This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts

for publication in the WCNC 2010 proceedings.

A Suburban Femtocell Model for Evaluating Signal

Quality Improvement in WiMAX Networks with
Femtocell Base Stations
Pei-Chen Lu, Kang-Ju Tsao, Cheng-Ru Huang, and Ting-Chao Hou
Department of Communications Engineering & Center for Telecommunications research
National Chung Cheng University, Chia-Yi, Taiwan
AbstractThe femtocell base station is a low power and low
cost small base station in the customer premise. It connects to
the core network via cable, DSL or similar backhaul technology.
In this paper, we focus on how received signal quality can be
improved by deploying femtocell base stations in a WiMAX
system. We propose a suburban femtocell model to analyze
the potential improvement on the SINR when using different
channel assignment schemes and different deployment densities
of femtocell base stations within a macrocell. The simulation
results show that we can enhance SINR by deploying large
number of femtocell base stations within macrocells, and the
Least-Interfered-First (LIF) scheme we proposed can be used
to significantly reduce the co-channel interference induced by
neighboring femtocell base stations.
Keywords - femtocell, WiMAX, SINR, suburban

I. I NTRODUCTION
The WiMAX (Worldwide Interoperability for Microwave
Access) network is a wireless metropolitan area network based
on IEEE 802.16 standards. IEEE 802.16a-d standards adopt
the cell structure and provide high speed transmission rate for
fixed users. For outdoor users, IEEE 802.16e is established
to support mobility and IEEE 802.16j is further specified
to enhance system capacity and coverage by incorporating
relay stations into the system. To improve indoor (home or
small office) users signal strength, which is usually attenuated
through walls, IEEE 802.16m SDD (System Description Document) [1] introduces the femtocell. Femtocell base station
is a low power, short-range indoor and easy-to-install base
station. Femtocell base station connects to the core network
by backhaul links, through which indoor users can access the
Internet.
The scarcity of available frequency spectrum has long been
an important design issue in wireless networks. Frequency
band can be spatially reused to increase the system capacity
but it also introduces co-channel interferences. In a WiMAX
network with both macrocells and femtocells, frequency bands
used by macrocells and femtocells are necessarily overlapped,
though with distance separation between cells to reduce the
degree of interference. How to minimize the interference
between macrocells and hundreds of thousands of femtocells
in a WiMAX network is certainly an important issue. In
this paper, we build a suburban model to understand how
the deployment of femtocells will affect the received signal
quality. We show how the femtocell density will improve the

Fig. 1.

Femtocell architecture.

SINR (Signal to Interference-plus-Noise Ratio). We also show

how a simple channel assignment scheme, Least-InterferedFirst (LIF), can effectively reduce the co-channel interference
when compared to a random channel assignment scheme.
The rest of the paper is organized as follows. In section
2, we give a brief background information on femtocells. In
section 3, the system architecture of the WiMAX network with
femtocells is described. In section 4, we show our simulation
results about the SINR improvement when femtocells are
included in the system. Section 5 is the conclusion.
II. F EMTOCELL BASICS
Femtocell base station is a low power wireless base station
(or access point) deployed by users at home. Femtocell base
station connects to the core network by cable, DSL, optical cable or similar backhaul technology. Moreover, Femtocell base
station operate on licensed band and it can work with various
wireless communication systems, such as UMTS (Universal
Mobile Telecommunications System), WCDMA (Wideband
CDMA), WiMAX, UMB (Ultra Mobile Broadband) or LTE
(Long Term Evolution).
From the technical and/or business point of view, femtocells
offer the following benefits:
A. Indoor coverage improvement
The greatest advantage of using a femtocell is that it improves indoor wireless signal quality. Year 2007 data showed
that 32 percent of mobile calls are made from indoors [2].
Therefore the communication quality offered to indoor users is
one of the main criteria that users would consider in choosing
a mobile service provider.

978-1-4244-6398-5/10/$26.00 2010 IEEE

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

Fig. 2.

WiMAX femtocell architecture.

Fig. 3.

Femtocell architecture.

B. System coverage/capacity enhancement

Femtocells not only can increase indoor signal strength, but
also can improve macrocell edge users signal quality and
thus extend the system coverage. Due to the shorter distance
between a user and his/her serving femtocell base station, a
higher SINR can be achieved and thus higher system capacity.
C. Relief to macrocells loading
Macrocells coverage radius is typically 1 to 5 kilometers,
covering hundreds or thousands of users. However, it is
difficult for a macrocell base station to serve hundreds of
users simultaneously because of the limitation of the available
frequency spectrum. If femtocells are deployed in a macrocellular network and reuse the same frequency band as those
used by macrocells, femtocells will not only reduce macrocells loading but also enhance the system capacity.
D. Low system deployment cost
Being a consumer device, femtocell base stations cost
is much lower than macrocell base station and is expected
to be equivalent to WiFi AP. Furthermore, femtocells allow
operators to deliver data services at a very low cost because the
traffic is connected to the core network over the households
existing broadband link [3].
WiMAX femtocell base station is a service access point
over WiMAX wireless communication system. In Figure 2,
WiMAX femtocell is just like a Customer Premise Equipment
(CPE) and provides a service access point for users. Users
connect to WiMAX Access Service Network Gateway (ASN
GW) through broadband link, and then access to WiMAX core
network.
III. S YSTEM A RCHITECTURE
We consider a 19-macrocell structure in a suburban area,
where the coverage of each macrocell is approximated by a
hexagon, as shown in Figure 3. The reason that a 19-macro
cell structure is chosen is because the co-channel interference
from macrocells beyond two tiers is negligible [4]. Since we
focus on the user SINR in a typical macrocell (the center
macrocell in Figure 3), we can disregard macrocells that are
three or more tiers away.
In the suburban area, streets form a grid structure and run in
the north-south or east-west direction. Houses are situated in

the square blocks separated by streets. There are 100 houses

within each block. Some houses have femtocell base stations
installed, some have not. Whether a house has femtocell base
station or not follows a uniform probability distribution. The
mean of the uniform distribution is also called the femtocell
density. Users are randomly placed in the area. A user may
be inside the house (indoors) or on the street (outdoors). A
user may communicate with the macrocell base station or
the femtocell base station, depending on the received signal
strength. The SINR is calculated by considering interference
from the other 18 macrocells and nearby femtocell base
stations.
A. System parameters
The WiMAX system parameters and OFDMA parameters
are chosen based on the IEEE802.16m Evaluation Methodology Document (EMD) [5] and IEEE802.16-2004 [6]. These
parameters are shown in Table 1 and Table 2.
TABLE I
S YSTEM PARAMETERS
Parameter
Macrocell Transmit Power [dBm]
Femtocell Transmit Power [dBm]
Macrocell Coverage Range [m]
Femtocell Coverage Range [m]
Macrocell Antenna Height [m]
Mobile Station Antenna Height [m]
Noise Figure [dB]
Noise Power [dBm]

Value
46.532 (45W)
10.02 (0.01W)
1000
30
12.5
1.5
5
-169.93+10log10 (BW)

Femtocell base stations will be deployed in a great quantity.

However, it is not necessary to calculate the interferences from
all femtocells to a user because femtocell base station is a
low power home device. Given the femtocell transmit power
(0.01W) in Table 1, a femtocell base station that is 200 meter
away will cause interference at least 30dB below the receiver
sensitivity. Therefore, when calculating the interference from
femtocells, we ignore the interference from femtocells that are
at least 200 meters away.

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

TABLE II
OFDMA PARAMETERS
Parameter
Carrier frequency [GHz]
Channel Bandwidth [MHz]
FFT size
Frame duration [ms]
OFDMA symbol duration [gs]
Permutation type
Number of subchannels
Number of subcarriers
Number of subchannells per MS

Value
2.5
10
2048
5
102.86
PUSC
60
28
2

The femtocell base station is assumed to be placed in the

middle of a house. Whether a house will have a femto BS or
not is determined randomly according to the desired femtocell
density. Since there are 100 houses in each block, if the
density is 0.1, there will be about 10 houses with femtocell
BS installed in each block. Figure 5 shows three snapshots
of femtocell deployment results when the density are 0.1, 0.5
and 0.9, respectively.

B. The block, house and street

In Figure 4, we show the relationship among houses, blocks
and streets. Each side of a square block is 200 meter long and
100 houses are located within each block. A house occupies
a 14x14 meter square area, centering within a 20x20 meter
square land. Therefore the distance between two neighboring
houses in the same block is 6 meters. The width of the street
is assumed to be 30 meters.
To speed up the simulation, we only consider users located
in the center macrocell in Figure 3 (also the inner hexagon
in Figure 4). The other 18 macrocells are mainly used for
calculating the interference to the users in the center macrocells. Besides interferences from other macrocells, there are
also interferences from nearby femtocells. In order to calculate
femtocell interferences to users at the edge of the center
macrocell, the exact distribution of femtocells in neighbor
macrocells needs to be taken into consideration. When the
femtocell transmission range is set at 30 meters (Table 1),
we have said that we only need to consider the interference
caused by the femtocell base stations within the 200-meter
distance because the femtocell transmission power is small.
As a result, in our study, we dont have to bookkeep all
femtocells in neighbor macrocells. We only have to consider
femtocells in the outer hexagon, which extends 200 meters
outward from the inner hexagon, in Figure 4. This greatly
reduces the complexity of the simulation without sacrificing
the accuracy.

Fig. 5.

Snapshots of femtocell deployment.

C. Frequency reuse - (1,3,3)

Since the femtocell is deployed within the coverage of
a macrocell and it is also within the interference range of
several neighboring macrocells and femtocells, it should try
to avoid using the same frequency channels as those used by
neighbor cells. On the other hand, there are limited frequency
channels for the wireless communication system. Frequency
reuse with minimized co-channel interference is common
practice to increase the system capacity. For frequency reuse at
the macrocell level, the following (c, s, n) notation is used [7].
The number c represents how many cells share the frequency
spectrum. It is also called the frequency reuse factor or the
cluster size. For example, if c is equal to 1, it represents that
all neighboring cells use the same set of frequency channels.
The number s represents the number of sectors in each cell.
If s is 3, it means that each cell is divided into 3 sectors and
the angle of each cell is 120 degrees. The number n indicates
how the sectors share the frequency channels. If n is 1, all
sectors use the same set of frequency channels. If n is 3, each
sector uses one third of the frequency channels available to
the cell. (c, s, n)=(1,3,3), (1,3,1) or (3,3,3) have been used in
frequency reuse studies. In this study, we consider the (1, 3,
3) frequency reuse pattern for the macrocell.
For the femtocell, we choose to allow it to use the frequency
channel that does not overlap with the channels that its sector
is allocated. For example, suppose the 60 channels of the
WiMAX system are divided into three groups, w1 (channels
0-19), w2 (channels 20-39), and w3 (channels 40-59), If a
sector is allocated the w1 group, then the femtocells within
this sector will be allocated channels from the w2 and w3
groups to reduce the co-channel interference. An example of
the resulting channel reuse pattern is shown in Figure 6.
D. Path loss model

Fig. 4.

Streets, blocks, and houses in a macrocell.

The IEEE 802.16m EMD document [5] has specified system

models and parameters for researchers/engineers to evaluate
the system performance on the same ground. The EMD

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

TABLE III
T HE AVERAGE SINR AND PROBABILITY OF CONNECTING TO
FEMTOCELLS WITH THE RANDOM CHANNEL ASSIGNMENT.

Femtocell density
0
0.1
0.5
0.9

Average SINR prob.

15.32
16.67
22.21
27.82

Prob. connecting to Femtocell

0
0.046
0.224
0.412

Fig. 6. (1, 3, 3) frequency reuse for macrocells and the resulting femtocell
frequency reuse pattern.

includes outdoors, indoors, indoor to outdoor, and outdoor

to indoor propagation models. In our suburban femtocell
model, depending whether the user is indoors or outdoors, and
whether it will connect to macrocell BS or femtocell BS, the
following four path loss models are used: (For the definition
of each term, please refer to [5].)
Outdoor macrocell user
P L[dB] = 38 log10 (d) + 21 log10 (f ) + 62.26

Macrocell user inside the house

P L[dB] = P Lb + P Ltw + P Lin ,
where
P Lb
P Ltw
P Lin

= 32.4418 + 20 log10 (f ) + 20 log10 (d)

+ 0.0174d + 20 log10 (max(0.013d/f ), 1),
= 14 + 15(1 cos )2 ,
=

0.5din

Indoor femtocell user

P L(dB) = 46.4 + 20 log10 (d) + 20 log10 (f /5.0) + 12nw

Femtocell user outside of the house

P L[dB] = 43.8+36.8 log10 (d)+20 log10 (f /5.0)+12nw
IV. R ECEIVED S IGNAL Q UALITY E VALUATION

In this section, we evaluate how the deployment of femtocells will improve the users received signal quality (SINR)
under different femtocell deployment densities. We wrote a
C++ simulation program incorporating the details discussed
in section III.
For a given femtocell density, femtocell base stations are
randomly and uniformly placed in houses inside the macrocell.
These randomly selected femtocell base stations are assumed
to be actively serving one user inside the house. After the
required numbers of femtocell base stations are placed, a
user is randomly placed within the center macrocell in Figure
4. This user may be inside or outside the house. We first
decide whether the user should connect to the macrocell BS or
femtocell BS by comparing the received signal strength (RSS)
from the macrocell BS and the RSS from the nearest femtocell
BS. Once the BS is determined and the channel is selected,
we then calculate the co-channel interference from the other

18 macrocells and surrounding femtocells to derive the SINR.

For each femtocell density, 10 random femtocell placements
are simulated, each with the random user placement being
repeated for 3000 times.
Under the (1,3,3) channel reuse assumption discussed in
subsection III.C, a femtocell is allowed to use a particular
group of channels based on its location. However, which
channel from the pool of available channels to use might have
significant performance ramification. We compare two femtocell channel assignment schemes. One is random, without
considering what neighboring femtocells are using. The other
does consider avoiding the channels that are being used by
neighboring femtocells.
In the real environment, a femtocell does not know the locations of surrounding femtocells. It can only infer the distance
by the strength of signals from the neighbor femtocell, or it can
request location or channel information from the macrocell,
if the macrocell maintains such information. In this study,
we assume that the femtocell is capable of measuring the
received signal strength of each channel by channel scanning.
The measured signal strength of each channel is in fact the
interference intensity if the femtocell were to use the same
channel. Therefore, in our study, we propose a femtocell
channel assignment scheme called Least-Interfered-First (LIF).
With the LIF scheme, a femtocell chooses the channel that it
measures the lowest signal level for its own use.
Note that the channel assignment issue for macrocell users is
less critical because the distance between co-channel users in
different macrocells is relatively long, as can be seen in Figure
6. The only channel assignment requirement is to make sure
that any two users in the same macrocells do not use the same
channels. In the following, we compare the random channel
assignment and LIF channel assignment for femtocell users in
terms of the average SINR and probability that an arbitrary
user will connect to the femtocell .
A. Random channel assignment
When a user connects to the femtocell, the femtocell randomly assigns a channel from the allowed groups of channels
to the user. Table 3 shows the average SINR and the probability
that an arbitrary user will choose to connect to the femtocell. It
can be seen that when the density of the femtocell base station
becomes higher, the probability of connecting to a femtocell
becomes higher, and the system average SINR improves more.
In the environment without the femtocell base station, the
average SINR is 15.32dB. When the femtocell density is 0.1,

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

Fig. 7.

MS moving in the macrocell.

Fig. 9.

The CDF of SINR with random channel assignment.

TABLE IV
T HE AVERAGE SINR WITH THE LIF SCHEME .

Femtocell density
0
0.1
0.5
0.9

Fig. 8.

The maximum SINR received by a user based on EMD model.

0.5, and 0.9, the average SINR improves by approximately

1.3dB, 7dB, and 12 dB, respectively.
We show how the SINR varies as the user (MS) moves in
and out of houses. In Figure 7, we assume an MS moves in
and out of 5 lands/houses, crosses a street, and in and out of
3 lands/houses, from left to right. The total moving distance
is 190m and the distance from the macrocell BS is roughly
780m. We calculate the received SINRs with respect to the
serving macrocell and femtocell (with the strongest RSS) as
the MS moves and plot the stronger of the two SINR in Figure
8. The stronger SINRs (in the 50-60dB range) occur mostly
when the MS is indoors. The MS connects to the femtocell BS
in this case. As the MS moves out of the house, the femtocell
BS signal attenuates severely through the wall so that the MS
will pick up the stronger signal (roughly 10dB in SINR) from
the Macrocell BS and thus choose to connect to macrocell BS.
Inside a house without the femtocell BS, the MS also receives
stronger SINR (around 2dB) from the macrocell BS than that
from the next-door femtocell BS. The above observation of
course might vary according to how close the MS is to the
macrocell BS.
Figure 9 plots the cumulative distribution function (CDF)
of the SINR. It can be seen that as more femtocells are
deployed, the probability of getting high SINR (larger than 50
dB) becomes higher. On the other hand, if we observe the low
5 percentile, we find only roughtly 1dB and 2dB improvement
for femtocell density at 0.5 and 0.9, respectively.
Figure 10 shows the SINR probability density function
(pdf) for femtocell density=0, 0.1, 0.5, and 0.9. Without

Average SINR prob.

15.32
16.78
22.80
29.80

Prob. connecting to femtocell

0
0.046
0.224
0.412

the femtocell base stations, the SINR pdf has a peak near
SINR=11dB. As the femtocell density increases, the peak at
11dB drops slightly while another peak near 56dB starts rising.
We also notice that a small spike at -1dB gets more obvious
as the femtocell density increases.
We found that the reason why SINR would be near -1dB
is because that there is an interference source of the same
channel nearby. It is illustrated in Figure 11. In the figure,
the user is connected to the center femtocell BS and it is
assigned the same channel as the one used by the upper
left femtocell. With the random channel assignment, a nonnegligible probability of assigning the same channel as the
one used by neighboring femtocell exists. This probability
becomes higher as the femtocell density becomes higher.
As the femtocell density increases, the SINR pdf slowly
transforms from a single mode (single peak) distribution
to a dual-mode (dual-peak) distribution. When there is no
femtocell, indoor and outdoor users connect to the macrocell
BS. The SINR is between -5dB and 50dB. As more femtocells
are deployed, more indoor users may connect to femtocell BS.
When a user connects to the femtocell BS via the line-of-sight
(LOS) path, the SINR will lie between 50 to 62 dB. Outdoor
users tend to remain connected to the macrocell BS, regardless
of the number of femtocells deployed. Therefore, the increase
in the 50-62dB range mainly comes from the decrease in the
10dB and below range.
B. LIF Channel Assignment
Table 4 shows the average SINR and the probability that
an arbitrary user will choose to connect to the femtocell
when the LIF mechanism is used. Similar to the random
channel assignment scheme, when the density of the femtocell
base station becomes higher, the probability of connecting to

This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2010 proceedings.

Fig. 10. The probability distribution function of SINR with the random
channel assignment scheme.

Fig. 12.

The CDF of SINR with LIF channel assignment scheme.

Fig. 11. Interference from nearby femtocells with random channel assignment
scheme.

a femtocell becomes higher, and the system average SINR

improves more. When the femtocell density is 0.1, 0.5, and
0.9, the average SINR improves by 1.5dB, 7.5dB, and 14.5dB,
respectively.
In comparison with the random channel assignment (Table
3), we see that the LIF scheme improves the average SINR
by up to 2dB (at femtocell density=0.9). This is not a great
deal of improvement by considering the amount of extra effort
needed for the LIF to work.
Figure 12 is the CDF of SINR when the femtocell uses the
LIF channel assignment scheme. It is very similar to the CDF
in Figure 7. The low 5 percentile is improved by about 3dB
when the femtocell density is 0.9, a little better than when the
random channel assignment scheme is used.
If we carefully compare the pdf curves of SINR shown in
Figure 13 for the LIF scheme with Figure 10 for the random
scheme, though they look quite similar, we can see that there
is a distinct difference. The spike at -1dB in Figure 10 no
longer appears in Figure 13. This clearly demonstrates that
the LIF channel assignment scheme can effectively eliminate
the close-range co-channel interference.
V. C ONCLUSIONS
This paper proposes a suburban femtocell model that is
suitable for evaluating the SINR improvement when femtocell base stations are deployed in a WiMAX network. The
model considers how femtocell base stations are placed, how
frequency reuse is configured, along with appropriate path loss
models for different user-BS communicating scenarios. With
the model, we can quantify how SINR is improved as different
densities of femtocells are deployed.

Fig. 13. The probability distribution function of SINR with the LIF channel
assignment scheme.

The major benefit of deploying femtocells is in its improvement of the indoor users SINR to the 50-60dB range. As
more femtocells are deployed, more indoor users connect to
the femtocell instead of the macro cell, greatly increasing the
average SINR.
On the other hand, as more femtocells are deployed, potential co-channel interference between femtocells also becomes
worse. In this case, proper interference mitigation schemes are
required in assigning channels to femtocells. The LIF (LeastInterfered-First) channel assignment scheme we proposed can
effectively eliminate the the co-channel interference between
nearby femtocells.
R EFERENCES
[1] IEEE 802.16 Broadband Wireless Access Working Group, IEEE
802.16m System Description Document [Draft], IEEE 802.16m08/003r1, 2009-02-07.
[2] Picochip, The Case For Home Base Stations, White Paper, April 2007.
[3] V. Chandrasekhar, J.G. Andrews, and A. Gatherer,, Femto Networks: A
Survey, IEEE Communications Magazine, September , 2008.
[4] I-K. Fu, A Dynamic Simulation Platform for Heterogeneous Multiple
Access Systems, Master thesis, National Chung Cheng University, Taiwan, 2002.
[5] IEEE 802.16m-08/004r5, IEEE 802.16m Evaluation Methodology Document (EMD), 2009-01-15.
[6] IEEE Std 802.16-2004, IEEE Standard for Local and Metropolitan Area
Networks - Part 16: Air Interface for Fixed Broadband Wireless Access
Systems. New York: IEEE, June 2004.
[7] WiMAX Forum, Requirements and Recommendations for WiMAX
Forum Mobility Profiles, Nov. 9, 2005.