2022 IEEE NIGERCON

2022 IEEE Nigeria 4th International Conference on Disruptive Technologies for Sustainable Development (NIGERCON) | 978-1-6654-7978-3/22/$31.00 ©2022 IEEE | DOI: 10.1109/NIGERCON54645.2022.9803130

LoRa Network Planning Using Empirical Path Loss

Models
Uyoata Etuk Uyoata
Department of Electrical and Electronics Engineering
Modibbo Adama University, Yola, Nigeria
[email protected]

Abstract—Novel wireless network technologies such as LoRa low duty cycle wireless connectivity technology designed to
can benefit from network planning. In this preliminary study use the ISM bands namely, 433 MHz, 868 MHz and 912
to investigate the suitability of Low Power Wide Area Network MHz. LoRaWAN is the MAC layer commonly used with
for IoT deployment in Nigeria, we present the performance
of various classical empirical path loss models for LoRa and LoRa. Architecture wise, LoRaWAN has a star of star topology
offer insight into the suitability of these models in various with a gateway serving multiple end nodes spread across a
environments. In this study, path loss simulation results are wide geographical area. End nodes require a single hop to
presented for Okumura-Hata, COST 231, COST-WI, Ericsson, transmit low bytes data to a gateway. The gateways connects
SUI and 3GPP models under urban, rural and suburban outdoor to a network server through a back-haul which could be
propagation environments. 433 MHz, 868 MHz and 912 MHz
carrier frequencies are considered for 1 and 3 metres of mobile WiFi or cable. Although Lora is designed for single hop
end node heights. Simulation results show that COST-WI and communication, multi-hop communication has been proposed
Okumura-Hata models are good candidates for network planning for LoRa as in [4] and a review of relaying in IoT is given in
in rural and sub-urban environments whereas Ericsson and 3GPP [5]. LoRaWAN enables flexible configuration of parameters
models join that list for urban environments. such as bandwidth, coding rate, spreading factor to achieve
Index Terms—Network planning, Empirical models, LoRa,
LPWAN, LoRaWAN, IoT desired performance in various deployment scenarios.
The performance of LoRa in specific environments have
I. I NTRODUCTION been reported in the literature using coverage, range, packet
loss, Time on Air as performance metrics. Furthermore the
Low Power Wide Area Networks (LPWAN) have been suitability of LoRa for specific applications has been con-
shown to satisfy the requirements of IoT applications that sidered in the literature. For example, Lora has been tested
need long range, long battery life, low data rate and low for river monitoring [6], smart farming [7], among others.
cost features. LoRa, SigFox, NB-IoT are front-line competitors Empirical path loss models have also been developed for LoRa
in the LPWAN arena and their physical layer performance in specific environments in cities like Lebanon [3] and Bonn
comparisons are documented, see [1]. One of the ways to [8] to mention a few. To the best of our knowledge, none such
have insight into the performance of wireless networks is work has been carried out in any city in Nigeria.
using path loss models. These models broadly categorised In this paper we take a first step toward characterising the
into empirical, deterministic and statistical have been studied empirical path loss measurements for LoRaWAN in selected
for various networks including LPWAN, see [2] and [3] and locations in Nigeria by carrying out theoretical path loss
references therein. Variations in the environments where LP- performance analysis of classical empirical path loss models
WAN are deployed often inspire environment specific channel to determine their suitability for LoRaWAN. In so doing
measurements that are compared to classical models such we provide a guide on choice of models that better suit
as Okumura, COST-Hata among others. This paper presents LoRaWAN. Furthermore our contribution includes theoretical
simulation results to demonstrate the suitability of various path loss performance analysis for urban, rural and suburban
classical empirical propagation models for various environ- environments for the three carrier frequencies of LoRa and for
ments using LoRa as the connectivity technology. various heights of the end node, specifically, 1 and 3 metres.
II. L O R A III. E MPIRICAL PATH L OSS M ODELS
LoRa is the physical layer protocol of a LoRaWAN which A. Okumura-Hata
is a non-proprietary Low Power Wide Area technology. LoRa This model is a valuable propagation tool derived from
is a spread spectrum technique built from the chirp spread improvements [9] made on the Okumura model [10] which
spectrum (CSS). While LoRa was developed by Semtech, was developed from extensive measurements in Tokyo. This
LoRaWAN was developed by LoRa Alliance as an open model holds for parameter delineations such as carrier fre-
source platform. LoRa is a low data rate, low bandwidth, quency (150Hz ≤ f ≤ 1500M Hz), transmitter - receiver

978-1-6654-7978-3/22/$31.00 ©2022 IEEE

Authorized licensed use limited to: Inje University. Downloaded on July 10,2024 at 10:21:51 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE NIGERCON

separation (1 − 20km) and base station height of (3 − 200m). where Lori is the correction factor from the street orientation
The propagation path loss for the Okumura-Hata model can and w is the street width in metres as earlier stated. Lori can
be expressed as [9]: be obtained thus:

Lb = 69.55 + 26.16log10 (f ) − 13.82log10 (hb ) − a(hm ) −10 + 0.354φ,
 0o ≤ φ < 35o
(1)
(44.9 − 6.55log10 (hb ))log10 (dm )(dB), Lori = 2.5 + 0.0075(φ − 35), 35o ≤ φ < 55o (10)
 o o
4.0 − 0.114(φ − 55), 55 ≤ φ ≤ 90 ,

where dm is the distance between the gateway and the end
node in metres, hb is the height of the gateway in meters and where φ is the street orientation angle. For b metres mean
hm is the height of the IoT end node. a(hm ) is a correction separation between buildings, the multi screen diffraction loss
factor that depends on f . For large cities and f ≥ 400MHz; (Lmsd ) given by:
a(hm ) = 3.2log10 (11.75hm )2 − 4.97, (2)
Lmsd = Lbsh + ka + kd log10 (d)
(11)
and for a medium - small city, +kf log10 (f ) − 9log10 (b),
a(hm ) = (1.1log10 f − 0.7)hm − (1.56log10 f − 0.8). (3) where the terms Lbsh , ka , kd and kf denote multi screen
diffraction loss contributions when the gateway antenna is
B. Cost 231-Hata
above the roof top, loss contribution when the antenna is
This propagation model is based on both the Okumura below the roof top, diffraction loss as a function of distance
Model and the Urban Hata model. It extends the Hata model and diffraction loss as a function of frequency respectively.
to cover up to 2000 MHz. Other parameter restrictions of the Expressions for theses loss contributions are given in equations
model include effective base station or gateway antenna height 12, 13 and 14 respectively where △hb = hb − hr .
of between 30 - 200 m, mobile unit’s height of between 1 - (
10 m, inter-device distance of 1 - 20 km. The path loss model −18log10 (1 + △hb ) f orhb > hr
Lbsh = (12)
is given by [11]: 0 f orhb ≤ hr .
L = 46.3 + 33.9log10 (f ) − 13.82log10 (ht ) − a(hm )+
(4) 
(44.9 − 6.55log10 (hr ))log10 (d) + cm [dB], −54
 hb > hr

where ht is the gateway height in metres, hr is the height of ka = 54 − 0.8hb , hb ≤ hr , d ≥ 0.5km (13)
the mobile end node. For rural and suburban environments, 54 − 0.8 △ hb d , hb ≤ hr , d < 0.5km,


0.5
a(hm ) = (1.1log10 (f ) − 0.7)hr − (1.56log10 (f ) − 0.8) (5) 
18, hb > hr ,
and cm = 3dB. For metropolitan centres, kd = △hb (14)
18 − 15 , hb ≤ hr ,
a(hm ) = (1.1log10 (f ) − 0.7)hr − (1.56log10 (f ) − 0.8) (6) hr
and cm = 0 dB.   
f (M Hz)
−4 + 0.7 − 1 , M edium cities


C. Cost 231-Walfisch-Ikegami Model 
 925
kf = 
This model combines the work by Walfisch-Bertoni and the f (M Hz)
−4 + 1.5 −1 , M etropolitan centres,


Ikegami model [12] making considerations for road width, 925
building heights, distance between buildings among other such (15)
parameters. Its restrictions include 800 -2000 MHz frequency For this work, k = 22 m, b = 8 m, hb = 18 m, ka = 54,
operating range f , 4 -50 m gateway height hb , 1 - 3 m end kd = 18, φ = 45 as in [11].
node height hr , 0.02 - 5 km of gateway to end device distance.
D. Ericsson Model
When there is no line of sight between the gateway and the
end device, the path loss given by [12]: This model extends the Okumura Hata model providing
corrections for urban, suburban and rural terrains. The path
L = L0 + Lrts + Lmsd , (7) loss using the Ericsson Model can be expressed as in [13]:
where L0 is the path loss contribution due to free space given L = a0 + a1 log10 (d) + a2 log10 (hb ) + a3 log10(hb )log10 (d)
by:
−3.2log10 (11.75hr )2 + g(f ),
L0 = 32.45 + 20log10 (d) + 20log10 (f ), (8) (16)
Lrts is the rooftop to street diffraction and scatter loss given where a0 , a1 are constants with values provided in Table I
by: and,
Lrts = −16.9 − 10log10 (w) + 10log10 (f )+ g(f ) = 44.49log10 (f ) − 4.78log10 (f )2 (17)
(9)
20log10 (hb − hr ) + Lori ,

Authorized licensed use limited to: Inje University. Downloaded on July 10,2024 at 10:21:51 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE NIGERCON

TABLE I TABLE II
E RICSSON M ODEL PARAMETERS SUI M ODEL PARAMETERS

Environment a0 a1 a2 a3 Model Parameter Category A Category B Category C

Urban 36.2 30.2 -12 0.1 a 4.6 4.0 3.6
Suburban 43.2 68.93 -12 0.1 b (m−1 ) 0.0075 0.0065 0.005
Rural 45.95 100.6 -12 0.1 c (m) 12.6 17.1 20

E. 3GPP Model 180

This model developed by the 3GPP applies to macro cells 160

in an urban and suburban environments. The path loss model
is rendered thus: [14]: 140
−3
L = 40(1 − 4 × 10 hb )log10(d) − 18log10 (hb )
(18) 120
+21log10 (f ) + 80 [dB],
100
where hb is the height of the base station/gateway above the
average roof-top height (in metres), f is the carrier frequency 80
and d is base station/gateway - end device separation in km.
60
F. Stanford University Interim Model
This statistical model was developed by Stanford University 40
1 2 3 4 5 6 7 8 9 10
an extension of the Erceg model [15] for below 11 GHz
frequency bands measurements carried out in selected cities
in USA. The model covers three terrain categories, A, B, Fig. 1. Path loss results for urban environment at 433 MHz
and C with A being at the maximum path loss end of the
spectrum and terrain C at the minimum path loss end of the
180
measurement spectrum. Terrain A is a hilly terrain associated
with the highest path loss. It has tree density that is moderate
160
to heavy. Terrain B is either mostly flat with moderate to heavy
tree density or it is mostly hilly with light foliage densities.
Terrain C is a mostly flat terrain characterised by light foliage 140

density. For the SUI model, the path loss is defined as:
   120
4πd ′

 20log10 f or d ≤ d0
λ  


 100
L = A + 10γlog d (19)
10
d

0


 ′ 80
+ △ Lbf + △Lbh f or d ≥ d ,


where d0 is the reference distance = 100m, γ is the path loss 60

1 2 3 4 5 6 7 8 9 10
exponent and
 
∆Lbf + ∆Lbh
− Fig. 2. Path loss results for urban environment at 868 MHZ
′
d0 = d0 10 10γ , (20)

IV. R ESULTS
 ′ 
4πd0 c In this section, the results of simulations are presented. The
A = 20log10 , γ = a − bhb + , (21)
λ hb general pattern of the graphs show a logarithmic increase in
path loss as the distance is increased, and this is consistent

f
  
h with results in the literature. In Figs. 1 - 3, the path loss
△Lbf = 6log10 and △ Lbh = −10log10 (22) performance of the models discussed in section III are shown
2000 3
for a maximum distance of 10 km between the gateway and the
Note that equation 22 holds for mobile end device antenna mobile end node. For these plots, the SUI model is the most
height, hr ≤ 3 metres. Other parameters are as given in optimistic whereas the COST-Hata model showed the most
Table II. pessimistic results. This trend is also seen for the suburban

Authorized licensed use limited to: Inje University. Downloaded on July 10,2024 at 10:21:51 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE NIGERCON

180 180

160
160

140
140

120
120
100

100
80

80
60

60 40
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Fig. 3. Path loss results for urban environment at 915 MHz Fig. 4. Path loss results for suburban environment at 433 MHz

180

environment. However, for the rural environment, the Ericsson 160

model takes over as the most pessimistic model. The effect of
receiver height is also obvious as there is a general drop in 140

path loss as the height of the mobile device is increased from

120
1-3 metres, the largest being given by the Ericsson path loss
which approximates to about 15dB difference. This is due to
100
the availability of the the Freznel zone as the height of the
mobile increases. The effect of variation of carrier frequency 80
can be observed as there is increased loss in the traversed
path for the same inter-distance as the frequency increases, 60
the highest being for variation from 433 MHz to 915 MHz
than for 868 MHz to 915 MHz. For LPWAN planning in an 40
1 2 3 4 5 6 7 8 9 10
urban environment, a mix of pessimistic and optimistic models
is an approach to follow. For example, choosing Okumura-
Hata, Cost-WI, 3GPP and Ericsson models for the 868 MHz Fig. 5. Path loss result for suburban environment at 868 MHz
carrier frequency in an urban environment will provide a wider
latitude for actual measurement comparison. Hence for IoT
applications such as smart city and smart transport, these V. C ONCLUSION
models can be useful in understanding path loss behaviour.
In this paper, we have shown the theoretical path loss
Figs. 4 - 6 show path loss results for the suburban envi- performance of LoRa using frequently used empirical path loss
ronments. The patterns remain similar to that in the urban models. In doing so, we have considered urban, suburban and
environment except for the jump in path loss of the Ericsson rural environments. The suitability of the path loss models for
model. Urban environments may show high path loss but considered environments were discussed as well. In our future
for suburban environment where less obstacles are prevalent, work, we plan to compare actual drive test measurements with
COST-WI, 3GPP and Okumura - Hata models are attractive. these theoretical results.
For a rural environments, Figures 7 - 9, show the path loss for
the three considered frequencies. Like in Figs. 1 - 9, for the R EFERENCES
3GPP curve there is no distinction for varying antenna height.
[1] K. Mekki, E. Bajic, F. Chaxel, and F. Meyer, “A comparative study of
This is so because in eqn. (18), there is no end node height lpwan technologies for large-scale iot deployment,” ICT express, vol. 5,
dependency. For network planning and model verification in no. 1, pp. 1–7, 2019.
rural environments, the results show that Okumura-Hata, SUI [2] G. Callebaut and L. Van der Perre, “Characterization of lora point-
to-point path loss: Measurement campaigns and modeling considering
and COST-WI models are attractive candidates. IoT applica- censored data,” IEEE Internet of Things Journal, vol. 7, no. 3, pp. 1910–
tions for rural environments could include smart agriculture. 1918, 2020.

Authorized licensed use limited to: Inje University. Downloaded on July 10,2024 at 10:21:51 UTC from IEEE Xplore. Restrictions apply.
 2022 IEEE NIGERCON

180 220

200
160

180
140
160

120 140

100 120

100
80
80

60
60

40 40
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Fig. 6. Path loss results for suburban environment at 915 MHz Fig. 8. Path loss results for rural environment at 868 MHz

220 220

200 200

180 180

160 160

140 140

120 120

100 100

80 80

60 60

40 40
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10

Fig. 7. Path loss results for rural environment at 433 MHz Fig. 9. Path loss results for rural environment at 915 MHz

[3] R. El Chall, S. Lahoud, and M. El Helou, “Lorawan network: Radio 1968.

propagation models and performance evaluation in various environments [11] E. Harinda, S. Hosseinzadeh, H. Larijani, and R. M. Gibson, “Compar-
in lebanon,” IEEE Internet of Things Journal, vol. 6, no. 2, pp. 2366– ative performance analysis of empirical propagation models for lorawan
2378, 2019. 868mhz in an urban scenario,” in 2019 IEEE 5th World Forum on
[4] M. Bor, J. E. Vidler, and U. Roedig, “Lora for the internet of things,” Internet of Things (WF-IoT), 2019, pp. 154–159.
2016. [12] F. Ikegami, T. Takeuchi, and S. Yoshida, “Theoretical prediction of mean
[5] U. Uyoata, J. Mwangama, and R. Adeogun, “Relaying in the internet of field strength for urban mobile radio,” IEEE Transactions on Antennas
things (iot): A survey,” IEEE Access, vol. 9, pp. 132 675–132 704, 2021. and Propagation, vol. 39, no. 3, pp. 299–302, 1991.
[6] J. Cecı́lio, P. M. Ferreira, and A. Casimiro, “Evaluation of lora tech- [13] M. Stusek, D. Moltchanov, P. Masek, K. Mikhaylov, O. Zeman,
nology in flooding prevention scenarios,” Sensors (Basel, Switzerland), M. Roubicek, Y. Koucheryavy, and J. Hosek, “Accuracy assessment and
vol. 20, 2020. cross-validation of lpwan propagation models in urban scenarios,” IEEE
[7] B. Citoni, F. Fioranelli, M. A. Imran, and Q. H. Abbasi, “Internet of Access, vol. 8, pp. 154 625–154 636, 2020.
things and lorawan-enabled future smart farming,” IEEE Internet of [14] ETSI, “Universal mobile telecommunications system (umts); selection
Things Magazine, vol. 2, no. 4, pp. 14–19, 2019. procedures for the choice of radio transmission technologies of the umts
[8] M. Rademacher, H. Linka, T. Horstmann, and M. Henze, “Path loss (umts 30.03 version 3.2.0),” 1998.
in urban lora networks: A large-scale measurement study,” ArXiv, vol. [15] V. Erceg, L. J. Greenstein, S. Y. Tjandra, S. R. Parkoff, A. Gupta,
abs/2109.07768, 2021. B. Kulic, A. A. Julius, and R. Bianchi, “An empirically based path loss
[9] M. Hata, “Empirical formula for propagation loss in land mobile radio model for wireless channels in suburban environments,” IEEE Journal
services,” IEEE transactions on Vehicular Technology, vol. 29, no. 3, on selected areas in communications, vol. 17, no. 7, pp. 1205–1211,
pp. 317–325, 1980. 1999.
[10] Y. Okumura, “Field strength and its variability in vhf and uhf land-
mobile radio service,” Rev. Electr. Commun. Lab., vol. 16, pp. 825–873,

Authorized licensed use limited to: Inje University. Downloaded on July 10,2024 at 10:21:51 UTC from IEEE Xplore. Restrictions apply.