Impact of macrocellular network densification on the capacity, energy and cost efficiency in dense urban environment

International Journal of Wireless & Mobile Networks (IJWMN) Vol. 5, No. 5, October 2013

IMPACT OF MACROCELLULAR NETWORK
DENSIFICATION ON THE CAPACITY, ENERGY AND
COST EFFICIENCY IN DENSE URBAN
ENVIRONMENT
S. F. Yunas, T. Isotalo, J. Niemelä and M. Valkama
Department of Electronics and Communications Engineering,
Tampere University of Technology, Tampere, Finland

ABSTRACT
This paper aims to show the effect of macrocellular network densification on the capacity, energy and cost
efficiency. The presented results are based on radio propagation simulations that consider macrocellular
network with different inter-site distances, i.e. different site densities, and also take into account the
presence of indoor receiver points by varying outdoor and indoor receiver distribution. It is observed that
as a result of densifying the network, the cell spectral efficiency reduces due to increasing level of inter-cell
interference. However, as a result of densification, the network area capacity can be improved since the
area spectral efficiency increases. Nevertheless, the densification efficiency decreases because of the
reduction of cell spectral efficiency, especially when indoor receiver points are taken into account. The
results hence indicate that densification of macrocellular network suffers from inefficiency which results in
higher energy and cost per bit per Hertz, and thus calls for alternative methods to deploy networks, or
alternatively, more sophisticated methods, such as base station coordination or inter-cell interference
cancellation techniques, to be implemented for future cellular networks.

KEYWORDS
Macrocell densification, area spectral efficiency, energy efficiency, cost efficiency, 3D ray tracing model

1. INTRODUCTION
Macrocellular networks have been and still continues to be the basis for cellular network
deployments globally. High power transmitters with highly elevated and directive antenna array
positions are superior in terms of wide-area coverage provisioning. They also play a major role in
fulfilling the mobility demands of cellular users, and hence, are assumed to maintain their
position in the future as well. Moreover, it is envisioned that macrocell networks will continue to
provide the outdoor coverage layer with small cells satisfying the local outdoor and indoor
capacity demands. Current cellular networks are inherently heterogeneous in terms of network
configuration and this trend is building towards an even denser heterogeneous configuration as
new small cell technology and different other indoor network solutions become more and more
common.
The global cellular traffic has been on a steady rise since the early days of 2G networks. With the
increasing popularity of mobile broadband subscriptions, the projected growth in the capacity
demand has been estimated to increase 15 folds in the upcoming five years [1]. Fundamentally,
the main mechanisms to increase the network capacity are increased link and radio resource
management efficiency together with utilization of wider bandwidth and the cell size, i.e., having
DOI : 10.5121/ijwmn.2013.5507

99


dense network configuration with small cells. The idea of enhancing the system capacity through
network densification can be dated back to late 1940s when the cellular concept was introduced
[2]. The initial adoption of the concept, however, was slow at first but started to gain serious
attention when 2G networks were introduced. Since then, network densification has been viewed
as a feasible pathway towards network evolution.
Lately, engineers and academic researchers have been studying and evaluating the concept with
different performance enhancing techniques. In [3], the performance of macrocellular
densification with different transmission schemes has been compared with a network employing
base station coordination algorithms. The study concentrates on various techniques that can
maximize the minimum spectral efficiency of the served users. In addition, a constant user
density, irrespective of the network size, has been assumed. This results in a partially loaded
system where some of the base stations are kept in sleep mode to avoid over provisioning of the
network capacity. The results show that the cell spectral efficiency increases as the network is
densified to a certain point and then saturates. In [4], the average cell spectral efficiency is shown
to increase linearly with network densification in partially loaded system. The impact of
macrocell densification on the network throughput and power consumption in both homogeneous
and heterogeneous network environments has been studied in [5]. The study considers a fully
loaded network, where all the base stations are continuously transmitting at full power. However,
the maximum transmit power per base station is varied as the network is densified. The findings
in [5] follow the outcomes of [4] i.e., in a homogeneous macrocell network, the cell spectral
efficiency tends to improve with increasing network density. In [6, 7], the performance of
homogeneous macrocellular network densification has been examined and compared with
different heterogeneous network deployment alternatives. The papers take a slightly different
approach by introducing variable traffic. Hence the system performance is evaluated in terms of
served area traffic during busy hour. Unlike in [3-5], where only an outdoor environment is
assumed, the studies in [6, 7], also take into account the indoor environment with buildings and
users distributed among different floors. Nevertheless, their results indicate increasing served area
traffic per busy hour as the network is densified.
While significant amount of time and effort in the last two decades was dedicated by the industry
and academia in improving the spectral efficiency of wireless networks, more recently, the focal
point of the industry has started to expand towards including energy and cost efficiency aspects
into its domain. To cope with the current rate of ‘exponentially’ increasing capacity demand,
deployment of several magnitudes more base stations will be required, which is considered by the
industry to be a feasible pathway. However, this strategy is known to significantly increase the
cost and energy consumption of the cellular networks. The ICT industry currently contributes
around 2% of the global greenhouse gas emission and this contribution is expected to increase to
4% by the year 2020 [8]. According to some studies conducted in 2007/2008, the radio access
networks alone had a share of around 0.3% - 0.5% in the global CO2 emissions [9, 10] and out of
this roughly 80% came from the base stations [11]. As the worldwide awareness regarding global
warming increases, political initiatives at the international level have started to put requirements
on the operators and manufacturers to lower the gas emissions of communications networks [12].
This has led the telecommunication industry, especially the standardization and regulatory bodies,
to focus their attention towards building 'greener' wireless networks. In the research community,
considerable number of studies have been conducted and published in the recent years focusing
on quantifying the energy consumption of the wireless networks by establishing different metrics
for evaluation of the energy efficiency, proposing power consumption models for different base
station types and ways to improving the power consumption of the networks while maintaining
decent quality of service and system throughput. Studies emphasizing on the importance of
having a holistic framework for evaluating the energy efficiency of the wireless networks have
been reported in [13, 14]. In [13], the authors discuss the importance of evaluating the energy
efficiency at each level of network hierarchy, namely component, link and network level, their
100


mutual dependencies and the need for optimizing the system as a whole, rather than just focusing
on improving only one aspect, for achieving an optimum system performance. Building up on the
foundation laid in [13], the authors propose an energy efficiency evaluation framework in [14] for
evaluating nationwide energy-efficiency of a mobile operator. The framework constitutes a power
consumption model for different LTE base station types together with a proposed large scale
deployment and long term traffic model. In [11], a new metric, area power consumption, is
proposed to evaluate and compare the energy efficiencies of networks with different cell ranges
(varying cell site densities/km2). A brief overview of the energy efficiency metrics at different
levels of a wireless network/system has been provided in [15]. The impact of cell size on the
power consumption has been studied in [4] and [16-18] for different deployment strategies.
However, the results in [16] differ with the other studies. The authors in [16] report that large cell
deployments are efficient in terms of area power consumption as compared to small cells while
the finding of [4], [17] and [18] claim otherwise. The contradictory results reported in the above
studies come from differences in the power consumption models utilized in the studies and using
different energy efficiency metrics that capture different statistics of the network [19]. The impact
of network densification on the energy efficiency in the wireless networks has been investigated
in [19]. Unlike the previous studies in [4] and [16-18], which fail to take into account the impact
of interference and system throughput while evaluating the energy efficiency, the studies in [19]
investigate the relation between energy efficiency, area capacity and cell size by taking into
consideration both the interference and noise, and takes relates the energy efficiency in terms of
system throughput. The paper refines the analytical power consumption model proposed in [11]
to include the backhaul power consumption as well which has significant impact when
considering dense deployments [20]. The studies reported in [15-19] have a common short come,
i.e. they all use a simple analytical model that combines the power consumption of components
into two or three parameters and hence do not accurately model the influence of each component,
within the base station, on the total power consumption. In [14], the power consumption of core
components that have significant impact on the total power consumption of the base station, have
been separately modelled. Nevertheless, it still fails to take into account the backhaul power
consumption in its model, which is a key in denser deployments, as mentioned in [20]. This
problem has been addressed in [21], where the proposed analytical model takes into account the
contribution of the core components within the base station as well as the influence of backhaul
transmission unit. The performance of the proposed power consumption model is further
validated by comparing it with measurements from a live 3G network.
In this paper, we study the downlink capacity performance of macrocellular network densification
in a full load condition, which is the worst case scenario and also a typical methodology that is
used for network capacity dimensioning. Unlike in [4] and [5], where the base station parameters
(antenna height, transmit power) are varied with respect to the cell size, we keep the antenna
height and transmit power unchanged; only the downtilt angle is varied as network is densified.
Further, an accurate deterministic 3D ray tracing model has been used in our study. For energyefficiency analysis we use the power consumption model given in [21]. The main target of the
paper is to evaluate impact of site densification on the average cell and area spectral efficiency for
outdoor and indoor receiver point locations, and also to evaluate whether macrocellular
densification is a feasible pathway, in terms of energy and cost efficiency, towards the evolution
of future cellular networks. The study aims to answer the question, how much system capacity
gain can we achieve through macrocellular network densification and whether the capacity gain
is enough lower the energy per bit and cost per bit to make pure macrocellular densification
energy and cost efficient solution? The rest of the paper is structured as follows. We introduce the
system model in Section 2 with the analysis methodology. Section 3 presents the results and
performance analysis for capacity efficiency. In Section 4, the power consumption model for the
base stations is introduced and the energy efficiency results and analyzed. Next, the cost
modelling methodology is discussed in Section 5 and the cost efficiency results are analyzed.
Finally, based on the analysis, concluding remarks are given in Section 6.
101


2. SYSTEM MODELLING
This section presents the system model that we have used in our macrocellular densification
studies. It starts with the introduction of the simulation environment and cell layouts. This is
followed by a brief description of the simulation tool and the propagation model that was used in
evaluations. Next, different network and site configuration parameters as antenna model together
with average inter-site distance (ISD) and corresponding tilt angles are described. Finally, a list of
general simulation parameters is provided.

2.1. Simulation environment and cell layout
To imitate a dense urban environment, we created a fictive Manhattan type grid city model
(Figure 1). A hexagonal layout was used as the basis of our macrocellular deployment strategy,
although the actual network layout determined by the Manhattan grid that does not allow a pure
hexagonal cell layout. Each building has dimensions of 110 m x 110 m, a height of 40 m and
comprises of 8 floors. The streets were selected to be 30 m wide. For indoor floor plan, an open
office layout was chosen. It renders a hall area with no rooms, i.e. no hard obstruction for signal
propagation except for the ceiling, floors and exterior walls.

Figure 1. Manhattan grid city model (aerial view). The arrows show sector antenna positions and
orientations (example case: ISD 297 m). Note that only the first interfering tier is illustrated. The
appropriate number of interfering tiers for each configuration used in the simulations is given in Table 2.

2.2. Propagation model
A commercial radio wave propagation tool (Wireless Insite), was used for the coverage prediction
simulations. The outdoor and indoor radio channels are modelled using a 3D ray based model.
The model employs a ray-launching technique based on 'Shooting and Bouncing Ray' (SBR)
method to find the propagation paths through the 3D building geometry between a transmitter and
receiver [22]. Rays are shot from the emitting source in discrete intervals and traced
correspondingly as they reflect, diffract and transmit (penetrate) through and around the
obstacles. Each ray is traced independently and the tracing continues until the maximum number
of reflections is reached. Once all the propagation paths have been computed and stored, the field
strength for each ray path is calculated using Uniform Theory of Diffraction (UTD) [23-25].
The accuracy of a 3D propagation model is dependent upon the input data and the total number of
reflections, transmissions (or wall penetrations) and diffractions a single ray can encounter. To
limit the calculation time, we used an empirical 'hit-and-trial' method, which involves simulating
with a smaller number of interactions, and then re-simulating the same scenario by steadily
increasing interactions and comparing the results. Once the results start to converge with
102


insignificant change, those settings were then selected. In our case, this was observed at 10
reflections, 1 diffraction and 1 transmission where providing sufficiently accurate results.
To account for the outdoor-to-indoor propagation loss, the exterior wall direct penetration loss
was chosen to be 25 dB. The corresponding electrical properties of the exterior wall were found
empirically by adjusting the conductivity σ , permittivity ε and thickness of the wall, and
observing the difference in the average signal level between several outdoor and indoor receiver
points.

2.3. Antenna models and positions
An extended 3GPP antenna model based on [26] was adopted for simulations. The proposed
version extends the original model of [13], which only considers the horizontal plane, and
includes a vertical antenna pattern model with an option to set the electrical downtilt. The
horizontal (azimuth) pattern, Gh , is given by:
2
  ϕ


Gh (ϕ ) = − min 12 
 , FBRh  + Gm
  HPBWh 




(1)

where, ϕ , −1800 ≤ ϕ ≤ 1800 , is the azimuth angle relative to the main beam direction, HPBWh is
the horizontal half power beam width [0], FBRh is the front-to-back ratio [dB] and Gm is the
maximum gain of the antenna [dBi]. The vertical (elevation) pattern, Gv , is given by:
2


 φ − φetilt 
Gv (φ ) = − max  −12 
 , SLLv 


 HPBWv 



(2)

where φ , −1800 ≤ φ ≤ 1800 , is the negative elevation angle relative to horizontal plane (i.e., φ =
−900 is the upward plane relative to the main beam, φ = 00 is along the main beam direction, and
φ = 900 is the downward plane relative to the main beam), φetilt is the electrical downtilt angle [0],
HPBWv is the vertical half power beam width [0], and SLLv is the side lobe level [0] relative to
the maximum gain. The antenna parameter values were adopted from [26] except for electrical tilt
angles which were based on the average inter-site distances.
The antennas were placed 2 m above the building roof, i.e. 42 m above the ground level. In order
to ensure that the transmitted signal is not obstructed by the roof, sector antennas were placed
either at corners of the buildings rather than at the center to ensure unobstructed propagation (see
arrows in Figure 1).

2.4. ISDs (cell density) and electrical tilt
The cell density depends upon the average inter site distance (ISD or d site ), which further
specifies the dominance area of a cell. In our study, we define dominance area as the region
where a cell provides highest signal level as compared to the rest of the cells. Altogether five
different ISDs were considered. These were calculated from the center of the building (except in
the average ISD of 170 m case, where it is calculated based on average inter cell distance owing
to the square layout of the buildings).
Assuming a regular hexagon cell, the dominance area of a cell, Acell is given by:
Acell  km 2  =



3
d site
6

(

)

2

(3)
103


The cell density, ρ cell , per km2 is defined as 1 / Acell .
In order to avoid unnecessary interference into neighboring cells, the sectors are required to be
down tilted. As mentioned earlier, the electrical tilt angle depends on the ISD as it defines the
maximum cell range. Knowing the base station (BS) antenna height ( hBS ), the mobile station
(MS) antenna height ( hMS ) and the cell range ( rcell ), the tilt angle was calculated geometrically as:
 hBS − hMS 

 rcell


φetilt = arctan 

(4)

The rest of the simulation parameters are gathered in Table 1. Note that the effective isotropic
radiated power (EIRP) in the maximum antenna gain direction is 43 dBm + 18 dBm = 61 dBm.
Moreover, for the receiver noise floor level calculation, a 20 MHz bandwidth was assumed
(nominal for long term evolution, LTE).
Table 1. General simulation parameters.

Parameter
Operating frequency
Bandwidth, W
Transmit power at the antenna, PTX
BS antenna beam width, HPBWh/v
MS antenna type
BS antenna gain
MS antenna gain
BS antenna height, hBS
MS antenna height, hMS
Receiver noise figure
Receiver noise floor, Pn
Propagation environment
Propagation model
Building dimensions
Building height
Street width
Indoor layout
Outdoor-to-indoor wall penetration loss

Unit
[MHz]
[MHz]
[dBm]
[degrees]

[dBi]
[dBi]
[m]
[m]
[dB]
[dBm]

[m]
[m]
[m]
[dB]

Value
2100
20
43
Directional (650/60)
Half-wave dipole
18
2
42
2
9
-92
Manhattan
3D ray tracing
110 x 110
40
30
Open office
25

2.4. Analysis methodology
Due to homogeneity of the environment, we only consider the receiver points from the dominance
area of the center cell site for statistical analysis and then normalize the analysis to 1 km2 area.
For simulating a continuous cellular network and analyzing its performance, it is necessary to
take into account all the interfering cells that contribute to the interference level in the dominance
area of a serving cell. In a realistic environment, the transmission path loss actually caps the
number of receivable interfering sources at the serving cell which in turn limits the total
interference level. However, in an ideal environment, like the Manhattan grid, the situation is
exacerbated by the street canyons. The signals travelling in a street canyon tend to travel further
as compared to the signal travelling in free space [27]. This is due to the tunneling effect caused
by the side walls which direct the signals into the alley. As a consequence, the effect of distant
104


interfering tiers especially the LOS (line of sight) tiers, which were negligible before, starts to
become more visible.
To estimate the number of interfering tiers, only those tiers that had significant impact on the
relative interference levels at the serving cell border (which is worst case scenario) are
considered. This hit and trial method provides a fast and reliable estimate of the
effective/dominant interfering tiers. Table 2 lists the average inter-site distance ( d site ) and the
corresponding electrical tilt angles, cell areas, cell densities (cells per km2), and the number of
interfering tiers used in the simulations.
Table 2. ISD,electrical tilt, cell area, cell density and interfering tiers.

d site
960 m
828 m
593 m
297 m
170 m

φetilt
o

3.5
4.1o
5.8o
11.4o
47.5o

Acell [km 2 ]

ρcell per km 2

0.26
0.2
0.1
0.03
0.008

3.8
5.1
9.9
39.3
119.9

Interfering tiers

2
2
3
4
4

3. CAPACITY EFFICIENCY ANALYSIS
3.1. SINR evaluation and mapping to Shannon capacity
The performance of any cellular system or layout in a certain environment is highly dependent
upon the radio propagation conditions. The quality of the radio link is determined by the coverage
and the interference conditions which set a cap on the maximum throughput/users per cell, as
defined by Shannon capacity bound, C:
C = W log 2 (1 + Γ )

(5)

where W is the bandwidth of the system, Γ is the signal-to-interference-noise ratio (SINR), which
defines the radio propagation condition. From (5) it is evident that the cell/area spectral efficiency
depends directly on Γ .
Assuming that there is no intra cell interference (typical assumption for frequency reuse 1
systems like orthogonal frequency division multiple access (OFDMA), where perfect
orthogonality is assumed between the users of the same cell), the SINR at a jth receiver point
(both outdoor and indoor) is calculated using the following relation:

Γj =

∑I

S j , own
j , other

+ Pn

(6)

where Sj,own is the received signal power from the own cell (serving cell) at jth receiver point, Ij,other
is the received interference power from the other cells at the jth receiver point, and Pn is the noise
floor level which includes the noise figure of the receiver as well.
In a multi-cellular scenario, a cell having the strongest signal level is considered as the serving
cell and the rest are treated as interferers. For a set of i cells reachable at the jth receiver, the best
serving signal can be found as:

105


S j = arg max ( Pr0 j , Pr1 j ,..., Prij )

(7)

i

where, Prij is the received signal power from the ith center cell site at jth receiver.

3.2. Cell spectral efficiency and area spectral efficiency
We consider a fully loaded scenario in our simulation, i.e. all the base stations are transmitting at
full power at all times. The cell spectral efficiency, ηcell , is defined as the maximum bit rate per
Hertz that a cell can support under certain radio propagation conditions. For an area with a cell
density of ρ cell and average cell spectral efficiency of η cell , the average area spectral efficiency,

η area , for a fully loaded system with constant interferers is defined as [28]:
η area bps / Hz per km 2  = ρ cell × η cell


where ηcell =

(8)

C
.
W

3.3. Capacity efficiency result and analysis
The general target of radio network planning is to design a network that provides sufficient
coverage and maximizes overall capacity of the network with minimal costs. One of the most
obvious methods for enhancing the network capacity is to increase the number of cells. However,
the achievable average SINR ratios that eventually define the capacities (or ’cell spectral
efficiency’) in the cell level depend heavily on the network configuration. In this chapter, the
results from the simulations, in terms of coverage, radio channel conditions and capacity, are
analyzed and discussed.
Figure 2 provides the radio coverage statistics for the center cell site for both a) outdoor and b)
indoor receiver points, for different ISDs. For the outdoor environment we can see that the signal
levels improve in the whole cell area as macrocellular network is densified. Further densifying the
network to ISD 170 m is shown to slightly degrade the coverage due to heavy downtilt angle
(47.5º). For the indoor environment, the coverage pattern follows the same as in the outdoor.
Again in the extreme densification scenario (ISD 170 m), the coverage performance degrades
even beyond ISD 969 m.
1

1

Outdoor statistics

0.8

Indoor statistics

0.9
0.8

Cummulative Probability


0.9

0.7
0.6
0.5

ISD 960 m
ISD 828 m
ISD 593 m
ISD 297 m
ISD 170 m

0.4
0.3
0.2

0.6
0.5
0.4
ISD 960 m
ISD 828 m
ISD 593 m
ISD 297 m
ISD 170 m

0.3
0.2

0.1
0

0.7

0.1
-120

-100

-80
-60
-40
Received power [dBm]

a)

-20

0

-120

-100

-80
-60
-40
Received power [dBm]

-20

b)

Figure 2. CDF distribution of received signal strength for a) outdoor receiver pointsa and b) indoor receiver
points.
106


Figure 3 gives the center cell site statistics for the radio channel conditions (SINR) for both the
outdoor (a) and indoor environment (b). In the outdoor environment, the radio channel conditions
does not improve much in the overall cell, rather the SINR performance starts to degrade as we
densify the network. The reduction in the SINR performance is more visible in the indoor
environment with ISD 170 m having the worst interference conditions of all due to close
proximity of the interfering cells.
1

1

Outdoor statistics

0.8

Indoor statistics

0.8



0.9

0.9

0.7
0.6
0.5
0.4
ISD 960 m
ISD 828 m
ISD 593 m
ISD 297 m
ISD 170 m

0.3
0.2
0.1
0
-30

-20

-10

0
10
SINR [dB]

a)

20

30

0.7
0.6
0.5
0.4
ISD 960 m
ISD 828 m
ISD 593 m
ISD 297 m
ISD 170 m

0.3
0.2
0.1

40

0
-30

-20

-10

0
10
SINR [dB]

20

30

40

b)

Figure 3. CDF distribution for SINR [dB] for a) outdoor receiver points and b) indoor receiver points.

From the overall cell site capacity perspective, the improvement in the cell edge performance is
of significance, as these regions, due to being away from the serving base station, experience
worse radio propagation conditions. Hence, it is extremely important that the cell edge conditions
in terms of cell capacities are as high as possible. The maximization of these edge conditions can
be achieved with a proper radio network deployment that eventually minimizes the inter-cell
interference caused by the overlap between adjacent or neighboring cells. Thus, in our analysis
we will focus on the lower 10th percentile values, which represent the conditions at the cell edge
regions.
Figure 4 shows the statistical 10th percentile values for the received signal levels (i.e., coverage)
and SINR, respectively, for the outdoor (a) and different indoor floor levels (b). The x-axis
indicates the cell density per km2 and y-axis the corresponding received signal strength [dBm] or
SINR [dB]. For analysis, we have grouped the indoor floor levels into three classes; the bottom
floors, middle floors and the top floors. The bottom floors bar presents the average of the 10th
percentile values on the 1st and the 2nd floor, the middle floors bar indicate the average of the 10th
percentile values on the 4th and the 5th floor, while the top floors bar shows the average of the 10th
percentile values on 7th and 8th floor. From Figure 4a) we can see that the outdoor receiver points
experience quite high signal levels from the very beginning as compared to the indoor floors. The
received signal levels are relative to receiver noise floor level which is at -92 dBm (as shown by
the dashed line). For less densified configurations, the receiver points in the lower floors
experience high signal losses as compared to ones on the top floors. However, as a result of
densification of the network, the overall coverage levels start to improve. The improvement in the
coverage level comes from the deployment of more base stations together with antenna down tilt
that results in smaller cell sizes, thereby reducing the path losses. Subsequent densification of the
network does not bring any further improvement in the indoor coverage, while the outdoor
receiver points experience a moderate improvement in the average signal levels. In the extreme
case of 120 cells/km2 (or average ISD of 170 m), the average signal levels saturate for receivers in
outdoor and top floors, whilst the signal levels in the middle and lower floors start to experience
107


coverage limitations. This is due to very high antenna tilt angles that cause signal losses in the
lower floors.
5
Outdoor
Bottom floors
Middle floors
Top floors

-20

Outdoor
Bottom floors
Middle floors
Top floors
SINR [dB] - 10 percentile

Received Signal Strength [dBm] - 10 percentile

0

-40

-60

-80

0

-5

-100

-120

3.8

5.1

9.9

39.3

119.9

-10

3.8

5.1

9.9

39.3

119.9

2

Cell density [per km2]

Cell density [per km ]

b)

a)

Figure 4. Cell edge values (the 10th percentile statistics) for a) received signal strength [dBm] and b) SINR
[dB].

Figure 4b) presents the 10th percentile values i.e., the statistics from the cell edge, for SINR in
outdoor and indoor environment for different cell densities. Although the coverage conditions on
the top floor are better than in the middle and lower floors, the SINR performance degrades quite
abruptly on top floor as the cell density increases. This is due to the rising interference conditions
that become more prominent on the top floors as the network is densified. On the other hand, as a
result of coverage improvement, the radio conditions in the lower and middle floor improve
slightly when the network is densified to the level of 5 cells/km2 (or average ISD of 828 m). For
more densified configurations, lower and middle floors start to become coverage and clearly
interference limited.
Table 2. Capacity performance at the cell border region (the 10th percentile values) for outdoor and different
indoor floor levels.

Cell spectral efficiency, ηcell , [bps/ Hz]
10 percentile
d site
969 m
828 m
593 m
297 m
170 m

ρcell per km 2
3.8
5.1
9.9
39.3
119.9

Outdoor
0.84
0.84
0.92
0.88
0.49

Bottom floors
0.76
0.97
0.88
0.75
0.26

Middle floors
0.96
1
0.86
0.73
0.27

Top floors
0.98
0.88
0.75
0.47
0.28

Table 4 provides the 10th percentile values for the cell spectral efficiency versus cell densities, for
the outdoor and different indoor floor levels. The SINR values under the dominance area of the
center site are directly mapped to the cell spectral efficiency. In a full load condition, the cell
efficiency is shown to decrease as the network is densified. Initially (3.8 cells/km2), the cell edge
spectral efficiency is at the level of 0.84 bps/Hz and reduces to the level of 0.49 bps/Hz for
outdoor locations when network is densified to the level of 120 cells/km2. For the indoor floor
levels, the overall cell edge efficiency is higher on the middle and top floors as compared to the
lower floor levels and even outdoor location. However, as the network is densified to 120
cells/km2, the cell spectral efficiency reduces to approximately 0.27 bps/Hz on all the floor levels.
108


The higher degree of resource reuse due to denser deployments results in an increase of the area
spectral efficiency as shown in Table 5. The impact of outdoor and indoor location on the area
spectral efficiency is observed to be quite marginal in the beginning (ISD of 969m and 828m), but
as the network is densified, the difference in the area capacity gain starts to become more visible.
For the cell spectral efficiency, the effect tends to get more recognizable when the network is
densified beyond the level of 5 cells/km2 (or average ISD of 828 m). This is attributed to the
rising indoor interference level, mostly on the top floors as shown in Figure 2b).
Table 3. Average cell and area spectral efficiency for different ISDs.

ηcell [bps / Hz]
d site
969 m
828 m
593 m
297 m
170 m

ρcell per km 2
3.8
5.1
9.9
39.3
119.9

Outdoor
2.7
2.65
2.57
2.09
1.65

Indoor
2.67
2.61
2.05
1.99
0.88

ηarea [bps / Hz / km 2 ]

Outdoor
15.1
22.42
36.05
92.81
289.2

Indoor
14.96
22.06
28.74
88.06
153.9

In mobile communications industry, it has been widely speculated that more than 70% of the
overall network traffic originates from indoor users. Hence, to properly dimension its network a
mobile operator has to consider service provisioning from the indoor perspective. However, the
results indicate that macrocellular network densification in urban Manhattan environment clearly
suffers from inefficiency indoors. If the radio network planning target is limited to coverage
provisioning for outdoor users only, the densification efficiency is higher (see Figure 3a). On the
other hand, if networks are planned for indoor coverage (as in practice), the efficiency is clearly
lower. To illustrate this for a practical outdoor/indoor user distribution, we consider a scenario
where majority of the receiver points are located indoors i.e., we assume the receiver distribution
ratio as 20% outdoor and 80 % indoor. To ensure statistically reliable results, the receiver points
were randomly selected, with several iterations, from both outdoor and indoor environment with
the intended ratio.
Figure 5 illustrates the capacity analysis in a slightly different way, where the relative area
spectral efficiency for a network with different cell densities per km2 has been depicted. The area
capacity values are relative with respect to nominal site density (3.8 cells/km2). The dashed line
illustrates 100% densification efficiency (ρeff) line, whereas the solid line shows the improvement
of the area spectral efficiency for 20/80 % outdoor/indoor receiver point distribution. For less
dense configuration, there can be observed a linearly increasing trend in the area spectral
efficiency. The densification efficiency is still roughly 0.8 for 9.9 cells/km2 (or average ISD of
597 m). However, beyond that point the efficiency can be observed to deteriorate significantly
due to increase of inter-cell interference resulting in from network densification, and abruptly
drops down to 0.38 for 119.9 cells/km2 scenario. These results clearly illustrate the inefficiency of
macrocellular network densification with a more practical user distribution.

109

3.5

30

3

25

2.5

20

2
ρeff=0.38

15

1.5

10

1

5

0.5

Average cell efficiency [bps/Hz]

Relative average area efficiency

35

20% Outdoor and 80 % Indoor
0
0

5
10
15
20
25
30
Relative cell density [ISD 960 m is set as reference]

0
35

Figure 5. Relative area efficiency ( ) and average cell efficiency ( ) vs. relative cell density (the dashed
line indicate a linearly increasing area efficiency curve in an ideal case).

4. ENERGY EFFICIENCY ANALYSIS
4.1. Power consumption modeling
In order to estimate the energy efficiency of a macrocellular network over a certain area, it is
important that the power consumption of a single base station (BS) site is modelled as accurately
as possible. A base station site comprises of a base station unit, also known as the base transceiver
station (BTS), which has the capability to transmit and receive radio signals to and from the
mobile subscribers. It acts an interface between a subscriber and the core network, enabling the
subscriber to wirelessly connect to the mobile operator’s network. A macrocellular base station
consists of various internal power consuming components that contribute to the overall energy
usage. The components are power amplifier, transceiver, digital signal processing (DSP) chips
and rectifier [14], [21]. Each sector of a base station has its own set of these components, hence
the total power consumption scale with the number of sectors per base station. Apart from the
internal components, there are some external equipment that also have significant contribution to
the overall power usage of a macro base station cell site. The base stations require these external
equipment for its smooth operation and connectivity to the mobile operator’s core network, for
instance, an air conditioning unit is required to maintain an ambient temperature (usually 25ºC)
inside the base station shelter which houses the macro base station unit. Moreover, for connecting
the base station to the backhaul network a transmission unit is installed. Note that the choice of
transmission unit has considerable impact not only on the power consumption but also on the
capital expenditure CAPEX (cost of equipment, deployment of backhaul network to the base
station site etc.) and operating cost OPEX (power consumption, leased line rent etc.). For
example, an optical transmission unit will consume less power and provides a very high capacity
than a microwave transmission unit, but the associated cost of leasing or deploying an optical
network increases the CAPEX considerably. In our study we consider a microwave link
transmission unit, which is still widely used solution for mobile backhaul connectivity. For
accurate estimation of the total network energy utilization, it is essential to also take into account
the contributions from these external equipment.
An analytical power consumption model for a macrocellular base station site is proposed in [13]
that takes into account the contribution from both internal base station components and external
site equipment. The model further takes into account the impact of hourly network load on the
total base station power utilization as well. As such the power consumption of a macro base
station site, Pconsumption/Macro, is given by [21]:
110


Pconsumption / Macro [Watt hour ] = Pconstant + Pload F
•

(9)

where Pconstant is the contribution from the internal components and external equipment whose
power consumption is network load independent, Pload is the contribution from the
components/equipment whose power consumption depends on the instantaneous load of the
network, and F is the load factor varying from 0 to 1, with 0 meaning ‘no load’ and 1 pertains to
‘high load/full load’. The F scales the power consumption of the load dependent base station
components according to the network load per hour. In our analysis we use F = 1, which is the
worst case scenario, i.e. the network is running on a 100 % load and all the load dependent
components are consuming high power.
The load independent power consumption components include; rectifier (Rect), microwave link
(MLink) and air-conditioning (Air-cond) unit. Thus, Pcontant is given by:

Pconstant [Watts ] = ( nsector PRect ) + PMLink + PAir-Cond
•

(10)

where PRect is the power consumption of the rectifier, PMLink is the power consumption of a
microwave link, and PAir-cond is the contribution from the air conditioning unit. As every sector of
a macro base station has a separate rectifier, the total power consumption of the rectifier is scaled
with the number of sectors, nsector, installed at a base station.
The network load dependent power consumption components include power amplifier,
transceiver and a baseband processing card (digital signal processing, DSP card). Thus, Pload is
given by:
Pload [Watts ] = nsector  n Ant ( PAmp + PTRX ) + PDSP 


•

(11)

where PAmp is the power consumption of the amplifier, PTRX is the power consumption of a
transceiver and PDSP is the power usage contribution from the DSP card. The power contribution
from these components is scaled with nsector. Note that in case of MIMO (multiple input multiple
output) configuration, a sector can have more than one antenna installed. Each antenna has a
separate amplifier and a transceiver. Hence, the number of antennas per sector, nAnt, also has an
effect on the total power consumption. In our analysis we assume a SISO (single input and single
output) configuration, i.e. 1 antenna per sector. The power consumption of the amplifier, PAmp,
depends on the input power requirement of the antenna, PTX, and efficiency of the power
amplifier, ηamp and is given by:
PAmp =

PTX

ηamp

(12)

Table 6 summarizes the input parameters for the base station power consumption model. The
parameters are approximate values of the power consumption of different base station
components and external equipment taken from [21], except for the power amplifier efficiency
value which is typically 30 % to 55% for the base stations. Using the input values in Table 3, the
power consumption of a 3-sectored macro cellular base station operating in a full load condition
is approximated to be 1338 Watts.

111

Table 6. Input parameters for the power consumption model.

Component/Equipment
Number of sectors, nsector
Number of antennas per sector, nAnt
Transmit power at the antennas, PTX
Power consumption of DSP chip, PDSP
Power Amplifier efficiency, ηAmp
Power consumption of Transceiver, PTRX
Power consumption of Rectifier, PRect
Power consumption of Air-conditioning unit, PAir-cond
Power consumption of Microwave-Link unit, PMLink

Unit

[Watts]

Value
3
1
20

[Watts]
[%]
[Watts]
[Watts]
[Watts]
[Watts]

100
45
100
100
225
80

4.2. Area power consumption and energy-efficiency
One of the most commonly used metric for assessing the energy efficiency of a network is by
evaluating the energy-to-bit ratio performance of the network, i.e. the amount of energy
consumed in transmitting one bit of information. On a network level, this relates to the total
power consumed by the network in providing an aggregate network capacity. This methodology
is appropriate for assessing the energy efficiency of a network operating under full load condition
[13].
For studying the impact of base station site densification on the energy efficiency of the network,
we first find the area power consumption by normalizing the total power consumption of a base
station given in (9) to 1 km2 area. The normalized area power consumption of a macro cellular
base station, Pconsumption/km2, is given by:
Pconsumption / km2 Watts / km  =


2

Pconsumption / Macro
Asite

(13)

where ASite is the area of a base station (in km2) and is defined as 3×Acell for a 3 sectored base
station. The area power consumption is deemed as an appropriate metric in a case where the
network is operating below its full load capacity, and the target is to minimize the power
consumption over an area [13]. Hence, we use the energy-bit-ratio metric in our analysis. Finally,
we define the energy-efficiency as the power consumed in transmitting one bps/Hz and is
calculated as following:
Eeff [ bps / Hz / kW ] =

η area
Pconsumption / km2

(14)

where η area is the average area spectral efficiency as defined in (8).

4.3. Energy efficiency results and analysis
Table 7 summarizes the energy efficiency results for different inter site distances (varying cell
densities) for outdoor and indoor receiver points. As we can note from the results, the power
consumption per km2 increases with the increase in the cell density. This is because the area
power consumption depends on the coverage area of the base station. We assume that the network
coverage is continuous, without any coverage holes, and that the base stations consume same
amount of power irrespective of the coverage area size. Hence, densification of the network leads
112


to increased power consumption per area proportionally with the increase in number of base
stations.
Table 7. Area power consumption and energy efficiency for different ISDs.

ηarea [bps / Hz / km ]
2

d site

969 m
828 m
593 m
297 m
170 m

ρcell

per km

3.8
5.1
9.9
39.3
119.9

2

Pconsumption / km2 [kW / / km ]
2

1.7
2.3
4.4
17.6
53.5

Outdoor
15.1
22.42
36.05
92.81
289.2

Indoor
14.96
22.06
28.74
88.06
153.9

Eeff [bps / Hz / kW ]
Outdoor
8.9
9.7
8.2
5.3
5.4

Indoor
8.8
9.6
6.5
5
2.9

By densifying the network, the spectrum resource are reused more frequently, which thereby
improves the area spectral efficiency. However, looking at the impact of site densification on the
energy efficiency of the network, it is noted that although increasing the number of bps/Hz/km2,
the energy needed to transmit 1 bps/Hz also increases as we densify our network, especially in the
indoor environment. As an example, consider the initial case of 3.8 cells/km2 (ISD 960 m), where
the average area spectral efficiency is the same for both outdoor and indoor environment. In this
case, the total power consumed per km2 is approximately 1.7 kW, which leads to energy
efficiency of approximately 8.9 bps/Hz/kW for outdoor and 8.8 bps/Hz/kW for indoor
environment. Upon decreasing the inter-site distance to 828 m (i.e., 5.1cells/km2), a slight
improvement can be observed in the energy efficiency (9.7 bps/Hz/kW for outdoor and 9.6
bps/Hz/kW for indoor). This improvement comes from the fact that in the initial stages of
densification, the macrocellular network is slightly coverage limited. Hence, by densifying the
network, the coverage levels improve in both outdoor and indoor environment, thereby improving
the radio channel conditions and hence permitting higher cell spectral efficiency. Subsequent
densification down to ISD 593 m and 297 m starts to degrade the energy efficiency performance
as the network becomes more and more interference limited. The impact of degradation is more
visible in the indoor environment due to relatively low rate of spectral efficiency improvement as
compared to the outdoor environment. Eventually, when we densify the network to an extreme
case (ISD 170 m case or 120 cells/km2), given approximately 32 times more cells/km2 as
compared to initial ISD 969 m case, the area power consumption increases significantly. However
at this stage, a slight improvement in the outdoor energy efficiency can be observed, but for the
indoor environment the degradation in the energy efficiency performance extends even further.
The reason is attributed to the capacity inefficiency in macrocellular networks.
As discussed in the previous section, the macrocellular network densification suffers from
capacity in-efficiency in the indoor environment. The relative indoor capacity gain that we can
achieve when we densify the network from 3.8 cell/km2 to approximately 120 cells/km2 (32 times
more cells) is only 38%, which means that although the power consumption per km2 is increasing
as we are increasing the number of cell density, the associated area capacity in the indoor
environment to offset this increase of power consumption is not enough, hence the Watts/bit
increases. Looking at the big picture, we conclude that pure macrocellular network densification
suffer from both outdoor and indoor energy inefficiency.

5. COST EFFICIENCY ANALYSIS
One of the most obvious ways to increase the capacity of a wireless network is by reusing the
existing allocated spectrum as frequently as possible throughout the network service area, in other
words by increasing the base station density. As such, the capacity of a cellular network is
113


considered to be proportional to the base station density. Unfortunately, the network
infrastructure cost also increases with the number of base stations which is a key concern for cost
aware mobile operators that are striving to provide better services at lower cost in a highly
competitive market. For any technological pathway to be feasible, the benefits must outweigh the
incurred costs. Hence, correct estimation of the benefits and costs is very crucial. Cost efficiency
analysis, or cost-benefit analysis, is one of the key methodologies that provide a general picture of
the cost structure of an evolutionary pathway for a certain technology or system and whether or
not it is a feasible option for investment. In this section we describe the cost modelling
methodology used in our analysis, and based on the cost model we evaluate the cost-bit ratio for
different ISDs (different site densities). Finally, we conclude the section by analyzing and
discussing the cost-efficiency results.

5.1. Cost modeling
The cost of deploying a macrocellular network can be broadly divided into two types:
i.
ii.

Investment cost or CAPEX (capital expenditure),
Running/operational costs or OPEX (operational expenditure)

The CAPEX consists of equipment costs like radio base station, transmission equipment,
antennas, cables, and site build out and installation cost. OPEX consists of site rental, electricity,
transmission or leased line, and OA&M (operation, administration & maintenance). In addition
to these, there can be cost components as such as radio network planning, core network and
marketing costs whose impact can be modeled and taken into account as part of the radio network
costs [29]. However, in the frame of this article, the scope is limited to items listed for CAPEX
and OPEX as they typically depend very strongly on the number of deployed radio components.
Combining CAPEX and OPEX gives the total cost of ownership (TCO) value of the deployed
network.
The total cost structure of a mobile operator is dominated by the accumulated running costs i.e.
the OPEX [30], which spans over the life-time of the network, while the CAPEX is considered
during the initial network roll-out phase or when the network is upgraded. Thus, in order to
account for both the CAPEX and OPEX in finding the ‘total cost per base station’ we use a
standard economical method known as discounted cash flow (DCF) analysis, which gives the net
present value (NPV) of the base station cost.
The net present value of the base station is simply found by summing up the discounted annual
cash flow expenditure for a given study period (in years) [30, 31]. Mathematically;
Y −1

BS NPV = ∑
i =0

ci

(1 + r )

i

(15)

where Y is the study period in years (typically 8 years for base stations value depreciation), ci is
the total annual expenditure per base station (total annual cost which includes running cost and
may include investment cost) in the ith year and r is the discount rate which is assumed to be equal
to 10%. Table 8 gives the various cost items related to CAPEX and OPEX and their approximate
values. The values have been adopted from [31-33].

114


To find the net present value of the base station we make the assumption that the mobile operator
is deploying its network as a Greenfield project1 and that the whole network is deployed in the
first year, so the CAPEX will be considered only for first year. Using the values in Table 8, the
total cost per base station is 93 k€.
In order to assess the cost viability of the macrocellular network densification as an evolutionary
pathway for future mobile broadband systems, we analyze the cost-bit ratio efficiency, or simply
the cost-efficiency metric for the different ISDs. The cost efficiency is defined as the cost
incurred in transmitting one bit/Hz and is calculated as following:
ceff [ bps / Hz / k € ] =

η area
Tcost/km2

(15)

where Tcost/km2 is the total area cost, i.e. the total cost of base stations over 1 km2 area and η area is
the average area spectral efficiency.
Table 8. CAPEX and OPEX related cost for a macro cellular base station.

CAPEX (Initial costs)
Macro base station equipment

Site deployment cost
Total CAPEX
OPEX (Running costs)
Site rent (lease)
Electricity (power consumption charges)2
Transmission line / Leased line rent3
Operation, Administration & Maintenance (OA&M)
Total OPEX per annum

10 k€
5 k€
15 k€

5 k€/year
2.25 k€/year
0 k€/year
5 k€/year
12.25 k€

5.2. Cost efficiency results and analysis
The cost efficiency analysis results for different macrocellular cell densities have been
summarized in Table 9. As evident, the total cost of deployment per km2 increases as we increase
the base station density. However, the important metric to investigate is not the aggregate cost but
the cost per bit efficiency, i.e., we are interested in the relative gain that we can achieve from
densification. In other words, we need to investigate whether the macrocellular densification can
provide decent capacity gain to offset the incurred cost of deployment and hence bring down the
cost per bit to make macrocellular densification a viable business case for investment.
In general, looking at the cost efficiency values in Table 9, it can be seen that it follows the
energy efficiency performance pattern. In the initial stages (increasing the cell density from 3.8 to
5.1 cells/km2), there is a slight improvement in the cost efficiency performance for both outdoor
and indoor environment. However, further densification not only degrades the cost efficiency
performance but also the difference between outdoor and indoor environment starts to become
more noticeable. This is, as mentioned previously, attributed to the inefficiency of pure
macrocellular network densification in the indoor environment. From (15) it can be seen that the
1

Any new network which is designed and deployed from scratch, i.e. operator has no prior deployment in that region.
The annual cost of electricity consumption is normally calculated based on total kilowatt hours (kWh) consumed
during a given year. The cost per electricity unit is assumed to be 0.2 €/kWh. The calculated annual cost of electricity
for macrocell base station given in Table 5 conforms quite well to the range given in [32].
3
We assume that the base stations are connected to the backhaul network via a microwave transmission link. Thus,
there is no leased line (E1/T1) rent or deployment cost of fiber optic network.
2

115


cost efficiency depends on the area spectral efficiency. The more the network is densified, the
smaller capacity gain in the indoor environment is achieved, which results in higher cost per bit.
For the outdoor environment, increasing the base station density to the extreme case (i.e., 120
cells/km2) results in relatively higher area capacity gain than the indoor environment, which
results in slightly improved cost efficiency. However, the efficiency still lags behind the cost
efficiency of ISD 969 m. Hence, it can be concluded that the pure macrocellular densification
suffers from cost inefficiency especially in indoor environment in dense urban area.
Table 9. Total area cost and cost efficiency results for different ISDs.

d site

ρcell per km 2

969 m
828 m
593 m
297 m
170 m

3.8
5.1
9.9
39.3
119.9

Tcost / km2 [ k € / km

117
157
305
1224
3720

2

]

ηarea [ bps / Hz / km
Outdoor
15.1
22.42
36.05
92.81
289.2

2

]

Indoor
14.96
22.06
28.74
88.06
153.9

ceff [ bps / Hz / k €]
Outdoor
0.13
0.14
0.12
0.08
0.08

Indoor
0.13
0.14
0.09
0.07
0.04

6. CONCLUSION
In this paper we have shown how macrocellular network densification reduces the cell spectral
efficiency under full load conditions and varying receiver point distributions (outdoor/indoor) in
dense urban environment using a 3D radio signal propagation model. As a result of reduction of
cell spectral efficiency, the area spectral efficiency starts to saturate and macrocellular network
densification becomes less efficient. However, if an operator targets only for outdoor coverage
with macrocellular network, it still might be sufficient to fulfill the capacity demands through
densification. From indoor coverage and capacity provisioning point of view, however,
macrocellular network densification is clearly less efficient (0.38 densification efficiency with
20/80% outdoor/indoor receiver point distribution). The reduction in the spectrum efficiency also
has a direct impact on the energy and cost efficiency of macrocellular network: lower spectrum
efficiency in the indoor environment results in higher energy consumption and cost per bit.
Moreover, from coverage point of view, in a dense urban environment with high rise buildings,
the macrocellular network is not efficient in providing good coverage to indoor floors. This
affects the attainable capacity in the indoor environment, and clearly further indicates that
increasing network capacity demand will require alternative deployment strategies, as
introduction of small cells (micro) or indoor (pico, femto), that will provide local indoor capacity
within the network. On the macrocellular network, however, alternative mechanisms as
interference mitigation techniques using smart antenna systems, base station transmission
coordination or interference cancellation mechanisms are clearly needed to increase the
densification efficiency.
Future work will concentrate on analyzing the coverage, capacity, costs and energy consumption
of small or microcell networks and comparing that with macrocellular networks. Moreover, we
will evaluate the effect of macrocellular network densification with base station coordination
from capacity, costs, and energy-efficiency point of views.

116


ACKNOWLEDGEMENTS
This research work has been financially supported by the Finnish Agency for Technology and
Innovation (Tekes), under the Sino-Finland collaboration project “Energy-Efficient Wireless
Networks and Connectivity Devices – Systems (EWINE-S)”.

REFERENCES
[1]

Traffic and Market Data Report, (2012), Annual report, Ericsson Inc.

[2]

Macdonald, V. H., (1979), “Advanced Mobile Phone Service: The Cellular Concept”, The Bell
System Technical Journal, Vol. 58.

[3]

Liang, Y., et al., (2008), “Evolution of Base Stations in Cellular Networks- Denser Deployment
versus Coordination”, IEEE International Conference on Communications.

[4]

Badic, B., Farrell T. O’, Loskot, P., He J., (2009), “Energy Efficient Radio Access Architectures
for Green Radio: Large versus Small Cell Size Deployment”, IEEE 70th Vehicular Technology
Conference (VTC).

[5]

Richter, F., Fettweis, G., (2010), “Cellular Mobile Network Densification Utilizing Micro Base
Stations”, IEEE International Conference on Communications.

[6]

Hiltunen, K., (2011), “Comparison of Different Network Densification Alternatives from the LTE
Uplink Performance Point of View”, IEEE 74th Vehicular Technology Conference (VTC).

[7]

Hiltunen, K., (2011), “Comparison of Different Network Densification Alternatives from the LTE
Downlink Performance Point of View”, IEEE International Symposium on Personal, Indoor and
Mobile Radio Communications (PIMRC).

[8]

ICT Sustainability Outlook: An Assessment of the Current State of Affairs and a Path Towards
Improved Sustainability for Public Policies, White paper, (2013), BIO Intelligence services and
Alcatel Lucent.

[9]

Sustainable Energy Use in Mobile Communications, (2007), White paper, Ericsson Inc.

[10]

Pickavet M., et al., (2008), “Worldwide Energy Needs for ICT: The Rise of Power-Aware
Networking”, International symposium on Advanced Networks and Telecommunications Systems.

[11]

Richter F., et al., (2009), “Energy Efficiency Aspects of Base Station Deployment Strategies for
Cellular Networks”, IEEE 70th Vehicular Technology Conference (VTC).

[12]

EU Commissioner calls on the ICT industry to reduce its carbon footprint by 20% as early as
2015, MEMO/09/140. Press release, EU Commission (www.europa.eu)

[13]

Correia, L.M., et al., (2010), “Challenges and Enabling Technologies for Energy Aware Mobile
Radio Networks”, IEEE Communications Magazine.

[14]

Auer G., et al., “Cellular Energy Efficiency Evaluation Framework”, IEEE 73rd Vehicular
Technology Conference (VTC).

[15]

Chen T., et al., (2010), “Energy Efficiency Metrics for Green Wireless Communications”, IEEE
International conference on Wireless Communications and Signal Processing (WCSP).

[16]

Ericson M., (2011), “Total Network Base Station Energy Cost vs. Deployment”, IEEE 73rd
Vehicular Technology Conference (VTC).

[17]

Le T., and Nakhai M., (2010), “Possible power-saving gains by dividing a cell into tiers of smaller
cells”, IET Electroncis Letters, Vol. 46, No. 16.

[18]

Leem H., Baek S. Y. and Sung D. K., (2010), “The Effects of Cell Size on Energy Saving, System
Capacity, and Per-Energy Capacity”, IEEE Wireless Communications and Networking
Conference.
117

[19]

Tombaz S., Sung K. W. and Zander J., (2012), “Impact of Densification on Energy Efficiency in
Wireless Access Networks”, IEEE Globecom Workshops.

[20]

Tombaz S., et al., (2011), “Impact of backhauling power consumption on the deployment of
heterogeneous mobile networks”, IEEE Globecom conference.

[21]

Deruyck M., Joseph W., and Martens L., (2012), “Power consumption model for macrocell and
microcell base stations”, Transactions on Emerging Telecommunications Technology.

[22]

Schuster, J. and Luebbers, R., (1996), “Shooting and Bouncing Rays: Calculating the RCS of an
arbitrarily shaped cavity”, IEEE Transactions on Antenna and Propagation, Vol. 1.

[23]

Keller, J. B., (1996), “Geometrical Theory of Diffraction”, Journal of Optical Society of America,
Vol. 52, No. 2.

[24]

Kouyoumjian, R. G., Pathak, P. H., (1974), “A Uniform Geometrical Theory of Diffraction for an
edge in a perfectly conducting surface”, IEEE Proceedings, Vol. 62, No. 11.

[25]

Balanis, C., (1989), Advanced Engineering Electromagnetics, John Wiley & Sons Inc.

[26]

Gunnarsson, F., et al., (2008), “Downtilted Base Station Antennas - A Simulation Model Proposal
and Impact on HSPA and LTE Performance”, IEEE 68th Vehicular Technology Conference (VTC).

[27]

Sridhara V., and Bohacek S., (2007), “Realistic propagation simulation of urban mesh networks”,
International Journal of Computer and Telecommunications network Computer Networks and
ISDN Systems (COMNET).

[28]

Alouni M., and Goldsmith A., (1999), “Area Spectral Efficiency of Cellular Mobile Radio
Systems”, IEEE Transactions on Vehicular Technology, Vol. 48, No. 4, pp. 1047-1066

[29]

Giles. T, et al., (2004), “Cost Drivers and Deployment Scenarios for Future Broadband Wireless
Networks”, IEEE 59th Vehicular Technology Conference (VTC).

[30]

Smura T., (2012), Techno-economic modelling of wireless network and industry architectures,
Doctoral dissertation, Dept. Communications and Networking, Aalto University.

[31]

Johansson K., et al., (2004), “Relation between base station characteristics and Cost structure in
Cellular systems”. IEEE International Symposium on Personal, Indoor and Mobile Radio
Communications (PIMRC).

[32]

Markendahl, J. and Ma kitalo, O., (2010), “A comparative study of deployment options, capacity
and cost structure for macrocellular and femtocells networks”, IEEE International Symposium on
Personal, Indoor and Mobile Radio Communications (PIMRC).

[33]

Niemela J., and Isotalo T., (2012), Coverage constrained Techno-economical comparison of
Macro and Small Cell Planning Strategies, Internal report, Dept. Electronics and Communications
Engineering., Tampere University of Technology.

118

Impact of macrocellular network densification on the capacity, energy and cost efficiency in dense urban environment

More Related Content

What's hot (20)

Viewers also liked (19)

Similar to Impact of macrocellular network densification on the capacity, energy and cost efficiency in dense urban environment (20)

Recently uploaded (20)

Impact of macrocellular network densification on the capacity, energy and cost efficiency in dense urban environment