2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE)
Energy and Spectral Efficiency in Cellular Networks
Considering Fading, Path Loss, and Interference
Abdulbaset M. Hamed∗§ and Raveendra K. Rao∗
∗ Faculty of Engineering, Department of Electrical and Computer Engineering
University of Western Ontario, London, Ontario, N6A 5B9. Canada
§ Faculty of Engineering, University of Tripoli, Tripoli, Libya
Email: {ahamed6, rrao}@uwo.ca
Abstract—Due to the increase of energy consumption in wireless
systems, energy-efficient cellular planning concept has become an
important concern in designing cellular networks. In this paper,
energy efficiency (EE) and spectral efficiency (SE) analysis for
single and multi-cell cellular systems are presented and investigated.
The efficiency analysis is studied under the influence of Nakagami-
m multi-path fading superimposed on path loss, and co-channel
interference for three base station (BS) antenna configurations
which are omni, and 120o and 60o directive antennas. The downlink
SE and EE are derived and simulated concerning the random
users’ location, normalized reuse distance, cell radius, and random
channel gain. Theoretical and simulation results show that, as
expected, single cell scenario provides higher efficiency than multi
cell; however, antenna directivity improves both efficiency metrics.
The analysis provides insight contribution for the SE-EE trade-
off issue in cellular networks for which most of the network and
propagation parameters are included in the analysis.
Keywords—Cellular Networks, Energy Efficiency, Spectral Effi-
ciency, Fading Channel, Co-channel Interference.
I. INTRODUCTION
With the massive deployment of mobile cellular networks,
corresponding energy consumption is also escalated. Wireless
base station (BS) is regarded as a huge contributer in the total
energy consumption of the cellular networks. D. Lister has
reported [1] that energy bill of BS accounts for about 18%
of the Operation Expenditure in European cellular market. This
rise of energy consumption leads to an increase in greenhouse
gas emission that ultimately contributes to the global CO2
emissions [2]. Studies in [3] and [4] have shown that power-per-
user in mobile networks is rapidly increasing and the demand
for electrical energy in wireless communication has annually
increased by 20%. In this sense, energy efficiency (EE) and
spectral efficiency (SE) have recently gained considerable at-
tention from government, academia, and industry not only to
reduce the energy bills, but also to minimize the effect of global
warming and increase sustainable development. Optimizing SE
was the main concern for most of the research efforts and EE was
not regarded as a significant performance metric until recently.
However, these efficiency metrics must be considered jointly
instead of separately in the efficient design of mobile cellular
systems. Maximizing one metric (EE or SE) contradicts with
other one and hence, balance between them is necessary for
designing future wireless networks. In [5], fundamental trade-off
between SE and EE was introduced for green wireless networks
in AWGN without considering the fading channel effects. Power
versus bandwidth efficiency analysis has been done for a simple
interference model and considering Rayleigh fading channel
in [6]. For interference-limited wireless network, the EE and
978-1-5090-5538-8/17/$31.00 ©2017 IEEE
SE trade-off problem was studied in [7] with iterative power
allocation algorithm; also, further analysis has been done for
the same problem considering the shadowing and frequency
reuse factor effects in [8]. The average SE and EE between two
user equipments in cellular network have been obtained under
Rayleigh fading channel; moreover, the optimal transmission
power was also derived in [9]. These metrics have been also
investigated in [10] for wireless body sensor network considering
the interference effect and distance power control strategy was
also implemented to improve the spatial energy efficiency. D.
Tsilimantos et al. have introduced a simple theoretical framework
for analyzing the SE-EE trade-off problem in cellular networks.
In their paper [11], an optimal resource allocation approach and
traffic repartition scheme have been presented to reduce the
complexity trade-off problem. To the best of our knowledge,
the aforementioned efficiency metrics are not extensively inves-
tigated when the transmission environment are fully considered.
In this paper, a thorough efficiency study for a cellular network
is done under the effect of wireless channel, path loss, and
co-channel interference. The study includes investigation and
evaluation of spectral efficiency and energy efficiency for single
and multi cells wireless networks. The remainder of this paper
is organized as follows. The statistical models for system and
channel are presented in Section II. The signal to interference
plus noise is also introduced. Section III addresses definitions
and derivations of spectral and energy efficiency for single and
multi-cell systems. Numerical results based on system simulation
and theoretical derivations are then presented and discussed in
Section IV. Finally, the work is concluded in Section V.
II. C ELLULAR N ETWORK AND C HANNEL M ODELS
The negative effects of interference and channel impairments
are manifested as the multi-cell network becomes huge. A large
scale cellular network is considered here which consists of
homogeneous macro-cells with hexagonal tessellation as shown
in Fig.1a. The base stations (BSs) are located on the center
of each cell and the users are assumed to be served by the
closest BS. The system resources such as power and bandwidth
are equally assigned to the all cells, which makes the SE-EE
analysis valid for the whole cellular network. The downlink, BS
to users, analysis is our interest in this work. Three different
BS antenna configurations are implemented in our analysis:
omni, and 120o and 60o directional antenna. Fig.1 shows an
example of the hexagonal cellular network with the main network
parameters included in our analysis and simulation which are
the cell radius (R), the frequency reused distance (D), and the
interfering distance (Dj,i ). As we can see, the same frequency is
reused for more than one cell and hence, the network experiences
 2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE)

5 '

' L ' L XVHU .

5
K

.
U .

' '5VLQθ

5
LWK XVHU %6
' L
MWK %6
' L
G GU
U
L
XVHUL
5 UG
K

L
θ
' L ' SO
' L ' 5 3 3UU
L L L
θ

'

'

(a) Omni antenna (b) 120o directivity (c) 60o directivity (d) Single cell

Fig. 1: Single cell and Multi-cell hexagonal tessellation with network parameters for omni and directional antennas
co-channel interference problem and its effect on the SE and EE where r is the user’s distance away from the BS and θ is its
is one of our concerns in this paper. angle with respect to a certain reference. These two variables
are uniformly distributed and their commonly used probability
A. Propagation and Interference Models density function are given by
For the cellular network shown in Fig.1a, the transmission 2r
fr (r) =,0  r  R
power of j th BS is Pj . This Lets mobile user i be served by the R2
closet BS. The power received by the ith mobile user at distance 1
rj,i away from the j th BS can be written as fθ (θ) = , 0  θ  2π (5)
2π
Using this model the user’s distance can be estimated and hence
−α 2
Pj,i = Pj rj,i hj,i , (1) the path-loss part in (1) and (2) can be directly calculated.

where rj,i is a random distance between the j th BS and ith

C. Channel Model
served user, α is the path-loss exponent, and hj,i is a random
variable representing the channel fading effects of the user and Wireless channel introduces various impairments and effects
BS path. The interference received power that aggregated by the including small-scale fading, which is due to multi-path, that
same frequency BSs of the user ith can be expressed as cause serious degradation of signal to noise ratio (SNR) leading
to poor wireless communication system performance. Several

NI
statistical models are available to describe these effects. The
−α 2
Ii = Pk Dk,i hk,i , (2)
received power mentioned in (1) and (2) are randomly scaled by
k=1
a factor h2 which represents the multi-path channel effect. Many
where Dk,i is the k th BS and ith user interference distance, distributions have been derived to model the multi-path fading
and NI the number of co-channel interfering cells. The generic factor, h2 . In this paper, Nakagami-m fading channel model is
expression for the instantaneous signal to interference plus noise used to model h and the probability density function (pdf) of
(SINR) of the ith user is expressed as h2 ∼ Gamma (m, h) is given as
m−1  
−α 2
mm γj,i mγj,i
Pj rj,i hj,i fγj,i (γj,i ) = m exp − , γj,i  0 (6)
I
γj,i = N I , (3) γ j,i Γ(m) γ j,i
−α 2
No B + k=1 Pk Dk,i hk,i
k=j Pj,i h2j,i
where m is the distribution parameter, γj,i = No B is the
where No stands for power spectral density of AWGN, and B Pj,i h2 j,i
is the user’s bandwidth. instantaneous SNR, γ j,i = is the average SNR, and
No B
−α
Pj,i = Pj rj,i is the i user’s path loss received power at certain
th

B. User’s Location Model
It is assumed that the users are randomly distributed across
the cell; hence, it is important to model the user’s location.
Here, the BS coverage area is approximated as a circle shape
with radius R. The user’s location is defined using the polar
coordinates system as
L (r, θ) = rejθ
distance rj,i for the j th BS and hence the actual average SNR
is γ j,i r−α .
III. SE AND EE FOR S INGLE AND M ULTI -C ELL S CENARIOS
Bandwidth and power are the two main concern resources in
cellular planning. Since the demand on high data rate has been
rapidly increasing, spectral efficiency oriented cellular network
(4) design causes a huge increase in energy consumption. Next
 2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE)

generation cellular networks should be planned not only in
terms of spectral efficiency but also energy efficiency metric.
In this section, the downlink energy and spectral efficiencies are
studied considering multi-path fading, co-channel interference,
and frequency reuse factor for single and multi cells scenarios.
A. Analysis of Single Cell Scenario
In this part, our analysis is based on single cell scenario where
the BS is located at the cell center. The cell coverage area is
assumed a circle with radius R. The BS serves a set of K users
and its resources are equally shared among all the users. Thus,
each user occupies B bandwidth and receives Pi power. The
 ⎤
users are randomly located according to model (5).
2) Energy Efficiency: The energy efficiency for a system is
generally described as a ratio between what the system delivers to
what it consumes. Thus, the efficiency function can be defined
as f (x)/x where x ∈ [0, X] indicates the system’s resources
constrained by X [14]. In our analysis for a cellular system, f (x)
represents the downlink data rate, which is the main out come of
the BS, and x denotes the signal to noise ratio, which constitutes
the BS’s transmission power, noise and channel effects. For a
single cell, the downlink energy efficiency can be thus formulated
as [11]
⎡
Pipl h2i 

 ⎢ log2 1 + No B  2 ⎥
K

1) Spectral Efficiency: The downlink spectral efficiency

d
ηE = E⎣ hi ⎦ , [bit/J] (11)
Pipl h2i 
quantifies the amount of BS delivered data rate to the users i=1 No B
over the assigned bandwidth. In case of fading channel where
the random fading coefficients are independent, the aggregated The expectation operation in (10) can be done by taking the
downlink spectral efficiency can be obtained as [7] average using the pdf of the channel assuming that the set of h2i
are i.i.d random variables. Hence, with aid of [12, eq. (8.4.6/5),
K    
Pi h2i −α eq. (8.4.3/1), eq. (2.24.3/1)], the downlink energy efficiency of a
ηs =
d
E log2 1 + ri |h2i , (bit/s/Hz) (7) single cell system over Nakagami-m fading channel is obtained
N o B
i=1 in closed form as function of downlink average SNR, users’
where K is the total users’ number and the expectation, E, is distance, and fading parameter as
taken over the channel gain, h2j,i . In this scenario, there is only
one BS; hence, the j notation is not used in the analysis. The
downlink received power is Pipl = Pi ri−α and then γipl = γi ri−α .
Spectral efficiency can also be obtained using the averaging (m)
mr K α(m+1)

concept over the fading channel distribution as

d
ηE (ri , α) = i
×
Γ(m) log 2 i=1 γ m+1
i
 ∞   
mriα  −m, 1 − m
K
ηsd = [log2 (1 + γi ) |γi ] fγi (γi )dγi (8) 3,1
G2,3  (12)
0 i=1 γ i  0, −m, −m
By substituting the pdf of γi , (6), in (8) and using [12, where G[.] is the Meijer’s G-function [12]. Since ri is a random
eq. (3.383.10), eq. (3.326.2)], we have variable as modelled in (5), energy efficiency as function of cell
radius can be obtained by averaging (12) over the pdf of ri .
1   γm riα
K m
Using [13, eq.(07.34.21.0003.01)], it is derived in closed form
ηsd (ri , α) = e i ×
log 2 i=1 as
k=1
  k−m
(m) Rα(m+1)−1  1
m K
m α γ i −α
Γ −m + k, ri r (9) d
ηE (R, α) = ×
γi m i αΓ(m) log 2 i=1 γ m+1
i
+∞ β−1 −t   
mRα 
1
where Γ (β, x) = Γ(β) x
t e dt is the complementary 3,2 −(m + 1/α), −m, −m
incomplete gamma function.The spectral efficiency is a function G3,4  (13)
γ i  0, −m, −m, −(m + 1 + 1/α)
of average SNR, users’ distance, fading parameter m, and
the path loss. Since the user’s distance R is modelled in (5),
ri can be eliminated from (9) using 0 ηsd (ri )fri (ri ) dri . By B. Analysis of Multi Cell Scenarios
substituting (9) and (5) in to the above integration, the downlink
spectral efficiency is obtained in closed

form by expressing
 the
incomplete gamma function term Γ −m + k, γ ri using [13,
m α The coverage of single BS is constrained by the power
j,i limitations and hence multiple BSs are required for covering
eq.(06.06.03.0009.01)] as given a large area. In this part, we consider a regular hexagonal
topology as shown in Fig.1 which is a very useful choice among
2    (m − k)!
K m m−k
ηsd (R, α) = × other network configurations due to its well known properties.
α log 2 i=1 l! However, this system experiences several issues such as co-
k=1 l=1
 k−m−l channel interference which is our concern. Furthermore, our
γi interest is still focused on the downlink spectral and energy
Rα(m−k+l) (10)
m efficiency analysis considering the cellular system and signal
It is clear that the downlink spectral efficiency is a function of propagation issues. The instantaneous SINR is given by (3)
Pj,i h2
cell’s radius, path loss exponent, and signal to noise ratio. which No Bj,i is the instantaneous SNR. Correspondingly, the
 2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE)

average SINR can be expressed as 1) The user’s position is randomly picked according to (5).
h2
Pj j,i −α
2) The distance of co-channel interference is obtained
No B rj,i from Fig.1.
γ Ij,i = NI , (14) 3) The interference term Ξ and γ Ij,i are calculated by
Pk h2 k,i −α
1+ k=1 No B Dk,i (15), (16), (17) and (18) according to the BS antenna
k=j
configuration and directivity.
Assuming all BSs have the same power transmission, and h2k,i 4) The SE and EE are calculated using (9) and (13).
Pj h2 j,i P h2
and h2j,i ∼ Gamma (m, h), this implies that N oB
and kNo Bk,i 5) Assuming that the all BSs transmit the same amount of
−α −α power.
are equivalent. Also, it is clear that γ j,i ∝ rj,i and γ k,i ∝ Dk,i
and then the average SINR can be then rewritten as [6] The following system parameters are used in the evaluation:
−α
γ j,i rj,i
γ Ij,i = , (15) TABLE I: Simulation Parameters
1 + γ j,i Ξ(NI , Dk , ri , α) Symbol Description Valua
NI
Dk,i −α R Cell radius 1 Km
where Ξ = k=1 rj,i is the interference term which α Path loss exponent 4
k=j γ Signal to noise ratio 0 − 20 dB
depends only on the network geometry. Our investigations are
based on the highest co-channel interference power where the Rd Normalized reuse distance 2−6
users are located on the cell boundary; i.e., rj,i = R. For the NI Number of interfering cells 6
three proposed cases of cell antenna configuration, the average θ User’s angle 0 − 2π, 2π π
3 , 3
SINR is modelled as a function of normalized reuse distance, For single and multi cells networks, the SE and EE are plotted as
Rd = D/R. As shown in Fig.1a where omni directional antenna a function of SNR for omni antenna, and 120o and 60o directivity
is implemented, six cells in the first tire create co-channel antenna systems over several Nakagami fading channels in Fig.2
interference, and hence the interference, Ξ, is then presented and Fig.3, respectively.
as


NI /3
π −α 
P 2PQLPXOWL

Ξ (θ, Rd , α) =
o
{ Rd − sin(θ + (−1)n ) P R PXOWL
3n  P  R PXOWL
6SHFWUDO(IILFLHQF\ELWV+]

n=1

π  −α 
P 2PQLPXOWL
P R PXOWL
−α
+ Rd + sin(θ + (−1)n + (Rd + (−1)n sin(θ)) }  P  R PXOWL
3n P VLQJOH
(16)  P VLQJOH

where θ is the user’s angle which takes values 0 < θ < 2π.

For directional antenna on the BS, two cell sectorization are
considered: three-sector where 120o directivity is used and six- 

sector with 60o directivity as shown in Figs.1b and 1c. The 

interference term for both cases is expressed as 

    
o

π −α 615G%
Ξ120 (θ, Rd , α) = Rd + sin(θ + ) +
6 Fig. 2: SE of single and multi cells networks with different
−α 2π antenna directivities over fading channel as a function of SNR
(Rd + sin θ) ,0 < θ < (17)
3 


o

π −α
Ξ60 (θ, Rd , α) = Rd + sin(θ + )
4
(QHUJ\(IILFLHQF\ELW-

π
,0 < θ <
(18)
3 R P PXOWL
By knowing the average SINR for each antenna case, the spectral  R P PXOWL
2PQLP PXOWL
and energy efficiencies of multi-cellular system can be obtained R P PXOWL
using (9) and (12) where the users are randomly located on the  R P PXOWL
2PQLP PXOWL
cell’s boundary. P VLQJOH
 P VLQJOH

IV. S IMULATION AND N UMERICAL R ESULTS

    
In this part, we present several numerical and simulation
results to evaluate the SE and EE of the single and multi-cellular
networks. For single cell the SE and EE are evaluated using (10)
and (13). Also, the investigation is done for multi-cellular by
simulating (9) and (12). Our simulation algorithm is composed
of the following steps:
615G%
Fig. 3: EE of single and multi cells networks with different
antenna directivities over fading channel as a function of SNR
It is shown that SE is an increasing function of SNR whereas the
EE decreases monotonically as SNR increases. We also observe
 2017 IEEE 30th Canadian Conference on Electrical and Computer Engineering (CCECE)

that single cell is superior in terms of SE and EE than multi cells V. CONCLUSION
network which is referred to the negative effect of co-channel
This paper has presented a general framework for investigat-
interference. However, directional antenna technique improves
ing the SE and EE in the cellular networks. An evaluation based
the efficiency metrics. For instance, 60o directivity approaches
on theoretical derivations and simulation has been done for both
to the single cell performance, and 120o is far better than the
single a multi-cell systems where omni and sectorized antenna
omni antenna particularly at high SNR range. In contrast, the
systems are implemented on the BSs. Our study has considered
overall SE of multi cells is greater than single cell when it is
multi path fading, signal propagation path loss, and co-channel
calculated for the whole network; however, our investigation
interference. The downlink SE and EE have been derived and
concerns the effect of fading channel and co-channel interfer-
analyzed in terms of cell radius, reused frequency distance, path
ence. As aforementioned in (16), the interference Ξ is mainly
loss exponent, users’ location, fading parameter, and signal to
the function of Rd ; therefore, the introduced efficiency indexes
noise ratio. The results have shown that, based on the worst
are also examined in terms of Rd for fixed SNR value and
interference case, single cell achieves higher SE and EE than
different antenna configuration. Fig.4 illustrates the SE of multi
the multi-cell. BS antenna sectorization technique improves the
cells system for different Nakagami-m channels and antenna
overall efficiencies of cellular system; however, more BS antenna
directivities as a function of Rd . It is clear that SE improves as
system complexity is expected.

R EFERENCES

[1] D. Lister, Vodafone Group Research and Development, “An operators view
6SHFWUDO(IILFLHQF\ELWV+]

 on green radio, presented at the Proc. IEEE Int. Workshop on Green
Communications, 2009.
 [2] A. Fehske, G. Fettweis, J. Malmodin, G. Biczok, “The global footprint of
mobile communications: the ecological and economic perspective”. IEEE

2PQLP  Commun Magazine 49(8), 5562 (2011)
  R P  [3] Kilper, D.C.; Atkinson, G.; Korotky, S.K.; Goyal, S.; Vetter, P.; Suvakovic,
R P 
D.; Blume, O., “Power Trends in Communication Networks,” Selected
 2PQLP 
 R P 
Topics in Quantum Electronics, IEEE Journal of , vol.17, no.2, pp.275,284,
R P 
March-April 2011

[4] EIA, “Summer net internal demand, capacity resources, and capacity
 margins, 2012 actual, 2013-2017 projected ,” Energy Information Admin-
        
5G
istration: Official Energy Statistics from the U.S. Government, May 2015.
[5] Chen Yan, S. Zhang, S. Xu, and G. Y. Li, Fundamental trade-offs on green
Fig. 4: Spectral efficiency as a function of Rd for different wireless networks, IEEE Commun. Mag., vol. 49, no. 6, pp. 3037, 2011.
antenna configuration and fading channel [6] Akhtman, J.; Hanzo, L., “Power Versus Bandwidth-Efficiency in Wireless
Communications: The Economic Perspective,” Vehicular Technology Con-
Rd increases which means low co-channel interference impact ference Fall (VTC 2009-Fall), 2009 IEEE 70th , vol., no., pp.1,5, 20-23
Sept. 2009.
on the system. As can be seen, 60o achieves better SE than the
[7] Y. Li, M. Sheng, C. Yang and X. Wang, “Energy Efficiency and Spectral
120o and the omni antenna cells. Also, it is evident that fading Efficiency Trade-off in Interference-Limited Wireless Networks,” in IEEE
channel impacts the SE in particular at high Rd for omni antenna. Communications Letters, vol. 17, no. 10, pp. 1924-1927, October 2013.
Moreover, the EE is also plotted in terms of Rd in Fig.5. Indeed, [8] A. M. Alam, P. Mary, J. Y. Baudais and X. Lagrange, “Energy Efficiency-
it is obvious that EE for all different antenna configuration is Spectral Efficiency Tradeoff in Interference-Limited Wireless Networks
decreased as Rd increases which means more transmission power with Shadowing,” 2015 IEEE 82nd Vehicular Technology Conference
is required. It is also noted that EE for each system is becoming (VTC2015-Fall), Boston, MA, 2015, pp. 1-5.
close together as Rd increases, which refers to less effect of [9] L. Wei, R. Q. Hu, Y. Qian and G. Wu, ”Energy Efficiency and Spectrum
Efficiency of Multihop Device-to-Device Communications Underlaying
interference. Cellular Networks,” in IEEE Transactions on Vehicular Technology, vol.
65, no. 1, pp. 367-380, Jan. 2016.

 R P 
[10] Liu, Ruixia; Wang, Yinglong; Wu, Shangbin; Wang, Cheng-Xiang; Zhang,
  R P  Wensheng, “ Energy efficiency and area spectral efficiency trade-off for
 RPQLP  coexisting wireless body sensor networks,” Science China Information
R P  Sciences Journal on , vol.59, no.12, pp.193-196, December 2016.

(QHUJ\(IILFLHQF\ELW-

 R P 
 2PQLP  [11] D. Tsilimantos, J. M. Gorce, K. Jaffrs-Runser and H. Vincent Poor,
“Spectral and Energy Efficiency Trade-offs in Cellular Networks,” in IEEE
 Transactions on Wireless Communications, vol. 15, no. 1, pp. 54-66, Jan.
2016.
 [12] A. P. Prudnikov, Yu. A. Brychkov, and O. I. Marichev, Integrals and
series. Vol. 3, Gordon and Breach Science Publishers, New York, 1990.
More special functions; Translated from the Russian by G. G. Gould. MR
 1054647 (91c:33001)
[13] “The Wolfram functions site for special mathematical functions.” Wolfram
         Research, Inc 2016. Available: http://functions.wolfram.com
5G [14] E. V. Belmega, S. Lasaulce and M. Debbah, “A survey on energy-efficient
communications,” 2010 IEEE 21st International Symposium on Personal,
Fig. 5: Energy efficiency as a function of Rd for different Indoor and Mobile Radio Communications Workshops, Instanbul, 2010,
antenna configuration and fading channel pp. 289-294.