Measurement and Prediction of Indoor Signal Propagation for ISM Band

Nusrat Tanzim, Khandkar M. Rashid, Shazzad Hosain

Department of Electrical Engineering and Computer Science, North South University, Dhaka, Bangladesh
E-mail: nusrat812@northsouth.edu, khandkar130@northsouth.edu, shazzad@northsouth.edu

ABSTRACT
This paper is focused on the measurement and
prediction of indoor signal propagation for ISM band
system in frequency bands 2.4 GHz and 5.3 GHz. In
this research, two basic radio propagation models
are studied and compared with theoretical and
practical data. This comparison result is
implemented on the test indoor wireless network.
Based on the consideration, this paper proposes an
enhancement to the path loss model in the indoor
environment for improved accuracy in the
relationship between distance and received signal
strength. The model can be used as a prediction
model that can be further developed to fit in other
indoor scenarios too.
KEY WORDS: WLAN, ISM
Propagation Model, Path Loss

Band,

Indoor

1. INTRODUCTION
The past decade has witnessed a phenomenal growth
in wireless communication. Over the last few years,
WLANs have gained strong popularity in a number
of vertical markets which have profited from the
productivity gains of using hand-held terminals and
notebook computers to transmit
real-time
information to centralized hosts for data processing.
Today WLANs are becoming more widely
recognized as a general-purpose connectivity
alternative for a broad range of business customers.
Indoor scenarios are usually very complicated and
due to people movement environment changes
rapidly. The need for an efficient way to evaluate
radio propagation in buildings is increasing. It is also
critical to optimize the locations of the base stations
required to ensure satisfactory system performance.
Consequently, radio-propagation prediction for
indoor environments, which forms the basis for
optimizing the location of the base stations, has
become an important research topic. Indoor radio
propagation is not influenced by weather conditions,
such as rain, snow, clouds etc as is outdoor
propagation, but it can be affected by the layout of a
building, and especially by the use of different
building materials. Owing to the reflection, refraction
and diffraction of radio waves by objects such as
walls, windows, doors and furniture inside the
building, the transmitted signal often reaches the
receiver through more than one path. Due to
multipath propagation, where several waves arrive at

the receiver via different paths and with different

phases, rapid variations (fading) of the received
signal envelop occur. Time variations of the received
signal and wide bandwidth of the transmission are
the reasons why the statistical evaluation of
measurement result is necessary.
There are many different ways how the signal
coverage in buildings can be determined. In our
research we focused on a processing of measured
values, the optimization of parameters for COST231
Multi-Wall model, which allow mean signal level
prediction for initial coverage planning.
For simulation and analysis of different propagation
model, number of test wireless networks has been
created to meet the requirement of basic propagation
mechanism. To observe and aid the simulation
process we used Frequency Spectrum Analyzer
(SPECTRAN HF-2025E). With spectrum analyzer we
took many test point data from our test-bed wireless
networks to identify different variations in our test
parameters. Test network mapping has been done
with the help of GIS mapping software to feed map
information into Ekahao Heat Mapper software,
which generates SNR information as gradient over
the map. For simulation hardware preparation, we
modified some commercially available ISM band
wireless antennas to serve our purpose and frequency
needs. Moreover, to generate user defined wireless
output power we used customized access point
firmware that enabled the facility to have more
precious control over transceiver device.
2. PROBLEM ANALYSIS
The existing challenge for indoor environments is
that the signal propagated from the transmitter
antenna will experience many different signal
transformations and paths with a small portion
reaching the receiver antenna. Awareness of this
process will assist the user in the process of better
understanding radio performance limitations.
The indoor propagation channel differs considerably
from the outdoor one. The distance between
transmitter and receiver is shorter due to high
attenuation caused by the internal walls and
furnitures. Most often, the lower transmitter power
is also a cause of it. The short distance implies
shorter delay of echoes and consequently a lower
delay spread. The path loss and the statistical
characteristics of the received signal envelop play an
important role in coverage planning applications.

The considered propagation models in the literature

are divided into two groups: empirical models and
deterministic models. Earlier models were expressed
in form of simple mathematical equations that give
the path loss as the output. The equations were
obtained by fitting the model measurements results at
2.451GHz.

3.2 Data Collection Preparation

Access point SSID was set as NSU-test-network,
operating at channel 11 (start frequency: 2451 MHz,
Center frequency: 2462 MHz, Stop frequency: 2473
Mhz) with access point output power: 150 mW.
At Spectrum analyzer, all the frequency parameters
were synchronized to the access points operating
frequency as shown in Fig. 4(a).

Fig. 1, Overlapping 2.4 GHz Channels in ISM band

Between these two groups, the empirical ones
minimize the computation time in order to give
results even if their accuracy is low. For this reason,
the empirical models have been adopted to be
introduced in the WLAN models. [1], [2], [3]
3. DATA COLLECTION AND MEASUREMENT
CAMPAIGN
3.1 Site Survey
Site survey campaign was performed on the 10th
floor of the North South University South Academic
Tower. It is a typical modern multi-floored building.
Many walls and partitions divide measured floor and
there are four metal lifts in the scenario as well (Fig.
2). Access point was mounted at the roof with two
Omni-directional 12 dB antennas whose radiation
pattern is shown in Fig. 3.

Fig. 4(a), Spectrum analyzer preparation

3.3 Narrowband Data Collection and Measurement
A narrowband system was developed at the
Department of Electrical Engineering and Computer
Science consisting of one Access Point, USB
transceiver, Spectrum Analyzer and two wire Omnidirectional ground-plane antennas. The system was
designed for 2.451 GHz ISM band. Full computer
control is made through the USB interface o f a
notebook computer. A value of the signal level was
taken automatically every second and saved in the
computer while the receiver was moving along a
measured path. More information about system and
measurement method can be found in [4], [5], [10].

Fig. 4(b), Measurement with SPECTRAN HF-2025E

spectrum analyzer

Fig. 2, Floor layout with signal level distribution

Fig. 3, Radiation Pattern of Omni Antenna

3.4 Wideband Data Collection and Measurement

A simple Wireless LAN peer-to-peer connection
was built up using two notebooks - both supplied
with identical standard WLAN USB cards [8].
Omni-directional antennas are integrated in the
cards. Ekahau Heat Mapper and Net Stumbler
software package was used to measure the signal
power level and QoS.
In the case of narrowband measurements, one of the
notebooks was situated at the same positions as the
transmitter. The second notebook was placed with
the help of tripod; so that it can be moved around
the azimuth in all 360 degrees may have been

scanned with 5-degrees step. At several locations,

the signal power level was measured for a few
hours. Three of the locations are shown in Fig. 14.
The main goal of the wideband measurement was to
find cumulative distribution functions of the received
signal level for each location. These functions
describe time variations of the received signal. It can
be used for calculations of the signal level above the
mean for an acceptable percentage of time. This way
the corresponding QoS is determined. Detailed
information on used calculation procedures can be
found in [6], [7].
4. MODELING METHODOLOGIES
Software planning (using a propagation model) is
much more convenient and cost-effective way to
deploy a wireless network than a site survey with lots
of measurements and empirical decisions. Using
simulations many different configurations of the
network can be tested with no expenses to find an
optimal solution.
As it was stated the indoor propagation modeling is
one of the most complicated tasks in this field. In
addition, a detailed description of an indoor scenario
including furnishing, doors, constitutive electrical
parameters of used materials etc. is almost
impossible.
To find a balanced trade-off between the model
complexities (computation time, requirements on
input data etc.) and reasonable accuracy is a
challenge. Quite a large number of indoor
propagation models can be found in literature; among
them we are using Cost 231 one slop and multi wall
model.

Where L(P) (dB) is the average loss based on the

position P only, and (dB) is random fading with a
zero-mean statistical distribution. The empirical and
semi-empirical models are able to predict the average
path loss L(P). The random fading has to be
considered as a fade margin in the power budget of a
wireless link.
4.1.1 One-Slope Model
The One-Slope Model (1SM) [11] is the easiest way
to compute the average signal level within a building
without detailed knowledge of the building layout.
The path loss in dB is a function of just a distance
between transmitter and receiver antennas:
(2)
Where L0 (dB) is a reference loss value for the
distance of 1 m, n is a power decay factor (path loss
exponent) defining slope, and d (m) is a distance. L0
and n are empirical parameters for a given
environment, which fully control the prediction. As
an example table. 1 presents a few values taken from
various reference points.
4.1.2 Multi-Wall Model
A semi-empirical Multi-Wall Model (MWM)
provides much better accuracy than 1SM. The results
are site-specific but at the same time floor plan
description is needed as an input.

Fig. 6, Multi Wall Model Geometry

The basic idea of MWM is illustrated in Fig. 6. The
path loss between a transmitter and receiver LMW is
given by
(3)

Fig 5, Experiment Geomentry

4.1 Empirical Modeling Approach
The empirical and semi-empirical models are
primarily based
on
statistically processed
representative measurements. One-Slope and MultiWall model are very easy and fast to apply because
the prediction is usually obtained from simple closed
expressions. Also requirements on the input
environment description are reasonable. But, at the
same time, only the propagation loss without great
site-specific accuracy can be predicted.
Total path loss LTOT (dB) can be expressed,
(1)

where LFSL (dB) is the free space loss for the distance
d (m) between transmitter and receiver antennas,
which is in fact1SM prediction with power decay
factor n = 2.0, k wi is a number of walls of i-th type
between transmitter and receiver antennas, Lwi (dB)
is attenuation factor for i-th wall type, N is a number
of wall types, kf is a number of floors between
transmitter and receiver and Lf (dB) is the floor
attenuation factor. Since the floor attenuation is not
dealt with in this paper the original MWM [11] floor
attenuation calculation was simplified in above
equation. Floor attenuation analysis can be found in
[4].

5. DISCUSSION OF RESULTS
5.1 One Slope Model Data Processing
It can be clearly seen that the value of the power
decay factor n is highly dependent on the type of
building or structure of the indoor environment and
so it has the major influence on the resulting
determination of the signal level coverage. A typical
example of a coverage prediction using 1SM is
shown in Fig. 7.
Comparing Fig. 2 and Fig. 7 it is apparent that 1SM
prediction considers only the change of the signal
level with distance between transmitter and receiver
regardless of the actual structure of the indoor
environment. The 1SM provide only a rough estimate
(standard deviation usually greater than 10 dB) and
the selection of proper power decay factor n is
crucial.

A transmitter was located on the left side in the

corridor so that strong wave-guiding effect along the
corridor can be observed in Fig. 8. Then two 1SM
predictions were performed and compared with the
measurement. For corridors n = 1.4 was utilized (Fig.
9) and very good agreement between measurement
and prediction was seen in the corridor. On the other
hand very poor prediction accuracy was provided for
the offices since the signal attenuation was much
stronger than in the corridor and the prediction was
overestimated. If the n = 4.0 suitable for office
environment was used instead (Fig. 10), the coverage
prediction was perfect in the offices but strongly
underestimated in the corridor. To handle this
problem either the different parameter n can be used
for corridor and offices or an averaged n value
between 1.4 and 4.0 can used as a tradeoff, then the
prediction will be valid for the whole floor but at the
lower accuracy.

Fig. 7, One Slope Model Coverage Prediction

f

Lo

[GHz]

[dB]

1.9

38.00

[-]

1.9

Comment

3.5

Office Space

38.00

Open Space

1.9

38.00

1.3

Corridor

2.45

40.2

1.2

Corridor

2.5

40.0

3.7

Office Space

2.45

40.2

4.2

Office Space

Fig. 8, Measured Signal Level

Fig. 9, Prediction error (difference between 1SM

prediction and measurement) for n=1.4

Table 1, One Slope Model Empirical Parameters

The values of the power decay factor n vary
depending on the type of building and indoor
environment. The value n =2 corresponds to the
propagation in free space. Values smaller than 2 are
utilized for prediction of the signal propagation in
corridors, where the decrease of the power decay
factor is caused by a wave-guiding effect. In an
office environment with walls and furniture n is
usually between 3 and 6. The 1SM gives the best
results for environment with more or less uniformly
distributed walls and obstacles.
The 1SM performance and the importance of proper
parameter n selection are demonstrated in Fig. 8, Fig.
9 and Fig. 10.

Fig. 10, Prediction error (difference between 1SM

prediction and
In spite of the strong dependence of 1SM on used
empirical parameters, it provides an excellent tool
when no information on an indoor scenario is
available or when a very fast draft design is needed.
5.2 Multi-Wall Model Data Processing
Fig. 11 presents an example of a coverage prediction
using MWM. The MWM can be marked as sitespecific since particular walls are considered during
the prediction. But still, it must be understood that
the MWM introduces only an estimate of the real

wave propagation. In Eq. 3, only walls and obstacles

located directly between transmitter and receiver are
considered with their attenuation factors. Particular
reflections and diffractions are not taken into account
so the accuracy is limited in certain cases. As an
example the wave-guiding effect of bending corridor
cannot be modeled in Fig. 11.

Fig. 11, Multi Wall Model Coverage Prediction

A comparison with the 1SM shows a significant
improvement of the site-specific accuracy; see Figs.
8, 9, 10, 12. No change of model parameters is
needed.
For good prediction accuracy the proper wall
attenuation factors Lw - empirical parameters for Eq.
3 must be used. The attenuation factors do not
represent actual physical attenuations of the walls but
statistical values obtained from representative
measurement campaigns. It means if the receiver is
hidden behind a metal wall with limited dimensions,
the prediction cannot result in an infinite attenuation,
even so metal itself can be considered as a total
reflector of the electromagnetic energy. But in the
real scenario the wave can find its way around the
metal obstacle due to reflection, diffraction and
diffuse scattering, while the MWM considers only
walls along a line connecting the transmitter and
receiver.

Lo

L1

L2

L3

[GHz]

[dB]

[dB]

[dB]

[dB]

[-]

1.9

38.0

2.1

4.4

13.6

1.9

38.0

0.5

4.2

2.45

40.2

5.9

8.0

2.45

40.2

6.0

2.5

40.0

5.4

Comment
Office
Space
Half Open
Space
Office
Space
Office
Space
Dry Wall

Table 2, Multi Wall Model Parameters

As a result
modeling is
approach for
scenario to
solutions.

the planning based on propagation

recognized as a highly preferable
designing large WLANs in an indoor
provide optimal and cost effective

Fig. 13, Multi Wall Model Data ProcessingPrediction Compared to Measurement

Fig. 12, MWM prediction error - difference between

MWM
Even though there are a lot of building materials, due
to the statistical nature of the wall attenuation factors
in Eq. 3, only a very few wall types are necessary to
define for MWM. In fact in [11] only two wall types
are considered: Light wall (L1) - a light wall or
partition, and Heavy wall (L2) a structural thick
wall. Of course more wall types can be introduced
for a specific application or software tool (metal
walls, glass, etc.). Some empirical parameters of
MWM for miscellaneous types of interiors are
summarized in Table 2.

Fig. 14: Minimum received signal power level for

probability of 50, 90 and 99% as functions of the
notebook location and orientation
The cumulative distribution functions of the
received signal power level in each location of the
observation points were investigated with respect to
the tripod top rotation. It defines minimum received

signal level for probability of 50, 90 and 99% as

functions of the measurement notebook location and
orientation. The influence of the notebook azimuthal
orientation on the standard WLAN transmission can
be nicely demonstrated. Using the results, the gain
of eventual angle diversity can be determined as
well.

REFERENCES
[1] COST (European Co-operation in the Field of
Scientific and technical Research), COST 231 Book,
Final report, Chapter 4, propagation Prediction
Models.

6. COMPARISON WITH LOG DISTANCE

PATH LOSS MODEL
Log distance path loss model [8] predicts the signal
strength without considering site environmental
factors and mainly focuses on outdoor models. So
outcome from this model becomes inapplicable and
inaccurate for no line of sight cases. This is
unsuitable for indoor 802.11 networks.

[3] Winprop Documentation: Software tool for the

Planning of Radio Communications Networks
(Indoor),

The One slope and Multi wall empirical propagation

model considers attenuation in an indoor site
environment by including various wall attenuation
factors. This provides increased accuracy in both line
of sight and non line of sight cases.
Variation between predicted signal strength by Log
Distance Path Loss model[9] that we considered is as
much as 9 dB for the case of non line of sight. So, the
accuracy of our prediction is much higher with a
variance of 10% compared to the Log Distance
Path Loss Model with a higher variance of around
20% between predicted and measured actual signal in
the indoor site environment.
7. CONLUSION AND FUTURE WORK
Measurements taken at North South University,
Dhaka were compared against predictions made by
two empirical propagation models.The cost 231
model, in general over estimated the path loss,
especially at greater antenna heights.
In our study, the 10th floor of NSU south academic
tower was investigated for the propagation prediction
to see how the signal will be affected in such
complex floor with different types of building
material. This study can be carried out in the future
to see how the multi-floor factor can affect the signal
by analyzing study for other multistory building.
In future, the signal propagation prediction can be
extend to larger area with a number of single and
multistory buildings and simulate for outdoor
transmitters to indoor receivers and vice versa.
Comparison between different indoor propagation
models can also be done. And from the study a new
channel model can be developed which could be
used for wireless networks to provide optimal
performance in a local indoor environment.

[2] J.S Lee and L.E Miller, CDMA Systems

Engineering Handbook, Boston: Artech House,1998

http://www.awe-communications.com
[4] PECHA, P., KLEPAL, M., ZVNOVEC S.,
Results of Indoor Propagation Measurement
Campaign at 1900 MH, Radioengineering, vol. 10,
no. 4, December 2001, pp. 2-4.
[5] ZVNOVEC, S. Pokryt pikobunk signlem,
GSM, Diploma Thesis,
Department of
Electromagnetic Field, CTU Prague, January 2002,
(in Czech).
[6] SAUNDERS, S. R., Antennas and
Propagation for Wireless Communication Systems,
John Willey&Sons, Ltd, 1999.
[7] PARSONS, J. D., The Mobile Propagation
Radio Channel, 2nd Edition, John Wiley and Sons,
London, 2000.
[8]
Theodore
S.
Rappaport.
Wireless
Communications: Principle and Practice, Prentice
Hall, 2nd edition 2002, Ch-4.
[9] A.R. Sandeep, Y. Shreyas, Shivam Seth, Rajat
Agarwal, and G. Sadashivappa, "Wireless Network
Visualization and Indoor Empirical Propagation
Model for a Campus WI-FI Network", World
Academy of Science, Engineering and Technology,
vol. 42, 2008.
[10] Y.Wang,X.Jia,H.K.Lee, An Indoor positioning
system based on wireless local area network
infrastructure , 6th international symposium,
SATNAV 2003.
[11] Empirical Propagation Model for Indoor
Scenario,
http://www.awecommunications.com/Propagation/In
door/Empirical/index.htm