Vous êtes sur la page 1sur 5

Downlink Relative Co-Channel Interference Powers in Cellular Radio Systems

Bo Hagerman
Radio Communication Systems Laboratory

Dept. of Teleinformatics, Royal Institute of Technology


ELECTRUM 204, S- 164 40 KISTA, SWEDEN
Email: bosseh@it.kth.se

Abstract
One method to achieve high capacity in cellular
radio systems involves interference cancellation techniques at the
receivers. To design and evaluate the performance of such
interference cancellation receivers, it is essential to use realistic
models of the co-channel interference. In the paper we study the
ordered statistic of the relative co-channel interferepce, i.e. the
ratio between the power level of i:th strongest interferer and the
total co-channel interference power. Results from Monte-Carlo
simulations show that for base stations on a symmetric grid,
hexagonal cells or half-square Manhattan like street cells, the
probability density functions (pdf) of the interference ratios are
almost independent of the cluster size and of the base station
activities. The results demonstrate the dominance of the
strongest interferer and show that this dominance is even more
pronounced in those situations when the signal to interference
ratio is low. The presented results provide a base for more
realistic models for performance evaluation in cellular systems
than the commonly used Gaussian interference model.

I. INTRODUCTION
In cellular radio systems, the limited available bandwidth is
one of the principal design constraints. For future high
subscriber density systems, frequency reuse is one of the
fundamental approaches for efficiency spectrum usage and for
achieving the demands of required capacity. The possibility to
reuse the same channels, the frequency bandwidth, in different
cells is limited by the amount of co-channel interference
between the cells. The minimum allowable distance between
nearby co-channel cells (or the maximum system capacity) is
based on the maximum tolerable co-channel interference at
the receivers in the system. Receivers resistant to co-channel
interference allow a dense geographical reuse of the spectrum
and thus a high system capacity. This can be achieved by
using interference cancellation techniques that take advantage
of the structure of the co-channel interference [l-31. Fully
centralized detection algorithms may be feasible in a base
station receiver since the active users signaling waveforms
can be distributed via a backbone network. Even though the
task requires a high degree of implementation complexity,
these type of algorithms may be acceptable in a base station
which demodulates information from a subset of all mobile
stations. However, the mobile station is the destination of
information from only one base station in the downlink case.
The implementation costs of centralized algorithms may be
unacceptably high for this case. In the downlink there may
also exist restrictions on the amount of information that can be

0-7803-2742-XI95 $4.00 0 1995 IEEE

distributed about the active users. One approach to alleviate


this problem is to consider only the strongest active cochannel users at the receiver. Other active users may be
neglected and regarded as background noise. To design and
evaluate the performance of receivers with the approach
described for the downlink above, it is essential to use realistic
models of the co-channel interference levels in the cellular
radio environment. In this paper, we investigate the ordered
statistic of the relative co-channel interference, i.e. the ratio
between the power level of i:th strongest interferer and the
total CO-channel interference power. We consider radio
systems consisting of a cluster of cells, each with a base
station which communicates at a fixed power level with a
subset of all mobile stations in the system. In Section I1 the
used symmetrical cell patterns are described. Hexagonal cells
models macrocellular systems and microcellular systems are
modelled with street covering cells in a so called Manhattan
environment, i.e. in a regular network of streets. The
propagation models for the different systems are presented in
Section 111. Further, in Section IV we define the ordered
statistic of the co-channel interference. Results from MonteCarlo simulations are shown in Section V. Finally, in Section
VI some conclusions are drawn.

366

11. CELLULAR
MODELS
In planning a cellular system, the whole service area is
divided into non overlapping cells that cover the area without
gaps [4].The cells are grouped into clusters, wherein the
available channels are not allowed to be reused. The cluster
size C, defined as the number of cells per cluster, determines
how many channel sets the available spectrum must form.
Smaller cluster sizes provide more channels per cell and
thereby offering more capacity per cell. Therefore, with fixed
cell size the system capacity is increased for a decreased
cluster size.
A. Hexagonal Cell Coverage Model

In macrocellular systems, where base station antennas are


placed at high locations, cells are large and of almost circular
shape. When designing such systems, cells are modelled as
hexagons and the cluster size is determined by [4],
c = i2 + l ..J + J .2 , l., J. > O .
(1)
Since i and j are integers, the cluster size can only take certain

Fig 1 A hexagonal cell pattem example The reused coverage of one channel
group for cluster size 7 is shown

Fig. 2. A half-square Manhattan cell pattern example. The reused coverage of


one channel group for cluster size 2is shown.

realizable values, { 1, 3,4,7,B, 12, .. } . The normalized


channel reuse distance, RH, and1 the cluster size required to
cover a fixed assignment plan is related as [4],

C. TrafJic Model

RH = ,.J%.

(2)

An example of channel reuse in a hexagonal layout is shown


in Figure 1.
B. Half-Square Manhattan Cowrage Model

A street microcellular system is defined in this paper as a


system with outdoor cells, each with the base station antenna
mounted at a height low compared to the heights of
surrounding buildings. The en\ ironment used, the so called
Manhattan model, is based on a ideal city modelled as a large
chessboard where each square corresponds to one block
with a regular network of streets between the blocks. We place
a base station antenna at every street crossing and the cell size
is assumed to be half a block in all directions. The cluster size
for this so called half-square cells, can only take the values [:5]

+j2, i,j>O,

(3)

to cover the whole service area in a symmetric cell plan. As


seen, the cluster size can take on only certain realizable
values, { 1, 2,4, 5, 8, 9, . ..} . An example of channel reuse in a
half-square Manhattan layout is shown in Figure 2. The
distance to the nearest geornetric co-channel neighbors,
normalized to the cellblock side length is given by [SI

RSG = f i .
(4)
An important distance in this kind of environment is the
distance to the nearest LOS (Line Of-Sight) co-channel
neighbor. Normalized to the ci:ll/block length, we can write
the nearest line of sight co-channel neighbor distance as [5]
C
( 5)
RSL = j
j
z
m
3
where gcd(i,j) denotes the grealest common divider of i and j .

We will in this paper assume that the studied cellular


systems are uniformly loaded, i.e. all base stations carry on
the average the same amount of traffic. The channel allocation
within a cell is assumed to be done at random and independent
of the channel allocation in all other cells. Thus, a specific
channel within a cell is used with probability P. P will be
called the base station activity factor.
111. PROPAGATION
MODELS
A. Macrocellular Propagation Model

The propagation attenuation for base station antennas


placed at high locations is generally modelled as the product
of the -a:th power of distance and with a log-normal
component representing shadowing losses [4]. Thus, for a user
at a distance r from a base station, link gain is proportional to

where y is the dB attenuation due to shadowing, with zero


mean and standard deviation (J. In this paper, we use the
standard deviation (J = 8 dB and a = 4 for the power law.

B. Microcellular Propagation Model


In a street environment for a system operating at 870 MHz,
the attenuation has been measured and modelled in [6]. The
presented empirical propagation model shows that along a
street with LOS propagation to the base station, the
attenuation basically corresponds to free-space loss in the
vicinity of the transmitter. At a distance of 100 - 400 meter
from the base station a breakpoint can be observed whereafter
the attenuation increases faster than in free-space. Let m, and
m2 represent the power law before and after the breakpoint xL,
respectively. Thus, the distance link gain along the LOS street
is modelled as,

367

9-

Activity 50%
Cluster sizes

11.3, 4. 7. 9. 121
-20

-30

-m -40 ,

-An\

c
S -50Y

-70 -80

-100

50

100

150

200
Distance [m]

250

300

350

400

Fig 3 The link g i n as function of distance in the Manhattan street


environment Distance between street crossings is assumed to be 100 m

a function of the signal to interference ratio (SIR) [2-31, which


is defined as
where the smoothness of the transition is determined by q. A
street corner was found to have the same impact on the
received power along a crossing NLOS (Non Line Of-Sight)
street, as a hypothetical transmitter at the corner transmitting
with the same power level as the received power at the corner.
The link gain along the NLOS street is equal to

which mathematically is identical with (7). yo in ( 8 ) may be


interpreted as the distance from the corner to the hypothetical
transmitter and (8) is not valid for y < yo. The NLOS
expression is always used in combination with the LOS
expression which form the aggregate NLOS distance gain as

G(xc,Y) = GL O S ( p N L 0 S O : )

(9)

In (9) xc is the distance from the base station to the corner and
y is the distance along the NLOS street. The results in [6]
suggest the following choices to represent a typical city
environment: inl = 2, m, = 5, xo = 1 m, xL = 200 m, nl = 2, n2
= 6,yo = 3.5 m, y L = 250 m and q = 4. In Figure 3. the distance
attenuation given in (7) and (9) with the above parameter
values is presented. A signal from a base station can reach a
NLOS receiver through many paths. Since a corner implies a
strong attenuation, we will only regard co-channel
interference that has passed maximum one corner. The
shadow fading was found in [6] to be well modelled as lognormal fading with a suggested standard deviation os= 3.5
dB. The total aggregate attenuation is then modelled as the
product of the distance attenuation and the log-normal
component.
IV. CO-CHANNEL
INTERFERENCE
In interference limited systems, the receiver performance is

where S denotes the desired signal power level at the receiver


and the compound co-channel interference power level is
denoted by 1. Without loss of generality it is assumed that the
individual co-channel interferers are ordered (numbered) such
that their power levels satisfy I , 2 I , 2 . .. 2 I,, . The commonly
used interference model for performance evaluations is based
on the central limit theorem by which I = X I , is
approximated by a Gaussian random variable. However, when
studying co-channel interference cancellation receivers which
only consider the strongest subset of the active co-channel
users. the question arises how well I =
can be
approximated by I,, (II + I,), ... in the cellular environments.
For this purpose we are interested in the statistic properties of
the relative co-channel interference components, i.e., I , / [ , the
ratio between the power level of the i:th strongest interferer
and the total co-channel interference power. This is done by
studying the pdf, the mean and standard deviation values of
the ordered relative co-channel interference.
V. NUMERICALRESULTS

The results presented in this paper were obtained by means


of Monte-Carlo simulations of the two system environment
cases studied, macro- and micro-cellular systems. The
simulated systems are set up such that a large amount of cells
of prospective co-channel interferers are placed around a cell
with the used wanted channel. To mitigate border effects in
the simulations we have, independent of the cluster size, used
7 2 and 68 prospective co-channel interferer cells for the
hexagonal and the half-square Manhattan cell patterns,
respectively. For all results presented, the mobile position is
uniformly distributed within the used cell coverage. For each

368

10
Base Activity ["/I

[17. 33. 50, 67, 83, 1001


8

6
5

Third strongest interferer

-",5

I
4

\ \h

Second strongest interferer

hird strongest interferer

d strongest interferer

.
I
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.3
0.5
0.6
0.7
0.8
0.9
1
-0

0.1

0
0

'

Ix I

0.1

0.2

0.4

IX / I

5. Examples of the pdfs for the three strongest interference ratios in ;a


hexagonal environment

Fig. 7. Examples of the pdf:s for the three strongest interference ratios in a
Manhattan street environment.

10

0.8,
Cluster size 3
Base Activity [%]
(17, 33, 50, 67, 83, 1001

Cluster sizes

[I 2.4, 5. 8, 91
I

061 P=0.17

d strongest interferer

7
6

nd strongest inlerferer

3
2
1

0
'

0.1

0.2

0.3

0.4

11.5
l</l

0.6

0.7

0.8

09

4
5
6
7
x [Interference strength number]

10

Fig. 6 . Examples of the pdfs for the three strongest interference ratios in a
Manhattan street environment.

Fig. 8. Examples of the mean and the standard deviation for the ten strongest
interference ratios in a hexagonal environment.

set of parameters, data was collected from 10,000 mobile


positions where we have assumed the shadow fading to be
independent between different runs. For each mobile position,
we also assumed that signals from different base stations are
exposed to independent fading. In Figures 4 and 5 , we show
results for a macro-cellular system with a hexagonal cell
pattern with a cell radius of 1 km.The pdf:s for the three
strongest co-channel interference ratios are shown. In Figure 4
the base station activity factor is 0.5 and a set of pdf:s are
presented for each of the different cluster sizes in [l, 3 , 4 , 7 , 9 ,
121. As seen in Figure 4, the p d t s of the interference ratios are
almost independent of the cluster size. The absolute values of
the amount of co-channel interference powers are of course
decreasing with increasing cluster size, e.g. increasing reuse
distance. The dominance of the strongest co-channel interferer
in the hexagonal cell pattern is demonstrated in the results
shown in Figures 4 and 5 . If all relative CO-chaninel
interference are of the same order of magnitude, the pdifs

would be concentrated in the leftmost part of the figures. In


Figure 5 , we can see that the strongest co-channel interferer is
more dominant for a small traffic load on the system.
However, the differences depending on the base station
activity factor is small and the basic shape of the p d f s remain
for the whole set of activity factors shown. Results are shown
in Figures 6 to 8 for a micro-cellular system using a halfsquare cell pattern in an ideal Manhattan city environment
with a cellblock length of 100 m. As seen from the presented
results, this case follows the previous. However, the
dominance of the strongest co-channel interferer is even more
pronuunced. The variations for different cluster sizes (Fig. 6)
and base station activity factors (Fig. 7) is larger here then in
the former case. This since in the street environment, the cochannel neighbors at LOS distance will cause more
interference than the nearest geometric neighbors. In Figure 8
and 9, we prescnt the mean and the standard deviation for the
ten strongest co-channel interference ratios in the hexagonal

369

0.8-[

[/.I

Activity 50%
Cluster size 3
SIR Threshold T [dB]
[Int.. 15, 9, 31

9-

Base Activity
(17. 33, 50. 67. 83,1001

8-

76-

T=lnf.

4~

3
0.2
2

Std(lx I
0.1

4
5
6
7
x [Interference strength number]

0
0

10

Fig. 9. Examples of the mean and the standard deviation for the ten strongest
interference ratios in a Manhattan street environment.

In this paper we have studied downlink co-channel


interference powers in symmetrical macro- and micro-cellular
systems. The results demonstrate the dominance of the
strongest co-channel interferer. A conclusion would be that
the central limit theorem (Gaussian assumption) is a poor
approximation in the studied environments. The presented
results may provide a base for more realistic models of the cochannel interference for performance evaluations of cellular
systems than the commonly used Gaussian interference
model. As shown, the basic shape of the pdf:s of the ordered
relative co-channel interference are almost independent of the
cluster size and the base station activity factor. It is shown by
utilizing receivers that take into account a subset of the
strongest co-channel interferers, that the amount of cochannel interference can be greatly reduced. With interference
cancellation of 1 or 2 of the strongest co-channel users, we
may reduce the mean value of the co-channel interference
with approximately 60 to 90%. The results indicate that the
improvement by using interference cancellation receivers may

0.2

0.3

0.4

0.5
1x11

0.6

0.7

0.8

0.9

10. Examples of the pdf:s for the two strongest interference ratios in a
hexagonal environment

and the half-square Manhattan street environments,


respectively. A set of curves are presented for each of the
different base station activity factors in [116, 113, 112, 213, 516,
9, 11 when the cluster size is 3 (Fig. 8) and 2 (Fig. 9),
respectively. The dominance of the strongest co-channel
interferer can be seen and if a receiver can take into account a
subset of the strongest interferers, the amount of co-channel
interference will be greatly reduced. In Figures 10 and 11, we
study the p d f s of the two strongest interferer ratios
conditioned on that the received SIR is less then a signal to
interference value threshold, i.e. SIR < T . As seen for both the
hexagonal (Fig. 10) and the Manhattan street (Fig. 11)
environments, for decreasing signal to interference threshold
the dominance of the strongest interferer is more emphasized.

VI. DISCUSSION

0.1

Activity 50%
Cluster size 2
SIR Threshold T [dB1
[Inf.. 15. 9. 31

98-

7.

Second strongest interferer

Stronqest interferer

Fig. 11. Examples of the p d f s for the.two strongest interference ratios in a


Manhattan street environment.

be even greater in those situations when the signal to


interference ratio is low.
REFERENCES
Verdu, S., Minimum Probability of Error for Asynchronous Gaussian
Multiple-Access Channels, IEEE Trans. Inform. Theory, vol. IT-32,
no.1, Jan. 1986.
Hagerman, B., Single-User Receivers for Partly Known Co-Channel
Interference, Liceniiute Thesis, TRITA-IT R 94:11, Royal Inst. of
Technology, 1994.
Hagerman, B., On the Detection of Antipodal Rayleigh Fading Signals
in Severe Co-Channel Interference, IEEE Vehicular Technologj
Conference, Stockholm, Sweden, June, 1994.
Jakes, W. C., Microwave Mobile Communication, New York: Wiley,
1974.
Gudmundson, M., Cell Planning in Manhattan Environments, IEEE
Vehicular Technology Conference, Denvel; CO,May, 1992.
Berg, J-E., Bownds, R., Lotse, E, Path Loss and Fading Models for
Microcells at 900 MHz, IEEE Vehicular Technology Conference,
Denver; CO,May, 1992.

370

Vous aimerez peut-être aussi