Interference Level Configuration in CDMA-based Cellular Networks1
K.TSAGKARIS, P.DEMESTICHAS, G.DIMITRAKOPOULOS, M.THEOLOGOU
National Technical University of Athens,
Electrical and Computer Engineering Department, Telecommunications Laboratory,
9 Heroon Polytechneiou Street, Zographou 15773, Athens
GREECE
E-mail: gdimitra@unipi.gr, pdemest.unipi.gr, ktsag@telecom.ntua.gr
Abstract: The accommodation of various traffic load situations in W-CDMA-based cellular systems requires the
engineering of the allowed interference levels per cell. This paper presents functionality that can complement
the design and management as well as the mechanisms required. The overall scheme is called Interference
Level Configuration (ILC). It relies on the solution of problems, which will be concisely defined, optimally
formulated and solved by computationally efficient algorithms. Numerical results will be presented.
Keywords: UMTS, QoS, Power allocation
1
Introduction
Wireless systems continue to attract immense
research and development effort [1]. One of the
main areas in this context is the evolution towards
the era of third generation (3G) cellular systems
[2], the main representative being the Universal
Mobile Telecommunications System (UMTS)
[3,4,5,6] .
Essentially, a cellular system is faced with a set of
traffic load scenarios. Each such scenario can be
corresponded to a demand pattern (vector) that
specifies a target number of transmissions, per
service and service area portion, which should be
simultaneously accommodated, so as to cope with
the offered traffic. In W-CDMA systems, one of
the main factors that may limit the system capacity
is the lack of feasible allocations of transmission
power to the connections.
So, an important (design and management) action,
for W-CDMA-based networks, is the proper
configuration (engineering) of the allowed
interference levels in the cells of the system.
This paper presents such (design or management)
mechanisms. The overall scheme will be called
Interference Level Configuration (ILC). It relies on
the solution of two sophisticated problems, which
are concisely (mathematically) defined, optimally
formulated, and solved by means of two new
computationally efficient algorithms. The objective
is to find the optimal feasible interference levels for
each cell and will be further analyzed.
The rest of this paper is organized as follows. The
ILCU and ILCD problems are described in more
detail in section 2. Sections 3 and 4 include the
optimal formulations and the solutions to the
problems. Section 5 includes numerical results and
finally, concluding remarks are presented in
Section 6.
2
Formal Description
Figure 1 provides the general descriptions of the
ILCU and ILCD problems.
Figure 1. General description of the ILCU and
ILCD problems.
The input provides information on the service area
layout, the propagation conditions, the services, the
system (namely, cell information and equipment
capabilities), and the demand pattern.
Service area layout. It is described through a graph
GP P, E P  . Each pixel p ( p  P ) corresponds to
a small part of the service area. In principle, a cell
1
This work is partially funded by the Commission of the European Communities, under the Fifth Framework Program, within the IST
project MONASIDRE (IST-2000-26144: Management of Networks and Services in Diversified Radio Environment).
1
will comprise several pixels. Edges of the E P set
reveal the connectivity between pixels.
Propagation
conditions.
The
propagation
conditions in the service area are described through
a
set
of
attenuation
values
AV 
2
a p, q  p, q   P . Each element a p, q 


provides the attenuation of a transmission that
originates from pixel p and terminates at pixel q .
Service aspects. The set of services is S . The QoS
requirements of s  S are expressed through the
minimum, uplink and downlink, signal-tointerference ratios, SIRu s  and SIRd s  ,
respectively. These derive from the characteristics
of s , namely, the bit-rate, service activity factor
(SAF) and the minimum required Eb I 0 ("energy
per bit divided by the interference spectral
density").
System description (cell information and equipment
capabilities). The set of cells is V . The following
information is required for each cell v  V : (i)
The set of pixels, Pv  , belonging to cell v ;
likewise, a function, c p  , provides the cell to
which a pixel p belongs. (ii) The location, l  v  , of
the Node-B of v . (iii) The orthogonality factor,
o d v  , that provides the proportion of intra-cell
interference, on the downlink, in v [7]. Its values
are in the range from 0 to 1. If od v   1 there is
full orthogonality and therefore no intra-cell
downlink interference.
Finally, the equipment capabilities specify the
maximum transmission powers of terminals and
Node - Bs' (base stations), which are denoted as
p mt s  and pnb , respectively.
Demand pattern. It is described through two
vectors, Du  d u s, p  s, p   S  P and
d d s, p 
element, d u s, p 
Dd 
s, p 
S  P .
Each
( d d s, p  ), is the number of
uplink (downlink) transmissions of service s ,
originating from (terminating to) pixel p .
The solutions to the ILCU and ILCD problems
should minimise the uplink and downlink
interference levels in the system. In this respect,
they compute allocations Au  pu s, p 
s, p  S  P
TP
tot , d
v v  V .
and
The
notation
Ad 
p u s , p 
represents the power that should be used by an
uplink transmission of s that originates from p .
TPtot,d v  represents the total
The notation
downlink power transmitted in cell by the Node B
of cell v .
The objective functions, which should be
minimised by allocations Au and Ad , are denoted
as OFu  Au  and OFd  Ad  , and are associated
with the resulting aggregate interference levels in
the system. Moreover, they should maintain the
QoS levels required by the transmissions of the
demand pattern, and ensure that the assigned
powers are compatible with the equipment
(terminal and Node B) capabilities.
3
Interference Level Configuration Uplink
3.1 Formulation
The formulation of the ILCU problem is the
following.
Minimise
OFu  Au  

 d u s, p   pu s, p 
sS pP
(1)
Subject to,
p s, p   a p, l v  I v   SIR s 
u
tot ,u
'
u
SIRu ( s)
1  SIRu ( s)
v, s, p   V  S  Pv  (2)
I tot ,u v   I own,u v   I oth ,u v   N w v  V
I own ,u v  

 d u s, p   pu s, p   a p, l v 
sS pP v 
v  V
I oth,u v  
(3)
(4)
   d u s, p   pu s, p   a p, l v 
wV v  sS pP  w 
v  V
pu s, p   pmt s 
s, p   S  P 
(5)
(6)
Relation (1) expresses the objective of minimizing
the uplink power of the transmissions in the
demand pattern. This leads to the minimisation of
the uplink interference levels in the system.
Relations (2) are introduced for preserving the QoS
requirements of the transmissions of the demand
pattern. The notation I tot,u v  corresponds to the
total uplink interference in cell v . Relations (3)
provide the components of the total interference of
2
each cell, which consist of the interference caused
by transmissions from this cell ( I own,u v  ) (4),
The termination criteria in step 3 depend on the
i
evolution of the I tot,
u v  values. When a feasible
interference caused by transmissions from the
neighbor cells ( I oth,u v  ) (5) and the noise power
3.2 Solution
solution exists, the above algorithm will converge
to the optimal values. Otherwise, the values will
tend to infinity. For this reason, the algorithm can
be terminated in one of the following cases. First, at
step
in
which
the
condition
i
i
i 1
i 1
I tot,u v  I tot,u v I tot,u v   (   1) is
By appropriately exploiting relations (2) and (4), as
well as (2) and (5), the following formulas are
obtained:
satisfied for every v V . Second, in case the
condition concerning the terminal power budget is
i
violated,
i.e.,
when
I tot,
u v  
N w . Relations (6) are introduced for preserving the
equipment capabilities.
I own ,u v   I tot ,u v  

 SIR s   d s, p
'
u
sS
pP v
min
pP v , sS
u
v  V
I oth,u v  
(4a)
a p, l v
I w   SIR s   d s, p  

 p, lw
a
  
 
w V  v
tot ,u
sS
'
u
pP w
u
v  V
(5a)
The exploitation of relations (2), (4a) and (5a) lead
us to the following set of iterative equations, which
provide the solution to the ILCU problem.
i
I tot
,u v  
i 1
I oth
.u v   N w
1   SIRu' s    d u s, p 
sS
i 1
I oth
,u v  

wV v
v  V
sS
p
mt
s   a p, l v 

SIRu' s  .
 d u s, p
pP w 
a p, l v 
a p, l w
(5b)
The set of equations above can lead us to an
iterative algorithm. Its formal description is as
follows.
termination criterion means that the ILCU
algorithm successfully accomplishes its task, by
converging to a feasible solution. The second
termination criterion means that the algorithm fails
to find an acceptable solution.
4
Interference Level Configuration Downlink
The formulation of the ILCD problem is the
following.
OFd  Ad  
Minimise
TPtot ,d v 

vV
TPtot ,d v  

 d d s, p   pd s, p   al v , p 
sS pP v 
v  V
Step 0: Initialise the algorithm iteration counter, i ,
and the initial interference values, i.e., set
0
i  1, I tot
,u v   0 for all v V .
the
i 1
I oth
,u v 
v V
the
I
i
tot,u
v
quantities through formulas (4b).
Step 3: Evaluate whether the termination criteria
are satisfied. If the termination criteria are
not satisfied increase the algorithm iteration
counter, i.e., set i  i  1, and go to step 1.
Step 4: Compute the optimal pu s, p  values by
using the I tot,u v  values and relation (2).
Step 5: End.
p s, p   al c p , p  I  p   SIR s 
d
tot , d
'
d
(8)
SIRd ( s)
1  SIRd ( s)
s, p   S  P  (9)
I tot ,d  p  
d d s, p  pd s, p  al c p, p 

sS
I ext,d  p   N w
quantities through formulas (5b).
Step 2: Compute for all
(7)
Subject to,
Algorithm for the ILCU problem
v V
first
4.1 Formulation
v  V
Step 1: Compute for all
The
(4b)
pP v 
I toti 1,u w  SIRu' s 

I ext, d  p  
p  P
(10)
  d ~s , ~p   p d ~s , ~p   al c ~p , p 
d
~
s S ~
p P  p 
p  P
p d s, p   p nb
s, p   S  P 
(11)
(12)
Relation (7) expresses the objective of minimizing
the downlink interference levels in the system. The
rest of equations are described in a similar manner
as in the uplink case.
3
4.2 Solution
By appropriately exploiting relations (9) and (10),
as well as (9) and (11), we are led to the following
set of iterative equations, which provide the
solution to the ILCD problem.
i
I tot
,d  p  
i 1
I ext
,d  p   N w
1   d d s, p   SIR d' s 
p  P
(10b)
sS
I
i 1
ext, d
~
 p     d d ~s , ~p   I toti 1,d  ~p   SIRd' ~s   al c ~p , ~p 
~
~
al c p , p 
s S p P   p 
(11b)
The set of equations above gives us an iterative
algorithm, identical with that of the ILCU
algorithm.
5
Results
This section provides indicative results on how the
ILCU and ILCD schemes can complement a design
or management process, by enhancing the details
on the anticipated interference levels in each cell.
The objective is to optimally allocate the
transmission power to the connections that
constitute the service demand vector. A macro-cell
test case and different scenarios for the demand
pattern’s accommodation are considered and
further analyzed in the sequence. Micro cell cases
have also been studied and are omitted here for
brevity reasons.
Figure 2. Cell area layout (a) and structure (b)
The service demand pattern consists of 4 services.
A speech service (s1), a 64/64kbps data service
(s2), a 144/144kbps data service (s3) and a high
data rate 384kbps service (s4) only in the forward
direction. The demand volume with the assumed
service characteristics is summarized in Table 1.
The connections referred, cause in each cell an
average loading factor of 46% in the uplink and
75% in the downlink.
5.1 Macro cell test case
Figure 2 depicts the cell area layout and structure.
It consists of 16 hexagonal macro-cells with
distance between the base stations (NodeBs) equal
to 1700m (cell radius equal to 1000m). NodeBs are
located in the center of each cell. It can be shown
the cell splitting into 48 pixels. The cell is also split
into 4 zones around the NodeB.
The Okumura – Hata propagation model is used.
Chip rate is set to 3,84Mcps. Moreover, it is
assumed that mobile terminals can transmit at
maximum 300mW(25dBm) for all the provided
services, while base stations can transmit at
maximum 20W(43dBm). The thermal noise density
is -174dBm/Hz corresponding to a noise power of 108,1dBm.
Table 1. Assumed service characteristics per cell
5.1.1
Scenario 1 - Users uniformly distributed
into pixels
In the first scenario, the users are uniformly
distributed between the pixels inside a cell. The
percentages of users in each cell of the service area
layout that are allocated in the different zones are:
6% in zone 1 for both UL and DL, 22% and 24%
for UL and DL in zone 2, 31% and 30%
respectively in zone 3 and 41% and 40% in zone 4..
It is obvious that in this scenario most of the users
are allocated near the edge of the cells, that is to
say the hardest load conditions are assumed. Figure
3 depicts the total uplink interference at each cell
site, evaluated by the ILCU scheme, and the
4
expected interference based on the loading factor.
Figure 4 depicts the total power transmitted by each
NodeB in the forward link, evaluated by the ILCD
scheme, respectively.
Since the majority of users are allocated near the
edge of the cells (zone4), the uplink interference
appears to be much higher than the expected.
Furthermore, cells towards the center of the area
layout experience higher interference levels in the
uplink while their NodeBs are transmitting at
higher power levels in the downlink.
iterations for the ILCD problem.
Focusing again on cell 6 (Figure 6), it is shown that
the reverse transmission power levels allocated per
service and cell area are reduced for about 2,2dB
comparing to scenario 1. Even for connections
originating from pixels located in the edge of the
cell the transmitted powers appear to be much
lower than the maximum value (25dBm).
Suitability for the management domain is indicated
through the low computational complexity
exhibited by the solution algorithms. Specifically,
the solution algorithm converges after 14 iterations
for the ILCU problem and after 145 iterations for
the ILCD problem.
Figure 5 focuses on a particular cell, and provides
additional information for demand handling. The
selected cell (cell 6) is located in the middle of the
coverage area and therefore senses higher levels of
interference. The figure depicts the reverse
transmission power levels allocated per service and
cell area. This ILCU and ILCD capability enables
the identification and potential re-engineering of
areas, in which the equipment capabilities are
stressed, e.g., zone 4. Therefore, outage situations
can be managed.
5.1.2
Figure 3. Total UL interference levels per cell site.
Scenario 2 - Users uniformly distributed
into cell zones
In the second scenario referred to macro-cells the
users are uniformly distributed between cell zones
within the cell. The percentages of users in each
cell of the service area layout that are allocated in
the different zones are: 25% in all zones for uplink,
while for downlink the percentages are 24% in
zones 1, 2 and 4 and 28% in zone 3. As before
Figure 2 depicts the total uplink interference at
each cell site and the expected interference based
on the loading factor, and Figure 3 depicts the total
power transmitted by each NodeB in the forward
link.
Since users are now allocated in a more uniform
manner between the cell zones, the total uplink
interference decreases (from 0,2 to 2,2dB) and
seems to evolve more normally around the
expected value in comparison with scenario 1. The
total base station power is also reduced
dramatically in the range of about 7 to 10dB.
Figure 4. Total DL transmitted power per Nodeb
Figure 5. Uplink transmitted power per service and
zone in cell 6 (Scenario 1)
The solution algorithm, in this case, converges after
9 iterations for the ILCU problem and after 22
5
interference situations that occur in the network.
Another extension in our work will examine the use
of additional carriers and its impact in the
interference levels and capacity per cell.
Figure 6. Uplink transmitted power per service and
zone in cell 6 (Scenario 2)
6
Conclusions
This paper addressed planning problems that are
important to the design and management of WCDMA-based cellular networks. The problems
aimed at configuring the allowed interference
levels per cell in order to keep up with various
traffic load situations. The problems were concisely
defined, mathematically formulated and solved by
computationally efficient algorithms. Finally, a set
of indicative numerical results was presented. An
interesting issue for future study is to extend the
scope of the test cases by applying the schemes into
different demand patterns and therefore, widen the
number of results on the dynamical handling of the
7
[1]
[2]
[3]
[4]
[5]
[6]
[7]
References
U. Varshney, “Recent advances in wireless networking”, IEEE Computer, Vol. 33, No. 6, June 2000
3rd Generation Partnership Project (3GPP) Web Site, www.3gpp.org, Nov. 2001
H.Holma, A.Toscala, “W-CDMA for UMTS”, J.Wiley&Sons, 2000
T.Ojanpera, R.Prasad, “Wideband CDMA for third generation mobile communications”, Artech House,
2001
J.Laiho, A.Wacker, T.Novosad (Editors), “Radio network planning and optimisation for UMTS”,
J.Wiley&Sons, 2002
A. Fukasawa, T. Sato, Y. Takizawa, T. Kato, M. Kawabe, R.E. Fisher, “Wideband CDMA system for
personal radio communications”, IEEE Commun. Mag., Vol. 34, No. 10, Oct. 1996
J.Laiho, A.Wacker, T.Novosad (Editors), “Radio network planning and optimisation for UMTS”,
J.Wiley&Sons, 2002
6