricultural IoT. Data collected
network to support those devices is necessary to successfully utilize from wirelessly connected sensors and agricultural equipment
those resources. Recently, low-power wide area network (LPWAN) can ease the difficulty of monitoring farmland and livestock,
have been recognized as an appropriate technology for agriculture and assist farmers' decision making.
use. LoRa is a representative network of low-power wide area To apply IoT technology in an agricultural environment,
network (LPWAN). It can be applied to IoT for agriculture due to its one must consider how these devices are connected. The
long range and low power capabilities. LoRa network performance
network system for agricultural IoT has to cover a wide area of
is highly affected by physical layer parameters and the environment.
While LPWANs have been around for a number of years, they are
farmland without relying on costly cellular networks. A low-
still considered an emerging technology and have had few scientific power wide area network (LPWAN) is a well-suited solution
studies performed. Currently, most studies have shown LoRa for this purpose. LPWAN is a type of wireless network designed
communication capabilities in urban, mountainous, and maritime to allow long range communications at a low bit-rate. End
areas, with little focus on agriculture use cases. Tree farming is a devices can communicate with base stations from longer
long-term investment, requiring careful monitoring to mitigate loss; distances than other wireless network systems while consuming
therefore, this paper provides an analysis about the impact of variant very less energy.
physical layer parameters on performance of LoRa networks in a tree
Some LPWANs available include Sigfox, Weightless-N,
farm. Overall, the LoRa communication range was smaller than the
theoretically expected range. Some PHY factors, spreading factor
NB-IoT and LoRa. Each network has its advantages and
and coding rate, showed a clear impact on LoRa performance. disadvantages. Though it is a fairly new technology, LoRa has
Actual data reliability was inconsistent at varying distances and PHY attracted a lot of attention, especially in terms of low costs and
configurations, unlike the consistency found in the RSSI reported by excellent battery performance. LoRa's modulation developed
the radios. Additional experiments show the Fresnel Zone still affects traditional spread-spectrum communications more robust by
LoRa networks. All experiments were performed in an Indiana rural generating chirp signal that changes frequency continuously
tree plantation in the United States. Each node was set differently by [14].
modulating spreading factor, bandwidth and coding rate using
915MHz and 13 dBm transmit power at variant distance.
Much of the research on LoRa is relevant to its potential for
covering large areas. Unfortunately, most of the studies have
Keywords—LoRa network, Tree farm, PHY factors, Fresnel been conducted in open or urban environments with only a few
zone, IoT, Precision Agriculture were related to agriculture [5][7]. To provide consistent and

978-1-5386-2092-2/18/$31.00 ©2018 IEEE


efficient management of large farmlands, measurements need radio frequency bands like 169 MHz, 433 MHz, 868 MHz
to be taken of LoRa network potential for those areas. (Europe) and 915 MHz (North America). Since LoRa uses
license-free frequency bands, network costs can be minimized
Especially in a tree farm, it takes at least 6 years and up to
[11].
several decades to harvest timber. Even minor failures can
result in wasting long-term efforts and large investments.
Therefore, to verify its potential for agriculture, we focused on B. PHY factors
experimenting changes in LoRa network performance, range, PHY factors underlie all of three classes of LoRaWAN. The
reliability, and RSSI, according to different physical layer configuration of LoRa can be tuned by controlling physical
parameters and how each affects it. In addition, by examining layer (PHY) settings, which results in performance changes.
traditional Fresnel zone limitations, the LoRa performance In this research, physical settings include spreading
affected could be measured. factor(SF), bandwidth(BW), and error correction rate(CR).
Background
LPWANs supplements short range wireless networks 1) Spreading Factor(SF)
such as Wi-Fi and Bluetooth with their large-scale coverage at
decent cost. Instead of narrowing bandwidth to increase signal- The basic premise of spread spectrum is that each bit of
to-noise ratio (SNR) like other LPWANs, LoRa uses a special information is encoded as multiple chips. Spreading factor(SF)
chirp spread spectrum modulation to cover large areas. is the relationship between the bit and chip [9]. A higher SF
increases the Signal to Noise Ratio(SNR), and thus sensitivity
Below are some details of LoRaWAN and physical factors and range, but also increases the air time of the packet [4].
consisting of LoRaWAN physical layer. Available SF's are from 6 to 12 which change LoRaWAN data
rates from 0.3 kbps to 27 kbps.
A. LoRaWAN
2) Bandwidth(BW)
LoRa is a chirp spread spectrum modulation scheme using
wideband linear frequency which improves receiver sensitivity Bandwidth (BW) is the range of frequencies in the
and tolerance to miscommunication between receiver and transmission band. LoRa can have 125 to 500 KHz of
transmitter [9]. bandwidth, typically set to 125, 250, or 500KHz. Changing BW
can influence message time on air which results in inverse
LoRaWAN technology mainly uses star network topology. influence on radio sensitivity. Larger BW gives a higher data
The gateways receive messages from multiple end-devices and rate (thus shorter time on air), but a lower sensitivity (due to
end-devices are connected to one or more gateways. End- integration of additional noise). A smaller BW gives a higher
devices can transmit data to gateways with own time interval sensitivity, but a lower data rate.
and data rate complying rules. These base stations are
connected to other network servers via standard IP connections 3) Coding Rate(CR)
[4]. The base station and end-devices can communicate bi-
Coding Rate(CR) is a form of Forward Error Correction
directionally with good coverage.
(FEC) that permits the recovery of bits of information due to
corruption by interference. This requires a small overhead of
LoRaWAN networks have three classes based on basic additional encoding of the data in the transmitted packet [9]. CR
features and additional features. Class A end-devices (basic can range from 4/5 to 4/8. A larger CR offers more reliability
LoRaWAN) provide bi-directional communications. Those by improving resilience to corrupted bits, but increases time on
communications consume the lowest power. Their downlink air and energy consumption [4].
communication occurs from the server following uplink
transmission. Class B additionally provides a scheduled receive
4) Transmission Power
slot trough opening extra receive windows. The servers know if
end-devices are listening. Class C provides bi-directional LoRa transmission power ranges from -4 to 20 dBm. Higher
communication with maximal receive slots for minimum transmission power causes an increase in energy consumption
latency [9]. and signal-to-noise ratio.

LoRa Technology has some key features. LoRa has C. Fresnel Zone
scalability in bandwidth. Additionally, LoRa can be used with
both narrowband and wideband service with a few The concept of Fresnel zone clearance is used to account
configuration changes. LoRa has great reliability as it resists the for interference by obstacles within the beam width of a radio
Doppler Effect, multipath, fading mechanisms, and other signal. The first zone must be kept largely free from
interference mechanisms [14]. Its protocol is designed obstructions to avoid interfering with the radio reception.
specifically for low power consumption extending battery However, some obstruction of the Fresnel zones can often be
lifetime up to 20 years. A single base station can penetrate into tolerated. As a rule of thumb the maximum obstruction
dense urban/indoor regions, and can connect rural areas up to allowable is 40%, but the recommended obstruction is 20% or
30 miles away. The last feature is low cost. LoRa reduces costs less. [8]
in three ways: infrastructure investment, operating expenses
and end-node sensors. LoRa uses license-free sub Gigahertz
II. RELATED WORK equipped with an SX1276 radio module that works in the 915
MHz band, and a 1/2 wave omni-directional antenna.
Before LPWAN became a popular network infrastructure
for agricultural IoT, mesh topology networks using smaller To find temperature and humidity data, a DHT11 sensor
range network devices such as Zigbee dominated wireless package was used; this is a composite sensor that contains a
sensor networks (WSN). However, this kind of implementation calibrated digital signal output to measure temperature and
had a serious problem with limited link budget due to high data humidity.
rates and a low link budget [15]. In contrast, LPWAN, using a
star network topology, has a distinct advantage with longer
range and battery life.
There has been a significant amount of research on outdoor
LoRa coverage and performance. Most were performed in
urban and maritime environments. Urban experiments showed
more than half of packets were successfully delivered from 5-
10km range [1], and a maximum 80% of packet delivery at a
distance of 2800 meters [11]. In maritime applications,
performance of 15-30km distance was observed.
However, unlike the prior mentioned research other
research showed only a few hundred meter LoRa network
coverage [5][7]. Also, from these experiments LoRa showed
distinct performance differences within various environments
such as indoor, outdoor, underground or with large amounts of
vegetation.
Even meteorological conditions such as temperature and
humidity affect network performance [11], and their impact on
LoRa performance was observed in [5].
Research also showed variant hardware physical settings
Fig. 1. Dragino IoT kit [6]
can have a great influence on LoRa network performance. With
increasing spreading factor LoRa had longer range with higher
reliability [11]. According to [7], BW was the largest factor in B. Experimental Setup
communication success. However, research by Cattani et al. [5] 1) Sites
showed that spreading factor, coding rate and then bandwidth
was the order that had the most effect on network reliability. Our experiments were performed three times at the same
tree farm. The farm is located at 13455 S 525 W, Romney, IN
It is important the transmitter and receiver use the same 47981, USA. It is primarily a flat meadow, with small hills in
spreading factor and bandwidth combination for successful the middle of the property. Mid-life maple, oak, and pine trees
packet delivery [13], which means that to experiment on various are planted in 8-foot separate rows on the property.
PHY settings, each device has to have the same settings. LoRa
radios with different coding rates can still communicate [5]. 2) PHY settings and Data Packet
Although prior research showed expected performance of TABLE I
LoRa network in many areas, little work has been done on ARDUINO NODE ID ACCORDING TO PHY SETTINGS
agricultural environments to evaluate LoRa performance. Since Arduino ID SF BW(kHz) CR TX Power (dBm)
LoRa is an appropriate technology to apply in this environment,
7-125-5 7 125 5 13
this paper focused on how PHY settings affect reliability and
coverage on an actual tree farm. Vegetation can also be an 7-125-8 7 125 8 13
obstacle to network performance, so it is expected that LoRa 9-125-5 9 125 5 13
will have lower performance than in other previously 9-125-8 9 125 8 13
researched environments.
11-125-5 11 125 5 13
III. EXPERIMENTS METHOD AND RESULTS 11-125-8 11 125 8 13
7-250-5 7 250 5 13
A. Hardware Setup 7-250-8 7 250 8 13
We used the LoRa IoT Development Kit from ‘Dragino’
which is a company in China offering LoRa shield based on The experiment was carried out by assigning an Arduino ID
Semtech SX1276 chip as can be seen in Figure 1. It also offers according to each PHY Setting. Table 1 shows these variant
the ‘LG01-P’ LoRa Gateway which is an open source single settings. Each three components of Arduino ID is spreading
channel LoRa Gateway. The kit offers two portable devices factor, bandwidth and coding rate.
SF can be set from 6 to 12. The SF is set to 7 (default), 9,
and 11. The CR is set to a minimum value of 5 and a maximum
value of 8. BW is set to 125kHz, aside from 7-250-5 and 7-250-
8. Only SF7 can be set to the maximum setting of 250kHz, so
SF7 was set in two cases of 125kHz and 250kHz. The value of
TX power was fixed at 13(default). To account for the 60%
clear 1st Fresnel Zone at 200m, the devices and antennas were
fixed 2.5m above the ground.
In accordance with necessary settings, 4 gateways were
used to match both the LoRaWAN spreading factor and
bandwidth. However, prior research showed LoRa shields with
different coding rates can communicate with the same gateway.
Therefore, both settings using same spreading factor and
bandwidth but a different coding rate, sent data packets to one
gateway.
Fig. 2. Experiment site (Tree Farm Land). Star shows the position of AP and
TABLE 2 each flag shows the position of LoRa shields.
DATA PACKET COMPOSITION
Data Value In addition, Fresnel zone tests were not done with different
settings. LoRa Shield's PHY setting conditions are set to default
data[0] ID[0]
values. SF was set to 7, BW was set to 125, and CR was set to
data[1] ID[1] 5. The experiment was conducted under the same conditions
data[2] ID[2] with the change of height only at 0m, 1m, 2m, and 3m at 200m.
data[3] Humidity
One hundred packets were used to determine tendency of
performance improvement accordance with height.
data[4] Temperature
data[5] Soil C. Metrics
data[6] Sun Upon this experiment’s completion, researchers calculated
data[7] Flame reliability and received signal strength indication (RSSI). The
RSSI was used to analyze data of different settings at different
data[8] Packet Number
distances. Reliability was calculated as the ratio of valid
received packets to the number of packets transmitted to the
The data structures of the payload used in transmitted data Packet Delivery Ratio (PDR). The PDR provided information
packets is character type and were 9 bytes in size. The first three about the communication reliability. [7] RSSI is a
bytes make it possible to distinguish which device is sending measurement of the power present in a received radio signal.
data. We sent data at 3 second intervals with a 0.5 second of [10]
delay between sensing. Each time, a gateway collected 150 data
!
packets at once. The transceivers were set to not to retransmit PDR = (1)
"
lost data and thus accurately checked packet delivery rate. If the
data packet wasn’t correct, this was considered a data loss and
D: Packets delivered successfully
skipped. The next data packet was then examined.
T: Total packets transmitted
Every end-device was tested at 100m, 150m, and 200m. The
first experiment was conducted on a day with a temperature of D. Experiment Results
25℃, and humidity of 80%. The second day had a temperature 1) Experiments with different PHY settings
24℃, humidity of 62%. The third experiment was done with a
temperature of 27℃, and humidity of 70%. Shields were put in TABLE 3
a box to minimize meteorological factors affecting experiment RESULTS OF EACH SETTING AT DIFFERENT DISTANCE
results. Every experiment was carried out for about three hours.
100m 150m 200m
Additionally, all the experiments were performed at a height of
Arduino ID PDR RSSI PDR RSSI PDR RSSI
2.5m, considering Fresnel zone of LoRa network frequency and
overall tree farm area. The result data was found through the 7-125-5 66% -71.74 81.77% -77.48 41.74% -92.66
average of the experiment results. 7-125-8 88.08% -69.22 91.78% -78.46 85.33% -90.64
9-125-5 95.11% -71.98 91.78% -84.5 78.66% -95.04
9-125-8 96.22% -76.27 92% -77.59 93.11% -90.97
11-125-5 99.11% -78.74 98% -80.67 82.67% -93.81
11-125-8 98% -75.74 97.33% -77.08 98.44% -90.77
7-250-5 78.67% -70.20 88.67% -72.35 73.11% -86.80
7-250-8 84.67% -69.99 99.11% -77.59 69.67% -88.38
Through Table 3 one can see that all settings but 7-125-5 1) Fresnel zone experiment
and 7-250-5 showed more than 90% reliability at 150m. Only TABLE 4
9-125-8 and 11-125-8 showed good reliability at 200m. RESULTS OF FRESNEL ZONE EXPERIMENT
Compared to related research, distances measured were much
smaller than expected, and at least setting spreading factor to 9 Distance Antenna Height Reliability RSSI
and coding rate to 8 ensures 200m communication range with 200m 0m 21.66% -102.188
good reliability.
200m 1m 85.33% -88.078
As expected, PHY settings followed predictions. In most
200m 2m 99.04% -89.896
cases, settings with increased spreading factor and coding rate
showed higher reliability at larger distances. 200m 3m 100.00% -89.205

Spreading factor clearly affected the readings. With higher


spreading factor, LoRa communicated with higher reliability at
larger distance. However, by combining spreading factor, PDR
didn't appear to linearly increase. Coding rate was impacted
when the distance grew further away. With a smaller distance
and reliable PDR (higher than 90%), changing coding rate
increased PDR.
Bandwidth had lowest impact on LoRa performance. As
shown from bandwidth 125 KHz (7-125-5 and 7-125-8) and
250 KHz (7-250-5 and 7-250-8) it was impossible to discover
relation from configuring bandwidth. However, as seen in
Figure 3, higher bandwidth impacted on RSSI as augmenting
bandwidth has a negative connection between data rates and
receiver sensitivity.
Fig. 5. PDR according to antenna height

Using Figure 5 as reference, height impacted LoRa network.


Performance in reliability surges along with antenna height
from 0m to 2m.
Considering that the Fresnel zone experiment only received
100 packets, the PDR data is different from the PHY setting
experiment results. This shows the tendency of LoRa
performance improvement with higher antenna placement.
However, this experiment alone does not reveal the difference
between 2m and 3m. Thus, it cannot be verified if the height
increase beyond the Fresnel Zone ensures better
communication. Further study needs to be done for clear
explanation.
II. CONCLUSION

Fig. 3. Average RSSI(dBm) according to Distance(m). RSSI values are This paper provides a basis on deploying LoRa within a tree
averaged by three experiments. farm. To complement research on applying LoRa to agricultural
IoT, this research has shown how PHY factors configurations
As Figure 4 shows, RSSI consistently decreased in every and different distances impact reliability of LoRa on tree farm.
setting as distance grew further, unlike PDR. Even when PDR Overall, LoRa communication range in tree farm is below
had erratic increase from 100m to 150m at spreading factor 7, expected values as specified commercially and in research.
RSSI showed a decrease. However, to secure high reliability in large distances,
However, an obvious relation was not determined between maintaining high spreading factor and coding rate is important.
RSSI and spreading factor. This correlates with other research Spreading factor and coding rate affected LoRa performance in
an expected way, but bandwidth did not conspicuously changed
[12]. Unlike Semtech's claims, RSSI seems to be affected by
performance.
distance, not spreading factor.
RSSI showed a consistent pattern from different PHY
factors configurations. Unlike inconsistent PDR changes in
different PHY configurations and distances, RSSI decreased
along with distances in every configuration.
However, we cannot determine what effect, if any, the tree [13] Bor Martin, John Edward Vidler, Utz Roeding, "LoRa for the Internet of
plantation had on LoRa network communication. In addition, Things." EWSN '16 Proceedings of the 2016 International Conference on
Embedded Wireless Systems and Networks, pp. 351-366, 2016.
this experiment not determine what effect complying with the
calculated Fresnel Zone had on LoRa communication in regards [14] AN1200.22 LoRa™ Modulation Basics 2015. [Online]. Available:
to antenna height and distance between nodes. http://www.semtech.com/images/datasheet/an1200.22.pdf

In the next paper, we plan to compare LoRa performance in [15] A comprehensive look at low power, wide area networks. 2016. [Online].
Available:http://cdn2.hubspot.net/hubfs/427771/LPWAN-Brochure-
open area and tree farm with the same PHY configurations to Interactive.pdf
measure how the tree plantation affect the network
performance. In future work, we will study other factors
influencing LoRa performance such as antenna height and
environmental factors.
ACKNOWLEDGEMENT
This research was supported by the MIST(Ministry of Science,
ICT & Future Planning), Korea, under the National Program
for Excellence in SW)(2015-0-00910) supervised by the
IITP(Institute for Information & communications Technology
Promotion)
REFERENCES
[1] Juha Petäjäjärvi, Konstantin Mikhaylov, Antti Roivainen, Tuomo
Hänninen, Marko Pettissalo. "On the Coverage of LPWANs: Range
Evaluation and Channel Attenuation Model for LoRa Technology."
Presented 2015 14th International Conference on ITS Telecommunications
(ITST). [Online]. Available: http://ieeexplore.ieee.org/document/7377400/
[2] Beser, Nurettin Burcak. "Operating cable modems in a low power mode."
U.S. Patent No. 7,389,528. 17 June 2008.
[3] Schwartzman, Alejandro, and Chrisanto Leano. "Methods and apparatus for
enabling and disabling cable modem receiver circuitry." U.S. Patent No.
7,587,746. 8 September 2009.
[4] N. Sornin (Semtech), M. Luis (Semtech), T. Eirich (IBM), T. Kramp (IBM),
O.Hersent (Actility), “LoRaWAN™ Specification,” V1.0.2, Jul. 2016.
[Online]. Available: https://www.lora-alliance.org/lorawan-for-developers
[5] Cattani Marco, Boano Carlo A. and Römer, Kay, “An Experimental
Evaluation of the Reliability of LoRa Long-Range Low-Power Wireless
Communication,” Journal of Sensor and Actuator Networks, vol. 6, no. 2,
p. 7, Jun. 2017.
[6] Dragino, Dragino IoT kit. 2017.
[7] O. Iova, A. L. Murphy, G. Pietro Picco, L. Ghiro, D. Molteni, F. Ossi and
F. Cagnacci, "LoRa from the City to the Mountains: Exploration of
Hardware and Environmental Factors", International Conference on
Embedded Wireless Systems and Networks (EWSN), pp. 317-322, 2017.
[8] Coleman, Westcott, David, David, " Radio Frequency Signal and Antenna
Concepts" in Certified Wireless Network Administrator Official Study
Guide(ebook), Wiley & Sons, Inc. p. 126.
[9] Semtech, SX1272/3/6/7/8: LoRa Modem Designer's Guide. 2013. [Online].
Available:
https://www.semtech.com/images/datasheet/LoraDesignGuide_STD.pdf
[10] Martin Sauter,"3.7.1 Mobility Management in the Cell-DCH State”
in From GSM to LTE: An Introduction to Mobile Networks and Mobile
Broadband(eBook), 1st edition, UK: John Wiley & Sons, 2011, p. 160.
[11] Augustin. Aloÿs, Yi. Jiazi, Clausen. Thomas, Townsley William M. "A
Study of LoRa: Long Range & Low Power Networks for the Internet of
Things." Sensors 16, no. 9: 1466. 9 September 2016.
[12] Luomala. J, & Hakala. I, "Effects of Temperature and Humidity on Radio
Signal Strength in Outdoor Wireless Sensor Networks." Proceedings of
the 2015 Federated Conference on Computer Science and Information
Systems, pp. 1247-1255. Annals of Computer Science and Information
Systems, 5. IEEE. 9 November 2015

You might also like