www.atdi.

com

LoRa Simulation using ATDI’s RF Engineering Solution

ABSTRACT Yahya Khaled / Telecom Engineer

This white paper discusses general overview of LoRa, and the ATDI South Pacific Pty Ltd
network planning approach in ATDI’s RF engineering tools. It
07. 2016
especially addresses the issues in dense urban environment with
high-rises impact in such network simulation. ATDI uses full set of
ahya Khaled / Telecom Engineer
3D GIS input and deterministic propagation models which are
proved in numerous use cases over three decades and validated ATDI South Pacific Pty Ltd
its accuracy in field measurements for urban environment. These
models are described as path-specific, which is a perfect candidate 07. 2016
for P-MP non-mobile scenarios, typically deployed for IoT
applications.
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

TABLE OF CONTENTS

LPWAN (LOW-POWER WIDE AREA NETWORK) 2

LORA® AND LORAWAN™ 3

LORA® PHYSICAL LAYER 3

NETWORK MODEL 3
KEY CHALLENGES 3

PROPAGATION MODELS 5

EMPIRICAL MODELS AND THEIR REQUESTED TUNNING 5

DETERMINISTIC MODELS AND PLANNING MARGINS 5
THE LIMITATION OF STATISTICAL MODELS WITH HR DATA 5
THE DIFFRACTION EFFECT 6
THE SUB-PATH ATTENUATION EFFECT 8
3D RAY TRACING MODELING IN ATDI PLANNING TOOL 10

ATDI’S LORA® NETWORK PLANNING APPROACH 11

END-POINT PATH SPECIFIC 11

BUILDING IMPACT PREDICTION 12
LINK ADAPTATION MODELING 13
CO-EXISTENCE ANALYSIS 15

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

1|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

LPWAN (LOW-POWER WIDE AREA NETWORK)

Low-power WAN (LPWAN) is a wireless wide area network technology that is specialized for
interconnecting devices with low-bandwidth connectivity, focusing on range and power efficiency.

It is designed for machine-to-machine (M2M) networking environments. Such network requires

decreased power, longer range and lower cost compare to commercial mobile network. The networks
can also support more devices over a larger coverage area than consumer mobile technologies and
have better bi-directionality.

LPWAN is a key factor for large-scale communication networks typically required in Smart-City
applications and Internet of Things, known as IoT. Recently, narrowband and spread-spectrum
technologies surfaced as cost-effective candidate technologies to fulfil low throughput and long-range
communication.

Bluetooth, ZigBee and Wi-Fi are adequate for consumer-level IoT implementations. The need for a
technology such as LPWAN is much greater in industrial IoT, civic and commercial applications. In
these environments, the huge numbers of connected devices can only be supported if
communications are efficient and power costs low.

The most critical factors in a LPWAN are 1:

- Network architecture

- Communication range
- Battery lifetime or low power

- Robustness to interference
- Network capacity (maximum number of nodes in a network)
- Network security

- One-way vs two-way communication

- Variety of applications served

Local Area Network (LAN) Low Power Wide Area Cellular Network
Short Range Communication Internet of Things Traditional M2M
Bluetooth, WiFi, ZigBee LoRa/NB GSM, 3G, 4G
Well established standard in Lowe power consumption, Low Existing coverage, High data
building cost positioning rate
Battery live, Provisioning Low data rate, Autonomy,
Network cost and Emerging standards Total cost of ownership is high
dependencies

1
Section 3, Technical Overview of LoRa® and LoRaWAN™, Technical Marketing Workgroup 1.0, Lora Alliance
Org. (Nov. 2015)

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

2|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

LORA® AND LORAWAN™

LoRaWAN™ is a LPWAN specification intended for wireless battery operated Things in regional,
national or global network. LoRaWAN target key requirements of internet of things such as secure bi-
directional communication, mobility and localization services. This standard has been widely spread in
IoT industries with its strong interoperability among smart Things with the benefit of low demand in
complex local installations.

LoRa® is the wireless modulation to create long range communication link. It is based on chirp
spread spectrum modulation, which maintains the same low power characteristics but significantly
increases the communication range.2

The main advantage of LoRa® is in its long range capability, which allows a single gateway or base
station to cover entire cities or hundreds of square kilometres. LoRa® and LoraWAN™ provides larger
link budget compare to other technologies in environment or obstruction of a given location. 3

LORA® PHYSICAL LAYER

LoRa is a spread-spectrum modulation scheme that that uses wideband linear frequency modulated
pulses whose frequency increases or decreases over a certain amount of time to encode information.

The main advantages of this approach are twofold: a substantial increase in receiver sensitivity due to
the processing gain of the spread spectrum technique and a high tolerance to TX and RX frequencies
misalign.

NETWORK MODEL

LoRa is simply a transmission modulation; the technology can be applied to any network model such
as Mesh, P2P and P-MP such as star topology. The network components are the Gateway and
Endpoints. The end points can be sensors for different applications.

KEY CHALLENGES

Regardless whether LoRa or NB wireless system - there are RF limitations inherited from the use-case
itself. The End-points (EP) can be deep-indoor, in the basement and even under-ground. Most
noticeable challenge is the EP and/or the gateway can be as little as 2m height pushing additional
pressure to RF simulators and link-level predictions.

END-POINT INSTALLATION LOCATION AND VARIABLE HEIGHT REQUIREMENT

Classical mobile coverage planning has typically addressed outdoor and indoor planning requirements
separately by deploying macro-sites, micro-sites and DAS for in-building. Accordingly, statistical
propagation models have evolved to assist in area planning. M2M communication networks are P-MP
in nature with sensors installed at different floors including basement. These requirements push

2
Section 2, Technical Overview of LoRa® and LoRaWAN™, Technical Marketing Workgroup 1.0, Lora Alliance
Org. (Nov. 2015)
3
Section 2, Technical Overview of LoRa® and LoRaWAN™, Technical Marketing Workgroup 1.0, Lora Alliance
Org. (Nov. 2015)

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

3|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

network planners to adopt 3D digital maps and apply path-specific full deterministic propagation
models.

BUILDING IMPACT IN URBAN ENVIRONMENT

Smart-City means large buildings and skyscrapers. Such environment has been hard to manage due
to its traffic density nature and short signal travel distance. Planning GW station location can be a
challenge due to vast obstructions in the vicinity of the transmitter leaving network planners
completely blind to the overall coverage footprint in such environment. Every candidate location can
produce a completely varying coverage. Hence, LoRa network simulation model must support high-
resolution 3D building layer and be able to predict outdoor diffraction and indoor signals with high
precision.

LINK ADAPTATION MODELLING

LoRa supports multiple spreading factors from 6 to 12. With 12 being the most robust but also the
longest in terms of air-time occupancy. Having large number of End-Points with such link mode can
completely exhaust the GW resources and hence deliver the wrong network design. Network
dimensioning should rely on accurate signal level prediction to work out the SF distribution amongst
target End-Points.

INTERFERENCE CALCULATIONS AND DIMENSIONING

Interference protection is not guaranteed with LoRa targeting unlicensed bands. “Polite transmission
protocols” are encouraged for all devices. With LoRa promising many kms of coverage – practical
range can be less than what is expected due to current band utilization and overall spectral noise
density.

CO-EXISTENCE WITH 2G/3G/4G

Lack of international and/or regional harmonization for ISM bands can be a threat to budget receiving
devices. Manufacturers have to support US, Europe, AUS and APAC markets. Improving receiver’s
selectivity for a specific market is an additional challenge for these devices before they undergo and
large scale deployment. For instance, US ISM band on 900 MHz extend to GSM900 in Singapore
creating an overlap that can potentially overload small devices even if operating on adjacent band
due to poor selectivity and high sensitivity.

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

4|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

PROPAGATION MODELS

EMPIRICAL MODELS AND THEIR REQUESTED TUNNING

Empirical models model the environment as a series of random variables. These models are the least
accurate but require the least information about the environment and use much less processing
power to generate predictions. An example of these types of model are the Stanford University
Interim (SUI) channel models developed under the Institute of Electrical and Electronic Engineers
(IEEE) 802.16 working group. These models are not available on purpose in ICS Telecom: medium
resolution cartography can indeed be processed very easily (from SRTM/Landsat data for instance),
making this propagation modeling without detailed cartography not accurate enough with regards to
the results that could be obtained using other models. Other examples of empirical models is COST-
231 Hata model. Although empirical propagation models for mobile systems have been
comprehensively validated (mainly macrocell 2G/3G planning, but not for IoT analysis), it has not
been established if they are appropriate for M2M systems.

These models are less dependent on the quality of the cartography: they try to re-create the urban
environment and the resulting mean path loss using typical inputs such as the distance, the average
building height (giving by the clutter file), the average street width…

The cartographic dataset loaded in ICS Telecom will differentiate the signal propagation between
downtown Hong Kong or in a medium size French city using a deterministic model, whereas it is the
tuning of the empirical model itself that will make the difference. Requiring less cartographic input is
a major asset for the empirical models, but their main drawback is the fact they require tuning in
order to be accurate. And this model-tuning phase cannot be achieved without accurate
measurements, that need to be performed according to the same technology and the same urban
environment as the one that will be simulated afterwards.

DETERMINISTIC MODELS AND PLANNING MARGINS

The deterministic models make use of the laws governing electromagnetic wave propagation to
determine the received signal power at a particular location. They require a 3-D map of the
propagation environment: the more compatible the accuracy of the cartography with a certain
technology to simulate, the better the coverage accuracy (for a given set of technical parameters for
the base stations / Terminals / CPEs). Typical examples are the ITU-R 525/526 models, used with
appropriate additional propagation effects (diffraction, sub-path attenuation and 3D ray tracing).

Depending on the type of technology to simulate, the receiver can be placed either above the urban
clutter codes (Fixed Wireless Access type of networking), or "dug" into the clutter. In this case,
attenuation associated to the signal strength received at each pixel will be attenuated based upon the
selected diffraction model.

THE LIMITATION OF ST ATISTICAL MODELS WIT H HR DATA

Empirical models are used in order to simulate by mathematical terms topographical characteristics
that are not available on the cartographic dataset used as a basis for the propagation calculation,
such as the average height of the buildings in the area, the width of the streets… All of these are
already available in a High Resolution cartographic dataset, making the characteristics of the

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

5|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

empirical model redundant with the cartographic dataset itself. The urban environment is described
as close to reality as possible, making deterministic models much more efficient in terms of accuracy
than empirical models when HR data is used.

THE DIFFRACTION EFFE CT

The diffraction effect is used in order to quantify the attenuation due to an obstruction of the direct
path between the Tx and the Rx by one or several obstacles.

In the Fresnel theory, the attenuation due to one single knife-edge obstacle located in the free space
path can be derived using Fresnel Integrals. As these integrals do not have any explicit solution, a
good approximation of this knife-edge diffraction loss can be :


Ld 6.920.log  0.1 1 0.1² 
where  h
r
The fraction h/r, called the clearance ratio, is the ratio of the algebraic height (positive upward) of the
edge above the line of sight over the radius of Fresnel ellipsoid at distance d from the Tx (see the
drawing next page).

Different diffraction methods offer specific ways to evaluate one  (single obstacle diffraction) or
several  (multiple obstacle diffraction) according to a path profile.
For example, Deygout proposed in 1966 a diffraction method that takes 2 obstacles into account: a
primary obstacle (obtained from the maximum clearance ratio 1 according to the entire line of sight
between Tx and Rx) and, if this primary obstacle exists (1>0), a secondary obstacle (obtained from
the maximum clearance ratio 2 according to :

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

6|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

 the line of sight between the Tx and the primary obstacle

 The line of sight between the primary obstacle and Rx.
The global diffraction loss is then given by Ld’=Ld(1)+Ld(2). This method
provided better estimations than the diffraction method proposed by
Bullington, but is still slightly optimistic.
In 1994, Deygout presented a generalized improvement of this method
using a potentially infinite number of edges. The search for the edges is
sequential : if the primary obstacle exists, two secondary obstacles (one
between Tx and the obstacle and the other one between the obstacle and
Rx) are searched .
Then, the same process is performed again on each side of the secondary obstacles possibly looking
for tertiary obstacles. This process is reiterated recursively (the n+1 obstacles depend on the n
obstacles) until no new obstacle is found. Then, the global diffraction loss is :

Ld Ld( i)
i

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

7|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

THE SUB-PATH ATTENUATION EFFECT

Selecting only a deterministic calculation method, corrected by a diffraction term generates a prediction
that is too optimistic. For deterministic predictions, an additional term is introduced, called the sub-
path attenuation effect. This correction term is directly derived from surface reflection modeling for
low incident angles. It is also called ground reflection attenuation Lgr. It represents the attenuation due
a partial obstruction of the Fresnel, whereas the Tx and the Rx are in visibility one with each other. It
is representative of a first attenuation method in order to take into account ground/building multi-path
effect, OFDM or not.

IN THE Z PLANE ONLY

A first option can be defined in order to take into account the sub-path attenuation effect in the Z plane
only.

Different methods can be selected in ICS Telecom, depending on the technology to simulate and the
cartography available. All of them quantify the sub-path effect by multi-reflections according to an
evaluation of the amount of terrain penetration into a fraction FZ of the Fresnel ellipsoid.

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

8|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

Such a calculation methodology is particularly suitable for fixed wireless type of technologies, where
the receivers are usually placed on the rooftops. The main source of penetration of the Fresnel ellipsoid
is then all the buildings located between the transmitter and the receiver.

CONSIDERING THE FRESNEL ZONE IN 3D

However, this sub-path methodology might be too pessimistic for Rx placed within buildings that are
taller than the one on which it is setup, or for mobile technologies, where is the receiver can be in the
street.
In that case, the ICS Telecom user has the choice to consider the potential penetration of the Fresnel
zone not only in the Z plane, but also in its "sides".

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

9|P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

3D RAY TRACING MODELING IN ATDI PLANNING TOOL

In ICS Telecom, different ray tracing modes are available in order to take into account the 3D reflected
paths along with the direct path.

REFLECTION TYPE
The radio-planner first selects what kind of reflected paths will have to be taken into account:

 The Specular mode takes into account the

direct path, and all reflections on building
sides. The reflected point radiates with the
same incidence angle as its input signal.

 The Lambertian mode takes into account

the direct path, and all reflections on building
sides. The reflected point radiates in all
directions, in 3D. Due to the amount of
calculations to perform, this mode is rather
slow.

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

10 | P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

ATDI’S LORA® NETWORK PLANNING APPROACH

END-POINT PATH SPECIFIC

ATDI adopts a set of full 3D and deterministic propagation models proven in case-studies and
validated by field measurements for urban environment. These models are described as path-specific.
Unlike classical models such as Hata which is typically used for macro-coverage predictions and street
level mobile receivers. Deterministic model such as ITU-R P.525 + Deygout94 is governed by rules of
physics and require full descriptive RF environment such as digital terrain model, clutter and
buildings.

Path specific models are the perfect candidate for P-MP non-mobile scenarios typically deployed for
the IoT.

Figure 1: Full 3D P2P outdoor to indoor prediction using path-specific propagation model

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

11 | P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

BUILDING IMPACT PRED ICTION

Buildings have two-fold impact; shadowing and absorption. Thanks to our dedicated building layer –
the tool is able to distinguish outdoor and indoor receivers and switch accordingly. Signal crossing
multiple buildings can undergo diffraction on few buildings and absorption on the last building where
the EP is installed. Such effect is only possible with fully-deterministic model and separate building
layer model. See Figure 1 for full predictions at every level. Separate clutter classes can be defined
for under-ground installations. Also see Figure 2 for building impact on outdoor signal level.

No signal

Figure 2: Building impact on LoRa spreading-factor performance using path-specific model

LoRa Gateway reserves SF12 mode for first transmitted packet and worst links to guarantee service
initiation and maintain the session for poor RF links. While this is tolerable for first few packets;
maintaining the link with higher SF means longer transmission duration, longer queues and poor
handling of resources. This will also effect the overall capacity of the serving Gateway. After first
transmission, the gateway signals the sensor to switch to better spreading-factors if permissible by
the RF link-level. The network design would ideally target lower spreading factors to increase
efficiency and to cater for the daily traffic demand. Urban environment, see Figure 2, can be heavily
shadowed outdoor with varying signal levels that would require high spreading-factors especially
indoor. With most of the sensor nodes being indoor or outdoor in fixed locations; it is of high
importance to predict the path-specific link level for every node taking into account all obstacles
within the path as described in the section “Deterministic propagation models”.

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

12 | P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

LINK ADAPTATION MODELING

LoRa overcome some of the urban challenges by adopting 7 different spreading factors (SF) and 4
coding-rate scheme (see AN1200.22). Varying SF result in different SNR and packet duration. While
such scheme has its merits and de-merits – it’s necessary to simulate this behaviour while trialling the
network in preparation to the full-scale deployment. Figure 3 and Figure 4 show coverage ranges and
adaptive link performance for suburban and urban environments respectively.

2km

1km

Figure 3: LoRa GW coverage in suburban environment (GW 5m AGL, EP 3m AGL, 24dBm EIRP)

Figure 3 shows LoRa spreading-factor adaptation outdoor and indoor in a suburban environment
taking into account 3D building layer. The Gateway is installed 5m AGL with End-Points being 3m
AGL. Coverage include interference margin and reliability calculations and outdoor to indoor
penetration when applicable.

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

13 | P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

1.5km

1km

Figure 4: LoRa GW coverage in heavy urban environment (GW 5m AGL, EP 3m AGL, 24dBm EIRP)

Figure 4shows LoRa spreading-factor adaptation outdoor and indoor in a heavy urban environment
taking into account 3D building layer. The Gateway is installed 5m AGL with End-Points being 3m
AGL. Coverage include interference margin and reliability calculations and outdoor to indoor
penetration when applicable.

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

14 | P a g e
WHITE PAPER 07/2016
LoRa Simulation using ATD’s RF Engineering solution

CO-EXISTENCE ANALYSIS
LoRa’s blocking immunity is roughly 82.5dB, 86.5dB and 89dB for 1, 2 and 10 MHz respectively (see
sx1272 datasheet). The immunity is measured in reference to LoRa’s receiver sensitivity plus 3dB.
The presence of strong near-by transmitter can compromise the performance of the receiver by
overloading the RF front. The receiver is not necessarily optimized for the hosting band-plan given
the discrepancy in band plan adoption amongst regions and countries.

See calculations below for LoRa 125kHz, SF=12 co-existing with 2G carrier being at +/- 2MHz:

 Interference threshold due to blocking = Lora Sensitivity + 3dB + LoRa_Blocking_Immunity

 Interference threshold due to blocking = -137dBm + 3dB + 86.5
= -47.5 dBm
Please refer to sx1272 datasheet for parameters above.

With GSM transmitting typically in the order of 100W EIRP with 86dB MCL for urban environment.
3GPP estimates 70dB MCL including antenna gain (4.5.1 3GPP TR 36.942). Hence GSM power
received is (50dBm - 86 dB) = -36dBm.

Additional protection required = -36dBm – (-47.5dB) = 11.5dB

LoRa Blocking Immunity failure due to GSM

Impacted locations with 2 MHz gap

Impacted locations with 1 MHz gap

Figure 5: Impact of GSM into LoRa (Urban GSM at 100W EIRP, 2 MHz away and LoRa at 125kHz BW and SF12)

Paris – London – Washington – Madrid – Moscow – Sydney – Kiev – Warsaw - Mumbai

15 | P a g e
www.atdi.com