You are on page 1of 5

Channel Models for IEEE 802.

11b
Indoor System Design
Andrea Borrelli, Cristiano Monti, Marco Vari Franco Mazzenga
Consorzio Università Industria Dipartimento di Ingegneria Elettronica
Laboratori di Radiocomunicazioni University of Rome Tor Vergata,
(RadioLabs) Via del Politecnico 1, 00133, Roma, Italy
Email: cristiano.monti@radiolabs.it, Email: mazzenga@ing.uniroma2.it
marco.vari@radiolabs.it

Abstract— Wireless local area networks (WLANs) based on affect the overall attenuation of the direct path. Modern office-
IEEE 802.11b standard are widely deployed to provide network like and classroom-like environments have been considered.
connectivity without being tethered off of a wired network. The developed channel models were subsequently used in a
An accurate planning of indoor radio networks and/or the
setup of localization procedures based on IEEE 802.11 requires computer tool for the clever positioning of the IEEE 802.11b
the characterization of the propagation channel. Starting from access points (APs) inside the buildings.
experimental data obtained in the campus area of the University The paper is organized as follows. In Section II we re-
of Rome Tor Vergata we evaluated the parameters for different view the indoor models adopted in this paper for path loss
multi-wall (MW) like path loss models. Differently from the characterization. In Section III we describe the measurement
standard MW model indicated in the literature, we included
losses due to doors and fire proof doors that according to their setup and the fitting procedure used to evaluate the models’
status, open or closed, may lead to significant contributions to parameters. In Section IV we provide the multi-wall channel
the overall attenuation. The developed channel model was used model parameters and finally Conclusions are drawn.
in a computer simulator for the optimal positioning of the access
points inside the campus’ buildings. II. I NDOOR PATH LOSS CHANNEL MODELS
The path loss models considered in this paper are summa-
I. I NTRODUCTION
rized in this Section. The one-slope model assumes a linear
Wireless local area networks (WLANs) based on IEEE dependence between the path loss (dB) and the logarithm of
802.11b are largely deployed to provide users with network the distance d between the transmitter and the receiver i.e.
connectivity without being tethered off of a wired network [1]-
LOS (d) = l0 + 10γ log(d), (dB) (1)
[3]. Wireless networks can provide nearly the same services
and capabilities commonly expected with wired networks. where l0 is the path loss at 1 meter distance, γ is the power
Like their wired counterparts, IEEE 802.11b has been devel- decay index or the path loss exponent. The model in (1)
oped to provide large bandwidth to users located in indoor is simple to use but it doesn’t adequately account for the
environments and are being studied as an alternative to the propagation characteristics of the environment. A generaliza-
high installation and maintenance costs incurred by traditional tion of the model in (1) is the two-slope model where the
additions, deletions, and changes experienced in wired LAN path loss exponent is assumed to change when the distance
infrastructures. Because of the unlicensed spectrum availability d between the transmitter and the receiver is greater than the
IEEE 802.11 WLAN devices operate in the ISM (Industrial break distance.
Scientific Medical) band at 2.4GHz. For an accurate planning A further generalization of the OS model in (1) consists in
of indoor radio networks the modelling of the propagation the addition to (1) of a further attenuation term due to losses
channel is required [4]. Channel models can also find appli- introduced by the walls and floors penetrated by the direct
cation in the setup of indoor localization procedures based on path between the transmitter and the receiver i.e.:
IEEE 802.11b such as those in [5],[6].
Starting from the experimental data collected during a mea- L(d) = LOS (d) + Mw (dB). (2)
surement campaign in the campus buildings of the University We refer to Mw as the multi-wall component that, for our
of Rome Tor Vergata we obtained the parameters for many purposes, is expressed as:
different indoor path loss models. The classical one slope (OS)
model and different formulations for the multi-wall (MW) 
I 
Nd 
Nf d

model have been considered. Differently from the MW model M w = lc + kwi li + χn l d + λ n lf d , (dB) (3)
in the current literature [7] we included the attenuation due to i=1 n=1 n=1

the presence of doors and, most important, of fire proof doors where lc is a constant, kwi is the number of penetrated walls
that according to their status, open or closed, can significantly of type i, li is the attenuation due to the wall of type i,

IEEE Communications Society 3701 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


i = 1, 2, ..., I; Nd and Nf d are the numbers of normal and
fire proof doors encountered by the direct path and χn (λn )
are binary variables accounting for the status of the n-th door
(n-th fire proof door),1 . The meaning of the parameters in (3)
are summarized in Table I. Since IEEE 802.11 terminals were
Mw model parameter Meaning
lc Constant factor (dB)
l1 Attenuation due to light wall (dB):
thickness (0,20] cm
l2 Attenuation due to medium wall
(dB): thickness (20,40] cm
l3 Attenuation due to heavy wall (dB):
thickness (40,60] cm
l4 Attenuation due to external building
wall (dB): thickness (60,80] cm Fig. 1. Department Office environment
ld Attenuation due to normal door
(dB)
lf d Attenuation due to fire proof door
(dB)
TABLE I
L IST AND THE MEANING OF THE MULTI - WALL MODEL PARAMETERS

located on the same floor of the AP, the attenuation term due
to propagation among different floors was not included2 in (3).
Starting from (2) and (3) we can obtain many different gen-
eralizations for the multi-wall formula in (2). In the classical
MW model, referred to as MWC (C: classic), [7] the path
loss is expressed as the free space loss added with Mw
i.e. LF reeSpace = 20 log10 (d) + l0 the free space loss and
l0 = 40.22dB for a center frequency of 2.45GHz. A first Fig. 2. Classroom environment
generalization for the MW model considers the path loss
exponent γ and the constant l0 in (2) as model parameters
available for fitting. The Mw component (3) is left unchanged. The measurement equipment consisted of: one IEEE 802.11b
We refer to this second model as MWE (E: extended). access point (AP) model Cisco Aironet 350 [8] equipped with
The third multi-wall model considered in this paper is: two vertical dipole antennas, a portable PC with an IEEE
L(d) = αd + l0 + Mw , (dB) (4) 802.11b card adapter. The values of the received power were
extracted from the PC card adapter using the software provided
where we assumed a linear dependence of the loss from the by the manufacturer. Before starting measurements we tested
distance d; α, l0 are two parameters and Mw is as in (3). We the available PC card adapters purchased by different man-
refer to this model as MWL (L: linear). The parameters of ufacturers [9]-[11]. The test consisted in fixing the positions
the channel models MWC, MWE and MWL will be obtained of the AP and of the card adapter and measuring the power
from fitting of experimental data as explained in the following transmitted on three different IEEE 802.11b channels (1,7 and
Section. 13). Results have been indicated in Fig.3 for three adapters and
III. M EASUREMENT SETUP for a fixed transmission power from the AP equal to 50mW
The characteristics of the measurement sites are depicted in (17dBm). As shown in Fig.3 the selection of the Enterasys
Fig.1 for the office environment and Fig.2 for the classroom RoamAbout [10] PC card adapter allows to reduce the bias
environment. The office environment is characterized by small on the measured power related to the selected transmission
rooms aligned along two parallel corridors. Offices are ac- channel.
cessed through fire proof doors shown in Fig.1. Small/medium During measurement campaign the AP was moved in different
size walls are dominant in this kind of environment. As positions that were selected in order to ensure line of sight
shown in Fig.2 the classroom environment includes large conditions for large areas inside the buildings. In Fig.1 and
rooms for didactics and mediums size walls are common. Fig.2 we indicate some of the points where measurements
were carried out. For each measurement we stored the position
1 open: χn = 0(λn = 0), closed: χn = 1(λn = 1) of the portable PC, the number of walls traversed by the direct
2 Inthe original multi-wall model formulation attenuation due to propagation
between floors is assumed to be a non-linear function of the number of path between the AP and the PC and the number of doors and
penetrated floors. fire proof doors intercepted by the direct path including their

IEEE Communications Society 3702 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


l1 l2 l3 l4 l0 ld lf d 10γ α
Model
αd + l0 61.5 1.9
OS 35.9 47.0
MWC 4.0 4.0 6.0 13.1 9.4 1.6 16.4 20.0
MWE 3.8 3.9 5.7 12.4 47.1 1.4 15.2 23.2
MWL 3.9 4.2 6.3 12.4 59.6 1.7 15.5 0.9
TABLE II
M ULTI - WALL PARAMETERS ( D B) - O FFICE E NVIRONMENT

l1 l2 l3 l4 l0 ld lf d 10γ α
Model
αd + l0 67.4 1.2
OS 56.2 23.7
MWC 3.1 0.0 10.5 0.0 9 7.4 0.0 20.0
MWE 4.6 0.0 11.0 0.0 53.0 7.8 0.0 15.4
MWL 4.6 0.0 11.0 0.0 60.0 0.8

Fig. 3. Measured power over three different IEEE 802.11b channels - TABLE III
channels 1,7 and 13 were considered M ULTI - WALL PARAMETERS ( D B) - C LASSROOM ENVIRONMENT

status (open/closed). In each measurement the orientation of


the embedded antenna with respect to the AP was randomly Table II and Table III the optimal parameters for the one slope
selected in order to avoid possible favorable positions. The model in (1) and for the linear channel model L(d) = αd + l0
antenna of the PC card adapter was located at 1.5m above the have been listed. As expected the results are dependent on the
ground. For each position of the PC in the area we collected 20 characteristics of the measurement site. In particular, it can be
values (samples) of the received power for a total measurement observed that in the classroom-like environment, where fire
time interval of 40s. The power variations observed in the proof doors are located only on the external building walls,
different samples of the same measure were mostly due to their contribution to the overall attenuation is null. In addition
the fast fading introduced by the movement of the obstacles wall attenuation is mainly due to the dominant presence of
(e.g. people) around the PC. Fast fading effects were averaged medium-size walls inside the building. The characteristics
out and the average power value was used for subsequent of the office-like environment are radically different as also
fitting. For each AP position in the area we collected up to 40 indicated from the results in Table II. It is important to observe
measurements and a total of 200 measurements were obtained the non negligible contribution due to fire proof doors that, in
for the office and the classroom environments. the office-like environment, can be located inside the building.
Starting from the measured data the model’ parameters indi- In Fig.4 and Fig.5 we plot the attenuation obtained from the
cated in Table I have been obtained through a least square model (continuous line) superimposed with the experimental
fitting procedure i.e. data (dots). Both environments and the multi-wall model in (2)
{γ, lc , li , ld , lf d }opt = were considered. To compare the goodness of the selected
M −1  models we calculated the standard deviation (in dB) of the
 model error evaluated with respect to the experimental data.
= arg min |Pm − (EIRP − L(dm ))|2 , Results have been reported in Table IV. As expected the
{γ,lc ,li ,ld ,lf d }
m=0
(5)
Office (dB) Classroom (dB)
where L(dm ) is the multi-wall model in (2), Pm is the Model
power at distance dm from the AP, M is the number of αd + l0 7.8 5.1
available measurements for the selected environment and an OS 7.4 5.1
EIRP = 20dBm (EIRP: Effective Isotropic Radiated Power) MWC 4.3 3.4
is the power transmitted by the AP. Since the detailed physical MWE 4.2 3.3
characteristics of the IEEE 802.11b modem devices were not MWL 4.4 3.3
provided by the manufacturers, the path loss models L(d) in TABLE IV
(2) with the parameters obtained as in (5) include losses and S TANDARD DEVIATION OF THE MODEL ERROR ( D B) - OFFICE AND
gains due to the subsystems in the modem front end and the CLASSROOM ENVIRONMENTS

gain of the receiving antenna.


IV. E XPERIMENTAL R ESULTS MW-like models provide errors lower than one slope and the
Multi wall model parameters obtained from fitting have been simplified linear models. This fact is further confirmed from
indicated in Table II and Table III for both environments. In the graphs in Fig.6 and Fig.7 where we plot the cumulative

IEEE Communications Society 3703 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


Fig. 4. Comparison between optimal MW classic model (continuous line) Fig. 6. Cumulative distribution of the model error - office environment
and measurements (dots) - office environment

Fig. 7. Cumulative distribution of the model error - classroom environment


Fig. 5. Comparison between optimal MW classic model (continuous line)
and measurements (dots) - classroom environment
localization algorithms based on IEEE 802.11b. In this paper
we used measured data to evaluate the parameters of many
distribution function of the absolute model error defined as path loss channel models some of them proposed in the current
em = |Pm /EIRP − L(dm )|, m = 1, 2, ..., M where L(dm ) literature. Both classical one slope models and different multi-
is the loss model and M is the number of measurements Pm . wall models were analyzed and results have been provided
From Fig.6 and Fig.7 it can be observed that multi-wall like for two different categories of environments: office-like and
models provide almost the same performance and no large classroom-like. Also the different MW-like models were com-
improvement is observed passing from M W C to M W E and pared and no particular improvements were observed when
also from M W E to M W L. It can be further observed that for passing from the classical MW model (MWC) to the enhanced
MW models the 90% of measurements is within an absolute multi-wall models M W E and to the M W L. Differently from
model error not greater than 7dB for the office while this value the results in the conventional literature, the MW models
increases up to 12 dB for the one slope model and the linear considered in this paper include the attenuation due to doors
loss models. For the classroom environment the previous limits and to fire-proof doors which, in accordance to their status,
are reduced to 5dB and to 7.5 dB, respectively. can significantly contribute to the overall attenuation.

V. C ONCLUSIONS R EFERENCES
[1] B. P. Crow, I. Widjaja, J. G. Kim, P. T. Sakai, ”IEEE 802.11 Wireless
The characterization of indoor path loss is an important Local Area Networks”, IEEE Communications Magazine, September
step in wireless network design and also in the setup of 1997, pp.116-126

IEEE Communications Society 3704 0-7803-8533-0/04/$20.00 (c) 2004 IEEE


[2] Wireless LAN: tecnologia e applicazioni M.Colonna, G. D’aria, No-
tiziario Tecnico Telecom Italia, Anno 11 n. 1, aprile 2002 (in Italian)
[3] The Institute of Electrical and Electronics Engineers, Inc IEEE Std
802.11b Wireless LAN Medium Access Control (MAC) and Physical
Layer (PHY) specifications, 1999 edition.
[4] D. Laselva, ”WLAN Indoor Radio Network planning”, HUT Comm.
Laboratory, S-72.333, Radio Network Planning and Radio Resource
Management for CDMA Systems
[5] On Indoor Position Location With Wireless Lans P. Prasithsangaree
1, P. Krishnamurthy 1, P.K. Chrysanthis, PIMRC 2002, The 13th
IEEE international symposium on personal indoor and mobile radio
communications Lisboa, Portugal september 15-18, 2002
[6] A probabilistic approach to wlan user location estimation Roos, P.
Myllymaki, H. Tirri, P. Misikangas, and J. Sievanen, A probabilistic
approach to wlan user location estimation, International Journal of
Wireless Information Networks, vol. 9, No. 3, July 2002.
[7] COST action 231, Digital mobile radio towards future generation
systems final report, European Commission Brussels 1999.
[8] http://www.cisco.com
[9] http://www.avaya.com
[10] http://www.enterasys.com
[11] http://www.orinocowireless.com

IEEE Communications Society 3705 0-7803-8533-0/04/$20.00 (c) 2004 IEEE

You might also like